Where For Art Thou Failure Rate Data

David C. Arner William C. Angstadt

Engineering Associate – Process Controls Asset Management Group
Air Products and Chemicals, Inc. Rohm & Haas Company
7201 Hamilton Blvd PO Box 584
Allentown, PA 18195-1501 Bristol, Pa. 19007

KEYWORDS

Loop, CCPS PERD, Failure Rate Data, Equipment Taxonomy

ABTRACT

With the advent of Safety Integrity Levels (SIL) representing the relative reliability of instrumented
protection systems in a quasi-quantitative manner, there has been a stirring interest by engineers seeking
to understand where quality failure rate data can be obtained. This paper addresses that issue in the
context of the Center for Chemical Process Safety (CCPS) initiative regarding operation of an industry
Process Equipment Reliability Database (PERD). The work performed to date regarding instrument
loops will demonstrate the overall project, the importance of failure mode definitions and what it takes
to obtain high quality trusted data that can be turned into useful information.

INTRODUCTION

It seems that today, everyone in industry is asking where failure rate data can be found, as process
control engineers seek to design and implement safety instrumented systems, SIS, to satisfy some safety
integrity level, SIL. The answers are often the OREDA5 data book, IEEE-5004, or the CCPS Guidelines
for Process Equipment Reliability Data1. Additional reliability data for portions of an SIS may be
inferred from third party certifications to a SIL level, if the equipment design is integrated into the
overall protective loop and maintained in accordance with the dictates of the certification. Invariably,
those answering the question indicate that actual plant data is the best, as it would be more applicable to
the specific application of interest, and definitively represent the design, operation and maintenance of
the system(s) in question. This response sounds good, but it strikes an ironical note. If the best data
exists within plants operated by those asking for the data; it begs the question, why is the data not
available?

Copyright © 2001 ISA, All Rights Reserved Page 1 of 10

The Center for Chemical Process Safety (CCPS), an AIChE industry technology alliance, has been
attempting to answer that question by facilitating an industry effort whose ultimate goal is to operate an
equipment reliability database making available high quality, valid, and useful data to the HPI and CPI;
enabling analyses to support availability, reliability, equipment design improvements, maintenance
strategies, support life cycle cost determinations, and provide better, more credible information for risk
analyses. While contributing to the advancement of knowledge in this area, participants learn what it
takes to accomplish this goal and to utilize the knowledge gained and apply it to their management
systems in a manner that adds value and lays the foundation for continued improvement.

Multiple companies from the oil, chemical, and industrial gas industries, as well as consultants,
insurance companies, and equipment manufacturers have banded together to achieve this aim. These
participants have come together under the aegis of CCPS which is a non profit organization dedicated to
technical advancement and knowledge. CCPS is providing a forum to facilitate development and
sharing of technical information with respect to an industry process equipment reliability database
(PERD)2.

WHO NEEDS DATA ANYWAY

Failure rate and reliability data is useful for many different analyses. The major focus of this paper will
discuss what it takes to validate and measure the performance of instrument loops with respect to SIL
levels in a true sense. Other activities requiring the use of credible data include; risk analysis, risk based
inspection decisions, reliability analyses, improved equipment selection, and protection system proof
test interval optimization.

THE KNOWLEDGE ADVANCEMENT PROCESS

In order to take advantage of plant data effectively, a management information system is needed that has
addressed the issues of what data fields must be collected, in addition to what format and structure.
Experience has shown that companies have not built fundamentally sound information systems that
support the analyses that need to be performed using actual plant data. Accessing plant data is too often
a costly exercise, often leaving the analyst guessing as to the relative quality of the data, since
interpretation is often required.

CCPS PERD has developed a rigorous taxonomy development work process that can be applied to any
type of equipment, which defines the following; the failure modes, inventory data (necessary for
analysis of different populations), and event data (necessary to infer specific failure modes). This is
accomplished in a consistent format and structure that allows industry to aggregate high quality data in a
cost effective manner. Figure 1 shows an overview of the work process.

Copyright © 2001 ISA, All Rights Reserved Page 2 of 10

 Perform/Modify Functional Analysis Start

Establish/Modify Failure Modes of Interest

Document/Edit Desired Calculations

Develop/Document/Modify Inventory Data Fields

No Inventory Data
Fields/Pick Lists OK?

Yes
Develop/Document/Modify Events/Event Data
Fields
Yes
No
Failure Modes Adequate?

Develop/Document/Modify Failure Mode Rule

Sets

No
Calculation Docs OK?

Yes

Failure Mode Rule No

Sets Reconcile to Event Data Forms

Yes

Package Sub-committee Deliverables for Submit

Programming Scope Documentation

FIGURE 1 – TAXONOMY DEVELOPMENT WORK PROCESS

Copyright © 2001 ISA, All Rights Reserved Page 3 of 10

The taxonomy development work process is at the core of the engineering foundation enabling the
development of an industry database. It provides excellent scope documentation in a consistent manner,
regardless of equipment type, for the database software programmer. This enables the programmer to
add new equipment taxonomies efficiently to the overall software application program. The basic
procedure uses a rigorous step-by-step methodology. It provides an understanding of the role of the
taxonomy and data field specifications in the database. The procedure takes advantage of basic
OREDA5 concepts and then uses the theoretical thought process expounded by Rausand and Oin6, into a
rigorous and practical methodology for creating equipment taxonomies.

A functional analysis begins the work process. The ultimate purpose of this step is to determine the
equipment’s finite list of applicable failure modes. Prior to documenting the functions, it is necessary to
develop the boundary diagram that explicitly shows what is included and what is excluded from
consideration when a failure occurs. Figure 2 shows an example PERD boundary diagram for an
instrument loop.

Signal Control Logic Signal

Conditioning Unit Conditioning

Sensing Final Control

Element Element

Dedicated Circuit
Utilities i.e. electrical
Power, Instrument Air
etc.

Main Distribution
Utilities i.e. electrical
Power, Instrument Air
etc.

FIGURE 2 – INSTRUMENT LOOP BOUNDARY DIAGRAM

Once the failure modes have been identified, the standard calculations necessary to provide the
reliability parameters applying to each failure mode can be documented. Following this, the data
inventory fields useful in the analysis of different populations are recorded. Standard industry
specifications generally provide the necessary information. To maximize usefulness, field validation
rules are established wherever possible. Table 1 represents some of the general inventory data
describing loops.

Copyright © 2001 ISA, All Rights Reserved Page 4 of 10

When selecting the type of loop, the choices of primary function are: process control, interlock,
indicator, or alarm. Operating modes can be: continuous operation, standby, cyclical operation, or
batch. As many systems are multifunctional, secondary loop functions may be specified. For each of
the primary functions, additional information may be recorded. Table 2 illustrates a sample of this for
interlocks. These tables and associated validation pick list tables provide the documentation a
programmer needs to either improve their information management systems or to assist electronic
extraction and translation of data from existing systems into a CCPS PERD compatible format.

Field Name Description/Comments PickList/Units

Equipment Loop Number (1) The property or tag number assigned
to the system by the subscriber..
Loop Name
Systemtype (1) Represents primary loop function Final choice of cascading menu
System ID (1) Same as loop number for instrument
loops
Normal Operating Mode (1) For the primary loop function PL_OperatingMode
Secondary Loop Function(s) Allow selection of all that apply. PL_Loop_Func_2ND
[Use the choices selected to enable Check boxes
tabs as appropriate]
Initial Commissioning Date
Subordinate to other system Check box
Parent System Only if subordinate checked yes
(1) CCPS PERD required data

TABLE 1 – GENERAL LOOP INVENTORY DATA

Interlock Function Input Form [Enabled if Primary Loop Function (Systemtype) = Interlock Protection OR Secondary Loop
Function = Interlock]
Field Name Description/Comments PickList/Units
Variable batch setpoints? Check box
MUST ALLOW FOR MULTIPLE
ENTRIES OF THE FOLLOWING
INFORMATION
Interlock Name
Interlock Type PL_IntlkType
Interlock set point % of scale
Initiation direction Radio Buttons • N/A
• Increasing
• Decreasing
Interlock Initiation By? Radio Buttons • N/A
• Deenergize
• Energize
• Hybrid

TABLE 2 – ADDITIONAL INTERLOCK INVENTORY INFORMATION

The real work comes when developing the event data fields and validation rules that allow the inference
of failure modes that have been identified and defined as part of the functional analysis. Examples of
some inference rules are shown in Table 3. Figure 3 shows how it is all tied together in a relational
structure. The figure shows the basic relationship used for any equipment, and has been annotated to
show how it would be used for instrument loops.

Copyright © 2001 ISA, All Rights Reserved Page 5 of 10

 Complete Failures Failure Mode Definition Input Form Criteria
Control Output Frozen 1. Final element does not change with change in measured Demand - Control Control Response Successful? = No
variable (i.e. failure may be due to controller or final Tab AND “Control Output Frozen”
element) selected
2. Control Fails to Respond

Proof Test - Control Response Successful? = No

Control Tab AND “Control Output Frozen”
selected
Maint/Repair Condition Found = Control Output
Frozen
Auto Control Reason for Manual Control = Control
Bypass Output Frozen
Tuning Condition Found = Control Output
Frozen
Interlock Spuriously Function 1. Interlock causes actuation of its function without Maint/Repair Condition Found = Interlock
measured variable exceeding its set point OR Spuriously Function
2. Startup not allowed to proceed even though safe state
exists OR
3. Shutdown occurs even though fault condition does not
exists
4. Shutdown occurs even though degraded state within
system hardware/software does not exist
Demand - Auto Did process reach/exceed auto start
Start Tab setpoint? Field = No AND
Premature Start? = Yes
Demand - If Process Interlock functioned? Field
Shutdown Tab = Yes AND Did process reach/exceed
interlock set point? Field = No,
THEN
Failure Mode = Interlock Spuriously
Function
Partial Failures
Interlock voting channel fail to Individual channel or point unable to accomplish its function Maint/Repair Condition Found = Interlock voting
function if measured variable exceeds set point. Loop as a whole channel fail to function
may still be able to accomplish its mission.

TABLE 3 – FAILURE MODE INFERENCE RULE EXCERPTS

Copyright © 2001 ISA, All Rights Reserved Page 6 of 10

 Inventory Record Potential Events

For Loops:
Application Data Fields Device Data Fields - Demands
- Proof Tests
Tag Number - Repairs Records
(For Loops = Loop No.) - Tuning
Equip ID
- Startups
- Equip Descriptors
- Shutdowns
- Interlock Bypassed
(For Loops:
- Auto Control in
Operating Operating Equip ID = Loop No.)
Manual
Conditions Mode

Equipment Function(s)

Failure Mode(s)

Relationship Legend

Many to Many

One to Many
Failure
Failure Cause(s)
Mechanism(s)
Many to One

One to One

FIGURE 3 - SYSTEM LEVEL TAXONOMY RELATIONSHIPS

TRIPPING OVER FAILURE MODES

Table 4 displays the failure modes that were determined following the functional analysis of instrument
loops. Complete failures represent 100% loss of one or more functions, while partial failures represent a
degraded or partial loss of one or more functions. The importance of identifying the complete list of
fundamental failure modes cannot be understated. Many of the industry standards today are using the
terms, fail danger, or fail-safe. These terms are only true when very specific assumptions are made,
which apply to a relatively small percentage of total loops within a plant. Whether a failure mode is
dangerous or not is generally a function of the process (initiation on increasing or decreasing signal) and

Copyright © 2001 ISA, All Rights Reserved Page 7 of 10

its design (on-off versus modulating signal, de-energize versus energize to initiate, etc.). The CCPS
PERD work allows for a rigorous accounting of failure modes so that those applicable to a given
analysis may be used.

Complete Failures Partial Failures

• Control Output > 100 % • Control Output High
• Control Output Frozen • Control Output Low
• Control Output < 0 % • Control Output Slow to Respond
• Process Variable Indication > 100 % • Control Output Too Fast
• Process Variable Indication Frozen • Control Output Erratic
• Process Variable Indication < 0 % • Auto Controller in Manual Mode
• Control Output Indication > 100 % • Process Variable Indication High
• Control Output Indication Frozen • Process Variable Indication Low
• Control Output Indication < 0 % • Process Variable Indication Erratic
• False Discrete Indication • Control Output Indication High
• Alarm Fail to Function • Control Output Indication Low
• Alarm Spuriously Function • Control Output Indication Erratic
• Interlock Fail to Function • Alarm Function Delayed
• Interlock Spuriously Function • Interlock Functions Early
• Interlock Function Delayed
• Interlock voting channel fail to function
• Interlock voting channel spuriously functions

TABLE 4 – INSTRUMENT LOOP FAILURE MODES

An example of how this rigorous methodology compares to some of the best references available today
can be seen in Table 5, comparing a level transmitter with CCPS PERD failure modes versus two
separate editions of OREDA data books. Comparing the failure modes, one can note that some failure
modes are not listed in the OREDA tables. As before, the CCPS effort is methodically reducing and
eliminating confusion by documenting the outcome of a rigorous approach, that if done in a fundamental
way, need only be done once and forever used for the benefit of industry.

Copyright © 2001 ISA, All Rights Reserved Page 8 of 10

OREDA -92 Databook OREDA -97 Databook CCPS CCPS

Level Sensors – Process Sensor - Level Transmitter - Level Switch - Level (Includes
Transducer (1) (2) (Includes Sensor) Sensor)
Critical Failure Modes Critical Failure Modes Complete Failure Modes Complete Failure Modes
• Fail to Operate on • Fail to Function on
Demand Demand
• Spurious Operation • Spuriously Functions
• Output > 100 %
• No Change of Output • Frozen Output
With Change of Input
• No Output
• Unknown

Degraded Failure Degraded Failure Degraded Failure Modes Degraded Failure Modes
Modes Modes
• High Output • Set Point High
• Low Output • Low Ouput • Set Point low
• Output Slow to Respond
• Output Too Fast
• Erratic Output • Abnormal Output • Output Erratic
• Unknown • Others

Incipient Incipient Incipient Conditions Incipient Conditions

• Contaminated • Minor in-service
• Unknown problems
Unknown
• Unknown

Table 5 Notes:
1. Taxonomy consists of (OREDA-92 pages 436 &437)

• Safety Systems
• Process Alarm Systems
• Level Sensors
• Transducers (analog signals out, electric)
• Application = Level measurement/alarm on process systems
• Operational mode = Continuous

2. The boundary definition shown in OREDA-97 Figure 15, page 320 comprises the sensing element and the
(local) electronics for signal conversion and transmission. The sensing element measures some process
parameter (e.g. pressure, level, temperature, flow, etc,) and the electronics connects the measurement to a
standard electric signal that is transmitted to a computer. Some sensors may be calibrated by adjusting a screw
at the electronic housing. Isolation valves (block valves) and associated pipe work are also included.

TABLE 5 - CCPS VS OREDA FAILURE MODES

Copyright © 2001 ISA, All Rights Reserved Page 9 of 10

 CONCLUSION
It can be seen a new day is dawning in the world of failure rate data. A methodical process has been
developed to make use of actual plant data from multiple companies. This consistent methodology
establishes a foundation of meaningful data that satisfies the needs of many different applications.
Participation in this initiative allows companies to achieve immediate financial benefits as information
systems can be engineered to performance criteria that adds value.

ACKNOWLEDGEMENTS

The authors would like to acknowledge Hal Thomas, our process safety mentor and PERD Initiative
Chairman, Det Norsk Veritas, the PERD phase one contractor and technical consultant, CCPS for its
support, and the PERD Instrumentation Subcommittee, consisting of George Bacak, Gary Helman, Dave
Arner, Chip Rauth, Hal Thomas and Bill Angstadt, Instrumentation Subcommittee Chairman .

REFERENCES

1. CCPS Guidelines for Process Equipment Reliability Data, AICHE, 1989

2. CCPS Guidelines for Improving Plant Reliability Through Data Collection and Analysis, 1998.
3. CCPS PERD Instrumentation Sub-committee, Loop Taxonomy Documentation, Unpublished.
4. IEEE Std 500-1984, IEEE Guide to the Collection and Presentation of Electrical, Electronic,
Sensing Component, and mechanical Equipment Reliability Data for Nuclear-Power Generating
Stations.
5. OREDA-1992, Offshore Reliability Data, DNV Technica, Høvik, Norway, 1992
6. Rausand M. & Øien, K., “The basic concepts of failure analysis, In Reliability Engineering and
System Safety” 53 (1996), pp. 73-83.
7. Weber, Bernard J.; “Integrating Risk Based Mechanical Integrity and ISA 84.01”, AIChE 35th
Annual Loss Prevention Symposium Proceedings, April 2001.

NOMENCLATURE

Failure - The termination of the ability of an item to perform a required function.

Failure cause - The circumstances during design, manufacture or use which have led to a failure.
Failure descriptor - The apparent, observed cause of failure.
Failure mechanism - The physical, chemical, or other process or combination of processes that has led
to a failure.
Taxonomy - A systematic classification of items into generic groups based on factors possibly
common to several of the items, e.g., functional type, medium handled.

Copyright © 2001 ISA, All Rights Reserved Page 10 of 10