Safety, Reliability and Risk Analysis: Beyond the Horizon – Steenbergen et al.

(Eds)
© 2014 Taylor & Francis Group, London, ISBN 978-1-138-00123-7

Applying RAMSSHEEP analysis for risk-driven maintenance

W. Wagner
DPI Consultancy, Haarlem, The Netherlands

P.H.A.J.M. van Gelder

TU-Delft, Delft, The Netherlands

ABSTRACT: The Ministry of Infrastructure and the Environment of The Netherlands and Rijkswa-
terstaat (RWS) aim to launch a risk-driven maintenance concept named RAMSSHEEP. Engineering and
consultancy companies are expected to develop the RAMSSHEEP concept in order to create better main-
tenance strategies. The objective is to investigate whether RAMSSHEEP can be applied as a risk-driven
maintenance tool for primary flood defence systems in The Netherlands. By applying a probabilistic
approach to a flooding problem, more insight is gained in the advantages and disadvantages of RAM-
SSHEEP. This paper addresses the problem which primary flood defence systems experience with respect
to maintenance strategies. Often external political pressure is applied to improve reliability and availability
and to reduce costs significantly. Therefore this paper suggests an approach that is able to find an optimal
maintenance strategy by finding the economic most beneficial maintenance interval.

1 Introduction More and more (large) companies have made an

inventory of their process system(s) and qualified
The Ministry of Infrastructure and the Environ- and quantified the possible risks that may endanger
ment and Rijkswaterstaat (RWS) aim to launch an optimal reliable performance of the system. The
a certain risk-driven maintenance concept named main purpose is based on controlling or even avoid-
RAMSSHEEP (Reliability, Availability, Main- ing risks during operating phase (Andrews & Moss
tainability, Safety, Security, Health, Environment, 2002). Therefore maintaining a system will play a
Economics and Politics) that will be further devel- crucial role in ensuring the reliability and availabil-
oped by engineering and consultancy companies. ity of the system is as high as possible against the
The objective of this research is to assess whether lowest possible costs (optimization of efficiency).
RAMSSHEEP can be applied as a risk-driven So by applying the concept of risk management it
maintenance tool, based on the results of the exist- is possible to get an indication of the reliability of
ing method of a probabilistic approach with ‘Life the system and how actions can be taken to ensure
Cycle Costs’ (LCC). By applying a probabilistic an optimal condition of the system (Zaal 2011).
approach more insight is gained in the advantages A well-known analysis to get an indication of
and disadvantages of RAMSSHEEP. the performance reliability (quality) of the func-
tioning of a system can be described by Reliability,
Availability, Maintainability and Safety (RAMS).
1.1 RAMS
The RAMS analysis can be seen as a risk concept
Nowadays the world exists of complex installations that describes the primary performance of all the
and processes. These installations and processes are functions of a system. A RAMS analysis can be
expected to be reliable and become more important used in every stage of the life cycle for the entire
due to the high-technical mechanization systems infrastructure in The Netherlands: road network,
(Moubray 1997). The installations and processes in major waterways and main waters. This analysis
society will be indicated as reliable and robust until can be applied on a complete system but also for
system failure occurs. Questions arise, like: small components within a network. Crucial infor-
mation for a management team is:
– What is the reason for system failure?;
– Who or what is responsible?; and – Is the system available?;
– How to prevent system failure from happening – Do the objects fulfill its purpose (function)?;
again? – What is the condition of the objects?;

703
– Is the system safe to use and to maintain?; and Before going into detail of the aspects, first
– When and how is maintenance applied to the the whole definition of RAMSSHEEP is given
system? ­(Rijkswaterstaat 2012):
These questions will be made comprehensible – Reliability—indicates the failure probability
by executing a RAMS analysis: consistency of of a system in which its functions cannot be
Reliability, Availability, Maintainability and Safety. fulfilled.
The RAMS analysis is an unambiguous method to – Availability—indicates the time duration in
estimate the risks of a system and it may result in which the system is functional and its functions
several measures that must be taken to fulfill the can be fulfilled.
system to its requirements. Therefore many users – Maintainability—the ease in which the system
can apply this analysis in the system, i.e. the client, can be maintained over time.
contractor, and etc. – Safety—the absence of human injuries during
using or maintaining the system.
– Security—a safe system with respect to vandal-
1.2 History and background
ism, terrorism and human errors.
In the seventies of last century the American – Health—the objective argument of good health
Defense industry in corporation with the civil air- with respect to the physical, mental and societal
craft industry developed a Reliability-Centered views.
Maintenance (RCM) approach whereby reliabil- – Environment—influence of the system on its
ity played a crucial role in the system (Moubray direct physical environment.
1997). In the late seventies a risk method has been – Economics—a serious reflection in terms of
developed called Failure Mode and Effect Analysis costs versus benefits (as well as direct and indi-
(FMEA) to analyze the Reliability, Availability and rect) to provide more insight for an economical
Maintainability of a technical system. The official responsible choice.
document ‘Military Standard 1629a’, which has – Politics—a rational decision on all the previous
been published in 1980, has been one of the most aspects.
important sources for the RAMSSHEEP theory
It may be obvious that not every aspect has the
nowadays (Van Gestel et al. 2004).
same influence or consequence on indicating the
In the late eighties the European Commission
performance level of the system. Many aspects
decided to divide the properties in the rail sector:
can be summarized under the economics aspect,
the infrastructure was managed by the nation and
like security, health, environment and politics.
the exploitation of the trains was managed by the
By expressing these aspects in a monetary way it
private sector. This led directly to a response in
become easier to analyze and decide whether or
the rail sector to create a method that considers
not an action is profitable. The SHEEP analysis
the performance of a system and how this can be
has many connections to the sustainability, which
tested. Therefore the RAMS analysis has been used
also considers the results of social, ecological and
which has been developed in the US (Nowlan &
economical aspects. Often these aspects are con-
Heap 1978). Over the years this analysis has been
sidered in a FMEA for object functions whereby
expanded and further developed based on the tech-
maintenance activities can be ranked for sustain-
nique in the major industrial sector. This method
ability (Van Gestel et al. 2004).
has been taken over by other infrastructural sec-
The most interesting aspects are the reliabil-
tors (rail sector, civil engineering sector, and etc.).
ity, availability and maintainability of a certain
In the 00’s the RAMS analysis has been expanded
system. These three describe important informa-
to more aspects that should be considered by the
tion on the performance level nowadays (ProRail
determination of the reliability of a system: the
2010).
so-called RAMSSHEEP analysis. Developments
in the risk concept of RAMS analysis lead to the
expansion on more societal aspects. 2.2 Reliability
To start with the reliability which is directly
related to the frequency of failure. The more often
2 RAMSSHEEP Analysis
a system fails, the lower the reliability. Thus the
reliability is the complement of the unreliabil-
2.1 Original definition
ity. Therefore the relation between the reliability
By introducing RAMSSHEEP in several tenders and probability of failure (unreliability) can be
the current market will be pushed to develop expressed by:
RAMSSHEEP further so the analysis can be taken
to a higher level (Van den Breemer et al. 2009). R( x ) = 1 − F ( x ) (1)

704
where R(x) = function of reliability; and F(x) = func-
tion of cumulative probability of failure.
The reliability is changing over time and there-
fore also depends on new elements in the system
or on maintenance activities/replacements of a
certain element. In line with this, the reliability
will decrease over time due to usage of the system.
This reliability can be calculated, for instance, by
using the First Order Reliability Method (FORM).
The limit state function will be used as a proba-
bilistic calculation for a failure mechanism model
­(Steenbergen et al. 2004) like:

Z = f (x) (2) Figure 1. The direct relation and influence on RAMS-

aspects.
When Z becomes lower than zero the system
fails and logically this means if Z is zero or larger
that this describes the limit state or non-failure.
failure of the system (unplanned unavailability).
As one can see in the example above that the link
2.3 Availability between availability and maintainability is very
Subsequently, the availability is a theoretical rate of strong and direct. Therefore the maintainability is
time of which a system is able to fulfill its function. related to Mean Time To Repair (MTTR) and a
A system can be down due to two causes: unavaila- good design (adaptions/measures to system may
bility due to planned maintenance or due to failure lead to better maintainability).
events (which of course lead to unplanned mainte-
nance). Planned (preventive) maintenance logically 3 Probabilistic Analysis
leads to a lower probability of failure and thus a
smaller chance of unavailability by unplanned 3.1 Risks
(corrective) maintenance. The main task is to find
an optimum balance that should lead to the larg- A well-known method to quantify risks is by using
est possible availability. Thus the availability is the a probabilistic analysis.
complement of the unavailability. Therefore the To make an indication of risks its probability of
relation between the availability and unavailability failure and consequences must be determined. The
can be expressed by: definition of a risk can be seen as the probability
of failure of a scenario multiplied by the conse-
A = 1 − U = 1 − (U unpl + U pl ) (3) quence of that scenario:

R = Pf ⋅ C f (4)
where Uunpl = unavailability due to unplanned main-
tenance; and Upl = unavailability due to planned
maintenance. where Pf = probability of failure of the system; and
There is a direct relation between the planned Cf = economical consequence.
and unplanned unavailability of a system. For The probability of failure of a system gives the
example, the planned unavailability (preventive chance of failure due to a failure mechanism or a
maintenance, inspection, and etc.) may lead to less combination of failure mechanisms.
unavailability due to failure events and vice versa. When the total probability of failure of the
In other words, a good management of preventive system has been calculated, the weak spots can be
maintenance activities on a system can lead to less determined. This gives an indication where and
interruptions and therefore less unplanned failure what actions must be taken to decrease this prob-
causes. This may result in a more controllable and ability of failure. The identification of the weak
predictable performance level of the system. spots within the system can be determined by com-
paring the current safety level with the required
safety level.
2.4 Maintainability
For lowering the risk not only determining the
At last the maintainability indicates the ease to probability of failure is important, but also the
maintain a system (1) to prevent the system from consequence of a failure is necessary. The calcula-
functional failing (planned unavailability) and tion of economic damage will be based on three
(2) the time to repair the system due to functional damage categories: (1) direct material damage;

705
(2) direct process damage; and (3) indirect process X, vector of random variables) and can therefore
damage. be used to optimize the variable.
The probability of failure of a system may be
3.2 Optimization evaluated by using system reliability theory, calcu-
lating reliability of all components first, followed
A tool that can be applied in economic analysis is by an evaluation of the system probability of fail-
the Cost-Benefit Analysis (CBA). Costs must be ure. Extending reliability evaluation with optimiza-
put on a balance against benefits. The costs and tion can be done in two ways:
benefits contain not only the costs and benefits
of the actions that must be taken (financially), • Top-down approach; and
but also costs and benefits with respect to nature, • Bottom-up approach.
environment, spatial quality, and etc. Benefits are In the top-down approach, lifetime costs of
mostly defined as the reduction ratio of the prob- the system are defined in the space of the design
ability of annual damage. The optimal safety level variables within the components. Well-known
in economical way can be achieved by implemen- optimization methods may then be used to find
tation of total costs of actions and the expected the minimum of lifetime costs. In practice, this
damage will thereby be minimalized (Eijgenraam approach leads to a high-dimensional optimiza-
et al. 2000). tion problem; number of design variables in order
The risk-based optimization is adopted as the of 100’s.
main principle on which to base an analysis of Since the tool is ultimately aimed at supporting
acceptable risk. The type of optimization has decision-making for the definition of acceptable
been applied successfully in several earlier studies risk levels, it is important that the tool is transpar-
in engineering. Reference is made to Van Dantzig ent for decision-makers and design engineers. It is
(1956), Burcharth et al. (1995), Vrijling et al. (1998) the opinion of many authors that the top-down
and Voortman et al. (1998). The basic form of risk- approach is too much of a black box approach.
based optimization is economic optimization that An alternative is the bottom-up approach. In
is aimed at minimization of the lifetime cost of the this approach lifetime costs are defined in the space
system (see Fig. 2): of failure probabilities of individual system com-
ponents. In this case, the number of dimensions of
Clife ( p, X ) = l ( p ) + R( p, X ), with the optimization problem is reduced of the number
(5)
R( p, X ) = Pf ( p, X ) ⋅ C f  of individual components in the system. Generally,
this means a reduction of the number of dimen-
sions in system optimization by a factor 10.
where I(p) = investment in the system (to decrease To obtain the optimization in this form, the
the probability of failure); R(p,X) = monetary risk; functions of the cost of an individual compo-
Pf (p,X) = probability of failure of the system; and nent have to be found. System reliability theory
Cf = economical consequence. All these parameters provides probability as a function of failure prob-
depend on which measure will be taken to lower abilities of components and the correlation matrix
the lifetime cost (p, vector of design variables and (Gijsbers 1987). The investment function and the
correlation matrix have to be found by a closer
analysis of every component (Vrouwenvelder &
Lenos 1987).
The result of the optimization should be inde-
pendent of the strategy used. It can be proven that
the minimum of the lifetime costs can be deter-
mined. This is only if the investment function is
defined as the minimum investment in the compo-
nent for a given failure probability. Therefore, the
investment function can be found by minimization
of the investments in the section for a number
of prescribed failure probabilities (Voortman &
­Vrijling 2001).

3.3 Level of economical decision

Figure 2. Schematic illustration of optimal safety in an The cost-benefit analysis can be calculated on
(simplified) economical risk and investment model based different levels. Mostly, there are three dividing
on the CBA (Vrijling et al. 2002). categories:

706
• Micro economy (firm); 4 Maintenance
• Macro economy (nation); and
• Political science. 4.1 Deterioration model
Each category has its own focus. Starting at the The life span of an object is the time that passes
smallest part ‘micro economy’, a cost-benefit anal- between realization and the failure of the object
ysis of a firm is relatively simple. The analysis con- (loss of function). It may sound logical that
tains an optimization of the process/improvement strength is often time dependent. The intersection
costs and the gains of the firm. Finally, this should of strength and load determines the moment of
result in maximization of the profit (or minimiza- failure.
tion of the losses). A deterioration model gives the relation between
The next level ‘macro economy’ is a bit more com- strength and time. The relation can be linear, expo-
plex. The cost-benefit analysis is acting on a higher nential, logarithmic, and etc. The deterioration
level of society and this especially relate to the bene- model thus determines strength at every point in
fits. The cost of a certain activity will be more or less time and is an approximation of reality. The input
the same, but the benefits of that activity can be low required for the model is the initial strength and
what result in a loss. In macro economy (nation) one usually a number of parameters that describe char-
may decide to continue this activity when it seems acteristics of the material or the object. The param-
that there are other benefits like growth, employ- eters that serve as input for the deterioration model
ment, currency, inflation, and etcetera. So, the cost- are usually determined from tests or observations.
benefit analysis has been analyzed on the bigger They rarely have a certain value and can usually be
picture instead of only the activities on the firm. best described by a random variable. This means
The highest level ‘political science’ is based on that strength at a certain time is a function of ran-
more immaterial aspects. The costs of a certain dom variables and is thus a random variable itself.
activity will however be the same, but benefits can The deterioration models give a stochastic
be analyzed by considering more than only activity description over the process of strength over time.
benefits like environment, culture, health, and etc. This describes the objects condition on a certain
These benefits can be classified as intangible. The moment of time. This means that it does not only
politicians may decide to accept the losses of a cer- gives information for determining risks over that
tain activity by considering the intangible aspects period, but also the expected value for the amount
worth this loss. of repair on the object. The deterioration model
forms therefore an important position in the main-
tenance planning. However, this piece of infor-
mation of the behavior of the object is often not
known. A mathematical description of the strength
process over time has to be found by measurements
of the object and by physical research. The model
research will be based on a certain time interval
what forms the basis for the deterioration model by
using curve fitting and extrapolation techniques.
This also introduces model uncertainty. The
input parameter in the model will have some
uncertainties (numbers based on tests, experiences
or intuition). These parameters can therefore be
described as stochastic variables with each an own
probability density function.
The deterioration process can best be modeled
as a stochastic process. The result gives an average
value and deviation of the objects strength for each
moment of time. The deviation will increase when
the uncertainties (in the mathematical model) of
the input parameters are larger.

4.2 Limit state function

Failure is generally defined as non-performance
Figure 3. The levels on which economic decisions can of what is requested or expected. The limit state
be based on. is a state, where strength of and load on the

707
c­ onstruction are equal. Two types of limit states c­ omfort-driven, and etc. First of all, it is proper
can be distinguished, namely: to define the term maintenance in just one sen-
tence (Vrijling & Van Gelder 2006): “All activities
1. Ultimate Limit State (ULS); and
aimed at retaining a structure’s technical state or
2. Serviceability Limit State (SLS).
at reverting it back to this state, which is consid-
When the ULS is exceeded, failure occurs as ered a necessary condition for the structure to
result of collapse under extreme loads. Examples carry out its function”.
of ULS are i.e. collapse of an earth body, deflec- These activities include both the repair of the
tion of the structure, and etc. When the SLS is structural strength, back to its original strength,
exceeded, the functional demands can no longer and executing inspections. The costs of main-
be met (for a certain moment of time). Deflec- tenance of civil engineering objects amounts to
tions of a floor, cracking in reinforced concrete, approximately 1% of the founding costs per year.
waves which are too high behind a breakwater For a life span around 100 years this means that
and so on, are examples of exceeding of SLS. It the maintenance costs are of the same magnitude
must be named that there is a significant differ- as the construction costs. The decline in i.e. new
ence in limit states between design period and housing development projects, maintenance costs
maintenance period of objects in risk manage- are clearly becoming an increasingly greater share
ment. Often this difference is forgotten or used of the expenses.
interchangeably. A direct consequence is the desire to minimize
Generally, failure can be schematized as exceed- maintenance costs. In order to realize this, the opti-
ing the load over the strength. The state of an object mal maintenance strategy has to be found. From
can be described using a limit state function: the mechanical engineering maintenance theory,
the following classifications in strategies are known
Z = R− S (6) (Vrijling & Van Gelder 2006):
1. Corrective maintenance (failure dependent
where R = strength; and S = load.
maintenance); or
If the strength and/or the load are described
2. Preventive maintenance (state dependent
with random variables, Z is also a random variable.
maintenance).
If Z < 0 the object fails. The probability of failure
can be calculated by the area of the joint probabil- In the first case of fault dependent maintenance,
ity density function whereby Z < 0. By using equa- an object is repaired or replaced when it can no
tion 6 the probability of failure can be given by: longer fulfill its function. Thus, repair takes place
after failure; therefore a failure intervention is
0 involved. The life of an object is fully exploited.
Pr{Z < 0} = ∫ f (ξ )dξ (7)
−∞ z Objects failure (and the associated costs) is
accepted. In engineering this type of maintenance
where fz(ξ) = probability density of Z; and ξ = real- is usually not acceptable, because generally the
ization of Z. accepted probability of failure is limited. This
The difficulty with the determination of the type of maintenance can, however, be applied to
failure probability is the fact that the distribu- non-integral construction parts (parts which do
tion function of Z usually cannot be determined not contribute to the stability of the entire object),
exactly. Only in a few cases, i.e. where all variables with modest consequences of failure (provided
have a known distribution, the distribution of Z reparation or replacement is not postponed for too
can be determined. long).
In the second case with condition dependent
maintenance the state of the object is determined
4.3 Maintenance strategies
at set intervals, by means of inspections. The
Maintenance activities have certain goals that decision whether to carry out repairs is based on
must be fulfilled. But one must first define the observations. The inspection intervals can be regu-
real function of maintenance within the system. lar or dependent on the condition of the object.
The core of maintenance has always been based In the latter case condition parameters, indicat-
on technical and economic aspects. Eventually a ing the condition of the object, have to be visible.
balance between technical and economic aspects The probability of failure in a period between two
make a good maintenance plan. Many compa- inspections has to be sufficiently small. Generally,
nies do have a profit-driven maintenance plan, the lifetime of the object can be better exploited
which describes a desire to make as high profit than with usage dependent maintenance, but the
as possible. Other strategies with maintenance costs of the inspections do have to be taken into
plans are availability-driven, reliability-driven, account (Ten Wolde & Ghobbar 2012).

708
4.4 Net Present Value (NPV)
The maintenance costs depend on the decrease of
strength and can therefore be seen as a function of
the amount of maintenance:

CM = g ( ∆R(t ) ) (8)

where g ( ∆R(t ) ) = maintenance costs function

dependent on change in strength over time.
When costs will be calculated over a certain
time line, one is able to assess and compare this
outcome with maintenance activities. The Net
Present Value (NPV) of the costs due to failure Figure 4. The net present value of the (maintenance,
has been calculated for a situation in which this risk and total) costs dependent on the maintenance time
risk annually occurs and will be discounted to the interval (Vrijling & Van Gelder 2006).
current situation. This can be calculated by using
equation 8:
The optimization of maintenance can be
CM determined by searching for the minimum of the
CNPV (t ) = (9) expected net present value of the total costs. This
(1 + r )t paper will only focus on the invariable deteriora-
tion process whereby the maintenance intervals are
where CM = maintenance costs; r = annual discount of the same length. The expected net present value
rate; and t = time. of the total costs depend of just one single vari-
The annual discount rate of 2.5% is applica- able; the maintenance interval:
ble whereby the inflation has not been taken into
account. Eventually this NPV can be implemented, E [ NPV ](t ) = f (t ) (12)
by using equation 9, in the expected net present
value of the maintenance costs (investment) over The function in equation 12 has a minimum for
its lifetime and the representing risk, by using a certain value of t in which:
equation 4, at that moment of time:
f ′(t ) = 0 and f ″(t ) > 0 (13)
N
CI
E [ NPV ]I ( ∆I ) = ∑ (10) For example, maintenance on a road will
n= 1 (1 + r )n⋅ ∆I
describe an invariable deterioration process
whereby all the parameters stay constant and only
T Pf (t ) ⋅ C f the maintenance interval is variable. The expected
E [ NPV ]R (t ) = ∑ (1 + r )t
(11) net present value will result in a constant mainte-
t=1
nance interval. However, this is only true when the
reconditioning will be done with the same proper-
where CI = investment costs; r = annual discount ties as the original condition.
rate; n = total amount of maintenance activi-
ties; ∆I = maintenance interval; Pf(t) = probabil-
ity of failure; Cf = economical consequence; and 5 Case Study
t = time.
Hereby Pf is the annual probability of failure, This section shows the basic probability theory in
which is based on the limit state function Z (see combination with the risk-driven RAMSSHEEP
also equation 7). These formulas can be used to analysis applied on ‘The Afsluitdijk’, a primary
determine the most (economically) optimal main- flood defense system in The Netherlands.
tenance interval over time. When the expected The gathered data for probability of flooding
investment costs (equation 10) summed with the has been based on annual maximum water levels
expected risk after these investment (equation 11) on 81 years from 1932 until now. Other supporting
are lower than the current risk, it is profitable data has been based on research and case studies
to take these investment measures. Otherwise, that was supported by RWS. All data assumptions
it helps to decide not to take any investments. that have been made are approved by RWS and
Hereby a certain optimum can be determined (see will therefore be credible and reliable enough to
Fig. 4). use in this case study.

709
5.1 Background Table 1. Overview of the division of functional risk.
The Dutch Deltacommittee advised the Dutch Annual monetary Part of
government on controlling the probability of Element/object risk total
flooding of the primary flood defence systems to
an economic and social optimum. This advice has Dike €6,087,000 67.8%
been based on the climate change what inter alia Discharge sluice €20,000 0.2%
will lead to water level rise (Deltacommissie 2008). Navigation lock €2,214,300 24.7%
One of the advices was that the dikes in the Road €657,000 7.3%
IJsselmeer area should be increased to fulfill the Total €8,978,300 100.0%
requirements in case of chance in failure. The Afs-
luitdijk should also be accounted for to increase its
crest level.
In this case study the complete system of the
Afsluitdijk has been analyzed and can be decom-
posed by: dike, road, navigation lock, drainage
sluice. Hereby, the focus will lay on the economical
optimum based on risk analysis. As the Deltacom-
mittee advised, this case study will give an opti-
mum solution for investing in the system and also
how and when it should be maintained.

5.2 Illustrated example

A risk-driven RAMSSHEEP analysis on the pri-
mary flood defense system ‘The Afsluitdijk’ must
be applied on the complete system, what includes
all its object functions: Figure 6. Schematically view of the contribution of the
1. Retaining of water (high/storm level); object functions to total net present value of the mon-
2. Discharging of surplus water from the rivers; etary after taking investments.
3. Navigating ships through locks; and
4. Road connection over the dike.
The risks determined in Table 5, should be
First of all, by making an economical probabil- reduced in the most efficient and economical way.
istic analysis of the current situation with the risk In other words, some measurements should be
formula (equation 4) one is able to analyze which taken to reduce the probability of occurrence of
function is dominantly present in the determina- a certain failure. These investments in the system
tion the total risk of the system. can be modeled on several parameters, like an
Figure 5 and Table 1 represent the present value extra road lane, higher crest of the dike, and etc.
of the monetary risk and points out that the dike On every investment measurement a cost-benefit
caries the largest risk as a part of the total. analysis will be made, to receive more insight in the
influence of the probability of occurrence on the
most efficient measure, based on equation 5.
Figure 6 and Table 2 represent the present
value of the monetary risk and points out that
the dike does no longer carry the largest risk, due
to the investment (increasing dike crest) (Kuijper
1992). The result of the calculation showed that
investment for the discharge sluice, navigation
lock and road was not profitable in compari-
son with its current risk; so no investment was
suggested.
The current situation makes it clear that the dike
had the highest monetary risk and was dominant
in the analysis. After the investment analysis has
been made, the choice of executing this investment
Figure 5. Schematically view of the contribution of the based on the economical most beneficial value
object functions to the total present value of the mon- changed this dominancy. If an investment has been
etary risk within the system. made, the expected costs should be decreased.

710
Table 2. Investments that lead to lower expected costs quality of available historical data. This is crucial
given by the present value. for an accurate reliability number.
The availability can be calculated by using the
Annual monetary risk reliability, maintenance interval and MTTR. When
Before After Part of these parameters have been implemented in equa-
Element/object investment investment total tion 3 the availability is known:

Dike €6,087,000 €557,207 16.2% A = 1 − U = 1 − (U unpl + U pl )

Discharge sluice €20,000 €20,000    0.5%
Navigation lock €2,214,300 €2,214,300 64.2%  1 MTTRc  1  MTTRp 
= 1−  ⋅ + 1 −  ⋅ (14)
Road €657,000 €657,000 19.1%  R ∆I M  R ∆I M 
Total €8,978,300 €3,448,507 100.0%
where R = reliability; MTTRc = corrective mean
time to repair; MTTRp = preventive mean time to
repair; and ∆IM = maintenance interval.
It is not remarkable that the reliability and avail-
ability of the dike are higher than the other object
functions (see results of Table 4). This is due to
the fact that the failure damage of this function is
extremely higher.
The fact that the dike does not need any regular
maintenance in this case, this has been indicated in
Table 3 by preventive MTTR, depends on the ini-
tial situation, but in general the maintenance inter-
vals increase because the subsoil becomes stiffer

Figure 7. Optimization of the crest height of the dike. Table 3. An overview of the preventive and corrective
MTTR in days to calculate the availability per function.

In Figure 7 the risk optimization of the height MTTR [d]

of the dike crest has been shown. The dashed line
represents the decrease of the risk for a higher Function Preventive Corrective Maintenance
crest level that comes together with a higher invest-
Dike 0 182 Heightening
ment cost (dotted line). According to equation 5 crest level
(the superposition of the dashed and dotted line) Discharge – – Repairing bed
the economical most beneficial crest height can be sluice protection
determined. Navigation 30 365 Vertical
Now when the monetary risk has been put to a lock cut-off wall
minimum it is also necessary to conserve the sys- Road 128 182 Renewing
tems’ quality for the coming years based on the asphalt
principle of LCC. This analysis led to a predic-
tive maintenance interval for as well as the navi-
gation lock and the road of respectively 14 and
Table 4. An overview of the reliability and availability
18 years. In this specific case the discharge sluice
numbers and the corresponding maintenance interval
and dike will not give any results, due to very slow and activity.
deterioration.
The analysis for the maintenance intervals has R A ΔIM
been calculated with equations 9, 10 and 11 based Function [RP in yr] [%] [yr] Maintenance
on the minimum costs over its lifetime that has
been given in equation 13. Dike 20,900 99.9 – Heightening
The results from this analysis can also be used crest level
to calculate the optimal reliability and availability Discharge 1000 99.9 – Repairing bed
what one may use to control the system (threshold sluice protection
level) and to steer on. The reliability and availabil- Navigation 86 98.5 14 Vertical
lock cut-off wall
ity have been given according to the equations 1
Road 80 97.6 18 Renewing
and 3. N.B. The practical approach for calculating asphalt
the reliability strongly depends on the amount and

711
 – The asphalt layer will be renewed every 18 years;
and
– The bed protection in front of the navigation
lock will be repaired every 14 years.
These maintenance activities describe a main-
tenance strategy that does not imply any (unfore-
seen) calamities. The annual savings, based on
risks, will be more than €5.5 million (of almost
€9.0 million) at the systems optimum for the
Afsluitdijk.
The risk-driven RAMSSHEEP approach is
mainly based on the economical driver to optimize
the reliability, availability and maintenance inter-
vals, so knowing this the type of analysis can best
Figure 8. The process of the strength of the road be renamed as EMAR: Economics, Maintenance,
over time, indicated by the length of cracks in meters Availability and Reliability. This risk-driven type
(Z-function). of analysis gives a broader, more simple and com-
plete basis for solving an infrastructural problem.
over time. The variable deterioration process of the It should be noted that an EMAR analysis should
dike, of course, can explain this (Kuijper 1992). be applied per object function within a system
Instead of the dike, the other object functions (decomposition and interfaces) and not per system
describe an invariable deterioration process and function. All the object functions together (inte-
therefore show a constant maintenance interval gration) will give a result for the EMAR analysis
over its life cycle; as we can see in Figure 8 the on top level (system level).
maintenance interval of the road is 18 years.
REFERENCEs
6 CONCLUSION
Andrews, J.D. & Moss, T.R. 2002. Reliability and Risk
This research paper answers the original research Assessment. London and Bury St. Edmunds: Profes-
sional Engineering Publishing Limited.
question, by doing a theoretical research and Burcharth, H.F., Dalsgaard Sørensen, J. & Christiani, E.
enriching this with an elaborated example. The 1995. Application of reliability analysis for optimal
research question was to assess (advantages and design of vertical wall breakwaters. Proceedings of
disadvantages) whether RAMSSHEEP can be the International Conference on Coastal and Port
applied as a risk-driven maintenance tool by apply- Engineering in Developing Countries (COPEDEC).
ing a probabilistic approach with LCC. The illus- Deltacommissie 2008. Samen werken met water: Een
trated example shows how to apply a risk-driven land dat leeft, bouwt aan zijn toekomst: Bevindingen
RAMSSHEEP analysis based on a probabilistic van de Deltacommissie 2008. Den Haag: Hollandia
approach with an economical driver to optimize. Printing.
Eijgenraam, C.J.J., Koopmans, C.C., Tang, P.J.G. &
This resulted in investment activities on the current Verster, A.C.P. 2000. Leidraad voor kosten-baten
situation and an optimal maintenance strategy. analyse—onderzoeks-programma Economisch Effecten
The RAMSSHEEP analysis is applicable to any Infrastructuur. Den Haag: Ministerie van Verkeer en
possible system within the infrastructural sector. Waterstaat & Ministerie van Economische Zaken.
This analysis determines the optimal maintenance Gijsbers, F.B.J. 1987. Optimisation of Maintenance
interval based on failure frequency and costs by TNO-IBBC report B-86-549. Delft: TNO.
using historical data. Not only RWS will benefit Kuijper, H.K.T. 1992. Maintenance in Hydraulic Engi-
from this probabilistic approach, but also users in neering, Economically Sound Planning of Mainte-
general could benefit due to a good balance of tax nance. Delft: Delft University of Technology.
Moubray, J.M. 1999. Reliability-centred Maintenance.
money and safety (maintenance). Oxford: Elsevier Ltd.
The RAMSSHEEP analysis is applied to a Nowlan, F.S. & Heap, H.F. 1978. Reliability-Centered
Dutch primary flood defense system. Several Maintenance. Washington, DC: United States Depart-
assumptions have been made in strength deterio- ment of Defense (Manpower, Reserve Affairs and
ration (failure frequency) and investment costs to Logistics).
model the system. According to the results of the ProRail 2010. Hand-out RAMS/LCC analyse: Leidraad
analysis, the costs are minimized when: voor Systems Engineering. NL: Prorail.
Rijkswaterstaat 2012. Leidraad RAMS—sturen op presta-
– The crest level of the dike will be increased by ties van systemen. Den Haag: Ministerie van Verkeer
0.81 m; en Waterstaat.

712
Steenbergen, H.M.G.M., Lassing, B.L., Vrouwenvelder, Voortman, H.G. & Vrijling, J.K. 2001. A Risk-based opti-
A.C.W.M. & Waarts, P.H. 2004. Reliability analysis mization strategy for large-scale flood defence systems.
of flood defence systems. HERON Volume 49 No. 1: Delft: Delft University of Technology.
51–73. Vrijling, J.K. & Van Gelder, P.H.A.J.M. 2006. Probabi-
Ten Wolde, M. & Ghobbar, A.A. 2012. Optimizing listic Design in Hydraulic Engineering. Delft: Delft
Inspection Intervals. Reliability Engineering & System ­University of Technology.
Safety Volume 114 June 2013: 137–147. Vrijling, J.K., Van Gelder, P.H.A.J.M., de Ridder, H.A.J.,
Van Dantzig, D. 1956. Economical decision problems for Kuijper, H.K.T., & Vrouwenvelder, A.C.W.M. 2002.
flood prevention. Econometrica No. 24: 276–287. Probability in Civil Engineering. Delft: Delft Univer-
Van den Breemer, J.J.A., Al-Jibouri, S.H.S., Veenvliet, K.T. & sity of Technology.
Heijmans, H.W.N. 2009. RAMS and LCC in the design Vrouwenvelder, A.C.W.M. & Lenos, S. 1987. Optimisa-
process of infrastructural construction projects: an tion of Maintenance—part II TNO-IBBC report
implementation case. Enschede: Twente University of B-87-717. Delft: TNO.
Technology. Zaal, T.M.E. 2011. Profit-Driven Maintenance for Physical
Van Gestel, P., Bouwman, E. & Reijnen, I. 2004. FMECA Assets. Geldermalsen: Maj Engineering Publishing.
Failure Mode Effect and Criticality Analysis. Deventer:
Sigma, Wolters Kluwer.
Voortman, H.G, Kuijper, H.K.T, Vrijling, J.K 1998.
Economic optimal design of vertical Breakwaters.
Proceedings 26th International Conference on Coastal
Engineering (ICCE): 2124–2137.

713