Reliability Design

Download as pdf or txt
Download as pdf or txt
You are on page 1of 107

NTNU

Norwegian University of
Science and Technology
Faculty of Engineering and Technology
Department of Production and Quality
Engineering

Masters
thesis

Student Ingrid Alms Berg

Design for Reliability


Applied to development
of subsea process
systems
Trondheim, June 14th, 2010

o t\Tt\u

1 o f3
Date
2010-01-11

Ourreference
MAR/LMS

Facultyof Engineering
Scienceand Technology
Department
of Productionand QualityEngineering

MASTER THESIS
2010
for
stud.techn.Ingrid Almis Berg

DESIGN FOR RELIABILITY. APPLIED TO DEVELOPMENT OF SUBSEA


PROCESSSYSTBMS
(Pilitelig design- anvendtpfl undervannsprosess-systemer)
The subsea power and process department at Aker Solutions is working to create innovative
processingsolutions within the subseaoil and gas industry. Reliability is one of the single most
important factors in the operation of subsea systems. It is therefore vital to address product
performance throughout the whole lifecycle of the equipment, from front-end design trough
installationand operation.
The main objective of this master thesis is to considerrelevant methodologiesand practisesfor
reliable productdesign;discussthe applicationof thesein the industry in generaland for the subsea
processindustry in particular.The thesis shall describethe methodsfor reliable designthroughout
the lifecycle of a product/system and develop an appropriate methodology for a subseaprocess
system.The identified methodologyshall then be appliedto the developmentof a specific process
sub-system.
The main objectiveis further divided into the following more specificsub-targetsor objectives:
l. Perform a literature study on the application of different methods for design for reliability.
The literature study should cover, and discuss,methods normally applied acrossrelevant
industries.
2. Familiarisewith systemsand main equipmentusedwithin subseaprocessindustry.
3. Based on learning from the literature study discuss and summarisethe main factors that
contributeto unreliability.Describechallengesfor the subseaprocessindustryin particular.

2of3
Master Thesis 2OlOior stud. techn.lngrid Berg AlmCs

Date

Our re{erence

2010-01-lI

MAR/LMS

4. Suggestand develop a methodologyfor reliability performanceand specificationthroughout


five defined stagesof a ProductLife Cycle (Front-end,Design,Development,Productionand
Post-Production)of a typical SubseaProcessSystem.Specialattentionshouldbe paid to the 3
first phases.
5. Perform a casestudy to evaluatethe applicabilityof the suggestedmethodologyfor a chosen
stage of the life cycle. The case study should be applied to a chosensub-systemwithin a
typical subseaprocesssystem.
Following agreementwith the supervisorand the responsibleprofessor,the various tasks may be
given different weights.
Within tree weeks after the date of the task handout, a pre-studyreport shall be prepared.The report
shallcoverthe following:
o
o
o

An analysis of the work task's content with specific emphasisof the areas where new
knowledgehasto be gained.
A description of the work packagesthat shall be performed. This description shall lead to a
clear definition of the scopeand extentof the total taskto be performed.
A time schedulefor the project. The plan shall comprise a Gantt diagram with specification
of the individual work packages,their scheduledstart and end datesand a specificationof
project milestones.

The pre-studyreport is a part of the total task reporting. It shall be included in the final report.
Progressreportsmadeduring the project period shall also be included in the final report.
The reportshouldbe editedas a researchreport with a summary,table of contents,conclusion,list of
reference,list of literature etc. The text should be clear and concise,and include the necessary
referencesto figures,tables,and diagrams. It is also importantthat exactreferencesare given to any
externalsourceusedin the text.
Equipmentand softwaredevelopedduring the project is a part of the fulfilment of the task. Unless
outsidepartieshaveexclusivepropertyrights oittti equipmlnt is physicallynon-moveable,it should
be handedin along with the final report.Suitabledocumentationfor the correctuse of suchmaterial
is alsorequiredaspart of the final report.

and copyingunlessotherwiseagreed.
The studentmust cover travel expenses,telecommunication,
If the candidateencountersunforeseendifficulties in the work, and if these difficulties warrant a
to the Department.
reformationof the task,theseproblemsshouldimmediatelybe addressed

\
/

3of3

MasterThesis2010for stud. techn.lngrid BergAlmAs

Ourreference
Date
2 0 1 0 - 0 1 - 1 1MAR/LMS

if these difficulties warrant a


If the candidate encountersunforeseendifficulties in the work, and
to the Department'
reformulation of the task, theseproblems should immediately be addressed
Deadline:June 14fr 2010.
required'
Two bound copies of the final report and one electronic (CD) version are

at NTNU:
professor/Supervisor
Responsible

ProfessorMarvin Rausand
Telephone:73 59 25 42
Mobile phone: 456 66265
@ntnu.no
E-mail: marvin.rausand

Supervisorat Aker Solutions:

Guro Rausand
SeniorSubseaEngineer
Power,Processing& Boosting
67 527417
Telephone:
Mobile phone:47234 442
@akersolutions.com
E-mail: guro.rausand

DEPARTMENT OF PRODUCTION
AND QUALITY ENGINEERING

/"gt-,

PerSchjPISrg . I
Associ ateProfessor/lleadof De\f,artment

,rr- r--rry
Marvin Raorun/
ResponsibleProfessor

NTNU

Design for Reliability

14.06.2010

Preface
This master thesis is written at the Department of Production and Quality Engineering, NTNU, during
the spring of 2010. The thesis is a part of the 5th year of the master program in Product Design and
Manufacturing.
The title of the thesis is Design for Reliability Applied to development of subsea process systems
and it is carried out in cooperation with the Subsea Power and Process department at Aker Solutions. It
is assumed that the reader has some prior knowledge of reliability analysis and probability, preferably
through the course TPK4120 Safety and Reliability Analysis. If a glossary is needed, it is suggested
that the reader uses System Reliability Theory (Rausand and Hyland 2004). The report should be
read according to the structure described in chapter 1.3, especially where appendices are concerned.
I would like to thank my supervisor Prof. Marvin Rausand and my co-supervisor Prof. Mary Ann
Lundteigen at NTNU, and my supervisor at Aker Solutions, Guro Rausand, for guidance,
contributions and advices throughout the semester. I also wish to thank the employees at the Subsea
Power and Process department for their contributions and support during my stay at Aker Solutions.
Trondheim, June 11th 2010.

Ingrid Alms Berg

Master Thesis, TPK 4900


Ingrid Alms Berg

NTNU

Design for Reliability

14.06.2010

Abstract
New products arrive on the market every day. Whether it is a computer, a cellular phone or a
car, the consumers are attacked by commercials enticing them to try the newest invention.
Getting a customer to buy a product may be a challenge, but trying to keep him or her
satisfied can be far more difficult. The regular western consumer is often well aware of what
he or she demands of a product. Depending on the product, a safe failure can be accepted with
a swift repair. What is not accepted is for a cellular phone to fail three times in seven months.
Every product has an expected reliability. If the product cannot answer to this reliability, the
manufacturer will be forced to repair products, answer to warranty claims and possibly suffer
a large economic loss.
Some industries demand very high reliability of the systems they use, especially as the
systems are expensive and failures could cause extreme harm. A countrys authorities have
regulations for most industries and especially strict regulations for high risk industries. An
example of a high risk industry is the petroleum industry. A normal requirement for new
subsea process systems in this industry is an availability of 97%. To achieve such a high
number, the reliability must be high as well. When a new Subsea Compression System (SCS)
is built for Norwegian oil and gas production, the laws and regulations of the Norwegian
authorities must be followed. These concern the maintainability and reliability of the system,
as well as the safety of the system, the environment and the employees of the operator.
Reliability is an increasingly popular subject to consider for manufacturers and authorities
around the world. A problem is that there are many different thoughts on how reliability can
be achieved. To ensure a common understanding between authorities, manufacturer and
client, several standards concerning reliability have been developed. These are often specific
to an industry and one example is ISO 14224; Petroleum and natural gas industries
Collection and exchange of reliability and maintenance data for equipment.
With a large number of specified reliability standards, one should think that considering the
product reliability throughout the product life cycle was normal. This is not necessarily the
case. In the early parts of the product life cycle, before the physical development commences,
many still believe that reliability activities are a waste of time, resources and money. What
they do not consider is the fact that alterations are easier to perform before a product is
manufactured than after.
A product life cycle is usually split into phases. For reliability activities in the product life
cycle, Murthy et al. (2008) suggests eight phases. The first three occur during the predevelopment, the two following take place during production, while the last three are part of
the post-development. The events taking place in these phases and the main tasks of a
reliability engineer are well described, and it is thus a good option for Design for Reliability.
Reliability engineering has existed as an engineering discipline for several decades now.
Especially the nuclear and the aerospace industries have studied and developed methods for
reliability. The methods can help detect and evaluate possible hazards and failure modes
experienced by the system or product during its operational life. Some of the methods study
how and why the hazards may occur, while other studies how a failure affects the overall
system. Through probability estimates, the reliability of a system is found. Examples of such
methods are HAZID, FMEA and FTA.

Master Thesis, TPK 4900


Ingrid Alms Berg

ii

NTNU

Design for Reliability

14.06.2010

Even though we have methods for reliability straight in front of our nose, many are unaware
of why they should be used. A system will never be 100% reliable, because of all the factors
contributing to a reduced reliability. In all eight life cycle phases such factors can be found.
Whether the manufacturer is in the car industry or the petroleum industry, many of the factors
will be the same. One important issue is uncertainty, especially of the epistemic type. This is
the uncertainty which we cannot know is there. It is hard to say whether the communication is
good enough or if the inputs to the reliability methods are acceptable. Another factor is the
human being who is unpredictable and thus unreliable. He or she partakes in every step of the
product development and is thus probable to contribute to the decreased reliability. However,
we cannot let the fear for such factors leading to less reliable products keep us from
developing something new. By being aware of them, we can use the factors to employ
reliability methods and thus increase the possibility that the methods are used correctly.
In using several methods to study the same product or system, we are more likely to get a full
picture of the hazards, failures and overall product reliability. One type of analysis including
several methods is the RAM analysis. This studies availability and maintainability together
with reliability. The three disciplines are highly connected and equally important for the
product performance. Although the RAM analysis is a very efficient tool, it cannot stand
alone. It is necessary to use other methods prior to it to obtain the input, as well as some
methods afterwards which can make use of the outputs.
Placing a reliability method at a random time during the design and development phases is not
considered good utilization of the method. It must be used when the necessary information is
available and the output can be of use. To combine and place the methods correctly, a
methodology has been developed. This is described at a level meant for entirely new products,
whether they are standard products or one-of-a-kind. The methodology is very general, but for
industries where the reliability is very important, they should be specified. The specification
can be according to the industry, but it would be even better according to organisation. A
specified example has been prepared for a subsea organisation. The purpose of both the
general and the specified methodology is to use them as a basis in the development of
reliability programmes.
Reliability programmes are established specifically for one product development project,
often for one phase at the time. It states why a method is chosen, when it should be used and
the responsibilities where the product reliability is concerned. It should be based on the
project risk, the project tasks and the available time and resources. Using a reliability
methodology to choose the methods and their combination will simplify the programme
development and ensure its quality.
For manufacturers who do not understand why reliability should be included in the design
process, or how this can be done, a website is useful. The internet is highly accessible, easy to
use and not time consuming. If a website for Design for Reliability first is developed, it can be
used as an educational tool by and for reliability engineers. It can also help in the
development of more specific methodologies and reliability programmes.
Reliability methods, methodologies and programmes will train a manufacturing organisation
in thinking differently when they develop new systems and products. The outcome of the
development will be products which are reliable, safe and functioning as they are supposed to.
This will again lead to more satisfied customers. No customer will accept that reliability
activities were left out of the product development when they stand with a failed product in
their hands.
Master Thesis, TPK 4900
Ingrid Alms Berg

iii

NTNU

Design for Reliability

14.06.2010

List of contents
Preface ......................................................................................................................................... i
Abstract ...................................................................................................................................... ii
List of contents .......................................................................................................................... iv
List of tables .............................................................................................................................. vi
List of figures ............................................................................................................................ vi
1.

Introduction ........................................................................................................................ 1
1.1. Objective ...................................................................................................................... 2
1.2. Scope and limitations................................................................................................... 2
1.3. Structure....................................................................................................................... 3

2.

Reliability ........................................................................................................................... 4
2.1. RAMS .......................................................................................................................... 5
2.2. Probability and reliability ............................................................................................ 7
2.3. Discussion .................................................................................................................. 11

3.

Subsea industry requirements........................................................................................... 12


3.1. Petroleum Safety Authority ....................................................................................... 12
3.2. Standards ................................................................................................................... 14
3.3. Further comments ...................................................................................................... 19

4.

Product life cycle .............................................................................................................. 21


4.1. Phases in a general perspective ................................................................................. 21
4.2. 8 Phases for reliability ............................................................................................... 23
4.3. Discussion .................................................................................................................. 26
5. Methods for reliability ...................................................................................................... 27
5.1. PHA and HAZID ....................................................................................................... 29
5.2. SWIFT ....................................................................................................................... 30
5.3. FMEA ........................................................................................................................ 31
5.4. FTA ............................................................................................................................ 33
5.5. RBD ........................................................................................................................... 34
5.6. HAZOP ...................................................................................................................... 36
5.7. Reliability Growth ..................................................................................................... 38
5.8. Accelerated testing .................................................................................................... 40
5.9. FRACAS .................................................................................................................... 42
5.10.
Discussion .............................................................................................................. 43

6.

RAM analysis ................................................................................................................... 44


6.1. Method ....................................................................................................................... 44
6.2. MIRIAM Regina........................................................................................................ 45
6.3. Use of outputs ............................................................................................................ 46

7.

Unreliability ..................................................................................................................... 48
Phase 1.............................................................................................................................. 49
Phase 2.............................................................................................................................. 51
Phase 3.............................................................................................................................. 53
Phase 4.............................................................................................................................. 54
Phase 5.............................................................................................................................. 56

Master Thesis, TPK 4900


Ingrid Alms Berg

iv

NTNU

Design for Reliability

14.06.2010

Phase 6.............................................................................................................................. 56
Phase 7.............................................................................................................................. 57
Phase 8.............................................................................................................................. 59
7.1. Discussion .................................................................................................................. 60
8.

Methodology for reliability in the product life cycle ....................................................... 62


8.1. Methodology in phases .............................................................................................. 62
8.2. Comparisons and discussion ...................................................................................... 77

9.

Case study ERMP.......................................................................................................... 79


9.1. ISO 20815, the methodology and the ERMP ............................................................ 79
9.2. Discussion .................................................................................................................. 82
10.
Design for reliability website ........................................................................................ 83
10.1
Further suggestions ................................................................................................ 85
11.
Concluding remarks ...................................................................................................... 87
11.1
Methodology and use ............................................................................................. 88
11.2
Recommendations for further work ....................................................................... 88

12.

References ..................................................................................................................... 90

Appendices ............................................................................................................................... 96

Master Thesis, TPK 4900


Ingrid Alms Berg

NTNU

Design for Reliability

14.06.2010

List of tables
Table 1: Standards used in phases ............................................................................................ 20
Table 2: Project risk categorisation (ISO 20815 2008) ............................................................ 80

List of figures
Figure 1: The bathtub curve ....................................................................................................... 8
Figure 2: The repetition in a frequentist approach experiment .................................................. 9
Figure 3: Experiment with previous knowledge in the Bayesian approach ............................. 10
Figure 4: The data collecting process (ISO 14224 2006)......................................................... 16
Figure 5: Design review process (IEC 61160 2006) ................................................................ 18
Figure 6: Design review and response process (IEC 61160 2006) .......................................... 18
Figure 7: Phases, Levels and Stages (Murthy et al. 2008) ....................................................... 24
Figure 8: Connection between life cycle phases (Murthy et al. 2008) ..................................... 25
Figure 9: SADT model (Rausand & Hyland 2004)................................................................ 28
Figure 10: SADT PHA/HAZID ............................................................................................... 30
Figure 11: SADT SWIFT ......................................................................................................... 31
Figure 12: SADT FMEA .......................................................................................................... 32
Figure 13: Fault Tree Process (NASA (A) 1999) .................................................................... 33
Figure 14: SADT FTA ............................................................................................................. 34
Figure 15: RBD ........................................................................................................................ 35
Figure 16: SADT RBD ............................................................................................................. 36
Figure 17: HAZOP process (IEC 61882) ................................................................................. 37
Figure 18: SADT HAZOP........................................................................................................ 38
Figure 19: Reliability Growth process ..................................................................................... 39
Figure 20: SADT Reliability Growth ....................................................................................... 40
Figure 21: SADT Accelerated testing ...................................................................................... 41
Figure 22: SADT FRACAS ..................................................................................................... 42
Figure 23: Unreliability - reliability ......................................................................................... 48
Figure 24: The whispering game .............................................................................................. 50
Figure 25: Baseball................................................................................................................... 61
Figure 26: Reliability methodology phase 1 ............................................................................ 63
Figure 27: Reliability methodology phase 2 ............................................................................ 65
Figure 28: Reliability methodology phase 3 ............................................................................ 67
Figure 29: Reliability methodology phase 4 ............................................................................ 69
Figure 30: Reliability methodology phase 5 ............................................................................ 71
Figure 31: Reliability methodology phase 6 ............................................................................ 73
Figure 32: Reliability methodology phase 7 ............................................................................ 75
Figure 33: Reliability methodology phase 8 ............................................................................ 76
Figure 34: Front page of website for Design for reliability (http://folk.ntnu.no/ingribe) ........ 83
Figure 35: Life cycle (http://folk.ntnu.no/ingribe/productlifecycle.jsp) .................................. 84
Figure 36: Phase 7 as opened from figure 35. .......................................................................... 85
Figure 37: Failure-reliability diagram ...................................................................................... 87

Master Thesis, TPK 4900


Ingrid Alms Berg

vi

NTNU

Design for Reliability

14.06.2010

1. Introduction
On April 20th this year, a blowout occurred while the Deepwater Horizon oil rig was drilling in the
Macondo Prospect oil field in the Gulf of Mexico. This led to an explosion which sunk the rig into the
sea, killed several crewmembers and gave way to an oil leak which has yet to be stopped six weeks
later (Cleveland 2010). The accident has spurred heavy criticism against the petroleum industry,
particularly as the accident is only one out of many. Today, the worlds population does not know how
to live without oil and gas. Until new technology provides acceptable amounts of energy, or the oil
and gas is spent, the petroleum industry will survive. We can only demand that the systems and
technology used are as safe and reliable as possible.
Failures occur in every industry and it seems we cannot avoid them. A customer wants the product to
perform as expected, being both reliable and safe. Minor failures are acceptable, but many and
dangerous failures are insupportable. Manufacturers loose money, clients and credibility if their
products fail prematurely. One solution is Design for Reliability, a discipline in which reliability
activities are performed concurrently with the design. It is believed that shortcomings in the design
affect the whole product (OConnor 2002). The feedback from reliability analyses are thus expected to
help the designers develop the most reliable system possible.
Design is a creative process, reliability analyses are not. Reliability activities are dependent on timing
and strict rules. Some believe that reliability activities in the design phase are a waste of money.
Testing of the materials and prototypes should be enough. However, this is likely to be a more
expensive approach if the design needs alterations. To alter the design after the physical development
is likely to cost more, both in time and money, than during the design process.
An example of a manufacturer who employs Design for Reliability in the product development is Aker
Solutions. This company is one of several suppliers to the petroleum industry. Currently their Subsea
Power and Process department is preparing a Subsea Compressions System (SCS) project for the
Midgard field on the Norwegian continental shelf. The pressure in wells decreases as oil and gas flow
out. To keep the flow rate above a critical minimum and maximise reservoir recovery, the SCS is a
solution. Both Norwegian authorities and the client demand high reliability from subsea process
systems. A failure in an SCS will not lead to a blowout, but possibly a loss in production or a minor
leakage. The former will affect the income of the operator. The latter is likely to cause environmental
damage. To prove to the client that Aker Solutions can fulfil the requirements, they need a reliability
programme for the design and development. Similar reliability programmes are now also demanded in
other industries such as the nuclear, the aerospace and the military industry.
Many methods for reliability have been developed throughout the years. They demand different inputs
and provide the analyst with several types of outputs. If used correctly, they will give valuable input to
the development and use of the products. To choose the best combination and timing of the methods, it
is necessary to know the factors contributing to unreliability and how the methods can prevent them. It
is of interest to establish a methodology based on the methods and the product life cycle. Can a
general reliability methodology be a basis for specified reliability programmes? If a methodology
simplifies the establishment of reliability programmes, can this be used to convince those who do not
use reliability activities during the design phase? Can an internet site on the subject be of help and can
it become a tool for reliability engineers in different industries?

Master Thesis, TPK 4900


Ingrid Alms Berg

NTNU

Design for Reliability

14.06.2010

1.1. Objective
The main objective of this master thesis is to study reliability methods and the factors contributing to
unreliability, and use this to develop a methodology for Design for Reliability. Further, the
methodology shall be used to develop a reliability programme applied for a specific phase of the
product life cycle of a Subsea Compression System developed by the Subsea Power and Process
department at Aker Solutions.
The objective is further divided into five main tasks:
1. How are the different methods used in Design for Reliability applied across the industries?
2. Prepare a system description of a system used within the subsea process industry.
3. Discuss and summarise the main factors that contribute to unreliability, specifically challenges for
the subsea process industry.
4. Develop and describe a methodology for reliability performance and specification throughout the
five defined stages of a Product Life Cycle of a typical subsea process system.
5. Perform a case study to evaluate the applicability of the suggested methodology for a chosen stage
of the life cycle, applied to a chosen sub-system within a subsea process system.

1.2. Scope and limitations


The thesis assignment was given by the Subsea Power and Process department at Aker Solutions, who
specialises on the subsea process industry. It was demanded that reliability in product design was
applied to the industry in general and further for the subsea process industry in particular. The thesis is
thus written with all industries in mind, while using the subsea process industry to exemplify the
further use of the reliability methodology, unreliability issues and in the case study.
The thesis was carried out during a project development of a Subsea Compression System for the
Midgard field. This field lies on the Norwegian continental shelf and is operated by Statoil. Any
standards and regulations used as a basis for reliability methods in this thesis are thus restricted to
those required by Statoil and followed by the Aker Solutions Subsea Power and Process department
for this specific project. This implies that only Norwegian laws and regulations are described.
The study of systems and equipment used within the subsea process industry has mainly been limited
to the items of interest to the Subsea Compression System. The more common types of equipment
such as X-mas trees, well-heads etc. have thus not been in the focus of this thesis.
The product life cycle is normally studied through phases, but the number of phases varies. Task 4 in
the assignment states that five phases; Front-End, Design, Development, Production and PostProduction are to be followed. Murthy et al. (2008) looks specifically into the subject of reliability in
the product phases and suggests that a framework with eight phases is more suitable. This gives
smaller gaps between the start and end of each phase and specifies whether the system is studied on a
component, product or business level. It is thus easier to evaluate the answers needed for reliability
purposes in each phase. After an agreement with the supervisors, it was decided that these eight phases
should be the basis of the methodology.
As stated in the preface to this thesis, the reader is assumed to have some prior knowledge of the
subjects reliability, statistics and probability. This limits the basic explanation of ordinary probability
distributions and how each reliability method is performed. The assumption also helps limit the time
spent before the main subjects of the report are addressed.

Master Thesis, TPK 4900


Ingrid Alms Berg

NTNU

Design for Reliability

14.06.2010

1.3. Structure
The first chapters study background issues for the methodology, beginning with an introduction to
reliability and related concepts in chapter 2. Chapter 3 looks at the subsea process industry
requirements concerned with reliability on the Norwegian continental shelf, and chapter 4 describes
the life cycle phases of relevance to the report.
The subsequent chapters are generally following the sub-objectives described in 1.1. Chapter 5 looks
into several methods created and used for reliability purposes. The methods are considered through the
use of the Structured Analysis and Design Technique (SADT), which is also presented in this chapter.
Chapter 6 describes the Reliability, Availability and Maintainability analysis (RAM analysis), which is
a collection of reliability methods and simulation tools. This type of analysis is often used for
development projects across industries and gives a good overview of a products availability and
reliability.
Factors contributing to unreliability are discussed in chapter 7. These are assumed to reduce the
possible reliability of the product and may arise in all phases of the life cycle. The subsea process
industry has been studied as an example in relation to the factors. The eight phases suggested by
Murthy et al. (2008) are used as the background for the discussions.
A general methodology for Design for Reliability is described for each of the eight phases in chapter
8. Methodology here refers to a simple set of methods related to other development activities. No time
frame is set, nor is any product specified. One figure for each phase is discussed and the main
reliability tasks summarised. The general idea is that these tasks are managed by one or more
reliability engineers, participating in the development together with other team members responsible
for design, economy etc. A methodology has been developed for Aker Solutions to show how a more
specific methodology for an organisation or industry can be prepared. This is based on the general
methodology and Aker Solutions Technical Qualification model. The Technical Qualification can be
found in appendix B and the Aker Solutions methodology in appendix C. These are placed in the
appendices for confidentiality reasons, but should be studied together with chapter 8.
Chapter 9 describes the case study, an Equipment Reliability Management Programme (ERMP)
developed for the Engineering, Procurement and Construction (EPC) phase of the Midgard SCS
project. The ERMP is based on ISO 20815 and the Aker Solutions project execution model (PEM),
found in appendix D. Appendix E contains the ERMP as it is presented to Statoil.
A website has been made in connection with this thesis, as a suggestion for how Design for Reliability
and reliability methodologies can be presented on new arenas. The website can be found on
http://folk.ntnu.no/ingribe and it is presented in chapter 10. A discussion on how it may be further
developed into a tool for reliability engineers is also made.
The main questions for this thesis were how Design for Reliability can be used in the product life
cycle through a methodology, and how a methodology can be specifically applied to several different
projects. The results of the thesis are discussed, together with other concluding remarks, in chapter 11.
Appendix A contains a system description of the Midgard SCS which can be used in order to
understand how the design has been developed prior to the start-up of the EPC phase. This is also a
part of the answer to sub-target 2 described in chapter 1.1.

Master Thesis, TPK 4900


Ingrid Alms Berg

NTNU

Design for Reliability

14.06.2010

2. Reliability
The English word reliability is believed coined by the English author Samuel Taylor Coleridge in
1816. As quoted by Selah and Marais (2006), this phrase is the eldest recorded use of the word:
He inflicts none of those small pains and discomforts which irregular men scatter about them and
which in the aggregate so often become formidable obstacles both to happiness and utility; while on
the contrary he bestows all the pleasures, and inspires all that ease of mind on those around him or
connected with him, with perfect consistency, and (if such a word might be framed) absolute
reliability.
Today, reliability is used to describe persons and objects, although the latter might be the more
popular. Most people have an understanding of what the term means and how to use it. A normal
customer expects a TV to function longer than five years in a normal living room. The same customer
would also understand that the TV cannot be expected to work for long outside in -50C and rain. If
the TV fails in the first case, the customer would complain about its reliability. A failure in the second
case would normally not lead to a complaint. Technical reliability is defined as (IEC 60050-191 1990):
The ability of an item to perform a required function under given conditions for a given time interval.
A customer is likely to base his or her possibility for complaining about the failure of the TV on this.
What the customer might not understand is how high product reliability is achieved. When new
technology is developed, it is its ability to function as required which is the point of departure. If it is
unlikely that the technology can perform as desired, there is little reason to consider under which
conditions and what amount of time it must function. Reliability is an engineering discipline, using
specifically developed standards and tools. Its objective is to help the design engineers and
maintenance organisations achieve and sustain reliable and safe systems. IEEE describes the discipline
in the following manner (IEEE 2009):
Reliability is a design engineering discipline which applies scientific knowledge to assure a product
will perform its intended function for the required duration within a given environment. This includes
designing in the ability to maintain, test, and support the product throughout its total life cycle.
Reliability is best described as product performance over time. This is accomplished concurrently with
other design disciplines by contributing to the selection of the system architecture, materials,
processes, and components -- both software and hardware; followed by verifying the selections made
by thorough analysis and test.
As IEC 60050-191 (1990) suggests, reliability is mainly concerned with a required function. An mp3player which plays music is supposedly answering the definition. If the same mp3-player is unable to
switch songs according to its owners wishes, is it still reliable? This depends on how the required
function is defined. A required function is (IEC 60050-191):
A function or a combination of functions of an item which is considered necessary to provide a given
service.
If the required function of the mp3-player includes the ability to switch a track according to its
owners wishes, alongside the ability to play music, it must do so. An important step in the
development of a reliable product is thus to make a thorough definition of the required function.

Master Thesis, TPK 4900


Ingrid Alms Berg

NTNU

Design for Reliability

14.06.2010

As explained by IEEE, reliability engineering also includes a consideration of maintenance. This hints
towards an umbrella discipline called Reliability, Availability, Maintainability and Safety (RAMS). It
suggests that these four disciplines are connected and affect one another. A product with high
reliability, based on a well defined required function will not only satisfy its user on a level of
functionality, it will also be safe and able to operate whenever demanded. All provided that the
operational conditions are acceptable.

2.1. RAMS
Reliability, availability, maintainability and safety are interrelated disciplines; they may all be
improved if one of them is improved. If focus is set on all four disciplines during design and
development, the reputation of the manufacturer is unlikely to be damaged by the product.
Requirements to the product can be given with RAMS as a basis. Such requirements can be split into
four categories (Lundteigen et al. 2009):
-

Functional safety and safety integrity requirements.


Product safety requirements.
Operational availability requirements.
Maintainability and maintenance support requirements.

To understand these types of requirements, reliability, availability, maintainability and safety must be
understood. Reliability was introduced in the previous section and will here only be discussed through
the other three subjects in the RAMS discipline.

Availability
IEC 60050-191 (IEC 60050-191) defines availability as:
The ability of an item to be in a state to perform a required function under given conditions at a given
instant of time or over a given time interval, assuming that the required external resources are
provided.
This definition focuses on repairable systems in the operational phase, stating that they are available
when they function or are able to operate. Any downtime due to failure, maintenance or repair will
contribute to a lower availability. Availability is the probability that the product will be in operation at
a given time (Rausand and Hyland 2004). Mathematically, average availability is defined as the ratio
between the uptime and the total time the product is meant to be in operation.

Aavg =

uptime
time in operation

If reliability is described as Mean time to failure (MTTF), availability is MTTF divided by the sum
of MTTF and the Mean time to repair (MTTR). MTTF and MTTR are defined mathematically in
Rausand and Hyland (2004).

A=

Master Thesis, TPK 4900


Ingrid Alms Berg

MTTF
MTTF + MTTR

NTNU

Design for Reliability

14.06.2010

This formula indicates that the availability of a product increases proportionally with the reliability. A
customer is interested in having the highest availability possible. If the downtime after a failure is very
long, the equation shows that the reliability needs to be high for the availability to stay high. This does
not mean that a product with low downtime can be produced without a thought to reliability, although
this might give an acceptable availability. A very low reliability would mean frequent failures, which
might even be more frustrating to the customer than the long downtime. Whether the availability is
high due to high reliability alone or in combination with a low downtime, it is evident that the
reliability is the key factor. It should therefore be in every manufacturers best interest to design for
reliability.

Maintainability
Maintainability is the ability of an item under given conditions of use, to be retained in, or restored to,
a state in which it can perform a required function, when maintenance is performed under given
conditions and using stated procedures and resources (IEC 60050-191 1990).
Maintainability combined with reliability determines the product availability. The maintainability is a
design feature that describes how easily a product or system can be maintained or repaired (IEC 60300
2003). For example, a subsea installation is placed on the seabed and it is only accessible through the
use of specific intervention vessels equipped with tools such as Remotely Operated Vehicle (ROV).
When a subsea process system fails, the simplicity of the repair operation depends on:
-

Simple interfaces to the system


Easily retrievable modules
Available spare parts

If the reliability of a product has been considered from the early life cycle phases, it is known how it
might fail. This information has probably also been used to establish an appropriate spare part
philosophy based on which subsystems generally need the most maintenance. The interfaces are the
keys to an easy system repair. The easier it is to reach the system, the easier it is to repair. The ability
to retrieve a module which cannot be repaired sub sea or must be replaced is also important for the
time to repair. Finally, the available spare parts decide how fast a repair may begin. If the part is
considered very important for the availability, it is possible that a spare part is placed on the seabed for
a lower repair time.
Good maintainability increases the availability and decreases the chances of leaving systems in a
failed condition for a long time. An unrepaired failure may not only lead to a long system downtime,
but it can also lead to problems with the safety of the system

Safety
IEC 61508 (2005) defines safety as:
Freedom from unacceptable risk.
When a reliability analysis is performed, it considers the probability of the occurrence of a hazardous
event leading to a function failure. The safety analysis will include this step and can therefore be
performed as a continuation of the reliability analysis. A reliability analysis evaluates the possibility of
failure and how to keep it from occurring. The safety analysis looks into the effects of a failure and

Master Thesis, TPK 4900


Ingrid Alms Berg

NTNU

Design for Reliability

14.06.2010

how to mitigate them. Together the two will give a full picture of how the initiation, development and
consequence of a failure can be avoided.
The reliability, availability, maintainability and safety of an undeveloped product can only be
considered through estimates. Based on previous knowledge, the probability that an event takes place
can be discovered and used in the design process.

2.2. Probability and reliability


During a lecture held in 1929, Bertrand Russell said (Der Kiureghian and Ditlevsen 2009):
Probability is the most important concept in modern science, especially as nobody has the slightest
notion of what it means.
Even today, 80 years after Russells lecture, there is no agreement on how to interpret probability.
What there is an agreement on is that probability is helpful in the description of the future. The future
is unknown and uncertain, but it is possible to predict upcoming events based on past experiences.
Historical events are certain, and an example is the knowledge about how a condemned car performed
its tasks. A new car can be expected to drive through the same areas, under the same conditions as the
old car. The historical performance data of the condemned car can thus be used to find data for the
new car. The events of the future are not known, but the information about what has happened may
give a pointer to what is to come. Reliability is based on the probability of future events and estimated
through the probability of how a system is expected to behave when experiencing them.
Data used for reliability estimates can be found through field experience, expert judgments and testing
(Rausand 2005). Field experience data are collected and stored in data bases, such as the Offshore
Reliability Database (OREDA). Expert judgements are derived by experts on the field who study
previous information, the system, environmental conditions etc to find answers. Testing is done under
specified and stable conditions. The three methods may also be considered together, but it is important
to remember that they all have uncertainties connected with their answers. Uncertainty will always be
found in relation to probability, especially as probability does not give any specific answer. For this
reason, one cannot be uncritical in the use of probability estimates (Berg 2009).
The OREDA project is used in the Norwegian petroleum industry and gathers failure rate information
from several companies. The database is used to make predictions and model the failure probabilities
for different types of equipment under development. A normal model for failure rates is the bathtub
curve.

Master Thesis, TPK 4900


Ingrid Alms Berg

Design for Reliability

14.06.2010

Failure rate

NTNU

Infant mortality

Wear-out phase

Normal life (Constant failure rate)

Time

Figure 1: The bathtub curve

This curve consits of three failure rate distributions, representing the three main phases of the
operational life; infant mortality, normal life and wear-out. In the first phase, the unknown defects
may become visible. During the normal life, the failure rate stays almost constant, while in the wearout phase the products age makes it more likely to fail. For a reliability engineer, this is an easy way
to show how the reliability of a function is connected with the items age. The probability distributions
thought to be the most representative of the products behaviour through its life are chosen. When a
curve is developed for the product life cycle, it can be used to explain why one failure rate is given for
the whole operational life of a component.
The failure rates in OREDA are found through assumptions about how the data from different fields fit
together. They are also considered constant through the assumption that infant mortality is removed
through testing, while the wear-out phase is avoided by replacement. For this, the bathtub curve is an
excellent explanatory model.

2.2.1. Probability approaches


As there is no agreement on what probability means or how the concept should be interpreted, many
theories can be found. The main difference between the theories is the discussion of whether
probability is a property of the object studied or something outside it seen only by the person studying
it. From the most dominating theories, the Frequentist and the subjective interpretations, two main
approaches are developed to find probability parameters. These are the Frequentist and the Bayesian
approaches. The two are thought to be incompatible and some would even claim that a true scientist
only accepts one of the approaches (Vallverd 2003).

Frequentist approach
This approach is tightly knit to the theory that the probability is a property of the object. Among the
theories related to this are the Classical probability theory, the Finite frequency theory, the Relative
frequency theory and the A priori theories (Berg 2009). The values of interest are found through
several trials performed under the same conditions over a long period of time. Probability is then the
long-term fraction of times that an event occurs, in a large number of trials (NUREG/CR-6823 2003).

Master Thesis, TPK 4900


Ingrid Alms Berg

NTNU

Design for Reliability

14.06.2010

A very important aspect of the Frequentist approach is that all personal believes must be put aside.
There is a real value for the object studied and the objective is to infer it. The probability of the real
value is hence not up for discussion, only for the estimated value (Hallinan 2009).
When a new hypothesis has been developed, the goal of the frequentist approach is to see if it can be
rejected. This is the normal approach often used in physics to decide whether a hypothesis is actually a
law. The question is (Hallinan 2009); Is the data probable given the null hypothesis? The null
hypothesis is then the alternative hypothesis which one uses to prove that the original hypothesis is
wrong. If nothing goes against the hypothesis, it can be accepted.
Repeatable experiments are used to find the true value. However, the frequentists accept that to
continue experiments for all eternity is unpractical. They therefore say that after a large amount of
trials, the value is close enough to the real value to be plausible. This value can be used to represent
the object studied in other studies.

Figure 2: The repetition in a frequentist approach experiment

Bayesian approach
The Bayesian or subjective approach desires to modify uncertainty through logic, believing that
probability is a quantification of a persons degree of belief. The probability is thus turned from a
personal opinion into something rational through logic (NUREG/CR-6823 2003). The approach is
based on theories such as the Subjective/Bayesian theory and the Impersonal theory.
While the Frequentist approach only considers information found through homogenous testing, the
Bayesian approach accepts all known information. It is acceptable to have a prior belief about what
one observes. This belief is then used to find the posterior value. As more information is obtained
about the object studied, the estimate is updated further. While the Frequentists whish to prove that a
hypothesis is correct by rejecting other hypotheses, the Bayesians prove correctness through the data
supporting it (Hallinan 2009).
The goal is to update the prior belief through Bayes theorem:

P ( A Br ) P ( Br )
P ( Br A )
for r 1, 2,..., k
=
=
k
k
P
B
A
P
A
B
P
B

(
)
(
)
(
)

i
i
i
=i 1 =i 1

P ( Br A )
=

Master Thesis, TPK 4900


Ingrid Alms Berg

NTNU

Design for Reliability

14.06.2010

Bayes theorem here shows P Br A as the probability that event Br occurs given that event A has
occurred. P ( Br A) is the probability that Br and A occurs. Bi is all the possible events B and

P( Br ) is the probability that Br occurs.

Figure 3: Experiment with previous knowledge in the Bayesian approach

Choose a side?
As the Frequency and the Bauesian approaches are based on such different foundations, which is the
better and should one take a side? What is best for reliability purposes?
The frequentist approach has a point in that it stays as objective as possible. It would be nice if the
values used to estimate reliability only were based on unbiased, proven findings. Sadly, this is almost
impossible. Reliability data are usually experience data collected from several similar systems and
products. These rarely operate under the exact same conditions. If experience data cannot be found,
expert judgement is used. This is hardly unbiased. The only possibility to obtain data suitable for the
frequentist approach is testing.
Although the Bayesian approach seems the obvious choice, it is not necessarily the best. The
calculations used to find a value are more extensive than for the Frequentist approach. Simplicity is
said to be the best value for scientific activity (Vallverd 2003). The fact that Bayesian estimations
often are biased is a definite downside to this approach. However, this downside is also what raises it
above the Frequentist approach. We cannot estimate every little value through homogenous
experiments or sampling. Nor is our main observational skill, seeing, unbiased. An experiment is
based on observations. When we observe with our eyes, our brain uses memories to classify what we
see. The classification then helps us understand what we just experienced. The same is done when we
smell or hear something.
To decide which approach is the better is not easy. They both have their advantages and
disadvantages. The question should maybe rather be which one is the best for the specific purpose of
the probability estimation? Medicine often needs to be exact, especially when the research tries to
confirm the bacteria leading to a certain disease. The Frequentist approach is then both suitable and

Master Thesis, TPK 4900


Ingrid Alms Berg

10

NTNU

Design for Reliability

14.06.2010

objective. For reliability we need more than we can get from the values suiting the Frequentist
approach. The best might be to choose the approach best befitting the value one is estimating.

2.3. Discussion
Reliability is a subject of high importance in any industry. If a product is unreliable, high repair and
warranty costs, dropping sales and a bad reputation can be the results. As reliability only is calculated
through measures of probability, the results will be uncertain. The better knowledge one has of the
system, probability approaches and reliability calculations, the easier it is to get an impression of the
systems reliability. Any estimation must be performed through careful consideration of how one
should approach the probability and find a parameter. Both the Frequentist and the Bayesian approach
can be used. In choosing an approach, it should be based on the type of information one has and how it
may best be utilised.
Design for Reliability could be considered useless and time consuming. What any critics claiming this
forget, is the fact that reliability does make a person look at the design through different eyes.
Reliability demands an optimal design to avoid failures. To design for reliability does not mean that
the reliability alone should be calculated and increased, but that the whole system should be reviewed
for solutions which will keep it functioning as long as possible. Given that the use of reliability in the
design process also leads to good maintainability and increased availability and safety, its importance
during the product development should be obvious.

Master Thesis, TPK 4900


Ingrid Alms Berg

11

NTNU

Design for Reliability

14.06.2010

3. Subsea industry requirements


When a manufacturer is involved in a new project, the first step is to define the requirements and
specifications relevant for that particular project. The requirements will typically be specified in the
laws and regulations of the countries involved, the standards stipulated in the customer requirements,
and internal standards and requirements within the organisation. All industries have some standards to
follow on the subject of reliability, whether they are specific or very general and stated by
organisations such as the International Organisation for Standardization (ISO) or the International
Electrotechnical Commission (IEC).
As an example of how regulations and standards within an industry may affect a manufacturer on the
subject of reliability, some of the requirements to the Norwegian subsea process industry are studied.
Even for this particular industry there are too many standards and regulations for all to be discussed. A
selection of requirements form Norwegian authorities, ISO and IEC have been made. The regulations
of the authorities in a country must always be the highest priority and followed closely. Standards
from organisations such as the ISO and IEC are often voluntary, but can be included in the client
demands or mandatory within the manufacturing organisation.

3.1. Petroleum Safety Authority


The Petroleum Safety Authority (PSA) is a subordinate organisation of the Norwegian Ministry of
Labour (PSA 2010). It is the regulator for technical and operational safety, including emergency
preparedness for the Norwegian petroleum activity. This includes the preparation and presentation of
regulations, as well as decisions on permits, orders, prohibitions and dispensations. For Aker Solutions
this means that the Midgard SCS must be manufactured in accordance with these regulations. The
installation and operation of the system must also be non-conflicting and although these phases are the
responsibility of Statoil, the Subsea Power and Process department at Aker Solutions is responsible for
making it possible.
The PSA has five regulations concerning operation on the Norwegian continental shelf (PSA 2010):
- The Framework Regulations
- The Management Regulations
- The Information Duty Regulations (Not of relevance to this thesis.)
- The Facilities Regulations
- The Activities Regulations
All of these must be followed by the operators and are meant to ensure that the operations are safe.
This safe includes the employees, the environment and the production system.

The Framework Regulations


The Framework Regulations are concerned with health, safety and environment in connection with
operations offshore (NPD (C) 2001). This includes the subsea part of operations and although no
direct connection with RAMS activities is made, the regulations affects the systems developed at Aker
Solutions. Based on demands for safe operations, the operators will set their requirements for the
system safety. With highly reliable systems, they are likely to have fewer accidents.

Master Thesis, TPK 4900


Ingrid Alms Berg

12

NTNU

Design for Reliability

14.06.2010

The Management Regulations


In risk reduction, the party responsible shall choose technical, operational and organisational
solutions which reduce the probability that failures and situations of hazard and accident will occur.
() The solutions and the barriers that have the greatest risk reducing effect shall be chosen based on
an individual as well as an overall evaluation. Collective protective measures shall be preferred over
protective measures aimed at individuals.
This is stated in the first section of the Management Regulations (NPD (D) 2001), and means that
Statoil in ordering an SCS from Aker Solutions must ensure that this is the best solution. This can be
done by ensuring that the system requirements follow the regulations, and that they are answered by
the system solution. It is thus evident that although Aker Solutions is not forced to follow the
regulations, they are more probable to sell their products if they use them in their planning. As a
reliability programme will be directly linked with the reduction of risk, the Management Regulations
may be called a good reason to develop one.

The Facilities Regulations


The Facilities Regulations is fully named Regulations relating to design and outfitting of facilities
etc. in the petroleum activities (NPD (B) 2001). This regulation has direct influence on the systems
developed for petroleum purposes on the Norwegian continental shelf. It may thus be the most
important and direct input to Aker Solutions. Any equipment designed by the Subsea Power and
Process department has to comply with these regulations. Several requirements are given which
concerns the demanded reliability of the system, especially for safety and failures. Section 4, design of
facilities states the following (NPD (B) 2001):
Facilities shall be based on robust and the simplest possible solutions and shall be designed so that
b) the major accident risk becomes as low as practically possible,
c) failure of a component, a system or one single mistake does not lead to unacceptable
consequences,
d) the main safety functions, as mentioned in Section 6 on main safety functions, are maintained,
e) transport and handling of materials can take place efficiently and safely, cf. Section 12 on
handling of materials and transport routes, access and evacuation routes,
i) provision is made for the lowest possible risk of pollution,
j) provision is made for fully satisfactory maintenance.
The choice of reliability activities used in the development of new systems must fulfil these and other
demands in the Facilities Regulations in addition to other requirements presented by the customer.

The Activities Regulations


The Activities Regulations states that the maintenance programme for an offshore installation shall be
based on a thorough identification of failures and their causes and mechanisms (NPD(A) 2001) The
probability of failure shall be the basis for a choice and classification of maintenance tasks. This
applies to the manufacturer through the agreement of which information is to be passed on at
handover. The necessary information fulfilling this regulation is found in the reliability analyses. The
reliability analyses are thus not only performed to show that the equipment can function as demanded,
but also to make a basis for the maintenance tasks. A RAM analysis can be very useful as it includes a
consideration of the systems maintainability. This is further discussed in chapter 6.

Master Thesis, TPK 4900


Ingrid Alms Berg

13

NTNU

Design for Reliability

14.06.2010

The PSA regulations are often referring to standards prepared especially for the Norwegian petroleum
industry. These were previously published by NORSOK, but are now the responsibility of Standard
Norge (PSA 2010). Among some of the standards relevant to reliability are:
-

Z-008, criticality analysis for maintenance purposes


Z-013, risk and emergency preparedness analysis

These standards are usually presented together with Z-016, which has now been turned into ISO
20815, Production assurance and reliability management.

3.2. Standards
Several other standards than those listed here are of interest, but these were considered the most useful
at Aker Solutions Subsea Power and Process. Through the use of standards and by referring to them in
documents, it is possible for all parties in a project to find and understand what is meant by a specific
term or description.
Standards ensure desirable characteristics of products and services such as quality, environmental
friendliness, safety, reliability, efficiency and interchangeability and at an economical cost (ISO
2010).

3.2.1. NORSOK
Z-008 Criticality analysis for maintenance purposes
Z-008 recommends that any preparations for a maintenance programme are started in the design phase
(NORSOK Z-008 2001). This means that the system still will be in the hands of the Subsea Power and
Process department, rather than with the future operator. The standard demands a consequence
classification of the potential failures which is then connected with the different functions. With a
proper RAM analysis included somewhere in the development process, it should be easy for the client
to develop a maintenance programme before the system is put to use.

Z-013 Risk and preparedness analysis


This standard links a risk analysis and the preparation and establishment of an action plan for
emergencies. It is demanded that a risk acceptance criteria is set, the hazards identified and studied and
that steps are taken in order secure the system according to the criteria (NORSOK Z-013 2008). Based
on the information retrieved and the solutions suggested for the risks, the emergency preparedness
programme will be established. Normally this standard is meant as an input throughout the life cycle,
including after the system is in operation. For the development of a new system, the Subsea Power and
Process department will do an extensive search for hazards as a response to the ordinary demands of a
client. Although the emergency preparedness is the responsibility of the client and operator, the risk
and reliability studies performed by Aker Solutions will be the main input when a new system is
handed over. Z-013 and Z-008 may not be compulsory for the engineers at Aker Solutions, but they
must be followed by their clients, and thus complied with.

Master Thesis, TPK 4900


Ingrid Alms Berg

14

NTNU

Design for Reliability

14.06.2010

3.2.2. ISO standards


The ISO standards are prepared by the International Organization for Standardization, which is a
network of standard institutes in 161 countries (ISO 2010). Some of the standards are specified for the
petroleum and natural gas industries and may therefore be looked upon as even more relevant than
other standards. ISO is a non-governmental organisation without any authority to enforce the
standards. To follow them is therefore a choice, unless otherwise is stated by the authorities in a
country. This is also applicable to the specified standards which the Subsea Power and Process
department may find itself subject to.

ISO 20815 - Production assurance and reliability management


This standard gives an outline of a programme for production assurance and reliability management. It
states that the programme could be concerned with one phase alone, or the entire product life cycle.
The main objective is that risk and reliability should be included as parts of the development process
from an earlier stage than what has been the habit. ISO 20815 (2008) observes that, in traditional
design processes technical safety and reliability aspects have generally not been considered until the
equipment or component verification. This is too late if one desires an optimal design.
A production assurance programme outline is given in Appendix A of ISO 20815 (2008). Among the
main objectives of such a programme are:
-

A statement of the performance objectives and measurements for the subject in development.
A description of the project risk explaining the extent to which activities are chosen for the
programme.
A description and distribution of the responsibilities and action management system for the
reliability.
An activity schedule including an overview of the activities to perform, when during the life
cycle they should be performed and how they are connected with each other.

Statoil specifically asks for a reliability management programme based on ISO 20815 in the Midgard
SCS development. This is further discussed in chapter 9.
Even without client demands for this standard to be followed, it may be very useful for a
manufacturer. Any early reliability analyses must be performed by the party responsible for the
development process. ISO 20815 (2008) indicates how different methods can be chosen and followed
up throughout the process, in early as well as late phases. It may simplify the discussion of the
allocation of time resources to different reliability tasks based on the project risk and the newness of
the technology.

ISO 13628, Design and Operation of Subsea Production Systems


This standard is not specifically concerned with reliability, but does affect the design of the systems
produced by the Subsea Power and Process department. It states requirements, recommendations and
guidance for the areas which require consideration when a subsea production system is developed
(ISO 13628 2005). Functional requirements do affect the reliability demands and it can therefore be
useful for a reliability engineer to know this standard. ISO 13628 (2005) covers all phases of the
product life cycle and several types of production systems in distinct parts.

Master Thesis, TPK 4900


Ingrid Alms Berg

15

NTNU

Design for Reliability

14.06.2010

ISO 14224 - Petroleum and natural gas industries Collection and exchange of
reliability and maintenance data for equipment
It is stated in this standard that its primary users are the owners and/or operators, who should find the
data to be collected, available in the operating facilities (ISO 14224 2006). Designers and
manufacturers are mentioned for the use of the reliability data, in connections such as lessons learned
and reliability estimation inputs. The standard can be quite useful for the understanding of what a data
collection truly is, its data and their uncertainties.
ISO 14224 emphasises that the following categories of data are to be collected (ISO 14224 2006):
-

Equipment data: taxonomy, attributes etc.


Failure data: cause, consequence etc.
Maintenance data: action, resources, consequence, down time.

If these data are collected with all the available information of interest, all RAMS disciplines may
easily be covered in analyses. All types of equipment and systems are of interest, whether permanently
or temporarily installed. The data collecting process is described in figure 4:
Design/
manufacturing

RAM
Analysis

Failure and
maintenance events

Operation and maintenance

Concept
Improvement

Adjustments
and
modifications

Loop

Data

Figure 4: The data collecting process (ISO 14224 2006)

High quality data is defined as the following (ISO 14224 2006):


-

Complete in relation to specification


Complying with definitions
Accurate
Properly stored
Sufficient in population
Relevant to the need of the user.

The standard also states how the data are verified and the limitations normally connected with data
collections. If this information is used properly by the reliability engineer, or any other user, it should
be easy to separate the useful data from those of low quality. It would also be possible to understand
whether the data verification is acceptable. Unacceptable data verification means care must be taken
when data are used.

Master Thesis, TPK 4900


Ingrid Alms Berg

16

NTNU

Design for Reliability

14.06.2010

ISO 17776 - Petroleum and natural gas industries Offshore production installations
Guidelines on tools and techniques for hazard identification and risk assessment
ISO 17776 is a useful standard for information about the principal tools and techniques commonly
used for the identification and assessment of hazards in the petroleum industry (ISO 17776 2000). It
explains how the tools and techniques can be employed in the development of strategies for
prevention, control and mitigation of possible hazardous events.
ISO 17776 (2000) defines the three main steps in hazard and risk assessment as:
1) Identification of hazard
2) Assessment of the risk
3) Elimination or reduction of the risk
The standard describes how the methods for hazard identification and risk assessment are to be
chosen, the role of experience and judgment, and how checklists, codes and standards can be used.
For a risk and reliability engineer the standard helps ensure a good risk management process. It can
also ensure the understanding of why a method has been chosen over another and why certain
decisions are labelled more important than other. ISO 17776 (2000) is highly descriptive and easy to
read and follow. The annexes explain several tools and what information they can obtain.

3.2.3. IEC
The international electrotechnical commission prepares and publishes standards for the electrical,
electronic and related industries (IEC 2010). Like the ISO it is a non-governmental organisation with
member organisations in several countries around the world. There is no enforcement of the standards,
but they are recommended as useful tools ensuring the standard of products.
For the subsea process industry these standards may give an input to how an activity shall be
performed and followed up. As electrical systems are becoming more common sub sea, these
standards are increasingly useful.

IEC 60300 - Dependability management


Dependability is a collective term used to describe the availability performance and its influencing
factors: reliability performance, maintainability performance and maintenance support performance
(IEC 60300 2003).
IEC 60300 (2003) explains how a dependability management system may be organised and
implemented in an organisation. It consists of several parts which explain the necessary steps, methods
and procedures needed to establish a dependability programme. Part 1 explains the fundamentals of
dependability management and provides the principals for organisations involved in the establishment
and use of such programmes (IEC 60300 2003). The second part provides the guidelines for a
dependability management programme, while part 3 introduces the suitable analysis techniques for
dependability.
For a reliability engineer in an organisation where reliability and dependability have been overlooked
and considered too expensive to use on a larger scale, the standard might be serviceable. It explains
why dependability management is important and how it can be implemented according to organisation
and product. It can be followed easily and employed without too many alterations or delays in normal
production processes.

Master Thesis, TPK 4900


Ingrid Alms Berg

17

NTNU

Design for Reliability

14.06.2010

IEC 61160 - Design review


A design review is a planned, documented independent review of an existing or proposed design (IEC
61160 2006).
Reviews of the design at predefined stages in the product development process can be useful to ensure
that product requirements are met with. IEC 61160 (2006) refers to design review as a repeatable,
confirmatory and refining procedure. This is shown in figure 5:

D
E
S
I
G
N
R
E
V
I
E
W

Design planning
Design change

Design input
Design process
Design verification

Design output
Design validation

Comlpeted design

Figure 5: Design review process (IEC 61160 2006)

A design review is not a task specifically meant for the reliability engineer, but is thought to help his
or her work in ensuring the product performance. The process requires planning, organising and
reporting (IEC 61160 2006). Although a review is a good measure of control for the design, it is
important that it stays exactly that. It is not meant to be a creative process or in any way lead to new
designs for the product. The only design alterations based on the review that are acceptable must be
due to unacceptable problems.
When the review process has been performed, it is up to the designers and other members of the
development team to follow up on it. Figure 6 shows how the design review process can be performed
and responded to.

Start

Meeting
notification
and agenda
6.5

Planning of
design review
6.2

Selection of
design review
team
6.3

Conducting
design review
meeting
6.6
Prepare and
distribute the
design review
minutes
6.7

Preparation of
input package
6.4

Review of design

Design manager
reponses to
actions and
recommendations
6.8

Follow up and
completion of
actions and
recommendations
6.9

Yes

No
All completed

Design manager
signs the minutes
closing the minutes
6.9

End

Response and completion of actions and recommendations

Figure 6: Design review and response process (IEC 61160 2006)

Master Thesis, TPK 4900


Ingrid Alms Berg

18

NTNU

Design for Reliability

14.06.2010

IEC 61508 - Functional safety of electrical/ electronic/ programmable electronic safetyrelated systems
The functional safety of a system depends on the inherent safety of the system and any extra protective
systems. Together these must answer to the safety requirements set by the manufacturer, clients and
authorities. IEC 61508 (2005) provides a method for the establishment of a safety requirements
specification, through an approach for the application of safety activities during the life cycle. Product
safety requirements concern the safety performance of products and are made both for intended use
and foreseeable misuse (Lundteigen et al. 2009). This ensures that the activities are dealt with
systematically.
IEC 61508 (2005) consists of seven parts which considers the general requirements, requirements for
E/E/PE safety-related systems, software requirements, definitions and abbreviations, examples of
methods for the determination of SIL, guidelines on application and an overview of techniques and
measures. The first four parts are basic safety publications, while the last three are help to the
implementation of the standard (IEC 61508 2005).
This standard is tightly connected with IEC 61511 and the two can therefore used together with
advantage to the functional safety. Alone, this standard can give input to the reliability process through
its allocation of safety requirements.

IEC 61511 - Functional safety safety instrumented systems for the process industry
sector
IEC 61511 (2003) is the process sector version of IEC 61508, but directed at system level designers,
integrators and end users rather than vendors developing new devices (Lundteigen 2009). The standard
follows the requirements given in IEC 61508, but they are modified to suit the practical situation,
concepts and terms in the process industry.
IEC 61511 (2003) consists of three parts focusing on framework, definitions, system, hardware and
software requirements, guidelines on application and determination of SIL. This confirms the
similarity to IEC 61508 and the two are often treated simultaneously. The Norwegian Oil Association
is one out of many who has created a guideline for the application of the two standards; the OLF 70
(Lundteigen 2009). This follows the Norwegian regulations and thus simplifies the approach for a
reliability engineer in the petroleum industry.

3.3. Further comments


The standards mentioned here is only a selection of those which can be of use for a reliability engineer
working within the subsea process industry. The ISO standards described might not be of specific
interest. However, they all have useful inputs to reliability engineering. Depending on the system
developed and the industry in question, the suitable standards should be studied and used according to
the phases in which they belong. These standards can be used in the different phases according to table
1. X means the standard is used directly, O means it can be used, and In use means that the results of
the standard having been used are applied.

Master Thesis, TPK 4900


Ingrid Alms Berg

19

NTNU

Design for Reliability

14.06.2010

Table 1: Standards used in phases

Phase

Standard
Z-008

Z-013

ISO 20815

ISO 13628

In use

In use

In use

In use

O
X

O/
in use

O/
in use

ISO 14224
ISO 17776

IEC 60300

X/
in use

O/
in use

O/
in use

O/
in use

O/
in use

O/
in use

IEC 61160

IEC 61508

IEC 61511

Standards as those described here are not the only documents which can be relevant to follow. Several
organisations developing new technology have defined levels for how far in the development process a
technology is. These are often called Technology Readiness Levels (TRLs). In many industries the
TRL documents are developed by specific organisations. Although each company may still have their
own TRL documents, a document for the whole industry in a country makes it easier for conformity
between the companies. In Norway, DNV has developed a document called Qualification Procedures
for New technology, DNV-RP-A203, specifically for the petroleum industry.
The American Petroleum Institute has developed a structured approach to manage uncertainty
throughout the product lifecycle, API-RP-17N, which is compatible with ISO 20815 (API-RP-17N
2009). This, as the DNV-RP-A203, is not a compulsory document, but highly recommended for the
reliability to be properly implemented. This recommended practice goes through all the phases, and
describes the most convenient reliability methods and tests.
Other interesting documents are those prepared by the Norwegian Oil Industry Organisation, OLF
(OLF 2010). These explain how other standards may be utilized and implemented to the best interest
of the petroleum industry.
The best way to identify useful standards is to follow the development in organisations such as IEC
and ISO, and the laws and regulations in a country. Which of these are needed in a project can be
found through the client demands, the concept and the requirements. In every organisation, it is up to
the responsible for reliability to follow the development and make sure the most recent regulations and
standards are known in a project.

Master Thesis, TPK 4900


Ingrid Alms Berg

20

NTNU

Design for Reliability

14.06.2010

4. Product life cycle


A product is usually only of interest to a consumer from the moment he or she considers buying it or is
provided with it, until it is thrown away. Environmentalists might concern themselves with how the
product is produced and later disposed of, as well as how it affects the environment while it is in use.
For the manufacturer, a product is born the day the need for it develops and it lives on until it has been
properly disposed of.
The lifespan of a product depends on what type of product it is. A normal toothbrush is only meant to
be in use for a few months and its total length of life as a physical product may be less than a year.
Equipments made for the subsea process industry are usually supposed to be operating for several
years. Ships and boats may even be in use for decades. While the lifespan of different products differ,
their life cycle is usually evaluated in the same way. A life cycle includes several phases concerned
with the concept development, design, production, sale, use, maintenance and disposal. How many
phases there are in a life cycle depends on the person looking at the product and his or her reason for
separating it into phases.
There are several existing models showing the phases of a product life cycle. In some cases the life
cycle is a closed loop where the disposal includes recycling and reuse. In other cases it is thought that
the product ends its life at the dumpsite. In general a product life cycle model depends on the intended
use of the model, whether it is developed for marketing, designers, or reliability engineers.

4.1. Phases in a general perspective


The main approach to how phases are defined is based on what the product goes through from it is
first thought of, to its disposal. The IEC has defined six life cycle phases (IEC 60300 2003):
-

Concept and definition


Design and development
Manufacturing
Installation
Operation and maintenance
Disposal

The first phase focuses on the development from an idea to a product outline. Research is made to see
if there is an opening for a new product on the market and what this opening demands from the
product. Different concepts are considered and the product requirements specified. Based on the
chosen concept, a detailed design process commences. This will establish the system architecture,
hardware and software. During manufacturing, the product components are produced or collected from
suppliers for assembly. The installation phase can be performed by the manufacturer, the operator or
both together. During operation and maintenance, the product is in use and maintained as far as
possible. When the maintenance costs go higher than the returns from the product while it is in use, it
is time to dispose of it. The disposal phase includes the removal, dismantling and destruction or
recycling of the product (IEC 60300 2003).
This life cycle model is suitable for most industries and companies, but the IEC 60300 (2003) does not
describe each phase with much depth. All organisations can choose which separation of phases they
prefer. An example is the very common five-phased model. This is employed by many organisations
and is separated into the following phases:

Master Thesis, TPK 4900


Ingrid Alms Berg

21

NTNU
-

Design for Reliability

14.06.2010

Front-end engineering
Design
Development
Production
Post-production

The five phases are used in connection with several project models and can be extended or divided
depending on the type of project. An example of an organisation supporting this product life cycle is
Aker Solutions. As the projects the Subsea Power and Process department currently work on concern
systems handed over to a client after manufacturing, the last two phases are of less interest than the
previous. They may have a say in how a system should be operated and disposed of, but it is not the
responsibility of the Subsea Power and Process department. The wish for the future is that the Subsea
Power and Process department can produce more standard products, in which case the phases might be
slightly altered.
A quick comparison to the six phases in IEC 60300 (2003), shows that these five phases are mainly
concerned with design and development. The post-production phase will last longer and possibly
include more tasks than any of the other phases, but it seems to be far less interesting. For a
manufacturer without responsibility after handover this makes sense.
Front-End
In the front-end phase an opportunity or a need for a new product has emerged. New technology may
give way for improved or new products that were impossible to produce earlier on. The need for a new
product idea can also be developed as a response to customers complaints or through competition
(Murthy et al. 2008).
When the new opportunities have been recognised, a filtering of ideas should be performed to find the
best suitable options for a new product. This filtering should answer the following questions (Murthy
et al. 2008):
1.
2.
3.
4.

Does the idea fit within the business market or technology focus area?
Are the business opportunities attractive (potential market size, growth etc.)?
Is it technically feasible to develop and produce the product?
Are there any potential hindrances that may stop the project (Intellectual property, legislative
and/or environmental issues)?

When a product idea has been chosen for further use, the product and business objectives must be
defined and customer requirements identified. A product concept is then developed. It defines the
main functions and sub-functions, as well as the relationships between the main components. The
development of a new product may be considered as a project. A project plan should therefore be
prepared, including a schedule, an allocation of responsibilities and resources, performance measures
and risk management plans.
Design
Based on the product concept and requirements found in the front-end phase, the internal arrangements
and interactions of the product can be determined. This concerns the sub-systems, assemblies, subassemblies and the components. A definition of the functional decomposition and relationships
between these is vital to the establishment of the product architecture.

Master Thesis, TPK 4900


Ingrid Alms Berg

22

NTNU

Design for Reliability

14.06.2010

As soon as a definition is ready, a more detailed design of the product can be prepared. Component
properties will now be set, detailed drawings made and a bill of materials written. According to Sim
and Duffy (2003), design activities can be classified as:
-

Design definition activities


Design evaluation activities
Design management activities

These activities should be performed concurrently to ensure an optimal design fitting the purpose and
concept of the product. Finally a review of the design should be carried out to verify the product
architecture (Murthy et al. 2008).
Development
The development phase is where the prototype is made and tested. This is meant to verify that the
product can be manufactured according to the planned design. When a product is custom-built, the
performance agreed upon between customer and manufacturer should be verified. The actual
performance must also be proven to be higher or equal to the desired performance. Through the tests
performed in this phase, problems can be discovered, understood and resolved (Murthy et al. 2008).
Production
The main challenge of the production phase is to uphold the designed-in performance throughout the
manufacturing process (Murthy et al. 2008). This is due to the inability of a production system to
produce exactly similar outputs. Variations in the production process, the materials used, and the
surrounding environment mean that special strategies must be used to keep up the quality of the end
product. Process control, inspection and product testing are methods used to ensure a high quality of
the end product.
Post-Production
When the product is ready for use, it is turned over to a customer. The customer is either found in a
large market, or the specific buyer of a custom-built product. In both cases the transportation and
product support are important parts of the post-production. When the customer is ignorant of the
products existence, the producer needs to include promotion, pricing and distribution channels in the
phase as well (Murthy et al. 2008). The product support is perhaps the most important part of the postproduction, as its focus is on keeping the customer content. Normal activities could include
installation, maintenance, warranties, repair and disposal.
Neither the six phases described in IEC 60300, nor the five phases used by Aker Solutions are
especially concerned with, or developed for, reliability purposes. A model comprising eight phases in
which reliability is the first objective is suggested by Murthy et al. (2008).

4.2. 8 Phases for reliability


The eight phase model has three different systems: stages, levels and phases. The three stages are
concerned with the physical development of the product, whether it is under planning, production or
released to a market. For the linking between business objectives, attributes and characteristics of the
product and its components, three levels are described. Finally the stages and levels are combined as
showed in figure 7, while the development of the product travels through the combinations in 8
different phases.

Master Thesis, TPK 4900


Ingrid Alms Berg

23

NTNU

Design for Reliability


Stage I
(Pre-development

Stage II
(Development)

14.06.2010
Stage III
(Post-development)

Level I
(Business)

Phase 1

Level II
(product)

Phase 2

Phase 5

Phase 7

Level III
(Component)

Phase 3

Phase 4

Phase 6

Phase 8

Figure 7: Phases, Levels and Stages (Murthy et al. 2008)

Stages I-III
-

Pre-Development
o Non-physical conceptualisation of the product with increasing level of detail.
Development
o Physical embodiment of the product through research, development and prototyping.
Post-Development
o The remainder of the product life cycle (production, sale, use, etc.).

Levels I-III
- Business level
o Linking the business objectives for a new product to desired product attributes.
- System, i.e., product level
o Linking product attributes to product characteristics the product is treated as a blackbox.
- Component level
o Linking product characteristics to lower level product characteristics, at an increasing
level of detail.

Phases 1-8
Phase 1: The need for a new product is identified and the decisions related to the product attributes are
made from an overall strategic management level of business. Data are collected and analysed in order
to see whether a new product is of interest to the consumers. Based on this, generation and screening
of ideas is done and a feasible option chosen. Finally the product concept and the performance
requirements are formed and evaluated.
Phase 2: Based on the requirements from phase 1 the desired performance is found. Overall product
reliability is derived and compared with the product requirements and the desired performance. This is
especially important as an evaluation if the product is custom-built. The comparison of the predicted
and desired performance decides whether or not the process may proceed to phase 3.
Phase 3: In this phase the detailed design is carried out through a functional analysis. The system is
decomposed into several sub-systems, down to the component level. It is important that the previously
set requirements are answered throughout the system. The functional analysis helps assign

Master Thesis, TPK 4900


Ingrid Alms Berg

24

NTNU

Design for Reliability

14.06.2010

requirements and functions to each component. From this, the predicted reliability of each component
can be calculated for the overall product, and compared with the desired reliability. If the comparison
shows an acceptable reliability, phase 4 may begin.
Phase 4: When this phase has been reached a physical development of the system begins. Starting on
the component level, tests are performed to ensure an appropriate production and that the desired
reliability is achieved. Testing continues up to the highest possible level and leads to a testing of the
overall product reliability if feasible. Any problems discovered through the tests must be corrected,
either through changes in design or choice of material. The testing is limited by the costs and available
time. If the product is custom-built, the testing is more important and operational tests should be
included.
Phase 5: In this phase it is suggested that a prototype is released to a limited number of customers
who can evaluate its features and test it in normal user circumstances. During the tests, logs must be
kept for later analysis of failures and success. If the results are positive phase, 6 may be initiated.
Phase 6: The first step in this phase is to design a production process for serial production. In order to
ensure the quality of the product and the production process, control limits are set and a quality control
plan prepared. This helps the follow-up of process variations. Such variations are unavoidable, but
may, with proper attention, be kept within acceptable limits. Tests should be performed during the
production to eliminate assembly errors, defects and early component failures. Before release of a
batch of products, a final sample testing could be done to ensure the reliability performance of the
finished product.
Phase 7: When the product is released to the market and is in use, it is interesting to look at the field
performance. Data collected through warranties, repairs and other feed-backs should be collected
continuously. Analyses could then be used to identify the actual product performance and reliability.
The analyses must take into account the variability in usage intensity, operating environment, and the
customer perspective. A comparison of actual and desired performance can be made and in case of a
mismatch, a cause and solution should be found for further production or new development projects.
Phase 8: Here the performance of the product released for sale is evaluated from an overall business
perspective. An analysis of all available data is performed and compared with the desired performance
from the first phase. If the actual and desired performances do not match, a backwards iteration
towards phase 1 must be done in order to find and implement the necessary actions. The results can be
used to evaluate the success of the product and as lessons learned in later development projects.
Figure 8 shows how the phases are connected through the three stages and can be compared with the
basic 5-phase product life cycle.
Phase 1

Phase 2

Phase 3

Pre-development

Front-end

Design

Phase 4

Phase 5

Development

Development

Phase 6 Phase 7

Phase 8

Post-development
Production

Post-production

Figure 8: Connection between life cycle phases (Murthy et al. 2008)

Master Thesis, TPK 4900


Ingrid Alms Berg

25

NTNU

Design for Reliability

14.06.2010

4.3. Discussion
The life cycle models presented here are only a few out of many. Both the five-phase and the eightphase models are mainly concerned with the events before the release and installation of the product.
To split the route from idea to finished product in many steps will possibly give more focus on the task
at hand and shorter time to a milestone is reached. For this reason, the eight-phase model is good.
However, it is also important that the increased number of phases does not lead to a decreased
overview of the process. For reliability purposes it can be argued that the eight-phase model is more
suitable. This model makes it easier to see which information has been found as the tasks of each
phase are more distinct. The model has been created with reliability performance in mind, while the
five-phase model is directed towards the production process. In the case that the five-phase model is
preferred, it may be possible to use the comparison in figure 8 to implement reliability at the correct
moment.
For later chapters in this thesis, the eight phases of Murthy et al. (2008) has been chosen as a basis for
both the discussion of how unreliability may occur, and to suggest a new methodology for design for
reliability. In general it can be said that everyone in need of separating the product life cycle into
phases must evaluate the purpose of the separation and the specific case. A decision should be based
on what simplifies the utilisation of the product life cycle the most.

Master Thesis, TPK 4900


Ingrid Alms Berg

26

NTNU

Design for Reliability

14.06.2010

5. Methods for reliability


In every industry the RAMS disciplines are viewed through different eyes. This has especially affected
the development of different methods used to analyse reliability and safety. The methods are not
necessarily very industry specific, but they have been developed from the specific need the industry
sees. In large parts, the industries can be separated in two; those who considers the aftermath of a
failure and those reviewing the events leading up to failure. The former is more likely to have been
responsible for the development and use of methods for safety, while the latter mostly is interested in
methods for reliability. A characteristic of the latter is that if a failure occurs, there is little which can
prevent the final catastrophic consequence to take place. Examples of such industries are the nuclear,
the aerospace and the military industries. If something goes wrong with a nuclear reactor, it is quite
possible that the consequence will involve far more than the destruction of the nuclear plant. A failure
during a space mission is highly likely to not enable the crew to return safely to the earth. Finally, if a
missile is detonated while it is still attached to a boat or an airplane, little can be done to prevent it
from destroying its host.
Among the industries where the aftermath of a failure has been given the most focus is the offshore
industry. This industry has existed for more than sixty years, successfully transporting oil from below
the seabed to onshore installations (Offshore 2009). Throughout most of this period, the main focus of
reliability and safety engineers has been on the aftermath of an accidental or hazardous event.
Equipment reliability has been considered through the use of separate analyses and these have not
always been performed early on in the product life cycle. Although the reliability of a product has
been considered important, the offshore industry has few standards for reliability methods and
programmes. One reason might be the actual possibility of preventing a failure from developing into a
catastrophe.
In a presentation on reliability strategy for the subsea process industry given in 2003, BP stated that
reliability analyses should be integrated to the design phase. They also found that no specific guidance
existed for the setting and allocation of reliability requirements in the subsea process industry. A goal
was set to allocate reliability down to the component level and educate more personnel on reliability
issues. (Williams 2003).
The methods used at a certain point of the life cycle will depend on how the product is manufactured,
its intended use and how far into the development it is. In the earliest development phases, preliminary
studies of the reliability can be performed, while in phases where the product is materialised, physical
tests become possible. For a product produced in large numbers, for instance an mp3-player, tests
leading to destruction may be used. Prototypes can be developed at little expense and depending on
the products size the tests may be performed by employees bringing the product home. When the
product is of such a scale or cost that it can only be made once, destructive testing is harder to apply.
Non-destructive testing is in use, but these rather prove the ability to perform than the failures. One
such test is factory acceptance test (FAT), whose main goal is to confirm that all parts of the system
functions as intended. Alterations may be done after this test in case of unacceptable results.
The subsea process industry is a very good example of how specific conditions might demand more of
the system reliability for custom made products than standard product. In this industry it is likely that
the products are of a very large scale, expensive and made to order. The environment of the product
when it is in use is less accessible to human beings and more demanding than in most industries.
Depending on the oil or gas fields calculated lifetime, a product can be required to operate for many

Master Thesis, TPK 4900


Ingrid Alms Berg

27

NTNU

Design for Reliability

14.06.2010

years. Due to new developments increasing the fields lifetime, a product can be demanded to last
even longer than originally planned. All these factors increase the importance of reliable designs. The
little accessible environment is a difficulty shared with the aerospace industry, while the extension of
planned time in use is a trait shared with the nuclear industry. An example is France where they
currently are extending the life of nuclear plants with ten years more, from the previous expected
twenty years. This can be both dangerous and press the systems towards their wear-out phases.
Thorough and precise methods for reliability are thus needed to confirm that the extended life is
acceptable.

Reliability methods and SADT


The more commonly applied methods for the analysis of reliability are not made to suit all phases of a
product life cycle. Nor can they include every aspect of interest to an analyst. All methods need certain
inputs and produce specific outputs. The outputs of one method may be the input of another. Some of
the methods are more overlapping, but necessary together for the output they give. Other can stand
alone at a specific point in the life cycle, but will not give the most useful outputs without the help of a
later method analysing them. The methods are not necessarily stuck to one single phase, but can be
updated throughout the life cycle. A good example is the failure mode and effects analysis (FMEA)
which exists in several formats, and can be useful in many phases.
Each method will be presented through the structured analysis and design technique (SADT).

Control

Input

Function

Output

Mechanism
Figure 9: SADT model (Rausand & Hyland 2004)

As shown in this figure, there are four main elements in the SADT model. A function is the method
which is to be performed. The inputs are the energy, materials and information needed to perform the
reliability method. Mechanisms are the people, facilities and equipment needed to perform the
function. Controlling elements for a reliability method are standards, requirements, demands and
budgets. Finally, the outputs are the results after the method has been used (Rausand & Hyland
2004).
The basic method will be presented first, then the required input, the mechanism, the control and lastly
the output.

Master Thesis, TPK 4900


Ingrid Alms Berg

28

NTNU

Design for Reliability

14.06.2010

5.1. PHA and HAZID


In the beginning of the product life cycle little information is available, unless the product is a new
generation of another product. The reliability analyses performed at this stage cannot demand too
much input. Their output, on the other hand, ought to give information which increases the knowledge
about the product features. An analysis which can be performed early is preliminary hazard analysis
(PHA). This method is a semi-quantitative analysis which helps identify potential hazards and
accidental events, ranks the events according to severity and considers hazard controls and follow-up
actions (Rausand(B) 2005). PHA was developed by the US army and has later spread through to other
industries. Another slightly different type of PHA, with the same objectives and need for inputs, is
HAZID (hazard identification).
The PHA procedure consists of four steps; PHA prerequisites, hazard identification, consequence and
frequency estimation, and risk ranking and follow-up actions (Rausand(B) 2005). All parts of a system
or product must at first be considered in a search for hazards. This includes failure modes, known
maintenance operations, safety systems and potential human errors. The frequencies and severities of
each hazard should be considered and classified in classes ranging from improbable to frequent
and from negligible to catastrophic (MIL-STD-882D 2000).
Inputs
Any information about the system; design plans, concept definitions etc. is useful input. Checklists can
be followed to remember all elements which are to be studied and what is to be studied about them.
Other inputs such as risk information and block diagrams may come from similar existing systems
(Stapelberg 2009).
Mechanisms
The mechanisms of this analysis are based on team work. The members should complement each other
with different backgrounds and knowledge, enabling a more complete review of the whole system. If
little information can be obtained, useful techniques for the retrieval of inputs can be what-if
techniques and brainstorming (Rausand(B) 2005). These techniques are based on imaginative thinking
and are generally thought efficient for team work.
Control
In some industries there are standards describing how the PHA is to be performed and which outputs
the final report should include. The main part of demands to the analysis will be given by the
organisation it is performed for. This will depend on the limitations of time and costs given to perform
the analysis. All product requirements which have been established can be considered, both as input to
the analysis, and as controlling the main focus of the analysis. Both PHA and HAZID are described in
ISO 17776 (2000), which is a standard especially prepared for the petroleum and natural gas
industries. This can thus be used as a means of control for the Subsea Power and Process department
at Aker Solutions.
Outputs
A PHA will produce a list of all hazards considered, including where they may originate from and a
ranking of how grave the events occurrence can be. The latter will be very useful for later reliability
analyses, if some hazards need to be prioritised above others (MIL-STD-882D 2000). To present the
outcome of the PHA, a worksheet is made. This worksheet is standardised in several industries and
will give a clear perspective of the different hazards connected to the system studied. There are few, or

Master Thesis, TPK 4900


Ingrid Alms Berg

29

NTNU

Design for Reliability

14.06.2010

sometimes no differences between the worksheets of the HAZID and the PHA. The worksheets will
mostly depend on the systems and the organisations that employ the analyses (degaard 2003).

Inputs
Basic system information
Previous knowledge similar systems
Operational/ Environmental conditions
Checklists

Control
Industry standards
Time/ Costs/ Resources
Product requirements

PHA/HAZID

Mechanisms
Team work
System knowledge
Worksheets

Outputs
List of hazards
Hazard information (origin, effect, etc.)

Figure 10: SADT PHA/HAZID

5.2. SWIFT
The structured what-if technique (SWIFT) is a brainstorming technique where a team search for
hazards through questions like What if...? and How could...? (Kritzinger 2007). It was originally
developed for the chemical process industry and is today also used in other industries, for instance the
aircraft industry. SWIFT mainly considers systems at higher levels, either the complete system as a
whole, or a sub-system. The technique is an alternative to HAZOP as it is less time consuming and
costly. These two methods do not operate on the same level of detail. It is thus possible to use them as
complementary studies (MoD 2010). SWIFT is easy to use through all phases and for many purposes,
for example as a study of a manufacturing process.
There is no specific standard for the SWIFT technique, but it is mentioned in the ISO/IEC 31010
(2009). The normal procedure includes a definition of the activity and problems of interest, a
generation of what-if questions and responses, and a report on the findings. If the method is performed
after a PHA or HAZID, many hazards will already be known. The SWIFT could then go further into
the question of how the hazard would affect the system, and whether the rankings made were correct.
A negative side of the SWIFT is that a problem is likely to stay unnoticed if the team fails to ask the
correct questions. Another problem is its inability to produce quantitative results which can answer the
more complex risk-related questions (Tech 482/535 2005).
Inputs
To perform a SWIFT analysis, it is necessary to have a basic system description, a definition of the
activity, and some problems of interest. These inputs, along with a checklist for which elements one
should study, would give all the information needed.
Mechanisms
The mechanisms of this method are the chosen team. It is very important for the success of the
analysis that the team members have sufficient experience with the system and the technique. They
must also be able to understand the output of other analyses performed before the SWIFT. The
technique itself is a mechanism through the questioning.

Master Thesis, TPK 4900


Ingrid Alms Berg

30

NTNU

Design for Reliability

14.06.2010

Control
Controlling factors are few for the SWIFT. It does not demand much time or cost to perform it, nor
has it been standardised. The main controlling factors would thus be the resources and whether any
political issues affect the questions asked. The latter may happen in most reliability analyses and is
troubling if it leads to the exclusion of failure considerations with dangerous effects. This method is
described in the ISO/IEC 31010 (2009) standard which also shows it against other techniques.
Outputs
Outputs from the SWIFT are hazards, their causes and possible effects. Together with other hazard
identification tools it can provide a good overview of the problems possibly faced by the system. This
can be used as inputs to FMEAs, FTAs and other reliability analysis methods.

Inputs
Basic system information
Previous knowledge similar systems
Operational/Environmetal conditions

Control
Resources
Politics
ISO/IEC 31010

SWIFT

Mechanisms
Team Work
What if? and How? questions

Outputs
List of hazards
Hazard information (origin, effect, etc.)

Figure 11: SADT SWIFT

5.3. FMEA
Failure modes and effects analysis (FMEA) is a systematic technique for the study of failures. It was
developed in the 1950s by reliability engineers for the study of military system failures (Hyland &
Rausand 2004). FMEA is meant to be an input to the design process, enabling alterations of problems
before the design is too settled (MIL-STD-1629A). In order for this to be possible it is suggested that
the FMEA is performed as early as possible in the development process. As the development
progresses, it is possible to keep filling in the failure information. This can be useful for other reasons
than reliability, for instance safety and maintenance. The FMEA can be extended to include a
criticality ranking of the different failures. In this case the criticality ranking is a combination of a
severity measure set for the failure mode and its frequency of occurrence (IEC 60812 2006). When the
FMEA is extended, it is called a Failure mode, effects and criticality analysis, FMECA.
The FMEA is usually a bottom-up analysis, where as many components, assemblies and subsystems as
possible are included. To begin the analysis, a study of how each part may fail is done (Rausand &
Hyland 2004). This should be followed by the study of why they occur, what their possible effects
are, and how they can be detected. It could also be studied how the failures might be compensated for
and whether they are dangerous or not.
Inputs
To perform the Failure modes and effects analysis, it is necessary to have an overview of the system.
The main inputs to the FMEA are system design drawings, functional breakdowns, and a list of the
possible hazards and failure modes found in previous analyses. If criticality is added, information
about the failure frequency should be obtained as well.

Master Thesis, TPK 4900


Ingrid Alms Berg

31

NTNU

Design for Reliability

14.06.2010

Mechanisms
The FMEA demands an understanding of the development and consequence of a failure. Worksheets
are used to keep the results of the analysis in a logic and comprehensible format. These can be adapted
to each FMEA, but will normally include the columns in table 5. Although the analysis can be
performed by one person alone, it is believed that team work is the most suitable. This is due to the
magnitude of possibilities in the design which may lead to failure (IEC 60812 2006).
Table 5: FMECA worksheet
Function

Operational
mode

Failure
mode

Failure
cause or
mechanism

Failure
detection

Effect
of
failure

Failure
frequency

Severity
ranking

Criticality

Risk
reducing
measures

Control
The FMEA has spread to almost all industries and has been developed for several other purposes than
product reliability, for example project management. Among the standards and handbooks explaining
the method and its application are MIL-STD-1629 (1980) and IEC60812 (2006). Based on the
industry and the relevant standard, it is possible for an organisation to develop its own guidelines.
Another possible means of control is to evaluate the impact the analysis has on the design. If the
analysis is performed untimely, its effect may be hard to notice, but some useful outputs can still come
of it (MIL-STD-1629 1980).
Outputs
The FMEA gives a good overview of the different issues to watch out for in the design and during the
systems operational phase. It can function as an input to several analyses, for example Fault Tree
Analysis and Reliability Block Diagram. The listing of the criticality of the failure modes may be used
to decide necessary alterations in the design and maintenance tasks during the operational life. With a
thorough analysis, the worksheets will give a clear overview of the system, its failure modes and their
effects. This will be useful throughout the product life cycle and in future projects.

Inputs
System design information
Previous knowledge
Operational/Environmetal conditions
Failure modes and identified hazards

Control
Industry standards
Organisation demands
Time/ Cost / Resources

Outputs
Failure mode information (Origin, effect, etc.)
Possible criticality ranking

FMEA

Mechanisms
System knowledge
FMEA worksheets

Figure 12: SADT FMEA

Although the FMEA is well incorporated in most industries, including the subsea process industry,
there are criticisms against it. The method itself is time consuming, especially if absolutely all
component failures are examined to the same level of detail (Rausand & Hyland, 2004). Another
problem is that the method only focuses on one failure mode at the time, leaving it unsuitable for the
study of dependent failures. It is thus a possibility that systems with a fair degree of redundancy are
insufficiently analysed.

Master Thesis, TPK 4900


Ingrid Alms Berg

32

NTNU

Design for Reliability

14.06.2010

5.4. FTA
"A Fault Tree is a graphic depiction or model of the rationally conceivable sequences of events within
a complex system that could lead ultimately to the observed failure or potential failure." (NASA (A)
1999)
The Fault tree analysis (FTA) is a deductive method where the analyst starts with the final hazardous
event and traces it back to the original failure (NUREG-0492 1981). It was first introduced for the
safety evaluation of the launching system for the Minuteman I missile in 1962 (Hyland & Rausand
2004). Later improvement has lead to a very extensive use in most industries where risk and reliability
studies are performed. One reason may be the methods ability to give an overview of an entire system
based on a few problems (NUREG-0492 1981).
For reliability purposes, it is important to start the FTA development as early in the product life cycle
as possible and update it concurrently with the design development (IEC 61025 2006). As the original
hazard is traced back through the possible contributing failure events which lead to it, the contributors
are put into boxes and connected. The connections are based on whether a failure can occur alone or
not in order to induce the next event. Underneath the failure, an and- or an or-gate is placed to describe
the connection between the contributors. The final level in the Fault Tree is normally at the component
level, but this is optional. Figure 13 shows how the procedure is performed and what the end product
looks like.
Fa u lt

1 . Id e n tify fa u lt o r fa ilu re

AND

3 . L in k f i r t-le v e l c o n trib u to rs
to fa u lt b y lo g ic g a te s
2 . Id e n tify p o te n tia l
f i r t-le v e l c o n trib u to rs

OR

AND

5 . lin k p o te n tia l s e c o n d le v e l c o n trib u to rs to f i r sle v e l b y lo g ic g a te s


4 . Id e n tify p o te n tia l
s e c o n d -le v e l c o n trib u to rs

B a s ic E v e n t
6 . R e p e a t/le v e ls /c o n tin u e
Figure 13: Fault Tree Process (NASA (A) 1999)

As the bottom-level is entered into the tree, the contributors failure rates or probability of occurrence
can be included. If all the necessary estimates for this level are obtained, it is possible to estimate the
failure rate or probability of the occurrence of the top event. This will be done by following the gates
from the bottom and up to the top. The method can be found in Rausand & Hyland (2004). FTA is
helpful when a system is complex with many potential failures leading to a larger problem (NASA (A)
1999). It is easy to read, while systematic in its approach.
Inputs
The top event will be chosen from the previous hazard identification analyses, while the system design
is needed to trace the contributors. As the design becomes more detailed, more contributors may be

Master Thesis, TPK 4900


Ingrid Alms Berg

33

NTNU

Design for Reliability

14.06.2010

discovered. To calculate failure rates and probabilities for the top event, estimates for the basic events
are required (IEC 61025 2006).
Mechanisms
The FTA can be performed by one analyst alone, but a very thorough knowledge of the system is
needed. In cases where the systems are too complex for one person, it is recommended that a team is
used (IEC 61025 2006). As the fault trees may become too large to be drawn up manually, it can be
very useful to work with computer tools, for example CARA Fault Tree (Rausand & Hyland 2004).
These tools are often also able to calculate the failure probabilities and failure rates for the top event,
and discover the minimal cut sets.
Control
There are several standards and guidelines existing for the FTA, for example IEC 61025 (2006) and
ISO 17776 (2000). Again the procedure should depend on what the organisation demands of it. How
far down it goes will therefore be decided by the organisations need. A reasonable manner in which it
is possible to see whether the fault tree is reasonable, is through comparison with system breakdown
structures.
Output
The output from the FTA can be used to improve the design, evaluate the possible preventive
measures against failures and give input to reliability block diagrams (RBD). As the fault trees show
the effect one undesirable event has on the system as a whole, it is possible to evaluate whether
preventive measures, or design alterations are the best. The failure rates and probabilities calculated in
the FTA can become useful for the estimation of MTTFs.

Inputs
Hazard events
Previous knowledge
System functional design
Failure event connections
Failure probabilities/ rates

Control
Industry standards
Organisation demands
Time/ Cost / Resources

FTA

Mechanisms
Choice of approach (top-down etc.)
Analysis knowledge

Outputs
Failure rates/ probabilities
Failure event connections
Basic event connections

Figure 14: SADT FTA

5.5. RBD
A reliability block diagram (RBD) presents the connection between the different components fulfilling
a particular system function. The purpose is to show how the system can function or fail depending on
the specific components. Where a fault tree has been made, a transformation to an RBD may be
possible, and vice versa (Hyland & Rausand 2004).

Master Thesis, TPK 4900


Ingrid Alms Berg

34

NTNU

Design for Reliability

14.06.2010

To create an RBD, three types of system information must be studied (NASA (C) 1999):
-

Functional systems architecture data


Component reliability data
Mission times/Operating times

When an analyst has these data, he or she can determine the relationships between the components as
either serial or parallel. The components placed in a k-out-of-n (k-o-o-n) relationship, where the
system functions even when some components fail, will also be noted. An RBD is easy to understand
through its graphical representation. When a diagram of this type is presented, it can prove why
parallel structures generally are considered stronger than series structures. In the former, the systems
ability to function depends on its strongest link, while in the latter it will depend on the weakest.
The creation of the block diagram can only start after the difference between success and failure has
been established (IEC 61078 2006). As there are possibilities for a system function to be in a state less
than 100%, but higher than 0%, one could say that success is above 80%. When this is done, the
system can be divided into blocks that are linked according to how the information passes through the
system. By applying reliability information to each block, pivotal decomposition may be used to
calculate the system reliability. This is also useful for the analysis of component importance through
methods such as Birnbaums measure. Pivotal decomposition and component importance is described
in Rausand and Hyland (2004) and will thus not be further discussed here.

Figure 15: RBD

Inputs
The inputs to the Reliability Block Diagram are system design, failure definitions, k-o-o-n
relationships and redundancy information. A fault tree can be turned into an RBD and the other way
around, but the RBD is the easiest to read with respect to the connections of components, assemblies
and subsystems (Rausand & Hyland 2004).
Mechanisms
In order to perform the RBD, knowledge about the system is needed, both concerning its composition
and how the elements work together. If the system is very complex, including k-o-o-n structures,
switches and parallel structures, it is useful to perform the analysis in a team. RBDs can become very
large and complex, and it is thus helpful to employ computer programs. Such a program will keep the
information separate and facilitate calculations such as component importance and overall system
reliability.
Control
What controls the method is the interest of the organisation ordering it and the design progress. The
decision of what a success is will be decided by the system requirements. Creating the RBD, the

Master Thesis, TPK 4900


Ingrid Alms Berg

35

NTNU

Design for Reliability

14.06.2010

design is the basis and it is thus also the main element which may show that the RBD is put together as
it should be. Any changes in the design should be a reason to update the RBD. The RBD is described
together with Boolean methods in IEC 61078, which may be used for control of the performed RBD.
Outputs
One useful type of output from the RBD is minimal cut sets showing the smallest number of failed
components leading to the loss of a system function. This can help the understanding of the weakest
links within a system. If the RBD is used to find the system reliability and component importance, it
could be entered into other analyses, or used to compare with previous studies. The downside with the
RBD is that it cannot be used for repairable systems (Rausand & Hyland 2004). These should be
analysed through Markov diagrams, which are described in Rausand and Hyland (2004).

Inputs
System design
Functional relations

Control
Industry standards
Organisation demands
Time/ Cost / Resources

RBD

Mechanisms
System knowledge
Block diagram construction
Pivotal decomposition

Outputs
Cut sets
System reliability
Component importance

Figure 16: SADT RBD

5.6. HAZOP
The hazard and operability study (HAZOP) was first published in the late 1960s for ICI Mond
Division and further developed and published for the ICI Petrochemicals Division in 1974 (Swann &
Preston 1995). It is a structured and systematic method, performed as a team work with members
specialised in different areas necessary to understand the defined system (IEC 61882 2001).
The objective is to identify the potential hazards in the system and the possible operability problems,
for instance causes for operational disturbances and deviations (IEC 61882 2001). An attempt is made
to find all possible deviations from the intended design of the system. The search is performed through
the use of a number of guide words triggering the imagination of the team members. An example is
the word More which is supposed to bring the thought to a problem where the deviation is an
increase compared with the intended design. If it is used for a component controlling the amount of
pressure, more could mean that a valve accepts a higher pressure than intended. Figure 17 shows
how the analysis may be performed.

Master Thesis, TPK 4900


Ingrid Alms Berg

36

NTNU

Design for Reliability

14.06.2010

Start
Explain overall design
Select a part
Examine and agree design intent
Identify relevant elements
Identify whether any of the elements can be usefully
sub-divided into characteristics
Select an element
(and characteristics if any)
Select a guide word
Apply the guide word to the selected elements (and
to each of its characteristics as relevant)
to obtain a specific interpretation
Is deviation credible?
No
No

Yes

Have all interpretations of the guide word and


element/characteristics combinations been applied?
Yes

No

Investigate cause,
consequences and
protection or
indication, and
document

Have all guidewords been applied to the


selected element?
Yes

No

Have all elements been examined?


Yes

No

Have all parts been examined?


Yes

Stop
Figure 17: HAZOP process (IEC 61882)

Inputs
The input to the HAZOP is the system design and the operational information for the system. If similar
systems exist, hints may be acquired for the guide words. A HAZOP study can be done in several life
cycle phases, but is recommended performed just before the design is fixed. The design should then be
detailed enough to contain the information needed to answer the guide words (IEC 61882 2001).

Master Thesis, TPK 4900


Ingrid Alms Berg

37

NTNU

Design for Reliability

14.06.2010

Mechanisms
In order to perform the HAZOP, team work is necessary. Together, the members should agree on the
guide words and how the system is to be studied. Little is needed to perform the HAZOP and
depending on the extent of the study, the analysis may be performed within a small time frame.
Control
To control the analysis, the IEC 61882 (2001) is a good instrument. Any specific demands for the
output of the analysis will also be of interest. The design drawings can be utilised as a proof of
whether the identified problems are realistic or not.
Outputs
When the HAZOP has been performed, the outputs are used for improvements in the design, and to
alter routines in the operation. HAZOP can be performed for several purposes, and is very useful for
the discovery of problems to watch out for in the manufacturing phases.

Inputs
System design
Process information
Previous knowledge similar systems
Operational/ Environmental
conditions

Control
Industry standards e.g. IEC 61882
Organisation demands
Time/ Cost / Resources

HAZOP

Mechanisms
System knowledge
Team work

Outputs
List of hazards
Hazard information
Process and system alteration ideas

Figure 18: SADT HAZOP

5.7. Reliability Growth


Reliability growth is a physical test method. The intention is to improve the reliability through a testanalyse-and-fix technique. A component will be tested under increasing amounts of stress until failure.
The test results are then analysed and if they do not satisfy the requirements, improvements will be
made. This process continues until the reliability requirements are met with (Murthy et al. 2008).
For products that are very expensive and built as one-of-a-kind, it can be hard to perform reliability
growth tests. Reliability growth is meant to lead to a failure, which cannot be accepted for the overall
product in this case. Sub-systems and components which are not expensive and time consuming to
develop can, on the other hand, be subject to the tests.
What a reliability engineer must be aware of is that the amount of runs which are performed in the
reliability growth process has an effect on the outcome. For every run, the component or system is
pushed further towards failure. This must be considered in the analysis as the previous load can mean
that the system fails in run three, although it can handle more stress than what it was subjected to.

Master Thesis, TPK 4900


Ingrid Alms Berg

38

NTNU

Design for Reliability


Test

Test results

Modifications

NO

14.06.2010
Analysis

Reliability targets
met with?
Yes

Yes

Further
improvements
needed?

No

Analysis
finished

Figure 19: Reliability Growth process

As reliability growth tests usually are performed as early as possible during the products physical life,
little knowledge is available. The results of the tests must therefore be studied carefully by analysts
with good understanding of the necessary physical and mathematical analyses, and an ability to make
sound judgments (Blischke & Murthy 2000). In the subsea process industry and other industries where
the products are the prototypes, reliability growth tests are restricted. The tests must then be performed
on components that can be tested to failure and are not too expensive for this purpose. What the
analysts have to remember in this case is that the results from component tests may not give all the
answers. When the component becomes a part of the complete system, the possibility of new failures
might occur that were not found through previous testing (Weibull.com (B) 2010).
Inputs
To perform the reliability growth test, prototypes of the components and the results of analyses
performed implying necessary testing will be needed. Other information will be the outer limits a type
of material can handle under stress, and how the tests are performed.
Mechanisms
To perform the tests, qualified personnel is needed. Proper equipment and laboratories with good
environmental conditions for the tests are also necessary mechanisms of this method. Some of the
more important factors for a successful reliability growth test are good management, tests that give
comprehensible answers, the ability to identify the root cause of failures, effective corrective actions
and valid reliability assessments (Weibull.com (B) 2010). The reliability found in this type of analysis
is based on how the system responded to the testing and corrections. It is therefore vital that the
measures taken are thought-through and sensible.
Control
Control of the reliability growth tests will depend on resources, time and money. The number of runs
will vary with time and resources. The organisation developing the system may also have some
demands concerning which tests the system should undergo and the conditions they should be
performed in, for instance specific temperatures. IEC 61014 (2003) suggests programmes for
reliability growth and may be followed both as a mechanism and as control.
Outputs
The main outputs will be how the tested elements handle the stresses they are subjected to and whether
alterations must be done to the production of the system. If the reliability growth tests are used on

Master Thesis, TPK 4900


Ingrid Alms Berg

39

NTNU

Design for Reliability

14.06.2010

materials in a very early phase, the output might be used as inputs to the design phase when the
materials are chosen.

Inputs
Tests
Previous knowledge similar systems
Operational/ Environmental conditions
Prototypes
Reliability
Materials information
Growth

Mechanisms
Qualified personnel
Test equipment
Analysis

Control
Organisation demands
Time/ Cost / Resources
Product operational requirements

Outputs
Stress handling
Need of alterations materials etc.
Improved equipment
Design alteration suggestions

Figure 20: SADT Reliability Growth

5.8. Accelerated testing


Accelerated life testing is a method where an object is forced to fail. Failures are induced by
submitting the product to stresses it will experience during its operational life, but at a larger scale. A
reliability engineer is interested in failure data for the operational life, for example failure rates. The
main problem with such data is the time it takes before a failure. Databases can be useful to obtain
failure data, but they rarely consider the differences between the operational conditions of the items
that have failed (Weibull.com (A) (2010)). Accelerated life testing can therefore be a solution both to
the environmental differences and the time it takes to obtain the data.
The tests for this method can generally be split into two categories; qualitative and quantitative tests.
The former is useful for the detection of failure modes and failures, while the latter gives input to life
predictions and failure data, for example MTTF (www.weibull.com). Qualitative tests are primarily
made to pressure the product to failure and provide results fast. Examples of such tests are the elephant
test, the torture test, HALT and shake and bake tests. These do not necessarily employ time as a
measure and it can thus be hard to say something about the life time of the product. Quantitative tests
can be divided in two; accelerated usage rate tests and overstress acceleration tests
(www.weibull.com). The first is convenient when the products are not in continuous use, such as
dishwashers and coffee machines. Possible testing methods would then include using the products
with an elevated frequency.
As with reliability growth, custom-built products might not suit this type of testing. The components
and sub-systems which are easy to produce and inexpensive can be tested instead. A very important
aspect of this method is that the reliability engineer must be aware of what type of product he or she
handles. A product which is in continuous use should not be subjected to accelerated usage tests. For a
product which is turned on and off, it might be interesting to use both accelerated usage tests and
overstress acceleration tests. For qualitative tests, it must be evaluated whether it is likely that the
product will be subject to increasing pressure over time.

Master Thesis, TPK 4900


Ingrid Alms Berg

40

NTNU

Design for Reliability

14.06.2010

Inputs
Accelerated life tests cannot be performed without knowledge about the materials in the system, the
operational and environmental conditions and the possible system failures. Based on this information,
it can be possible to deduce which tests are the most appropriate.
Mechanisms
As this method is physical, it is very important to have the appropriate resources. For the test
operators, it is necessary with knowledge about the materials used in the product and its mechanical
construction. Otherwise the test may be performed incorrectly or the results misinterpreted. The
analysts will also need this knowledge, in order to subtract the correct results.
Control
The main constraints and controls for this type of analysis are time, money, legislations and the
product type. Time issues affect the tests available, especially when they demand much time for
preparation and execution. Depending on the tests chosen and the materials used in the product, it may
be very expensive to perform tests. If the materials are expensive to obtain, it is not desirable to use it
for destruction alone. Legislations can affect the tests through the time an operator is allowed to work
or the emissions they might lead to. When products only are produced once or in very few numbers,
the accelerated life testing can be difficult to employ. Among the standards suitable for the control of
this method are IEC 62059 (2008) and IEC 60068 (1995).
Outputs
The outputs from the accelerated life tests may be used to confirm previously predicted reliability
estimates. They can also help improve the manufacturing process and the design. If the tests are
performed after the design is frozen, alterations will be difficult. A possible use of the outputs could be
for maintenance and check-ups on specific modules and sub-systems which seems more probable to
fail. If it is desirable to test before the design is frozen, it is possible to do testing on items which
already exists. When the results are analysed and given as failure data for a longer life, they can be
stored for later development projects.

Inputs
Predicted failure scenarios
Operational/ Environmental conditions
Materials information

Control
Industrty standards
Organisation demands
Time/ Cost / Resources
Product options

Accelerated
testing

Mechanisms
System and test knowledge
Test equipment
Results reporting

Outputs
Failure information
Maintenance input

Figure 21: SADT Accelerated testing

Master Thesis, TPK 4900


Ingrid Alms Berg

41

NTNU

Design for Reliability

14.06.2010

5.9. FRACAS
Failure reporting analysis and corrective action system (FRACAS) is a closed-loop method where the
supplier and customer work together to study the product reliability. The main purpose is to have all
failures in both hardware and software systems reported, analysed and understood. All information
concerning a failure will be recorded, identificying the failed items, symptoms, operating times and
time of failure (MIL-HDBK-2155 1995). The verification of the failure and the successful corrective
actions are important to prevent the failure from recurring (MIL-HDBK-2155 1995).
Inputs
For reliability purposes, FRACAS is best used together with other analyses. The intention of this
method is to use it as a tool while the product or system is in use. Other methods suited for earlier
phases should be used as inputs. An example of how a combination of FRACAS and another method
can be favourable is the coupling of FRACAS with FMEA/FMECA (MIL-HDBK-2155 1995).
FMEA/FMECA can give input to FRACAS by descriptions of failure modes which are encountered.
A failure reported through FRACAS can be brought back to complete the FMEA/FMECA.
Mechanisms
FRACAS is dependent on accurate input data, well prioritised goals, and time and resources (Hallquist
& Schick 2004). A reporting system should be agreed on between the manufacturer and the client, and
used as a mechanism for the reporting. The people reporting and analysing the failures needs thorough
knowledge of the system and the nature of the failures affecting it.
Control
In FRACAS, the extent of the failure is a control parameter. A failure leading to a destructive
consequence takes time and money to investigate. If the destructions are extreme, it might not even be
possible to find any answers. Another control issue is the agreement between the manufacturer and the
client. This will regulate how the reporting is done and what information the parties are sharing.
FRACAS is shortly described in IEC 60300 (2003) and SEMATECH (Villacourt and Goviul 1994)
which may be used to follow up the planning and use of the method.
Outputs
Outputs of the FRACAS method are of use to later development projects. All lessons learned will lead
to less repetition of failures and history might therefore not repeat itself.

Inputs
Results from reliability methods
System design
System usage
Operational/ Environmental conditions
Reporting agreements

Control
Organisation demands
Time/ Cost / Resources
Reporing agreement
Extent of failure
IEC 60300/ SEMATECH

FRACAS

Mechanisms
System knowledge
Team work
Worksheets
Manufacturer-client communication

Outputs
List of failure events
Failure information
Repair information
Lessons learned

Figure 22: SADT FRACAS

Master Thesis, TPK 4900


Ingrid Alms Berg

42

NTNU

Design for Reliability

14.06.2010

5.10. Discussion
The methods described in this chapter are some of the more well-known tools used by reliability
engineers. All of them are applied across most industries, depending on product type. Which method
one ends up with should depend on the life cycle phase, the desired outputs, the product and the
available resources. A methodology for Design for Reliability must contain a selection of different
reliability methods based on their ability to complete each other.
Early in the life cycle, the method choice will be constrained by the existing information about the
product. Many parts may not be specified and changes will probably occur later on. During later
phases, the input to the analyses can be found from other similar products, methods performed in the
previous phases and tests. Depending on the cost, size and utilisation of an item, certain methods may
not be applicable at all.
Using SADT as a tool to evaluate the applicability of a method can be preferable when there is a need
to act in accordance with specific constraints and demands for the desired outputs. If a reliability
programme is desired, the SADT can show which methods may follow one another to complete each
other and the reliability picture. It is for example evident in this chapter that FMECA and FTA are
useful together in order to show both simple and dependent failures.
Some of the methods are already used together in specific types of analyses, in order to develop a
specific overview of a product. One such method is the reliability, availability and maintainability
(RAM) analysis. This uses FTA and RBD among other methods.

Master Thesis, TPK 4900


Ingrid Alms Berg

43

NTNU

Design for Reliability

14.06.2010

6. RAM analysis
A RAM analysis considers the reliability, availability and maintainability of a system. As mentioned
in chapter 2, reliability and availability are tightly connected through their definitions. Maintainability
is an important parameter as it describes the downtime in case of a failure; if the maintainability is
poor the availability will decrease. Reliability and maintainability defines, in large parts, the
availability of a system.
The RAM analysis should be used in early design phases in order to contribute to the optimisation of
the design of a system. The RAM analysis can give valuable information with regard to recommended
sparing and repair philosophies, as well as advice in connection with the optimal redundancy
introduced into a design. In later phases of a project the RAM analysis is used mainly as a verification
activity to show compliance with requirements.

6.1. Method
RAM analyses use information about expected time to failure of components, downtime in case of a
failure, spare part philosophies, and planned maintenance in order to study the overall availability of a
system. Due to the amount of information handled, it is not uncommon to use computer tools. These
will then use the input information, relate it according to the system design and operation, and
simulate the events of the operational life to obtain an overall availability.
In order to build a model of a system in a RAM analysis computer tool, the following inputs are
required:
-

Definition of system boundaries, assumptions and limitations


Identification of main failure modes and corresponding effects
Definition of reliability input data on component/module level
Definition of repair requirements

After the study boundaries for the system have been identified, FMECA can be a useful input. The
FMECA provides possible failures for each system component, the cause and criticality of each
failure, and detection and repair methods. If possible to determine, the detection time of a failure may
also be useful.
To establish detailed reliability data for each component or sub-system it is often useful to prepare
RBDs or FTAs. These methods allow for detailed assessments of each component in order to establish
failure rates and MTTF estimates.
The downtime caused by a failure includes the period from detection of a failure until the repair has
been carried out and the system is in operation. This is an important subject which can be studied by
RAM analyses. For some systems the downtime will be more extensive than others. While a radio
antenna which cannot receive signals only needs to be changed for the radio to work, a subsea process
system requires far more specific equipment for repair.
If a module in a subsea process system needs to be changed, a replacement must be brought from
storage to the specific location of the system. When the sub-systems and modules weigh several tons,
it is evident that diving is not an option for installation, sub sea repair and retrieval. Vessels capable of
carrying the load and lowering them into the sea and place them correctly are needed, as well as
technology such as ROVs (Remotely Operated Vehicles). The ROVs can connect the different

Master Thesis, TPK 4900


Ingrid Alms Berg

44

NTNU

Design for Reliability

14.06.2010

modules and operate manual valves on the system. Information about which vessels are required along
with the mobilization time of these vessels is required as an input to a RAM analysis of a subsea
process system whose downtime must be found.
For the intervention vessels, the period defined as the mobilization time includes the period from
detection of failure until the vessel is located and properly operational at the repair site, including:
-

Diagnostics of failure and recommendation of repair activity.


Preparation and approval of work order issued to Service Company.
Spare part and tool handling, as well as possible transportation
Identification and organization of appropriate intervention resources, clarification
of responsibilities, crew familiarizing with the failure/problem.
Preparation and approval of mission procedures
Manpower/resource transportation to field, preparation for intervention (including
vessel demobilisation from other work, transit to shore , mobilisation of dedicated
equipment for intervention on the site and transit to field).

The active repair time, mean time to repair (MTTR), is the time from the intervention vessel has
arrived at the Midgard field, to the time when the fault has been rectified (by repair or replacement of
equipment).
A reduced downtime can be obtained through an effective sparing philosophy. If a failure occurs in a
component, the downtime will depend upon whether the component has to be repaired, or whether
replacement by a spare is sufficient. If no spares are available the mean time to repair will increase
substantially. When a spare is available the expected active repair time will be reduced to a couple of
days (retrieve and replace on site). If no spare is available the active repair time may vary from a
couple of months to a year.
Capital spares should be available for the most critical items in a system; this will significantly
improve the availability of a system. The location of the spare parts and the time required to load them
onto a vessel will also be important factors with regard to the maintainability of a system.
When all the necessary input parameters have been determined, the computer model of the system can
be built. The input parameters are logged for each item, including failure rates, repair times,
mobilization times of vessel, spare parts required, and amount of production lost upon a failure etc.
Factors such as weather conditions will also be included in the analysis.
All possible combinations of events, and repair operations will be randomly simulated in the computer
program and the overall availability calculated from it.

6.2. MIRIAM Regina


MIRIAM Regina is a commercial software package for reliability and availability analysis. It was
developed in close cooperation with Norwegian oil companies and has been used for several field
development studies. In the Subsea Power and Process department it is used to evaluate the
operational performance of subsea process systems, considering the availability, production capability
and maintenance resource requirements (Rausand 2005). The results are based on stochastic
simulations of the inputs sent through a model describing the system.

Master Thesis, TPK 4900


Ingrid Alms Berg

45

NTNU

Design for Reliability

14.06.2010

The model s defined through boundary points, process stages and storage points. Boundary points are
where the inputs enter and the outputs exits the system. Process stages are components or functional
units as they are in real life, often broken down to sub-component stages. The storage points are the
buffers of the system which can be used to contain or release flow depending on restrictions in the
system (Rausand 2005). The entry point, functional units and exit point are linked together through
connections, describing the shape and flow paths in the model.
Before MIRIAM Regina is used, the following is done:
-

Definition of study objective, approach and metrics


Definition of study boundary, assumptions and limitations
Identification of main failure modes and corresponding effects (often through
previous FMECA)
Definition of reliability input data on component/module level (failure rates and
MTTF, can be found through detailed assessments of each component through
FMECA, FTA and RBD - see descriptions below)
Definition of repair requirements, repair time and intervention vessel mobilization
time

When all inputs and necessary data are entered into the program, simulations including sensitivities for
different concepts can be run. MIRIAM Regina uses the Monte Carlo simulation technique which
generates random values for the uncertain variables, based on the probability distributions describing
the inputs (Berg 2009). A number of simulation runs are made in order to obtain a full picture of all
events which may occur, and how they affect the systems availability. The Monte Carlo simulation
technique is also referred to as Monte Carlo next-event simulations and can be studied further in Berg
(2009), Rausand (2005), and Rausand and Hyland (2004). The results of the simulation runs are
transported to excel sheets where they can be further studied. These results include production
availability and deliverability, subsystem criticalities, resource and spare part usage etc. (Rausand
2005). The outputs of most interest to Aker Solutions and their clients are the availability and the
productiveness, which together describes the production regularity.
Regularity is a term used to describe the capability of a production system of meeting demands for
deliveries and performance (Rausand & Hyland 2004).

6.3. Use of outputs


The outputs from MIRIAM Regina and RAM analyses can be used as foundation for decisions
concerning necessary design alterations, maintenance programmes and operation of systems. Among
the outputs there should be a critical components list. The critical component list can be used as input
to suggest design changes, such as increased redundancy, that will improve the reliability of the most
critical components, or to suggest an optimal maintenance and sparing philosophy.
The availability of a system is in a large part affected by spare part and intervention philosophies. The
outcome of the RAM analysis can show where capital spares are required, the number of spare parts
required for each item, and it can also give input to the appropriate storage of these spare parts.
The mobilisation time for spares should be as short as possible. If it is shown that some spare parts
need far more time to be obtained and transported than is desirable, it can be decided that one spare
part should be installed on the seabed or moved to an easier location.

Master Thesis, TPK 4900


Ingrid Alms Berg

46

NTNU

Design for Reliability

14.06.2010

The RAM analysis gives an overview of the number of interventions required per year for the system.
An overview of the intervention frequency can be used for operational expenditure (OPEX)
calculations with regard to vessel costs etc. and to outline an effective maintenance philosophy.
As with outputs from any methods where probability is used, there are uncertainties. MIRIAM Regina
is no exception, and basically there are two types of uncertainties which an analyst must be aware of;
parameter uncertainty and model uncertainty. The input parameters may have errors in them, either the
chosen distributions are wrong or the inputs are incomplete or erroneous in themselves. This is
parameter uncertainty. Model uncertainty is directly connected with any models chosen or made for
the simulation. These may have errors in them or not be extensive enough to explain the real system.
Other types of uncertainty exist and can be found in the RAM analysis, but these are of less
importance.
Rausand (2005) described a problem with MIRIAM Regina concerning the lacking ability to model
any other probability failure distributions than for the time to failure. There are no possibilities for
implementation of probability distributions on other reliability parameters such as failure rates.
Uncertainty for reliability parameters is almost impossible to implement and the thought of their
quantification was not even a concern when MIRIAM Regina was developed. In the Norwegian
petroleum industry, the main inputs often come from OREDA. The opening chapters of the OREDA
handbook mention the assumptions made and that there are uncertainties in the data (OREDA 2002).
This, along with the fact that new systems rarely have much experience behind them, should make it
evident that more considerations of uncertainty could be desirable in MIRIAM Regina. Problems and
limitations are further discussed by Rausand (2005).
Although there are uncertainties associated with the tools for RAM analyses, the outputs can be of
great importance. To obtain an exact calculation of the availability is not the most important
application of the RAM analysis tool. It is rather more interesting to identify the most critical
components in a system, and to use the tool as an input for optimisation of design, sparing and
intervention

Master Thesis, TPK 4900


Ingrid Alms Berg

47

NTNU

Design for Reliability

14.06.2010

7. Unreliability
A search for synonyms for the word unreliability returns inaccuracy, instability, credibility gap and
insecureness among other answers (Thesaurus.com 2010). These are all good substitutions although
they might be considered too specific when they are used to describe an item. Unreliability is defined
as not reliable or not dependable. This would mean that it is the exact opposite of reliability, but
the origin of unreliability might as well be the same as for reliability.
Before starting a discussion of factors which may contribute to unreliability, it is useful to understand
what is meant by the term unreliability. Reliability is considered in terms of probability, thus it is
always between zero percent and 100 percent. One can think of unreliability as 0 percent on the scale,
being the opposite of reliability. But, another option is to think of 0 percent as no reliability
expected. This would render the word unreliability useless. If a product is 100 percent reliable, one
knows that it cannot fail to perform its intended function. When the product is believed to have no
reliability at all, one does not expect the product to perform the function one considers using it for.
This implies that the product is outside of the perspective of reliability as reliability depends on an
intended function. Unreliability can be the opposite of reliability, but when we are considering a given
product, we have expectancy of its ability to function. Keeping unreliability as a term for when the
product is expected to function, but fails to do so, we can look at unreliability as something which
decreases the reliability. Factors contributing to either reliability or unreliability could determine
whether a product moves up or down the probability scale for its ability to function. Factors
contributing to unreliability are negative, while factors contributing to reliability are positive.

Neutral

No ability to function

Absolute reliability

Unreliability

Reliability
System

Figure 23: Unreliability - reliability

The calculations and predictions of a systems reliability are based on assumptions, reliability methods
and previous experiences. If any of these are unreliable, unsuitable or uncertain, the actual reliability
will be different from the predicted reliability. If the inputs to the reliability analysis are unreliable or
unsuitable themselves, and known to be so, they can easily be excluded. The problem is if this is not
known. In that case, unreliable or unsuitable factors are basically the same as uncertain factors.
Uncertainty is a problem where the lack of knowledge leads to the wrong inclusion or exclusion of
information.
Uncertainty usually arises from a lack of knowledge or randomness in samples (Pat-Cornell 1996),
where the former is referred to as epistemic uncertainty and the latter as aleatory uncertainty.
Unreliability arising from uncertainty can be affected by both of these categories and if the category is
known, there may be a possibility of reduction as well. Aleatory uncertainty is unpredictable and is
accepted through the assumption that a phenomenon has an intrinsic randomness (Kiureghian &

Master Thesis, TPK 4900


Ingrid Alms Berg

48

NTNU

Design for Reliability

14.06.2010

Ditlevsen 2009). More knowledge about the phenomenon is not believed to alter this type of
uncertainty, but it is thought possible to study the randomness through uncertainty probability
distributions. These are used to show the outcome of the randomness and give an indication on how
the uncertainty of the input parameters is distributed in the output. Epistemic uncertainty is reducible
through increased knowledge. It can be split into either model, parameter or data uncertainty and be
studied according to what type of uncertainty it is (Drouin, Parry et al. 2009). The concept of
uncertainty and how it can be specifically studied and understood is further discussed in Berg (2009).
The reliability of a product can also be altered by the events and thoughts which where not brought to
the table during its design and development. Whether such events are positive or negative to the
reliability cannot be known in advance, but due to the possibility of a negative outcome, care should
be taken to avoid it throughout the process. The basic factors may be discussed and defined by their
origins, but to understand how they develop and affect a system, it is perhaps more useful to look at
them through the phase they arise in. The eight phases presented by Murthy et al. (2008) are used here.
In each phase the factors will be discussed in a general and a subsea perspective.

Phase 1
During this phase a customer presents a request proposal to possible manufacturers who reply with bid
proposals (Murthy et al. 2008). The request proposal will include information on what functions the
product must be capable of performing, performance requirements and general information about
environment etc. A bid proposal will indicate how the product can perform the desired functions and
be realised. It should also indicate the performance levels and the costs. After one or more rounds
where the possible manufacturers are narrowed down to one, a contract is signed and the main control
of the project handed over from the customer. This phase ends when a decision is made whether to
proceed with a development project or not.
There are four main factors which can have an undesirable effect on the reliability in this phase. These
are:
-

Bad communication
Lack of knowledge
Lack of information/ data
Misunderstanding/ misinterpretation

Communication is the main activity during the first phase and if it is not executed properly, it will be
hard to know what the desires and requirements truly represent. All the information is given through
communication channels of some kind and it is how these channels are handled that is of importance.
A well known childrens game is the whispering game, where one child starts out whispering one
phrase to another child. The second child whispers it to a third child who continues the chain. This
ends when the last child in the group has been given the phrase and speaks it out loud. A very normal
effect of the whispering chain is that the end result turns out quite differently from what the first child
whispered. This is an example of how communication with many segments and different
understandings can be unfortunate.

Master Thesis, TPK 4900


Ingrid Alms Berg

49

NTNU
L is a h a s a c a t
n a m e d F re d

Design for Reliability


L is a h a s a h a t
n a m e d F re d

14.06.2010
L is a h a s a h a t
m a d e o f b re a d

Figure 24: The whispering game

Another problem occurring with communication is when the basic definitions and knowledge used by
the communicators is different. This can be exemplified through the use of technology readiness levels
(TRLs) in the industry. While a client has one standard they follow, the manufacturer might use
another. Normally this would be investigated and the best solution chosen, but in the case that it is not
discovered, a product might be approved on the wrong grounds in a later phase. A possibility is that
those using the documents are of different nationalities, with different mother tongues. Their
understanding of documents, such as TRLs, might then lead to problems which are not discovered.
Communication cannot be avoided in this phase or any other phase, but it can be improved through an
elimination of the number of elements in the communication line. Language difficulties will always be
present, but this should be a good reason to ask more questions in order to ensure that the
understanding is the same.
Lack of knowledge and lack of information are here separated and given different meaning. The first is
meant as a lack of knowledge and education of the users of the information, while the latter indicates
that all the available information has not been obtained. Lacking knowledge can lead to misuse and
misinterpretation of analysis methods and their results. In the first phase this means that any studies
done for the establishment of a concept can give a wrong belief of what is a feasible product. The
problem will exist throughout all phases of the life cycle as new analyses are done and new possible
failures and hazards discovered. The lacking knowledge will also affect the communication if one
party does not have the ability to understand the other.
Lack of information is an epistemic uncertainty. It can be improved through a search for more
information and data, but if it is not known that something is missing, it will be hard to see the reason
to do more research. Any epistemic uncertainty leads to shortcomings in the inputs to reliability
analyses and will in the first phase render possible decisions on the wrong grounds. If Statoil does not
have appropriate information about the pressure in a well, the functional design specification for an
SCS will be wrong and in the end the system may fail to work as intended.
Misunderstandings and misinterpretations of data can happen at any time. In the first phase it will have
the same effect as lack of knowledge and information. It is an epistemic type of uncertainty, were the
information is available and easily altered for the correct use. It should hence be obvious that a second
evaluation of any interpretation of data must be done. Lacking information takes time to retrieve and
may be overcome if other information can indicate the correct answer. Misinterpretation, on the other
hand, is easily done and easily corrected.

Master Thesis, TPK 4900


Ingrid Alms Berg

50

NTNU

Design for Reliability

14.06.2010

Many problems may arise in the first phase and any factors known to contribute to uncertainty must be
studied thoroughly. If an error arises in this phase and is allowed to live on, it will continue to affect
the system throughout the other phases. While reliability analyses can overlap each other and remove
problems in later phases, no analysis can alter the problems and errors in the requirements set in the
first phase. These must be correct to begin with.
Subsea Process Industry
During this first phase, the Subsea Power and Process department will do research, including
reliability estimates, to find out whether the project is feasible and worth doing. Normally little has
been done of reliability studies this early, but in order to win the project, it is often important to show
that a certain reliability or availability can be met with. This will be done through estimations based on
previous projects and the OREDA handbooks.
The OREDA handbooks are based on data collected from the offshore installations of operators in
Norway. All data collections will have associated uncertainties. The uncertainties are due to the data
being collected in different locations with different operational conditions (OREDA 2002). The failure
rate data are assumed constant, as explained for the bathtub curve in chapter 2. Any assumptions and
misunderstandings connected with the appropriate use of these data can lead to the impression that a
reliability requirement can be reached more easily than in reality. The effect of this will mainly be that
more work has to be done later in order to achieve the desired reliability. Mitigation can be possible
through thorough knowledge of ISO 14224 (ISO 14224 2006)

Phase 2
When the focus of the product life cycle turns to technical specification at product level, one enters
into phase 2. First, the desired product reliability will be derived from the requirements set in phase 1.
The specifications and customer requirements can include variables, such as availability, which can be
used to obtain this. During phase 2, alternative system architectures for the product are reviewed and
the best response to the desired performance is chosen. An important feature of phase 2 is how the
engineering of the product is included in the product development. Before this stage, the product is
more of an idea, than on the paper with realistic measurements and technologies supporting it. As the
system architecture is decided, models can be used to find the overall predicted reliability and product
performance. FTA, RBD and FMECA may be used and the two first ones can be particularly useful to
find the best architecture. Any results will be measured against the desired reliability and performance.
Only when these are reached can phase 2 be completed.
The main factors which may lead to a decrease in the actual reliability during this phase are:
-

Wrong assumptions and interpretations


Miscalculations
Parameter uncertainties
Wrong time and cost limits

The first two are epistemic uncertainties, while the latter can be both epistemic and aleatory
uncertainty. Parameter uncertainty can both be a wrong choice of parameter, and randomness
connected with the chosen parameter.
If the desired product reliability is interpreted wrong, or the assumptions concerning the system
architecture is badly thought through, the reliability may be allocated erroneously. While allocating the
reliability, calculations will be made which can go wrong. Nothing should be accepted too lightly at

Master Thesis, TPK 4900


Ingrid Alms Berg

51

NTNU

Design for Reliability

14.06.2010

this stage. A wrong allocation can lead to too low reliability being demanded from one or more parts
of a system. Another possibility is that a part which actually can have lower reliability is set higher
than one which needs high reliability. Both possibilities will lead to a wrong understanding of the
system throughout the project. If the desired product reliability is wrong, either in itself or for the
components and sub-systems, a manufacturer might release a system which should not have been
accepted for release.
Even if the allocation is correct, the system architecture can be wrong. Poorly chosen system
architecture can lead to more specific product characteristics being left out. This will also decrease the
chance of reaching the highest reliability possible.
When methods for reliability are used, it is important that the analysts have expert knowledge of the
method, and a good understanding of the inputs to get relevant outputs. When the architecture of a
system or sub-system is based on the results from an FTA, the mistakes made during the analysis will
have an impact. Possible factors contributing to unreliability may therefore be wrong use of the
method, the inputs or the outputs of an analysis. Wrong use is epistemic uncertainty, as is a wrong
interpretation of the outputs. If the input parameters have some randomness in them, aleatory
uncertainty will propagate through the fault tree into the output. As said at the Subsea Power and
Process department: Shit in equals shit out.
This phase, as most other phases, will be constrained by limitations of time and costs. With a project
management unable to understand the content of a phase, the limits can be wrong or poorly exploited.
If it is discovered that a technology is unacceptable while an FTA or FMECA is performed, it will take
time to find more information or other alternatives. If there is no time available, the results will be
meek, showing only that the system is not good enough. The same can be the result if there are no
resources available to perform further analyses. This might lead to the closure of a perfectly feasible
project. Such problems can occur and should be accounted for when the phase is planned.
The main factors which contribute to reduced reliability in this phase are related to uncertainty about
the preferred characteristics/configurations for a product. Uncertainty about the allocation of the
reliability requirements to these configurations will also be important to follow. The main problem is
to make the correct choices, both for specifications and design. A bad choice will persist and possibly
accumulate throughout the whole product life cycle if no changes are made. The specific challenge of
phase 2 is to have the correct understanding of the design and notice how the reliability of one item
truly affects the whole system.
Subsea process industry
When a subsea process system development reaches this point, the system architecture will be
developed by a team of engineers with different backgrounds. If anyone among them assumes that a
decision is good enough without knowing how it affects the system reliability, weaknesses in the
system design may be the result. It is therefore important to involve a reliability engineer in the design
process to give input to, for example, choice of a parallel structure over a series configuration.
A subsea project faces all the challenges discussed for this phase. Phase 2 will contain estimates
which, unlike phase 1, have a direct impact on the choices made for the system design. No
assumptions can be made without good reasoning and the OREDA database uncertainties must be
understood to the fullest. The parameters from OREDA will contain aleatory uncertainties and must
thus be treated as such. Any miscalculations can make a large impact on the results. As the technology
is advanced, often complex and maybe even new, the time and cost frames for the design phases must

Master Thesis, TPK 4900


Ingrid Alms Berg

52

NTNU

Design for Reliability

14.06.2010

be decided with care. A rushed design will not benefit the reliability of the product. During these early
phases, reliability issues can be discussed freely and problems be eliminated without physical
interventions, and at low cost.

Phase 3
This is the phase where the detailed design takes place. Based on the decisions made in phases 1 and
2, the engineers must decompose the system into sub-systems and all the way down to the component
level. From the component level, the engineers should go upwards and find the total predicted
reliability based on the new information, and decide whether it is acceptable or not. The reliability
predictions are based on existing knowledge about other similar components. A design should be
accepted if proven to have an acceptable reliability. Through the reliability allocated in phase 2, the
proof can be found faster than if no allocation has been done. Methods for estimation of reliability that
have been applied in the previous phase should be updated as more information becomes available.
When the engineers and designers work in this phase, it is very important that their choices are
thoroughly considered and that all previous decisions are included. The product should not waver from
the product requirements, or from the decisions made in phase 2. By this point, all the basic
information about the operational conditions and the system functionalities should be known. The
factors discussed in phase 2 are also applicable to this phase:
-

Wrong assumptions and interpretations


Miscalculations
Parameter uncertainties
Wrong time and cost limits

A problematic factor could be the wrong interpretation of how an error in one component affects the
rest of the sub-system, or how the other sub-systems react to it. This might even be forgotten in the
process of predicting the reliability. Possible effects between sub-systems can be discovered through
the application of reliability block diagrams and similar models, but even these methods are
incomplete. This will be due to a lacking understanding of the system, or wrong assumption about how
it works.
As a part of phase 3, the methods for reliability evaluation and prediction are applied at a detailed
level. The amount of necessary inputs will be high and it is up to the analyst to know what they are
and how to use them. The attributes which are to be examined must be chosen, as well as the depth of
examination and the analytical approach (NASA (B) 1999). The completeness and suitability of the
analysis will thus only depend on the assumptions and choices of the analyst. If some of the basic
assumptions are unreliable, the final answer will be less reliable than imagined.
Miscalculations, parameter uncertainties and the wrong choice of time and cost limitations may arise
as in phase 2. The main difference from phase 2, where unreliability problems are concerned, is that
after phase 3 the design is frozen. This means that unreliability issues in the assumptions and
calculations which have not been discovered at the end of phase 3 will not be implemented in the
design.

Master Thesis, TPK 4900


Ingrid Alms Berg

53

NTNU

Design for Reliability

14.06.2010

Subsea process industry


This subsea process industry is to a large extent affected by all the factors discussed for general
products in phase 3. Estimations of the product reliability will be done, the final elements in the
system will be decided and the design is frozen. No assumptions can be made lightly and all choices
must be studied thoroughly before they are made.
The OREDA handbook will now be used a final time. This is also a phase where difficulties
concerning documents such as technology readiness levels (TRLs) can become important. These
documents are often internal to the companies, but contain the same terms and types of information.
They can therefore have different interpretations of the same elements. A TRL document describes the
maturity a technology, whether it is an idea, a proven concept, has been tested or used for a decade. As
the design is frozen, the system ought to have reached a level described as proven concept which is
demonstrated by the analyses performed.
Aker Solutions have eight technology readiness levels (ref [76]). This number is quite normal, but
what a level means may change within an industry. In order to ensure a common understanding
between the companies, it is necessary to decide which document to follow. The main problem with
assumptions concerning how far the development has come is the belief that a project goes by faster or
slower than it does. This could mean a more relaxed or stressed environment in the project
organisation. Other documents differing between a manufacturer and a client can lead to similar
problems according to the type of document.

Phase 4
When the reliability of the design has been predicted, the physical product must be manufactured in
line with the production. During phase 4, the development starts with the components and continues
on higher levels. A good technique in this phase is the TAAF, Test, Analyse and Fix, where a new
component may be developed and tested several times until it has reached the required abilities.
Among the tests performed are stress tests and accelerated tests. If the test results are acceptable, the
component is accepted. This can be done to all components and sub-systems, and if the final system is
a prototype not intended for later use, this can be tested as well. For some manufacturers it may be
hard to perform many destructive tests, especially if the prototype can is the same as the final
deliverance.
Physical testing shall be performed to prove that a product has been manufactured as specified during
phases 1 to 3. Shortcomings or errors made during the previous phases can be revealed through testing
and mitigating actions may be implemented.
Among the problems which can affect the reliability negatively are:
-

Lacking knowledge of production/manufacturing facilities


Lacking resources
Misinterpretations
Time/cost issues
Construction system problems

An important part of this phase is the planning of the manufacturing and assembly of the components,
sub-systems and system. If the planning team does not have the required competence to understand
how the facilities enable the production, it might leave steps out of their plans. The production must be
properly prepared and performed. Otherwise the different parts of the system will have errors in them,

Master Thesis, TPK 4900


Ingrid Alms Berg

54

NTNU

Design for Reliability

14.06.2010

or be assembled in a manner which is not desirable. Another reason why lacking knowledge on this
point is difficult is the testing possibilities. If the potential for the use of the facilities for testing is not
seen, tests which ought to have been performed might be ruled out. Although it is reasonable to
believe that there is enough knowledge about the facilities, it is the combination of this and the
understanding of which tests are needed that can be the greatest problem of the phase. If a necessary
test is left out, an unacceptable item might be put to use, creating unreliability in the system.
Another reason why a test can be left out is a lack of resources able to perform it. In some cases the
personnel who know the materials and the tests can be unable to oversee the procedure. The correct
personnel are vital for the tests to be performed correctly. They will be able to help the analysis
through their knowledge of how the test results should be interpreted. Misinterpretations can lead to an
unacceptable component being accepted or an acceptable item being left out. The first will be
problematic to the reliability while the latter will affect the time and costs available for other tests.
Time and costs are the constraints of the tests in phase 4. These factors may limit the number of tests
performed, the time to analyse the results and the extent to which a test may be performed.
The sub-contractors could also be a factor if diseases, strikes or production problems lead to
components being less reliable than they otherwise would have been. If the problems are not
discovered early enough and the parts considered acceptable without testing by the manufacturer, they
could contribute to a lower actual reliability than predicted. This affects the time and cost of the
project, and is a management issue.
The final factor of importance in this phase is problems with the construction systems. If the
technology for the construction is in a bad shape, the parts produced will be affected. Dimensional
problems are likely to be discovered, but internal cracks and alterations in metals and other materials
can stay unnoticed. This can severely reduce the reliability of the final product.
If the manufacturers product is custom-built and the prototype is handed over to the customer, phase
4 is the last phase. If the tests are accepted and verify the predicted performance and reliability of the
system, handover will be the next step. In this case phase 5 is of little or no interest. It is thus
imperative that any problems which lead to a decreased reliability in this phase are studied, understood
and an endeavour is made to avoid it. Problems with the manufacturing process are discussed under
phase 6.
Subsea process industry
This phase is as important for the subsea process industry as for any other industry. The physical
development of a system is expensive and must be planned very well. In some cases, all destructive
testing is undesirable for a product and it is then even more important that the planning has considered
all errors which may occur. However, factory acceptance tests showing whether the different system
functions as supposed to or not, will be performed. A subsea process system is unlikely to be tested as
heavily as a car, but some checks should be done to see that the materials are not damaged or weakens
the overall system.

Master Thesis, TPK 4900


Ingrid Alms Berg

55

NTNU

Design for Reliability

14.06.2010

Phase 5
When the products are meant to be produced on such a scale that the costs of prototypes are small
compared to the production costs for final products, it can be useful to let consumers test them before
a release to the market. Such tests will need good follow-up and good information to the consumers
about how they shall document their usage. This phase will return information on the reliability of the
prototypes, but the utilization will differ from one consumer to another. The analysts must therefore
consider the differences carefully in order to reach a proper conclusion from the experiment.
Problematic factors in the phase are related to the information given out and gathered about the
product:
-

Lacking information to test consumers


Erroneous reporting

A possible factor is the information given about the test. If it is badly explained or given in such a way
that the consumers try to find ways to break the prototype rather than using it normally, the outcome
of the test may be invalid. This testing is the last to be performed before the large scale production and
release to the market. It is therefore even more important that the prototype testing is comparable with
the actual use during the operational life.
Erroneous reporting from the consumers can be a direct effect of lacking information given before the
experiment started. If the reporting is inadequate or incorrect, the analysis of the reports will be based
on the wrong grounds as well. This can lead to changes in the production process that are for the worse
rather than the better. Bad reporting could also lead to measures not being taken to improve the
product.
Subsea process industry
This phase is not of specific interest to the subsea process industry. What may be said is that if any
end-of-development-tests are performed here, the same factors as discussed in phase 4 can be
considered.

Phase 6
Phase 6 is mainly concerned with production and is only relevant in cases where more than one
product is manufactured. If the product is custom-built and delivered after phase four, then this phase
is of little relevance. Nevertheless, the factors of decreased reliability may be relevant during the
production of custom-built products as well. The main factors are:
-

Deviations in the production process


Human interaction/handling

No process can be exactly the same every time and this is especially true when a large number of the
same product is manufactured. For this reason the production process must be followed carefully to
avoid deviations outside any acceptable limits, for example 2-3 mm diameter for a pipe. An
undiscovered deviation can lead to reduced or altered inherent reliability. When the deviations are
created by mechanical equipment, it is possible to make adjustments or repairs that erase the problem.
Even though most parts of a manufacturing process may be done mechanically, there are still humans
involved. The human being is considered to be very unpredictable and human errors are easily made.
If a bolt is supposed to be tightened by a person during the production process, he or she might tighten

Master Thesis, TPK 4900


Ingrid Alms Berg

56

NTNU

Design for Reliability

14.06.2010

it too hard or loose without knowing it. This can affect the reliability of the product. To adjust such a
problem is not the same as with a machine, especially as it can be harder to discover.
Any handling which has not been considered in an analysis can alter the reliability negatively. To
foresee human handling, an analyst must know how the process will be designed and how the
operation is performed. The tightening of a bolt is a rather small part of an operation and to consider
all errors that can take place will be difficult. Human handling is generally a large factor which cannot
be overseen. It is the most unpredictable, but it may also be the easiest to instruct. Through the use of
methods such as HAZOP or SWIFT, it can be possible to study the operational hazards in the
production process. If all the human handling has been thoroughly gone trough beforehand, both
during the production process design and when the workers are instructed, the risk of mistakes can be
reduced.
Subsea process industry
Human handling in the product development and assembly is an important issue for the subsea process
industry. A bolt which is not tight enough can mean that a system is unable to handle the pressure
from its inside or the water on the seabed. A bolt which is too tight can create cracks in the
construction. The little things that might not be a problem above water can create extreme problems
sub sea. Small deviations, erroneously tightened bolts and nuts etc. are nearly impossible to correct as
soon as a system has been installed sub sea.
The best manner in which problems inflicted by human handling can be corrected is through repeated
learning and clear instructions. Human beings cannot be taken out of the process, but they can be
given the task of learning how the errors arise, their effect and how they can be avoided.
Aker Solutions do have some series production and is currently looking into the possibility of
developing more standard shelf ware for subsea petroleum purposes. With this in mind, the Subsea
Power and Process department should look carefully into the mass production and review all the
difficulties which may arise in this phase. One thing is to evaluate the production of one item, another
thing is series production. Any of the problems discussed for this phase will then be even more
relevant, and more care must be taken to study the deviations in the production process.

Phase 7
Phases 7 and 8 are post-sale and the product is therefore no longer in the hands of the developers. A
system can be operated by the same company that manufactured it, but it will now be in use and only
maintenance can have an impact on the actual reliability. Phase 7 is placed on the product level and
therefore concerns the reliability performance of the product alone. For a standard product produced
for many customers, the reasons for a decreased reliability will be different for each item. The
conditions they are used in and how they are handled will depend on the consumers, as the products
are left entirely to them. The goal of the manufacturer in this phase is to collect the information about
how soon and why a product fails. For standard products this means that the complaints and the repair
information during the warranty period are the only sources of information.
The factors which can be affected by the manufacturer in this phase are associated with the transport,
installation and basic user information:
-

Unconsidered transport environment


Stresses during transport and installation
Lacking information to customer

Master Thesis, TPK 4900


Ingrid Alms Berg

57

NTNU
-

Design for Reliability

14.06.2010

Misinterpretation of failure

In handing over a product to a client or a market salesman, transport is the main step taken. During
transportation the product can be exposed to environmental conditions that it was not built to handle.
Unexpected stresses and vibrations may occur if it is not handled carefully. At installation, the
procedure would normally be well prepared in advance, but one cannot always get perfect conditions
during the operation. If the weather is an important factor for the installation to run smoothly, sudden
changes can give way for stresses to occur.
While a product is in use it may be exposed to mishandling which can affect its ability to perform an
intended function. A wooden box is usually meant to be a dry place to store things. Normally one
would say that for this purpose it would be rather reliable. If the box was used as a stool to get a glass
down from a highly placed shelf, its reliability could be altered. A box which has been used as a stool
may get cracks in it. The cracks can open the box to air and water leading to the content not being kept
in its desired state. Altered or decreased reliability from unintended use is hard to consider for an
analyst, as the discovery might depend more on imagination than knowledge. In order to avoid misuse
or complaints about how a product functions after such mishandling, the information given to the
customer must be well thought through. It is a known case that many customers avoid reading
manuals, but this does not mean that they are unnecessary. The manufacturer has a duty to inform the
customers. Any problems which occur due to a lack of information can reflect poorly on the
manufacturer and lead to warranty claims.
If a product has failed and repair is demanded, a maintenance programme is normally prepared some
time in advance. It should therefore not be a problem to return the product to normal operational state
fast. However, neither a maintenance programme, nor prepared repair actions can help the avoidance
of misinterpreted failures. This will lead to a wrong choice of maintenance activities and possibly ruin
the product. If the product is returned with a fault, it can be dangerous for the customer to use.
The main objective of this phase, for a reliability engineer, is to obtain information on how the product
performs while in use. Information about the failure in standard products will normally be found
through warranty cases and repair information. For the custom-built product it can be easier to keep in
touch with the customer, but it might be more difficult to obtain the repair information unless the
manufacturer performs the repair. The root cause of a failure is much needed information for the
manufacturer, if the overall reliability performance is to be evaluated. This does not affect the
reliability of the product, but may indirectly affect the reliability of a future product. However, it is
hard to do anything to prevent the lack of returned information. The only option is for the
manufacturer to go actively into the task of getting updates from the buyers.
Subsea process industry
When Statoil approaches Aker Solutions, they ask for custom-built products. Such products will meet
with the same factors decreasing their reliability as the standard ones. The main differences are that the
operational conditions are well accounted for, and there is thus less chance of mishandling during
operation. However, mishandling cannot be left out entirely as long as human beings are involved. The
systems are normally made on demand and the information concerning its use is well communicated
between the producer and the user.
We cannot look away from the possibility of bolts being tightened too hard or too loose, nor can we
think that transport in the subsea process industry is safe. Although systems have been transported to
subsea locations for several decades, the transportation is not perfect. A system transported from

Master Thesis, TPK 4900


Ingrid Alms Berg

58

NTNU

Design for Reliability

14.06.2010

manufacturing at Tranby outside Oslo to testing at Nyhamna in Mre and Romsdal will be subject to
several stresses. Some parts of the transportation will go by land, other parts by sea. Loading a system
from a stable position onto a means of transportation can subject it to rotations and forces which have
not been accounted for. The systems are usually quite heavy and will be pulled quite heavily in order
to be moved. During transportation, there will be movements forcing the system backwards, forwards
and sideways in its container. If it is not will secured, it can crash with walls and other elements. The
opposite problem may occur if another element is badly secured and crashes into the system. All of
these possibilities can affect the system in a negative way, rendering it less reliable.
Systems developed by the Subsea Power and Process department are meant to be installed sub sea.
This means that the installation is especially difficult and that the operations must be performed by
machines. The crew during the installation will usually be placed on a boat while an ROV performs
the tasks sub sea. If the weather and the sea suddenly became very challenging, the operations would
have to stop. Due to uncertain weather conditions, stresses changing the actual reliability can occur
during installation.
During operation, the FRACAS system can be used for reporting of failures, maintenance etc. If the
agreement is badly written between Aker Solutions and the client, Aker Solutions risks bad reporting.
Although the returned information will be easy to handle and confirm, any lacking information can
affect later projects. This is especially important to be aware of if the Subsea Power and Process
department wishes to develop shelfware products.

Phase 8
Phase 8, as 7, demands input from the users and operators of the products and systems. This phase also
needs input about the performance on the market based on price, market share and competition. The
main difficulty is the gathering of enough information and then the proper application of this
information. A reliability engineer might not see this phase as very useful, especially as it is used to
analyse the business success of the product. However, the analyst needs to use the information found
here to see which reliability tasks should be performed in later development projects. It can also be
studied how important the reliability is to the customer. Just because a product did well, it should not
be thought that the reliability programme was unnecessary. It should rather be analysed what
contribution it made.
The factors which can be a problem here are:
-

The missing information about product performance


Wrong analysis methods
Relaxed relationship to the usefulness of the output for reliability

As in phase 7, the information gathered for analyses of the product performance will depend on the
information returned to the manufacturer by the customers. As the overall performance is desired,
sales can be included in the analyses. The products success, if it is a standard product, can be found
from information returned by the salesmen and the manufacturers sales department. However, there
might still be some clients who do not return the product in case of failure. The satisfaction of the
customer is also hard to be informed about without questionnaires or the information salesmen might
have obtained. The estimated product performance depends on so many inputs being obtained that it
can be a highly uncertain estimate. For a custom-built product, this estimate should be easier to get
correct.

Master Thesis, TPK 4900


Ingrid Alms Berg

59

NTNU

Design for Reliability

14.06.2010

The methods used to analyse the returned information can be highly speculative as they will depend on
an interpretation of a persons opinion. Some could claim that OK means good performance, while
other would call it barely satisfactory. Alone or together with a too relaxed relationship to the
usefulness of the product performance analysis for reliability purposes, the information could be
rendered useless for future products.
Subsea process industry
This phase is where the manufacturer must go through the total performance of the product. A subsea
process system will have only one client and performance data should be easily retrieved. FRACAS
can again be a very useful tool. It is now even more important that misinterpretations of the
satisfaction of the client are expressed in a manner which is understood by the reliability engineer. The
main influence on reliability will mainly be noticed if the analyses are used as lessons learned in later
projects. If the client is misinterpreted, the future project can be built on a system which is not
satisfactory after all. Finally, the stored results must be easily accessible in the future.

7.1. Discussion
Most of these factors may not be possible to consider through estimates and normal analysis methods.
They depend on some sort of interaction, either between people or directly with the system. If the
specifications in phase 1 are not properly thought through, or have errors or holes in them, the whole
design and development process may go wrong. In phase 2, the allocation of functions and reliability
will decide the outline of the system. If this is not performed correctly, the outcome can be a system
which does not function optimally, or has a lower overall reliability than it is believed to have. Phase 3
is concerned with the very detailed level, before the design is fixed. Accepting an unacceptable design
is a possibility and although the root cause can be found in phase 1 or 2, it is not impossible that a
poorly done research job in phase 3 is the reason for the acceptance. From phase 4 and through to the
transportation in phase 7, direct human interaction with the system becomes an issue. The human
being is called both uncertain and unreliable. How a person performs a task is rarely possible to
foresee. The effects can be minor or major, but to discover the cause is hard. Finally, in phase 7 and 8,
the reliability of the system lies in the hand of its user. The manufacturer is responsible for the
information brought to the customer, but he or she cannot supervise how the product or system is
handled. In these last phases, the reliability engineer will be concerned with the gathering and analysis
of information about the systems performance. These tasks can be highly important for later projects
and should not be taken lightly. Any errors can lead to the wrong inputs being passed on to later
development projects.
What can be done to diminish the occurrence or effect of the factors contributing to unreliability, is to
think through the process from the product is designed, through its production, transportation and
installation, to the operation in the field. Each phase ought to be gone through by the team working in
the specific phases. The team members must consider the tasks performed in each phase and what the
product must endure while in development and use. In this way they will be aware of the pitfalls they
must avoid and work more consciously throughout the phase.
A what-if analysis may easily detect problems such as those discussed in this chapter. Whether they
are discovered or not will then depend on which processes one involves in the analysis and how the
questions are asked. If the main factors have been thought through in advance, it is more likely that
they can be avoided. A simple checklist for each phase should be easy to prepare and equally simple to
use. What one must be aware of is that the uncertainties and possible factors discussed here will not
always give a negative output. It is hence important to consider whether one is negative or positive to

Master Thesis, TPK 4900


Ingrid Alms Berg

60

NTNU

Design for Reliability

14.06.2010

the uncertainty brought into the project, possibly accepting one uncertainty and avoiding another. As
long as the possibility of negative factors does not become a reason to fear that all new developments
will be a failure, being aware of them can be of help.

Never let the fear of striking out keep you from playing the game (A Cinderella Story).

Figure 25: Baseball

Master Thesis, TPK 4900


Ingrid Alms Berg

61

NTNU

Design for Reliability

14.06.2010

8. Methodology for reliability in the product life cycle


The factors contributing to unreliability are basically the same for all production industries, whether
they specialise in custom-built products or shelfware. Some of them are directly tied up to errors made
during analyses which are meant to improve the reliability. Others arise with the manufacturing,
transportation and physical use.
All of the methods for reliability discussed in chapter 5 and 6 are available to all industries, only
limited by the product type. Although there are reasons to doubt the effectiveness of these methods, it
is more likely that problems are discovered with them than without them. If they are put together
according to the project phases and tasks, the overall output could prove to be quite useful. With this
in mind, it should be possible to make a general methodology for reliability, independent of industry,
product or organisation. If the methodology is properly implemented, and thoughts are made in each
phase about the possible factors reducing the reliability, it should be easier to mitigate some of the
problems. It should also be more likely that the estimated reliability for the final product is correct.
If we want a product with high reliability, it is necessary to prioritise reliability during the design and
development, as well as during later phases. With a proper methodology, it should be easier to analyse
and evaluate the reliability on a regular basis. A general methodology can be implemented in all
industries, but it should never be done without precaution. The methodology must rather be considered
as a starting point for each industry or organisation to develop their own, more specific,
methodologies.

8.1. Methodology in phases


In this chapter, one general methodology will be suggested. This has then been applied to the Aker
Solutions project execution model for technical qualification (Appendix B), for the development of a
more specific methodology for the Subsea Power and Process department, which is described in
appendix C. The purpose of this is to show how a general methodology can be implemented by any
industry or organization. Both methodologies are described through flow-diagrams for each of the
phases. As the three final phases are not in the technical qualification, the three last phases in the Aker
Solutions methodology are only described through text.
The methodologies are based on the following two questions:
-

What has to be done in this phase?


Which methods can be employed?

As shown in figure 8, the five product life cycle stages (Front-end, Design, Development, Production
and Post-Production) may be split in eight. As the Subsea Power and Process department at Aker
Solutions mostly develop custom-built products, the Front-end, Design and Development phases are
the most important. In the eight-phased model these three phases are phases 1 to 5. As shown in
appendix C, the phases 6 to 8 do not have the same focus as the first 5. In the general methodology, all
phases are equally important. Except for in phase 5 where the products are tested, there are few
differences between standard and custom-built products. They all go through the same phases and are
subject to many similar project activities. The methodology is thus made for a very general
perspective, including all kinds of production industries.

Master Thesis, TPK 4900


Ingrid Alms Berg

62

NTNU

Design for Reliability

14.06.2010

Phase 1
Phase 1
Project tasks

Project reliability tasks

Opening for a
new product

Customer
requirements/
market desires

QFD

Product
specifications

Reliablity, availabllity and


maintainability reqiurements

Desired reliability
Generating ideas

Concept chosen
and defined

Yes

GAP analysis
Evaluate previous projects
Analyse needs
Define gap

HAZID, SWIFT,
Early FMEA
Recommence
ceoncept
definition?

No

No

Concept
accepted?

Create reliability
programme

Yes

Terminate
project
Reliability report
phase 1

Client check-up for


acceptance

No

Client
acceptance?

Yes

Phase 2

Figure 26: Reliability methodology phase 1

Figure 26 shows the main development tasks during this phase of the project. The opening for a new
product is either found through a gap which appears in the market, or through the proposal given to the
manufacturer by a client. The main problems discussed for this phase in chapter 7 concerns
communication, knowledge and available data. While a client will provide requirements, studies must
be performed to derive all the desires and requests in a market. To understand what the client or a
market analysis means, a QFD, quality functional deployment, can be useful. This is a method in
which the customer expectations are identified and translated into technical characteristics (Yang
2007). A customer usually wants high reliability as this implies that the system is able to perform the
desired function for a certain amount of time. What the QFD can do is to identify where this reliability

Master Thesis, TPK 4900


Ingrid Alms Berg

63

NTNU

Design for Reliability

14.06.2010

is demanded and from that the reliability targets could be set. QFD is not a tool specifically made for
reliability engineering and was therefore not discussed in chapter 5. More information on the method
can be found in Yang (2007). From the QFD, information can be exported back to the study of the
customer requirements and onwards to derive more specific requirements. These can again be used in
the product specifications.
Phase 1 is the time to ensure a proper specification. This should include a definition of failure in
relation to the product function, a description of the environments the product will be exposed to and a
statement of the reliability requirement where critical failures and effects have a low probability of
occurrence (OConnor 2002). When stating the reliability requirement, it should be verifiable and
sensible according to the use of the product. Reliability requirements can be specified according to
time, failures or a success ratio, as long as they seem achievable, logic and useful.
When the project team during phase 1 generates ideas which may fulfil the product specifications, they
will decide to go further into those which they believe are feasible. A GAP study should be performed
to see what the manufacturer already has and is able to do and how far it is to the goal. A GAP
analysis is a technique where the questions Where are we? and Where do we want to be? are
answered (IfM 2010). A team with multiple backgrounds should be able to look at what is demanded
of the new product, compared with what the current technology may do. If entirely new technology
must be developed, the GAP analysis becomes less interesting. The most important outcome of a GAP
analysis is an overview of the technology needed. A reliability engineer could be present in the GAP
analysis team and look into the reliability of current technology compared with the reliability required
by a customer. When the GAP analysis is done, it will be easier to choose a concept based on the
width of the technology gap. GAP analysis can be further studied on Federal Agencies (2010).
To aid the decision of whether this concept is acceptable, early analyses should be done to evaluate the
possible reliability. Possible methods are HAZID, SWIFT and FMEA which do not need too much
input information. Information of use is found in studies and lessons learned from previous projects of
a similar type. If the reliability seems to be in the area of the desired reliability, it is be reasonable to
accept the concept.
When a concept has been accepted and a decision made to proceed with the project, it is recommended
to create a reliability programme. A reliability programme is here a combination of methods which are
connected with specific project tasks. The reason why a set of methods are chosen is given, and timing
for the project specified. The reliability programme should be based on the requirements and targets
for the reliability of the product (Yang 2007). Planning for reliability will increase the probability that
the final product is reliable, and diminish the costs of sudden reliability checks and unexpected
failures. An early investment in reliability activities will also decrease later costs due to lacking
investigation of failures. For this reason the programme should stress that reliability tasks are
performed concurrently with and throughout the development (MIL-STD-785B 1980).
When the concept is accepted and the programme prepared, the methodology suggests that the client is
involved for a final acceptance. This is only relevant if a client is involved in the project. Together, the
client and the development team should go through the information they have received from each
other and the decisions made. While doing so, they shall evaluate whether the concept is based on a
common understanding or not. If there are any unanswered questions or misunderstandings, the team
should return to a previous part of the phase and trace the problem to find a solution. This final checkup can be helpful to diminish the problems described for this phase in chapter 7.

Master Thesis, TPK 4900


Ingrid Alms Berg

64

NTNU

Design for Reliability

14.06.2010

Summary
In phase 1 the following reliability activities should take place:
-

A QFD translating what the customer wants and requirements for reliability.
A GAP analysis to decide what the new product is missing to fulfil the requirements.
Early reliability studies; HAZID, SWIFT and FMEA.
Planning of a programme which establishes the reliability throughout the life cycle.
Check-up with the customer.

Phase 2
Phase 2
Project tasks

Project reliability tasks


Input from reports phase
1, e.g. HAZID, SWIFT
and FMEA results

System break-down
into sub-systems, assemblies and
components

Reliability allocation

Update FMEA
Define system architecture
FTA and/or RBD

RAM analysis

Create overview of reliabilty


predictions

Compare predictions with desired


reliability and requirements

No
System architecture
acceptable?

Document
reliability results
etc.

Yes
Phase 3

Figure 27: Reliability methodology phase 2

Murthy et al. (2008) describes phase 2 as the phase where the system architecture is defined. This
includes a break-down of the system in order to describe the sub-systems in more detail. What is not
clarified is the detailed design of how the components are linked together and other specific details,
this will be done during phase 3. Based on the requirements and specifications set in the first phase, it
is now desirable to predict the overall product reliability. The hope is to find an architecture whose
predicted reliability matches the desired reliability from phase 1.
To start the system break-down, the results from the previous phase are used. The concept will include
information about how a function can be carried out and what the system must be able to do.
Simultaneously with the system break-down, it is possible to allocate the reliability. The desired
reliability is usually given as an overall requirement. To obtain this reliability, it is necessary for all
the sub-systems and components to have a desired reliability as well. A reliability allocation means

Master Thesis, TPK 4900


Ingrid Alms Berg

65

NTNU

Design for Reliability

14.06.2010

that a desired reliability is assigned to each part, depending on what is thought feasible and how this
fulfils the overall reliability.
As the system is better described through sub-systems and components, it can be useful to update the
FMEA. If no FMEA has been performed, this is the time to start one. As soon as the sub-systems and
their functions have been defined, one could start questioning their possible failures, which are more
specific than those of the system as a whole. The information could then be implemented in an FMEA
and studied further. As far as possible, the FMEA should be developed into an FMECA. With
criticality included, it can become a very useful input to a RAM analysis.
Phase 2 is the phase where the reliability truly can be built into the product (Yang 2007). As the
product architecture comes together, it is time to develop RBDs and FTAs. These methods may point
out the weak links in the system structure. When the components start falling into place, more
information should be retrieved on failure rates and these methods can then provide the engineers with
the overall system failure rates. If they are performed together with a RAM analysis, they may even
lead to a prediction of the overall availability. The outcome of the analyses should be used in an
overview of all the predictions of interest to the reliability.
The predicted overall reliability and availability should be compared with the requirements and used to
decide whether phase 3 may commence. When a system architecture is chosen, it must be proved
acceptable. If it is not suitable for the concept, a new evaluation of the system break-down must be
performed.
Useful tools both in phase 2 and phase 3 are computer-aided engineering methods. These can show the
different options for system architectures and the placement of components. Simulations may be made
on how the system reacts to stresses and failure modes can be evaluated. How far one can go in phase
2 depends on which level one studies the system at and the computer programs demands. The best
options for simulations will be found in phase 3 when the detailed design takes place. A good choice if
such tools are used is to start preparing the simulations in phase 2 and complete them in phase 3.
Simulations for reliability purposes are mainly connected with RAM analyses, but may also be used to
study the development of fault propagation through a system. The latter is of great interest to safety
where barriers must be considered. No computer tool is able to include all aspects concerning a system
and one must therefore be aware of the limitations not to be misled into trusting the accuracy and
completeness of the software models (OConnor 2002).

Summary
During this phase the following activities should be performed:
-

Reliability allocation to all sub-systems and known components.


Identification of hazards connected with the sub-systems and functions through an updated
FMEA.
Preparation FTA and RBD which can be used with a RAM analysis or alone.
Prediction of an overall reliability and comparison with desired reliability.

Master Thesis, TPK 4900


Ingrid Alms Berg

66

NTNU

Design for Reliability

14.06.2010

Phase 3
Phase 3
Project tasks

Project reliability tasks


Input from reports
phase 2

Functional analysis

Decompose for reliability

Assign requirements and


functions to components,
assemblies and sub-systems

Set dimensions
and links between
components

HAZOP

No

Update FMEA, FTA and RBD

Are requirements and


functions met with?

Document
reliability results
etc.

Yes
Phase 4

Figure 28: Reliability methodology phase 3

Phase 3 is where the very detailed design is made and the final design is frozen before the physical
development commences in phase 4. The main task of a reliability engineer in this phase is to ensure
that the reliability predictions from phase 2 are made more accurate as the component specifications
are fixed. This is the last phase in which it is possible to make alterations without a direct intervention
into the physical realisation of the product.
Based on the architecture established in phase 2, a functional analysis of the system should be done.
This will help assigning specific functions to each component, assembly and sub-system. Although
some similar work might have been done in phase 2, this is the absolute definition of how the main
function of the system is provided. Concurrently with the functional analysis, decomposition for
reliability should be performed. The latter would be a continuation of the allocation from phase 2, but
based more specifically on the functions being fixed.
As the design is about to be frozen, it is useful to perform an extensive HAZOP study. The design is
now sufficiently detailed for the questioning mechanisms of the HAZOP to produce meaningful
answers (IEC 61882 2001). The HAZOP can also be used as new input to methods which are already

Master Thesis, TPK 4900


Ingrid Alms Berg

67

NTNU

Design for Reliability

14.06.2010

used in the project. An FMECA should be completed with more details about the failures which may
occur in a component. The FTA or RBD made in phase 2 should be completed with more specific
failure rates and possible new connections of the functions. Where it is possible, criticality estimations
performed as part of the RAM analysis in phase 2 should be updated. Through criticality classification
of different components and sub-systems, planning for future maintenance and testing can be
commenced. This will increase the possibility of maintaining the reliability and availability of the
product while it is in use.
At the end of phase 3, it is important that all requirements are met with. The design can only be frozen
when it is acceptable according to the specifications from phase 1. Upon entering phase 4, there should
be no doubt that the product predicted reliability is the best it can be. Only issues connected to
production, transport, installation and wear are now controlled for the reliability not to be altered. To
omit such alterations, the problems evaluated for transport and installation ought to have been studied
along with other hazards. Any difficulties with production may be tackled when the manufacturing is
planned.

Summary
In phase 3 it is important to follow up the reliability requirements and specifications, as well as
predictions made in previous phases. The main steps to take are:
-

To follow the detailed design in settling the specifications which give the best reliability.
To perform a HAZOP based on the new details and specifications.
To update any reliability tools used in phase 2 based on new information.
To study the criticality of each component and consider these for future maintenance actions
and testing.

Master Thesis, TPK 4900


Ingrid Alms Berg

68

NTNU

Design for Reliability

14.06.2010

Phase 4
Phase 4
Project tasks

Project reliability tasks


Reliability analysis for
procurement where necessary

Plan for procurement

Prepare production plan

Process FMEA

SWIFT/HAZID
for production

Human factors
analysis

Test planning for materials


and prototypes

Prepare for analysis of test results and


plan for accelerated life tests and
reliability growth

Perform pre-production tests

Analyse test results

Production

Follow-up of deviations
Prepare FAT

Prepare for hand-over/ release


for test of prototype
Prepare for customer testing

Phase 4

Document
reliability results
etc.

Figure 29: Reliability methodology phase 4

According to figure 7, we reach stage II, development, when phase 4 begins (Murthy et al., 2008).
This stage lasts through phases 4 and 5 where production planning, testing and prototype development
take place. Phase 4 suits both custom-built and standard products, although custom-built products tend
to be the prototype whereas the prototype stays a prototype for the standard products.
All the specifications, requirements and design drawings will now be used to develop a plan for the
production. Before this planning is done, a thought to whether some items may be procured or not
should be given. If it is possible, a plan could be made for the enquiry, receipt and testing of the items.
The testing might only be necessary where the subcontractors cannot provide acceptable information
themselves. The reason for the testing is to see whether the procured items match the predicted
reliability (Murthy et al. 2008). If they do not match, more research must be done in order to develop
suitable components.
For new technology, plans must be made for the development. The planning of the manufacturing
should not only include when and where an operation takes place, but also which parameters it must
stay within. A reliability engineer may not have much knowledge of how the manufacturing process is
best performed, but can look into previous problems occurring in this phase. The information retrieved
should be used to see where the planners must take the operations under extra strict observation.
Useful inputs to such plans are FMEA, SWIFTs and HAZIDs. As humans will be involved in the
development process, human factors potentially affecting the process negatively should also be
analysed.

Master Thesis, TPK 4900


Ingrid Alms Berg

69

NTNU

Design for Reliability

14.06.2010

When the production plan has been made, a plan for testing of materials and prototypes should be
prepared. The reason for preparing the production plan before this test plan, is that the production plan
will include information about the stresses and temperatures the materials will be subjected to. Among
the factors contributing to unreliability discussed in chapter 7, are the misinterpretation and wrong use
of tests. An important question to ask when failures occur in tests is Will they occur in use? To
answer this it is necessary to investigate the actual physical or chemical cause of failure (OConnor
2002). A test can easily be misleading if it is performed on the wrong grounds. By this stage it is
evident that the product will be able to perform its intended function, the question now is what may
stop it from doing that. A test should therefore not be performed to demonstrate the successful
achievement, but the failure causes.
While the tests are planned and executed, the reliability engineer must prepare for reception of the
results and how they are to be analysed. As the development should not take more time than intended,
it is necessary that the analysis results are ready as fast as feasible. If possible, the analyst should give
input to which tests are needed, for example accelerated life tests and reliability growth. The tests must
be prepared with the component, sub-system or system in mind. Whether the reason for the test is to
discover new failures or evaluate failure rates, it must be accordingly.
While the prototype is produced, there is little other for the reliability engineer to do than to follow up
on any deviations in the production machinery. Such deviations should be analysed to see how they
affect the product, but could also be used in the preparation of factory acceptance tests (FAT) and
customer tests. A full-scale test where the system is evaluated to see if it functions as planned. FATs
are used in several industries when the systems are ready for use. The tests can lead to new alterations
if necessary, but the hope is that it only confirms that a system is ready for operation. Full-scale testing
can be performed together with the client and should therefore be placed in phase 5, straight before
handover.
Even with many tests performed, the true value of a life parameter will stay unknown. We cannot say
when a failure will occur, but we may be able to find the distribution of an expected value (OConnor
2002).

Summary
Among the steps to take for reliability in this phase are:
-

Include reliability in the production planning.


Compare the reliability performance of existing components with the predicted reliability from
phases 2 and 3.
Prepare for tests which demonstrate possible failures.
Study the outcome of the tests and see if a part of the production must be repeated.

Master Thesis, TPK 4900


Ingrid Alms Berg

70

NTNU

Design for Reliability

14.06.2010

Phase 5
Phase 5
Project tasks

Project reliability tasks

FAT

Analysis of results

Suggest alterations where


reliability is unacceptable

Estimate overall reliability

Alterations
For larger market/
series production

For dedicated client/


custom built products

Release to testing
customers

Prepare for
hand-over

Gather customer
test reports

Hand-over

For larger market/


series production

For dedicated client/


custom built products

Prepare for
customer testing

Prepare FRACAS

Document
reliability results
etc.

Phase 7

Analyse reports

Analyse test
results

Suggest alterations where


results are unacceptable
Prepare phases 6,
7 and 8.
Prepare for series
production
Document
reliability results
etc.
Phase 6

Figure 30: Reliability methodology phase 5

The progression to a failed state is time-variant (OConnor 2002). The main objective of phase 5 is to
obtain information on field performance through operational testing (Murthy et al. 2008).Testing in
the field will last longer than accelerated tests and give answers as to how the product will be used and
how it operates under different conditions. The progression to failure and the failures will hence be
more realistic.
Before the prototypes are handed over to a customer or a client, the FAT must be performed. This will
check that the system functions as intended, in an environment similar to the actual operational
environment. The results of such tests can be used to suggest alterations if necessary and for the
estimation of an overall inherent reliability. As the latter might be hard, it could be considered through
an updated FMEA or RAM analysis.

Master Thesis, TPK 4900


Ingrid Alms Berg

71

NTNU

Design for Reliability

14.06.2010

Phase 5 is split into two categories after the FAT; larger market/series production and dedicated
client/custom-built products. If the system is handed over after the FAT, little more than preparing a
FRACAS and all the reliability documents is necessary for the reliability engineer. The next phase is
then number 7. For products intended produced on a larger scale, the customer testing will commence
after the FAT. As suggested in chapter 7, there may be some issues with the testers understanding of
what is to be reported and how it should be done. It is thus critical that the reliability engineer prepares
easily understood documents containing this information.
The next step is to gather and analyse the reports. Any specific test results which are not positive
should lead to alterations, or in the worst case that the project is terminated. Alterations which are
feasible can be performed with the mass production in phase 6. Information from the analysis of the
test reports can be used to update FMEAs, FTAs etc. for later use in other projects.
The main objectives for the reliability engineer in this phase are (Murthy et al. 2008):
-

Verification of the reliability studies conducted during the project.


Providing appropriate data for justifying operational procedures and policies
Modifications with respect to reliability and maintainability.
Providing appropriate data to be utilised in subsequent phases.

As phase 5 ends the reliability engineer should prepare for the following phases where the product is
in operation and relying on its actual reliability and ability to perform. When the product is handed
over to a customer there is little left for the manufacturer except repairs, warranty-claims and followup. The reliability analyst should now study the actual performance of the product. If preparations for
this are made, it might be easier both to obtain the necessary information, and to analyse and use it for
future projects.

Summary
In this phase the main reliability activities are:
-

Follow-up of FAT and customer tests.


If possible update existing reliability information.
Plan alterations if found necessary after operational tests.
Prepare for the following phases.

Master Thesis, TPK 4900


Ingrid Alms Berg

72

NTNU

Design for Reliability

14.06.2010

Phase 6
Phase 6
Project tasks

Project reliability tasks


Follow-up of analyses of
production from phase 4.

Follow-up of production
devations

Analyse deviations

Sample testing of produced


batches

Analyse sample tests

Suggest process alterations


where necessary

SWIFT/ HAZID / HAZOP for


transportation

Transport and release

Suggest solutions to
problems possibly
encountered during
transportation

Phase 7

Document
reliability results
etc.

Figure 31: Reliability methodology phase 6

The last three phases in the model belongs in stage III, which is post-development. Phase 6 is
concerned with the production as the product is ready to be sent out on the market. For custom-built
products of which there is only one of a kind, this phase is of little or no interest. Some parts
concerning the production of components and the general production process are of interest, and
should be implemented in phase 4. Custom-built products which there are more than one of, for
example trains and ships, can be considered to go through this phase.
The main concerns discussed for this phase in chapter 7, are the deviations due to variations in the
production process. Some human handling will also occur, mostly due to operations which cannot be
performed by machines. Human handling here includes both mechanical operations and transportation.
In general the engineers who are not directly involved in the production may have little to do in this
phase. The main concern of a reliability engineer in this phase would be to ensure that the items are
within the acceptable limits for conforming items (Murthy et al. 2008). The production personnel can
check the items for conformity, leaving the reliability and design engineers to decide the limits of
acceptability. Any testing of the produced items will depend on what the items are tested for and how
strict the reliability requirements are. If the product requires very high reliability or is very expensive
to produce and repair, 100% testing is desirable.
The output of tests in this phase can be useful for reliability and design engineers in later projects. Any
information acquired about the production process may be used to decide in which way a product
should be produced. It can also be employed to decide how strict the testing must be to retain the
desired design specifications. Finally, the information about the production process can say something

Master Thesis, TPK 4900


Ingrid Alms Berg

73

NTNU

Design for Reliability

14.06.2010

about the expected costs of the erroneous items produced. A useful tool would be root cause analysis
studying the reasons for any problems encountered during the production. Root cause analysis can be
further studied in (NASA (E) 2010).
As the products are prepared for transport and release, the reliability engineer should perform analyses
of the hazards which may affect the product. This is a very important part of the product life cycle, due
to the possibility of destruction of very sensitive items. Even though the transport only includes plastic
outdoor chairs, the possibility that something might break is present. Given that the transport will
depend on the road, as well as how the product is tightened in a vehicle, the main reliability tools
available are SWIFT, HAZID and HAZOP. These can cover a large number of possible hazards, while
not being too specific about the functions that are damaged. Any problems should be resolved through
suggestions of how they can be prevented.

Summary
In phase 6 the following could be considered reliability activities:
-

Prepare for testing of produced items.


Gather and analyse test results. Perform a root cause analysis.
Report the problems and store information for later use.
Study the hazards of transportation and suggest preventive measures.

Master Thesis, TPK 4900


Ingrid Alms Berg

74

NTNU

Design for Reliability

14.06.2010

Phase 7
Phase 7
Project tasks
Warranties and repairs

Project reliability tasks


Gathering of
information

Reception of
FRACAS reports

Customer/ Client follow-up

Root Cause analysis of new


failures

Phase 8

Analysis of performance and


estimation of actual reliability

Document
reliability results
etc.

Figure 32: Reliability methodology phase 7

Phases 7 and 8 are both dependent on the customers for information for reliability purposes. Only
when the manufacturer is the operator of the product, can all existing information about the actual
product reliability and performance be gathered. Phase 7 is placed on level II which is the product
level. This means that it is the reliability performance of the product in the field that is studied.
When the product is handed over from the manufacturer, nothing but the actual reliability is of
interest. The regular consumer will not be impressed by the predicted reliability, but by what the
product gives in return for the money spent. A good design, where reliability was taken into account
during development, will have considered the possibility of various operational environments (Murthy
et al. 2008). When failures occur, customers can choose to use the product warranty and ask for repair
or a refund. If the product is custom-built, the reporting and repair depends on the contract between the
manufacturer and the client. It is the choice of complaining about the product which will give
information to the manufacturer on the performance of the product. Without such information it would
be nearly impossible for a product to be analysed for actual reliability.
For both standard and custom-built products the information retrieved about failures in the field should
be analysed for a root cause. This is where the FRACAS system is of great use. Based on the root
cause analysis and the general information, the reliability engineer should try to estimate the actual
performance and reliability of the product. Most of the information and feed-back will be negative and
it is therefore hard to see whether the actual reliability truly is a good estimate. However, for the sake
of future projects, it is useful to have some thought on the product performance. If positive data are
wanted, questionnaires might be the easiest way to obtain them. As it is unlikely that all the
information about a product or system in its operational life can be retrieved, the actual reliability will
be shown as a distribution around the actual value.

Master Thesis, TPK 4900


Ingrid Alms Berg

75

NTNU

Design for Reliability

14.06.2010

Summary
The tasks to be performed for reliability purposes in this phase are:
-

Gathering of field performance, failures, complaints, positive feedback etc.


Root cause analysis of failures and operational problems.
Analysis of the actual reliability.
Storing of information for later use.

Phase 8
Phase 8
Project tasks
Sales follow-up

Analysis of earnings

Project reliability tasks


Root Cause analysis of
failures discovered from sales
follow-up
Analysis of overall
performance

End project
Documentation and
lessons learned from
entire project

Figure 33: Reliability methodology phase 8

The questions that should be answered in this phase are the ones that truly show the market success of
the product. This phase belongs on the business level and is from a reliability perspective concerned
with how the product reliability affected the business objectives (Murthy et al. 2008). These objectives
were defined as early as phase 1 where they described the desired performance of the product.
Phase 7 and 8 are very similar as the data must be collected from outside the manufacturer, analysed
for root causes and reported for use in new processes, preferably in phase 1. The main difference is
that the sale, market share and revenue must be included as well (Murthy et al. 2008). The popularity
of the product will not only depend on the amount of consumers that buy it, but how the product
corresponded to their wants and desires. For reliability reasons it is of interest to see whether the actual
reliability of the product affected the sales, complaints and market shares. The manner in which the
overall success of the product was affected may tell the manufacturer what reliability demands the
customers have. For the custom-built products the main objective of this phase is to decide which
improvements can be made for the execution of future jobs (Murthy et al. 2008).
For the project as a whole, it is important that this phase is used to gather information from the shops
which sold the products, and the manufacturers sales department. This might not be a part of phase 7
and is therefore useful now. In some cases a system returns more than what it was worth, other times
less. Often information of this type is more likely to be given to a vendor than the manufacturer. An
analysis of the overall performance of the product and how it was affected by the reliability must be
based on all available information, not just the failures.

Master Thesis, TPK 4900


Ingrid Alms Berg

76

NTNU

Design for Reliability

14.06.2010

For both standard and custom-built products, phase 8 closes when the products market performance
has been analysed and nothing more can be gained for future developments. In all phases the
documentation of the reliability is demanded. In this final phase, this should be used for the review of
the project and then stored for later use. Lessons learned from the project can stop history from
repeating itself and are thus highly important to avoid unnecessary difficulties in future project and
create feasible reliability programmes.

Summary
The reliability tasks in this phase are:
-

Gather data on the product performance not found in phase 7.


Perform a root cause analysis of new data.
Analyse the new data together with the output from phase 7.
Separate the information which may be useful in the future and store it.
Close the reliability part of the life cycle.

8.2. Comparisons and discussion


All development projects are special and whether it prepares a standard or custom-built product, the
best choice of reliability tools and methods will change from time to time. Any choice should be based
on which findings the reliability engineer needs and the existing knowledge about the product. Even
when the outcome only intends for a new generation to be made, there is a need for fresh input. Just as
the variation in the manufacturing process never will produce two exact matches, no phase will ever be
repeated in the exact same way.
To develop new concepts and designs, a large amount of creativity is needed. Forcing creative work
into frames with too many rules is not effective. The creative parts of the product life cycle are
therefore best suited to be controlled in such a manner that the outcome is suitable for the project and
kept within a certain time limit. As the creative parts of a product life cycle are mostly taking place at
the same time as the desired and predicted reliabilities are estimated, there will be a conflict. A
reliability programme demands discipline, control and the use of rules and guidelines (OConnor
2002). The output of the creative processes must therefore be exact enough for use. The synergy of the
designer and a reliability engineer demands time for that feedback between the two parties. This
feedback will help the decision of whether the creative process is going in the right direction or not.
To relax the activities connected with reliability estimations can lead to reduced reliability and it is
therefore better to have a strict design process.
A reliability programme must correlate with the design and development programme and concurrent
engineering is the only option. In this chapter the reliability activities suitable for each phase of the
product life cycle have been discussed. They are considered an integrated part of the development
programme and, as shown in the diagrams, the output from reliability methods should be used as input
to the decision points where a phase is concluded. When the reliability tasks and responsibilities are
given as mandatory in the procedure, the reliability will become a part of the design.
The main differences between the general methodology and the one for Aker Solutions, are dependent
on the tasks performed in the phases. As the technology qualification does not follow the same phases
as used for the general methodology, there are bound to be some methods which appear in different
places. What is interesting is that the methodologies have more similarities than differences. This is
assumed to be the effect of the general methodology being prepared for all types of products. As the

Master Thesis, TPK 4900


Ingrid Alms Berg

77

NTNU

Design for Reliability

14.06.2010

products pass through the same phases and are subject to many of the same project tasks, there is little
reason while the general methodology would not be similar to the specified one.
An important aspect of these methodologies is that they are prepared for entirely new technology and
therefore includes as many tasks as thought necessary. The case study will look into whether it is
possible to adjust them for systems where the technology is known and the project risk is somewhat
smaller than for new technology. It will also show how a methodology can become the basis of a
stricter programme with limits, project risk categorisation and a specific goal.

Master Thesis, TPK 4900


Ingrid Alms Berg

78

NTNU

Design for Reliability

14.06.2010

9. Case study ERMP


The objective of the case study was to prepare an ERMP (Equipment Reliability Management
Programme) for the EPC phase of the Midgard Subsea Compression System (SCS) development. This
should be based on the methodology suggested in chapter 8, and adapted for the Aker Solutions
methodology in appendix C. One criterion to the ERMP was a demand that it should be in accordance
with the ISO 20815 standard. It was also asked that the EPC phase followed the Project Execution
Model (PEM appendix D) used for development projects in Aker Solutions. The ERMP report can be
found in appendix E. A system description of the SCS is given in appendix A.
Among the differences from the general methodology and the one suggested for Aker Solutions are:
-

The PEM does not follow the eight phases of Murthy et al. (2008)
The technology qualification is not used as a basis
The technology is not entirely new
The system studied is specific for the ERMP

Although the EPC phase does not directly correspond with the eight phases, many of the steps follow
in the same order as in phases 2 to 5. It is therefore considered that the tasks suggested for these
phases in the methodology can be applicable in the EPC phase. The main reason for not using the
technology qualification in the case study is connected with the newness of the technology. A similar
system has already been designed and is currently in development. The Midgard SCS is thus not
entirely new technology.
The ERMP is written for the Midgard SCS alone and thus based on work done in previous phases of
the project. Some reliability activities have already been performed, however, as the design develops
and the current technology is further studied, there may be alterations on the Midgard SCS and new
reliability estimates will be needed. This is the main reason for the implementation of new reliability
allocations and analyses in the EPC phase. The previous analyses are then updated and more thorough
reliability studies of the sub-systems and assemblies performed.
Among the main constraints for the ERMP are time, other project activities and resources. The time
limit for the EPC phase is four years and it should therefore not be problematic to perform the
reliability activities. Nevertheless, time may be a constraint in connection with the other project
activities and any delay in these will delay the information flow to the reliability activities. The
reliability programme must be performed concurrently with the other development tasks. Finally, the
resources necessary for a reliability activity will depend on its inputs and the mechanisms needed to
perform them. Some activities may only need one person, while others give better results with a team,
for instance HAZID and HAZOP.

9.1. ISO 20815, the methodology and the ERMP


Technical safety and reliability aspects are often not considered until a verification or testing point for
equipment (ISO 20815 2008). This will usually be at a late stage of the development process, when the
design is near to fixed or after manufacturing has started.

Master Thesis, TPK 4900


Ingrid Alms Berg

79

NTNU

9.1.1.

Design for Reliability

14.06.2010

ISO 20815

Appendix A in ISO 20815 (2008) gives the outline for a Product Assurance Programme which has
been used in the development of the ERMP. It lists the following as parts in the programme:
-

Terms of reference
Production-Assurance philosophy and performance objectives
Project risk categorisation
Organisation and responsibilities
Activity schedule

In order to implement a reliability or product assurance programme in a project, a scope, philosophy,


objective and justification must be given for it. A risk categorisation will help determine the activities
chosen and to what extent a reliability programme is necessary. If the project is categorised as low
risk, with well known technology, there is little reason to have a very costly reliability program. The
organisation and responsibilities describe who will perform the reliability activities and the responsible
for follow-up and acceptance. Finally, the activity schedule includes the reliability programme and a
methods description.
The risk categorisation for projects is done according to three levels:
-

Low
Medium
High

Table 3 shows how the project risk categorisation is suggested defined by ISO 20815 (2008):
Table 2: Project risk categorisation (ISO 20815 2008)
Technology

Operating
envelope

Technical
system scale
and complexity

Organisational
scale and
complexity

Risk
class

Description

Mature technology

Typical
operating
conditions

Small scale, low


complexity,
minimal change
of system
configuration

Small and
consistent
organisation, low
complexity

Low

Low-budget, low-risk project


using field-proven equipment
in the same configuration and
with the same team under
operating condition similar to
previous projects.

Mature technology

Typical
operating
conditions

Moderate scale
and complexity

Small to medium
organisation,
moderate
complexity

Low or
medium

Novel or non-mature
technology for a new
or extended operating
environment

New,
extended or
aggressive
operating
environment

Large scale,
high complexity

Large
organisation,
high complexity

Medium
or high

Low- to moderate-risk project


using field-proven equipment
in n operating envelope
similar to previous projects
but with some system and
organisational complexity.
Moderate- to high-risk project
using either novel or nonmature equipment or with new
or extended operating
conditions. Project involves
large, complex systems and
management organisations.

Master Thesis, TPK 4900


Ingrid Alms Berg

80

NTNU

Design for Reliability

14.06.2010

In the ERMP, it is considered acceptable to rate the risk between these levels as low to medium and
medium to high. This has been done in order to avoid misunderstandings of how a level should be
interpreted. Low to medium implies that although the risk is low, it is not to be ignored. Medium to
high signifies that even though the risk is high, everything is not unknown and new.

9.1.2.

The methodologies

The tasks in the ERMP are based on the risk categorisation and chosen from the general and the Aker
Solutions methodology. As the EPC phase is not the first phase in the Midgard SCS project, several
hazards have already been identified and reliability estimates calculated. These can now be used when
the design is specified further. Even with previous information, changes in the design and new
operational requirements will lead to a need for new hazard identification processes. In the ERMP it is
considered that such processes will be performed in several formats throughout the EPC phase until
the design is frozen. Some of the methods are suggested as updates while others may return for new
purposes.
The suggested methodology in relation to chapter 8 is in general more thorough than the ERMP for the
EPC phase activities. There are several reasons for this, among them the specific system and the other
concurrent activities. It is also important to note that it is easier to suggest a large amount of activities
than to perform them all. In a different project, a larger number of the methods, or different methods
from the methodology may be chosen for a reliability programme.
In appendix C it has been suggested that the FMEA in particular should be updated as often as
possible, even after the design is frozen. For the Midgard SCS project, the FMEA is last updated at
before the system design basis is completed and the major interfaces frozen. This is logic as there is
little possibility of altering the design any more, but as there is a possibility of new failure modes
being revealed in testing it might have been better to keep an opening for updates. The FMEA is
probably not used much in the next project phases, but it can be useful in future projects.

9.1.3.

Combining ISO 20815 and the Aker Solutions methodology

The standard and the methodology do not go against each other, but rather work as supplements. To
choose the correct methods and implement them at the right place is not evident from the combination.
Although the methodology states where a method should be placed according to a project activity,
there is nothing stating which methods are to be preferred. As ISO 20815 does not give much input on
this point, the reliability engineer is left to figure it out according to the activities in the specific
project. By comparing them with the methodology, this should be achievable. However, it might have
been desirable to have an indication of which tasks are imperative for the different risk categorization
levels.
What might be done is to prepare a document in relation to either the methodology or ISO 20815,
stating which methods are the most basic and which are only needed for high risk projects. This would
be based on the assumption that a low risk project only needs a few methods performed, while a high
risk project needs as many as possible. Such an assumption is not necessarily the best, but it does
follow the need a reliability engineer has for new information in projects. A low risk project, as shown
in table 2, already has an extensive amount of input information. It is therefore not in need of many
new analyses before the lacking information is obtained. For a high risk project, the opposite is the
case. Although such information could be given with ISO 20815 and the Aker Solutions methodology,
it would still be up to the reliability engineer to know which outputs, and thus which methods, he or
she needs.

Master Thesis, TPK 4900


Ingrid Alms Berg

81

NTNU

Design for Reliability

14.06.2010

9.2. Discussion
Although the methodology suggested in chapter 8 is not perfectly applied to the ERMP in the case
study, little suggests that the methodology is wrong. The ERMP is written for a specific system and
actual project, while the methodology is theoretical. The Subsea Power and Process department has
the PEM and the requirements of the client to follow. In addition, the existing manner in which
reliability activities are performed is proven through other projects. Reality is rarely the same as
theory, but it can confirm its relevance.
This ERMP rather confirms than demands an alteration in the methodology. What may be done is to
remove some of the repetition of methods and testing activities, but this might again suit a different
project perfectly. As long as the methodology is not intended to be applied directly, but as a basis for
the creation of project specific reliability programmes, it can be accepted as it is. In such a case it is
necessary that it is more thorough than required as the projects will ask for different inputs and
amounts of reliability activities. If it is not thorough enough, demanding projects can be negatively
affected through lacking methods having been suggested in advance.
What were not defined in this ERMP were the standards which are to be followed specifically for the
project. However, as this is for one phase alone, the overall documents concerning the project can
include this and it might therefore be considered unnecessary. If the ERMP was meant for the entire
project, the standards should maybe have been further specified and implemented. To include the
standards would not be likely to alter how the ERMP was used, but it could have an affect on the
engineers while they are searching for the best manner in which the chosen methods are performed.
Anything which increases the awareness of why reliability is important is positive for the product or
system developed.

Master Thesis, TPK 4900


Ingrid Alms Berg

82

NTNU

Design for Reliability

14.06.2010

10. Design for reliability website


If a manufacturer has a difficult time understanding why reliability should be a part of the design
process or how it should be implemented, a website can be of use. We often use the internet to obtain
new information and it has become one of the main sources when we wish to study a new subject.
A website has been created to study how a tool for a reliability engineer can be developed. It could be
used either to explain reliability to a business executive, or to obtain tips and useful information for
daily work. The website is based on the findings described in this report. The first page is directed at
those who are not familiar with design for reliability or why it is a useful tool. The text links the reader
on to a basic page which describes the product life cycle and a page describing different methods for
reliability.

Figure 34: Front page of website for Design for reliability (http://folk.ntnu.no/ingribe)

Master Thesis, TPK 4900


Ingrid Alms Berg

83

NTNU

Design for Reliability

14.06.2010

The page describing the product life cycle is based on the model suggested by Murthy et al. (2008).
From the table describing the process, similar to that of figure 3, links to each phase may be found.
The phases all have one page each, including a description and the belonging methodology.

Figure 35: Life cycle (http://folk.ntnu.no/ingribe/productlifecycle.jsp)

Master Thesis, TPK 4900


Ingrid Alms Berg

84

NTNU

Design for Reliability

14.06.2010
Clicking on this
link opens the
page as shown in
figure 32.

Figure 36: Phase 7 as opened from figure 35.

All the relevant methods suggested for a phase are linked to the description of the methods for
reliability. As the RAM analysis consists of more than one method, this has been given its own page.
For those who wish to skip some pages, a menu is found on the top of each webpage.

10.1

Further suggestions

Currently the website is a very basic description of how design for reliability may be used in the
different phases. It suggests a methodology and it is up to the users to decide how this can be

Master Thesis, TPK 4900


Ingrid Alms Berg

85

NTNU

Design for Reliability

14.06.2010

implemented for their processes. Someone who is sceptic to reliability activities being mixed with the
design process can study the purpose and how it should be done. Although this website is at an early
stage, it can be developed into a better tool through the following suggestions:
-

The methodology figures can be given links to the methods within themselves.
Examples of how the methods are used could be included.
A project could be used as an example in each of the phases.
A form could be developed where the phase and design tasks are filled in, returning a
reliability programme suggestion.

Many other suggestions could be given as well, but these are thought the most helpful to the users.
Links within the figures can be easier to use, while examples of the utilisation of the methods and
phases will improve the understanding. Especially the last point would be of interest if the page is to
become an interactive tool. For someone who wishes to develop a reliability programme for their
project, without much experience, this could be an easy solution. By plotting in the phase the
programme is needed for and the planned project tasks, a corresponding set of reliability methods
could be developed. This only requires a rather simple computer programme. If it is desirable to
specify the project even more, a risk categorisation according to ISO 20815 can be implemented in the
form as well. The main constraint of the tool would be the understanding of the user. A reliability
programme suggested by the tool cannot be used without care. The user must therefore study the
suggestion and choose whether the methods are suitable to the particular project or not.
The website is rather unspecific as all industries are meant to use it. For more specific use, a computer
program could be developed. This would include the same information as the website, but could be
more extensive in the explanations. The methods ought to be well explained, possibly with
information on which standards they are described in. A possibility of uploading experience from
previous projects should also be included. This information would provide the user with comments on
how the method was performed and the results. As long as the information is used as input, not as an
answer, this could work rather well.
An option for the company buying the program would be that their own specified methodology could
be implemented. The reliability programmes returned by the tool would then be far more specific and
possibly easier to use in the product development. Many further solutions could also be added, but this
should depend on the interest of the potential buyers.

Master Thesis, TPK 4900


Ingrid Alms Berg

86

NTNU

Design for Reliability

14.06.2010

11. Concluding remarks


To perform reliability studies of an unfinished design can seem unreasonable for several reasons. It
does cost money, demand extra resources and forces the designers to follow specific schedules which
might limit their creativity. This can be said about every attempt to control the development process.
Nevertheless, manufacturers do have to develop new products within time frames and according to
performance demands. Reliability engineering is a case of being precautionary rather than wise after
the event. It is at its best when the results are invisible and no failures occur. New products are
introduced to the consumers all the time. In order to win a place in the market and keep the customers
happy, the manufacturer must develop new products quickly without affecting the reliability
negatively. Design for Reliability seems to be a good solution.
The first reliability methods were developed to study the possible failures in a system, but they were
also responsible for making the design engineers more conscious about how they developed a new
system. The reliability and safety became far more important than low costs and efficient production.
Reliability analyses can provide the designers with hazardous scenarios met by the product in the
future. This can either lead to new alterations or increased confidence in the existing design.
Chapter 7 is dedicated to factors contributing to unreliability. Although several more factors can be
discussed, the selection in this chapter shows that there are reasons to wonder how good reliability it is
possible to achieve. The vision is absolute reliability, where the systems never fail while in use.
Absolute reliability cannot be called impossible, but it is impossible to know that it is absolute before
the product reaches its end of life. We do not have the ability to foresee the future, but through the use
of reliability methods, the product reliability can be increased. If the methods are used correctly, at the
right time and with other overlapping methods, it might be possible to achieve absolute reliability
without being conscious of it.

Optimum

Reliability costs

Failure costs

The goal of a manufacturer could be to reach the highest reliability possible, but this might not be the
most beneficial in the long run. To perform many reliability methods will increase the costs of the
product. This will again render the product expensive and possibly turn customers against investing in
the competitors products. The reliability methodology includes a great amount of methods and so
does the ERMP. It will be up to each manufacturer, after consultation with the reliability engineer, to
consider the number of methods that shall be employed. Projects involving a custom-built product are
likely to afford more reliability methods than projects for standard products. The clients of custombuilt products are often more willing to pay for high reliability than a regular customer at a sports
store. The optimal number of reliability methods is hard to find, but a diagram showing costs of
failures and costs of reliability methods can be used. This is shown in figure 37.

Figure 37: Failure-reliability diagram

Master Thesis, TPK 4900


Ingrid Alms Berg

87

NTNU

Design for Reliability

14.06.2010

Without any money spent on reliability methods, it is likely that the costs of failures will be high. With
many reliability methods, the failures will be fewer. To set the costs of failures against those of
reliability will, as shown in the diagram, give an optimum. This is a useful tool when the requirements
to the reliability are not set in advance. Thus, standard products in particular can benefit from this
diagram.

11.1

Methodology and use

The reliability methodology presented in this thesis is based on the assumption that reliability
activities throughout the product life cycle will give increased reliability. It is further assumed that
activities can overlap each other and together give a good overview of the product reliability. The
intention is for the methodology to be a tool for the development of reliability programmes. Here the
eight phases are defined through their main project tasks, but as shown with the Aker Solutions
reliability methodology this can be specified according to organisation.
Standards such as ISO 20815 prove the possibility of developing reliability programmes based on the
project risk. If this standard is used together with the methodology, as in the case study, it will take
less time to define why the programme is necessary and which methods are to be used. If the
methodology is used as a tool that any organisation could alter to suit its project execution model, it
could easily be adapted to all new projects. The condition which then has to be fulfilled is that the
organisations methodology is suitable for very high risk projects with entirely new technology. It is
easier to leave out a reliability activity than including one more. This is the reason why Aker
Solutions technical qualification was used for the development of the Aker Solutions methodology.
ISO 20815 and other standards for reliability are great input to organisations wishing to use Design for
Reliability in their projects. Seeing how useful it was to have the methodology ready before the case
study, it could be suggested that a standardised methodology should be developed as well. This
methodology would have to be open for specification according to industry and organisation. Its
objective must be to ensure that reliability methods are shown according to which phase they suit and
how they ought to follow each other. This would increase the possibility of a correct implementation
of the methods and hence increase the confidence in their outputs.
Chapter 10 described a website developed based on this thesis. It shows that there are possible ways in
which Design for Reliability and a general methodology can be explained without too much cost. The
chapter also suggests how a website can be turned into a tool used by those learning how to implement
reliability in their design. As reliability becomes a more interesting topic to manufacturers in all
industries, such a tool is of high interest.

11.2

Recommendations for further work

The time constraints of the project are such that the methodology has not been studied as intensively
as could be desired. One suggestion is that it could be applied to several projects in different industries
and with varying risk levels. In comparing how the methodology is applied and followed up by the
different organisations, the adequacy of the methodology can be analysed. Alterations might be
suggested for further development.
If it was thought reasonable to prepare a reliability methodology for a standard, either alone or as an
addition to another standard, this should be studied. Although it is quite hard to make a tool like this
suit every industry and organisation, it is not impossible. An enquiry into what the relevant
organisations needs must be done. This enquiry should include questions about the project tasks

Master Thesis, TPK 4900


Ingrid Alms Berg

88

NTNU

Design for Reliability

14.06.2010

performed during the product life cycle, the reliability demands the organisation must answer to and
which reliability methods they already employ. In using this as a basis, a methodology could be
developed.
Design for Reliability has become a household concept with many manufacturers around the world,
but there are still those who doubt its usefulness. It would be interesting to perform two projects
starting from the same concept but with only one using Design for Reliability. The outcome of tests
performed by customers can show if there is a difference. If the products are fixed to achieve the
same reliability, it becomes evident which were the less costly. The success of such an experiment will
depend on the designers awareness of reliability issues and the tests performed during manufacturing.
It is possible that the outcomes are so different that a comparison will be difficult.
Comparative studies could also be done for reliability programmes. In this case it is reasonable to
suggest that the same project is performed a set of times but with small alterations in the programmes.
The alterations would be the methods chosen and the number of methods. This could possibly prove
the use of extensive reliability programmes in high risk projects and thinner programmes in low risk
projects. Also here there can be difficulties proving the effect if the projects turn out very different
from one another.

Master Thesis, TPK 4900


Ingrid Alms Berg

89

NTNU

Design for Reliability

14.06.2010

12. References
1.

A Cinderella Story by Leigh Dunlap. (2004) Film. Directed by Mark Rosman. USA:
Warner Borthers Pictures.

2.

API-RP-17N. (2009). Recommended Practice for Subsea Production System Reliability


and Technical risk Management. Washington D.C., American Petroleum Institute.

3.

Berg, I., (2009) Uncertainty in Risk and Reliability Studies. Project Thesis. Department of
Production and Quality Engineering, NTNU, Norway

4.

Blischke, W. and Murthy P., (2000) Reliability Modeling, Prediction, and Optimization.
New York, John Wiley & Sons.

5. Der Kiureghian, A. and Ditlevsen, O. (2009) Aleatory or Epistemic? Does it matter?


Structural Safety, 31, 105-112.
6.

Cleveland, C.J. (2010) Deepwater Horizon Oil Spill. The Encyclopaedia of Earth,
http://www.eoearth.org/article/Deepwater_Horizon_oil_spill [viewed 29/05/2010].

7. Federal Agencies (2010) GAP analysis. Federal Agencies Digitization Guidelines


Initiative. [Online] Available from:
http://www.digitizationguidelines.gov/stillimages/documents/Gap_Analysis.pdf
8.

Hallinan, N.M. (2009) Likelihood: Frequentists vs. Bayesian Reasoning. Integrative


Biology 200B. Spring 2009. [Lecture Notes]University of California, Berkeley. Available
from: http://ib.berkeley.edu/courses/ib200b/Lectures/Nat/Likelihood%20lecture.pdf

9.

Hallquist, E.J. and Schick, T. (2004) Best practices for a FRACAS implementation in
Reliability and Maintainability Symposium. Proceedings of the 2004 Annual Reliability
and Maintainability Symposium, IEEE, New York. pp 663-667.

10. IEC (2010) International Electrotechnical Commission www.iec.ch


11. IEC 60050-191. (1990). International Electrotechnical Vocabulary. Chapter 191:
Dependability and quality of service. Geneva, International Electrotechnical Commission.
12. IEC 60068-2-67 (1995). Environmental testing - Part 2: Tests - Test Cy: Damp heat,
steady state, accelerated test primarily intended for components. Geneva, International
Electrotechnical Commission.
13. IEC 60300 (2003) Dependability management. Geneva, International Electrotechnical
Commission.
14. IEC 60812 (2006) Analysis techniques for system reliability - Procedure for failure mode
and effects analysis (FMEA). Geneva, International Electrotechnical Commission.

Master Thesis, TPK 4900


Ingrid Alms Berg

90

NTNU

Design for Reliability

14.06.2010

15. IEC 61014 (2003) Programmes for reliability growth. Geneva, International
Electrotechnical Commission.
16. IEC 61025 (2006) Fault tree analysis (FTA). Geneva, International Electrotechnical
Commission.
17. IEC 61078 (2006) Analysis techniques for dependability - Reliability block diagram and
boolean methods. Geneva, International Electrotechnical Commission.
18. IEC 61160 (2006) Design review. Geneva, International Electrotechnical Commission.
19. IEC 61508 (2005) Functional safety of electrical/electronic/programmable electronic
safety-related systems. Geneva, International Electrotechnical Commission.
20. IEC 61511(2003) Functional safety - Safety instrumented systems for the process industry
sector. Geneva, International Electrotechnical Commission.
21. IEC 61882 (2001) Hazard and operability studies (HAZOP studies) - Application guide.
Geneva, International Electrotechnical Commission.
22. IEC 62059. (2008). Electricity metering equipment Dependability. Part 31-1:Accelerated
reliability testing, Elevated temperature and humidity. Geneva, International
Electrotechnical Commission.
23. IEEE (2009) IEEE Reliability Society - Reliability Engineering.
http://www.ieee.org/portal/site/relsoc/menuitem.e3d19081e6eb2578fb2275875bac26c8/in
dex.jsp?&pName=relsoc_level1&path=relsoc/Reliability_Engineering&file=index.xml&x
sl=generic.xsl [viewed 9/02/2010]
24. IfM (2010) Gap analysis. Institute for Nanufacturing, University of Cambridge. [Online]
http://www.ifm.eng.cam.ac.uk/dstools/choosing/gapana.html
25. ISO (2010) About ISO http://www.iso.org/iso/about.htm [viewed 04/05/2010]
26. ISO 13628 (2005) Petroleum and natural gas industries -- Design and operation of subsea
production systems. Brussels, International Organisation for Standardization
27. ISO 14224 (2006) Petroleum, petrochemical and natural gas industries -- Collection and
exchange of reliability and maintenance data for equipment. Brussels, International
Organisation for Standardization
28. ISO 17776 (2000) Petroleum and natural gas industries -- Offshore production
installations -- Guidelines on tools and techniques for hazard identification and risk
assessment. Brussels, International Organisation for Standardization
29. ISO 20815 (2008) Petroleum, petrochemical and natural gas industries -- Production
assurance and reliability management. Brussels, International Organisation for
Standardization

Master Thesis, TPK 4900


Ingrid Alms Berg

91

NTNU

Design for Reliability

14.06.2010

30. ISO/IEC 31010 (2009) Risk management -- Risk assessment techniques. Geneva,
International Organisation for Standardization and International Electrotechnical
Commission
31. Kritzinger, Duane. (2007) Aircraft System Safety: Tools and Techniques
www.aircraftsystemsafety.com/Information/ToolsandTechniques/tabid/59/currentpage/16/
Default.aspx
32. Lundteigen, M.A. (2009) Safety instrumented systems in the oil and gas industry: concepts
and methods for reliability assessments in design and operation. PhD Thesis. Department
of production and Quality Engineering, NTNU, Norway
33. Lundteigen, M.A., Rausand, M. and Utne, I.B. (2009) Integrating RAMS engineering and
management with the safety life cycle of IEC 61508. Reliability Engineering and System
Safety, 94, 1894-1903
34. MIL-HDBK-2155. (1995) Failure reporting, analysis and corrective action taken.
Washington D.C., Department of Defense, United States of America.
35. MIL-STD-785B (1980) Reliability Program for Systems and Equipment Development and
Production. Washington D.C., Department of Defense, United States of America.
36. MIL-STD-882D (2000) Standard Practice for System Safety. Washington D.C.,
Department of Defense, United States of America.
37. MIL-STD-1629(1980) Procedures for performing a Failure Mode, Effects and Criticality
Analysis. A Washington D.C., Department of Defense, United States of America.
38. Ministry of Defence. (2010) Safety and Environmental Protection
http://www.aof.mod.uk/aofcontent/tactical/safety/content/techniques.htm
39. Murthy, D.N.P., Rausand, M. and sters, T. (2008) Product Reliability Specification
and Performance. London, Springer
40. NASA (A) (1999) Public Lessons Learned Entry: 0757 - The team approach to FaultTree Analysis. http://www.nasa.gov/offices/oce/llis/0757.html
41. NASA (B) (1999) Public Lessons Learned Entry: 0786 Independent Review of
Reliability Analyses. http://www.nasa.gov/offices/oce/llis/0786.html
42. NASA (C) (1999) Public Lessons Learned Entry: 0825 System Reliability Assessment
Using Block Diagraming Methods. http://www.nasa.gov/offices/oce/llis/0825.html
43. NASA (D) (2010) GAP analysis. NASA Process Control. [Online] Available from:
http://process.nasa.gov/documents/RootCauseAnalysis.pdf

Master Thesis, TPK 4900


Ingrid Alms Berg

92

NTNU

Design for Reliability

14.06.2010

44. NORSOK Z-008 (2001) Criticality analysis for maintenance purposes. Oslo, Norwegian
Technology centre
45. NORSOK Z-013 (2008) Risk and emergency preparedness analysis. Oslo, Norwegian
Technology Centre
46. NPD(A), (2001), Regulations relating to conduct of activities in the petroleum activities
(The Activities Regulations), [Online]. Available from:
http://www.ptil.no/activities/category399.html
47. NPD(B). (2001) Regulations relating to design and outfitting of facilities etc. in the
petroleum activities(The Facilities Regulations). [Online]. Available from:
http://www.ptil.no/facilities/category400.html
48. NPD(C). (2001) Regulations relating to health, environment and safety in the petroleum
activities (The Framework Regulations). [Online]. Available from:
http://www.ptil.no/framework-hse/category403.html
49. NPD(D). (2001) Regulations relating to management in the petroleum activities (The
Management Regulations). [Online]. Available from:
http://www.ptil.no/management/category401.html
50. NUREG-0492 (Vesely, W.E., Goldberg, F.F., Roberts, N.H. and Haasl, D.F.) (1981) Fault
Tree Handbook. Washington D.C., U.S. nuclear Regulatory Commission
51. NUREG-1855. (Drouin, M., Parry, G., Lehner, J., Martinez-Guridi, G., LaChance, J and
Wheeler, T.) (2009) Guidance on the Treatment of Uncertainties Associated with PRAs in
Risk-Informed Decision Making, main report. Washington D.C., U.S. Nuclear Regulatory
Commission.
52. NUREG/CR-6832 (Atwood, C.L., LaChance, J.L., Martz, H.F., Anderson, D.J.,
Englehardt, M., Whitehead, D. And Wheeler, T.) (2003) Handbook of Parameter
Estimation for Probabilistic Risk Assessment. Washington D.C., U.S. Nuclear regulatory
Commission.
53. OConnor, P.D.T. (2002) Practical Reliability Engineering. Chichester: John Wiley &
Sons
54. Offshore (2009) History of the Offshore Industry. [Online]
http://www.offshore-mag.com/history/index.cfm
55. OLF (2010) The Norwegian Oil Industry Association. http://www.olf.no
56. OREDA (2002) Offshore Reliability Data handbook. 4th Edition. Hvik, DNV
57. Pat-Cornell, M.E. (1996) Uncertainties in risk analysis: Six levels of treatment.
Reliability Engineering and System Safety. 54, 95-111.

Master Thesis, TPK 4900


Ingrid Alms Berg

93

NTNU

Design for Reliability

14.06.2010

58. Perfors. A (2006) Bayesian vs. Frequentist in cogsci


http://www.iq.harvard.edu/blog/sss/archives/2006/02/bayesian_vs_fre.shtml [viewed
27/05/2010]
59. PSA (2010) Norwegian Petroleum Safety Authority. www.ptil.no
60. Rausand, G. (2005) Uncertainty Management in Reliability Analyses. Master Thesis.
Department of Production and Quality Engineering, NTNU, Norway.
61. Rausand, M. and Hyland, A. (2004) System Reliability Theory Models, Statistical
Methods and Applications. New Jersey, John Wiley & Sons.
62. Rausand (B), M. (2005) Preliminary Hazard Analysis [Presentation by M. Rausand]
[Online] Available from: www.ntnu.no/ross/srt/slides/pha.pdf
63. Selah, J.H. and Marais, K. (2006) Highlights from the early (and pre-) history of reliability
engineering. Reliability Engineering and System Safety. 91, 249-256
64. Sim, S.K. and Duffy, A.H.B. (2003). Towards an ontology of generic engineering design
activities. Research in Engineering Design, 14, 200-223.
65. Stapelberg, R.F. (2009) Handbook of Reliability, Availability, Maintainability and Safety
in Engineering Design. London, Springer

66. Swann, C.D. and Preston, M.L. (1995) Twenty-five years of HAZOPs. Journal of Loss
Prevention in the Process Industries, 6, 349-353.
67. Tech 482/535, (2005) What if Analysis. Industrial Safety Engineering Analysis, Northern
Illinois University [Class notes presented by ASSE students] Available from:
www.ceet.niu.edu/depts/tech/asse/tech482/what_if_analysis.doc
68. Thesaurus.com (2010) Unreliability. [Online] Available from:
thesaurus.com/browse/unreliability
69. Vallverd, J. (2003) The False Dilemma: Bayesian vs. Frequentist. Electronic Journal for
Philosophy [Online] Available from: http://nb.vse.cz/kfil/elogos/science/vallverdu08.pdf
70. Villacourt, M. and Goviul, P. (1994) Failure Reporting, Analysis and Corrective Action
System. SEMATECH [Online] Available from:
http://www.favoweb.com/doc/fracas_sematech.pdf
71. Weibull.com (A) (2010) Accelerated Life Testing
www.weibull.com/AccelTestWeb/acceltestweb.htm
72. Weibull.com (B) (2010) Reliability Growth Analysis www.weibull.com/basics/growth.htm

Master Thesis, TPK 4900


Ingrid Alms Berg

94

NTNU

Design for Reliability

14.06.2010

73. Williams, K. (2003) A Reliability Strategy for the Subsea Industry - an Operators
Impressions. [shown at SUT evening seminar, Aberdeen] [Online] Available from:
www.sut.org.uk/powerpoint/BPreliability2_files/frame.htm
74. Yang, G. (2007) Life Cycle Reliability Engineering. New Jersey, John Wiley & Sons
75. degaard, D.A.S. (2003) Reliability Management for new Subsea system developments.
Master Thesis. Department of Production and Quality Engineering, NTNU, Norway.
76. Confidential references

Master Thesis, TPK 4900


Ingrid Alms Berg

95

NTNU

Design for Reliability

14.06.2010

Appendices
There are 6 appendices to this report numbered from A to F. Except for the preliminary report
the appendices are confidential and therefore left out of the main report. The appendices are
meant to be read together with the main report and a description of how this should be done is
given in chapter 1.3.
The appendices are the following:
A) System description of the Midgard Subsea Compression System
B) Project execution model for Technical Qualification in Aker Solutions
C) Aker Solutions methodology for reliability
D) Project Execution Model (PEM) for Aker Solutions
E) Equipment Reliability Management Programme (ERMP) for the Midgard SCS
F) Preliminary report

Master Thesis, TPK 4900


Ingrid Alms Berg

96

You might also like