Handbook of Reliability, Availability, Maintainability and Safety in Engineering Design - Part 41 ppsx

4.2.3 Theoretical Overview of Availability and Maintainability Evaluation in Detail Design Availability and maintainability evaluation determines the measures of time that are subject to

• Equipment and/or system utilisation.

• Failure occurrence in the equipment.

• Failure mode of the failed component.

• Failure consequence and severity.

• Number of similar parts or components.

• Frequency of preventive maintenance replacement.

Although seemingly problematic from the perspective of complexity, the multiplic-ity of similar parts in each component, with usually a large number of similar

com-ponents within each system, is in fact beneficial in characterising the demand for different kinds of spares It validates the application of classical limit theory

con-cerning the maintenance renewal process This is illustrated by the following theo-rem (Drenick 1960):

given N components, indexed by i = N, K, 1, of which the failure processes are independent renewal processes, let F i (t) be the distribution for the time between failures of component i Furthermore,λiis the expected number of renewals per time unit, so that its reciprocal, 1/λi, is the expected time between failures of

component i.

Let G N (t) be the distribution of the time between failures across all components If:

(i) lim

N →∞λi/∑N

i=1λi= 0

(ii) F i (t) ≤ Atσand A > 0,σ> 0 as t → 0 ∀i

then

lim

N →∞G N



t /∑N

i=1λi



= 1 − e −λt for t > 0 (4.115) Consequently, Drenick’s theorem states that, under the above assumptions, the pooled output will approach a Poisson process as the number of failures increase Condition (i) is non-restrictive Condition (ii) is satisfied by all failure distributions commonly used—for example, the Weibull distribution Thus, when the demand for

a spare is the result of several component failure processes (which it normally is), the demand tends to be approximated by a Poisson distribution—that is, the demand rate is constant, irrespective of whether the individual components have arbitrary failure characteristics

There are only a few quantitative methods available when determining spares

requirements These are identified as analytical methods based on constant demand rates, analytical methods based on renewal theory, as well as simulation models.

Analytical methods based on constant demand rates tend to be the most applicable for spares requirements modelling

Renewal theory describes component failure by the renewal process that is char-acterised by a distribution for the time between renewals denoted F (t) If the distri-bution F (t) = 1 − e −λt, then the renewal process is a Poisson process with rateλ Hence, the renewal process is usually a generalisation of the Poisson distribution

However, the renewal process does not include several properties of the Poisson distribution Most importantly, the result of two independent renewal processes is not a renewal process unless both processes are Poisson processes Furthermore, the probabilistic split of a renewal process does not yield independent renewal pro-cesses

As indicated previously, when modelling for spares requirements, the demand is ultimately dependent upon several factors Spares demand is in most circumstances the result of the component failure characteristics If the component failure is mod-elled as a renewal process, the spare demand is not a renewal process In effect, models based on renewal theory have limited applicability in terms of spares op-timisation Such models are limited to a single process—that is, a single system, single component, and single part situation, which is very rare when determining an optimum spares requirements strategy for a real-world engineering design

Simulation models are generally impractical for spares optimisation (or, in fact,

any kind of optimisation) Event-driven simulation can be applied to analyse basi-cally any stochastic system or process In terms of optimisation, however, it is not applicable The reason for this is the relatively extensive time required for a single function evaluation Any optimisation algorithm iteratively evaluates an objective function and/or its derivatives numerous times in order to establish the optimal so-lution If each function evaluation takes time, the optimisation algorithm soon be-comes impractical Function evaluation is generally much faster, and optimisation feasible with analytical models based on Poisson demand (constant demand rate)

An analytical method for spares requirements based on a Poisson demand, or con-stant demand rate, which is approximated by the concon-stant failure rate, can thus

be developed (with a sufficient degree of acceptance) as the probability of having

a spare when required Such a probability takes into consideration the constant fail-ure rate of an item (component or part) that is intended to have a spare, the number

of items in the equipment and/or system that are intended to have spares (critical items), and the number of items in the system as a whole The following model can

be used to determine the spares requirement quantity (Blanchard et al 1995):

SP=∑m

i=0[(−1)ln(e −nλt)]ie−nλt /i! (4.116) where:

SP = the probability of having a spare when required

m = the number of items in the system as a whole

n = the number of items intended to have spares

t = period of time in which an item is likely to fail

λ = the constant failure rate of an item intended to have a spare

4.2.3 Theoretical Overview of Availability and Maintainability Evaluation in Detail Design

Availability and maintainability evaluation determines the measures of time that are subject to equipment failure, particularly known values of failure rates and repair rates for each individual item of equipment at the lower systems levels of the

sys-tems breakdown structure Availability and maintainability evaluation is considered

in the detail design phase of the engineering design process, with determination of the rates and frequencies that component failures occur and are repaired over a

spec-ified period of time The most applicable methodology for availability and maintain-ability evaluation in the detail design phase includes basic concepts of mathematical modelling such as:

i Dependability modelling for availability and maintainability

ii Operational availability modelling subject to logistic support

iii Maintainability evaluation and built-in or non-destructive testing

iv Specific application modelling of availability and maintainability.

Due to the increasing complexity of engineering processes, it is unrealistic to ex-pect that standard specifications covering the operational evaluation of a system are adequate for detail engineering designs The problem in the specification of the op-erational process is complexity Potential deviations from the expected opop-erational behaviour can be caused by unexpected failures in a complex system environment,

or by the complex integration of several systems To challenge the problems of complexity, all possible operational sequences must be considered in an operational specification, essential for modelling a complex system in its expected operational state, or at least according to a predetermined level of abstraction of such an oper-ational state This form of modelling, which incorporates operoper-ational specifications during the detail design phase of the engineering design process, is often termed

operational modelling The aim of operational modelling is to determine the

op-erational view of an engineering design, and to integrate it with opop-erational and technical specifications to guarantee model consistency Various operational models are considered, including a graphical formalism appropriate for modelling concur-rent processes, and thus for describing the operational view of complex integrated systems

4.2.3.1 Dependability Modelling for Design Availability and Maintainability

Dependability is the measure of a system’s condition during operation, provided that

it is available for operation at the beginning of its application (i.e operational avail-ability, which will be considered in detail in the following section) Dependability

can also be described as the probability that a system will accomplish its intended application (or mission), provided that it was available for operation from the begin-ning (Dhillon 1999b) Dependability models used for the evaluation of performance

of an engineering design are considered from a twofold meaning of the concept of dependability (Zakarian et al 1997):

• System operational integrity

(reliability, availability and maintainability)

• System performance

(dependence on the performance of equipment)

A dependability model that considers the operational integrity of a process

engi-neering system, where the system is considered to be operational as long as its

functional requirements are satisfied, includes the measures of operational integrity (operational reliability Ro, operational availability Ao, and operational

maintainabil-ity Mo) A dependability model that considers system performance includes mea-sures of the process characteristics In other words, a process system is assumed to

function properly if it is able to achieve the required level of performance where the

process capability, as given in Eq (4.17), exceeds a given lower bound of a

partic-ular process characteristic Careful consideration of these concepts of dependability

of a process engineering system during the engineering design stage can definitely improve system dependability

Dependability Ds, considering system operational integrity, is modelled as

Ds= Mo(1 − Ro) + Ao(Ro) (4.117) where:

Ro = operational reliability as fraction/percentage

Ao = operational availability as fraction/percentage

Mo = operational maintainability as fraction/percentage

Expressing system dependability in performance measures for operational relia-bility, availability and maintainability would include the measures of MTTR and MTBF In this case, system dependability is the sum of the ratios of system uptime

to total cycle time, and system repair time to total downtime

It is therefore an indication of the fraction of time that a system is available in

a cycle of system operation and failure, plus the fraction of time that the system

is repairable when it is down (i.e the ability of being used when it is up plus the

ability of being repaired when it is down) Thus

In the case where the performance measure of operational availability can be ex-pressed as

Ao= MTBF

where:

MDT= expected mean downtime

MDT= Tp m+ Tc m+ T

Tp m = preventive maintenance downtime

Tc m = corrective maintenance downtime

Tl d = logistics and administrative downtime

then

Ds= MTBF

MTTR

In the case where the expected mean downtime includes only preventive

mainte-nance downtime, the availability performance measure becomes inherent availabil-ity, and Dsis expressed as

Ds= MTBF

MTTR

4.2.3.2 Operational Availability (Ao ) Modelling with Logistic Support

Operational availability, unlike inherent availability or achieved availability,

cov-ers all segments of time that the system’s equipment is intended to be operational (total time in Fig 4.1) The same uptime and downtime relationship exists, except that it has been expanded Uptime now includes operating time plus non-operating

(standby) time when the equipment is assumed to be operable Downtime has been

expanded to include preventive and corrective maintenance and the associated ad-ministrative and logistics lead time All are normally measured in clock time This relationship is intended to provide a realistic measure of equipment availability when the equipment has been installed and is functioning in an operational envi-ronment Operational availability is used to support operational testing assessment and life-cycle costing

Operational availability is the most desirable form of availability to be used in evaluating the operational potential of equipment, and is an important measure of system effectiveness because it relates the system’s equipment, logistic support and environment characteristics into one meaningful parameter—an index depicting the state of equipment at the beginning of its operation in an engineered installation Be-cause it is an effectiveness-related index, operational availability is used as a starting point for nearly all system effectiveness and sizing analyses during the later stages

of the engineering design process

One significant problem associated with evaluating operational availability is that

it becomes costly and time-consuming to define all the various parameters, espe-cially during the detail engineering design phase when all equipment (assemblies and components) are being identified For instance, defining administrative and lo-gistics downtime per equipment per specified period, and total preventive mainte-nance under normal operational conditions is very difficult and not feasible in many cases Nevertheless, evaluating operational availability does provide an accepted methodology of relating standard reliability and maintainability characteristics into

a single effectiveness-oriented parameter As such, it is an essential tool for deter-mining the integrity of engineering design An important aspect to take note of when

evaluating operational availability is that it is affected by equipment usage or util-isation rate The less an item is used in a given period, the higher the operational

availability will be

Therefore, when defining the ‘total time’ period, it is important to exclude lengthy periods during which little or no system usage is anticipated One other ex-pression for operational availability is when standby time is assumed to be zero, typ-ical of single stream processes with no equipment redundancy While maintenance-oriented, this form of operational availability still retains consideration of the same basic time elements The downtime interval includes corrective and preventive maintenance, as well as administrative and logistics downtime This form of op-erational availability would generally prove more useful in support of defining pre-ventive maintenance requirements and logistic support analysis during the detail design phase of the engineering design process The general mathematical model for operational availability is (Conlon et al 1982):

where:

OT = operating time

ST = standby time

TCM = total corrective maintenance

TPM = total preventive maintenance

ALDT= administrative and logistics downtime

Inherent availability looks at availability from a design perspective, whereas

op-erational availability considers system effectiveness and the opop-erational potential

of equipment, and is used for analysing the sizing of equipment during the later stages of the engineering design process Thus, more encompassing maintainabil-ity measures of mean time between maintenance and mean downtime are used in the operational availability equation Operational availability is, in effect, a model

of maintainability measures in which downtime resulting from both corrective and

preventive maintenance is considered Aois thus a smaller availability value than Ai Operational availability can thus be mathematically expressed as

where:

MTBM= mean time between maintenance

The mean time between maintenance (MTBM) includes all corrective and

preven-tive actions (compared to MTBF, which accounts for failures—in contrast to the

concept of Ao for dependability in Eq (4.119)) The mean downtime (MDT)

in-cludes all time associated with the system being down for corrective maintenance including delays (compared to MTTR, which addresses only repair time), including downtime for preventive maintenance (PM), plus administrative and logistics down-time Although it is preferred to design equipment for which most PM actions can

be performed while the equipment is operating (such as built-in testing, BIT), PM

in this context implies a certain downtime

The uptime and downtime concepts for constant values of availability indicate the relative difficulty of increasing availability at higher percentages, compared to improving availability at lower percentages This is illustrated by the fact that in-creasing availability from 99 to 99.9% requires an increase in MTBM by one order

of magnitude or a decrease in MDT by one order of magnitude, whereas increasing availability from 85 to 90% requires improving MTBM by less than 1/2 order of

magnitude or decreasing MDT by 3/4 order of magnitude.

a) General Approach for Evaluating Operational Availability

The operational and maintenance concepts associated with system utilisation must

be defined in detail using terminology compatible with all involved in the design

of engineered installations Using these definitions, a time-line availability model is constructed that reflects the availability parameters, as illustrated in Fig 4.9 (Conlon

et al 1982)

Figure 4.9 displays elements of availability, particularly standby times (STW) and (STC), which are included in quantitative operational availability

The up or down status of a system during preventive maintenance must be closely examined because, generally, a portion of the preventive maintenance period may

be considered as uptime Standby time must also be examined closely before deter-mining system up or down status during this period With the aid of the time-line model, all time elements that represent uptime and downtime are determined For example, a maintenance strategy may be defined so that the equipment is maintained

in a committable or up-state during the performance of preventive maintenance Additionally, for multi-mode systems, it will be necessary to determine uptimes and downtimes as a function of each mode This generally will require the use of

a separate time-line model for each identifiable operational mode Separate time-line

Total time (TT)

Operating time

(OT)

Standby time

Standby (cold)

Fig 4.9 Operational availability time-line model—generalised format (DoD 3235.1-H 1982)

models are generally required to support the availability analyses of systems that experience significantly different continuous, periodic, and surge utilisation rates Quantitative values for the individual time-line models are determined and coordi-nated with the engineering design project management baselines Time elements are computed and availability evaluated, using the definitions of operational availabil-ity appropriate for the detail design phase Availabilavailabil-ity model status is continually checked and updated as required The model is updated as the operational, mainte-nance and logistics support concepts progressively become defined and quantifiable

b) System Availability Evaluation Considerations

As indicated previously, the quantitative evaluation of availability must be carefully and accurately tailored to each system However, there are certain general concepts

that will apply to different types of process engineering systems, such as recovery time Normally, availability measures imply that every hour has equal value from

the viewpoint of operations and maintenance/logistics activities The operational concept requires the system to function only for selected periods The remaining time is traditionally referred to as ‘off-time’ during which no activity is conducted

An alternative to ‘off-time’ or ‘cold standby’ is the use of the term ‘recovery time’ Recovery time represents an interval of time during which the system may be up or down (Fig 4.10) Recovery time, RT, does not appear in the operational availability calculation that is based only on the total time period TT Significantly, corrective maintenance time TCM is found in both TT and RT time intervals

Corrective maintenance performed during the TT period is maintenance required

to keep the system in an operational available status Corrective maintenance per-formed during the RT period generally addresses malfunctions that do not result in

a downtime status

The principal advantage of using recovery time analysis is that it can provide

a meaningful availability evaluation for systems with operational availability that

is predictable, and preventive maintenance that constitutes a significant portion of maintenance time The recovery time calculation technique concentrates availability calculation during the operational time period, thereby focusing attention on critical uptime and downtime elements

(RT)

Operating time

(OT)

Standby time

Recovery

Fig 4.10 Operational availability time-line model—recovery time format (DoD 3235.1-H 1982)

4.2.3.3 Maintainability Evaluation and Built-In or Non-destructive Testing

Maintainability has been defined as a characteristic of design and installation It is this inherent characteristic of a completed engineering design that determines the type and amount of maintenance required to restore or retain it in a specified dition Where maintainability is a design consideration, maintenance is the con-sequence of the design It is thus apparent that the ability and need to perform maintenance actions is the underlying consideration when evaluating maintainabil-ity The consideration of maintenance when designing engineering systems is not new There have been very successful efforts in the development of design for

acces-sibility, built-in testing, etc What is new is the emphasis on quantitative assessment

and evaluation that results in a complete change in engineering design philosophy, methodology and management In the past, design for maximum or optimum reli-ability and maintainreli-ability was emphasised However, all this resulted in was un-known reliability and maintainability from the design stage through to installation New techniques and methods allow design integrity judgment to be quantitatively

measured, as in the case of maintainability evaluation Maintainability evaluation is

the determination of design considerations and testing, intended to evaluate system maintainability characteristics that are based on quantitative measures or indices In addition to evaluating these characteristics, maintainability evaluation should also address the impact of physical design features on system maintenance and mainte-nance action frequency

There are various mathematical indices used to evaluate system maintainability characteristics These indices must be composed of measurable quantities, provide effectiveness-oriented data, and must be readily obtainable from applicable

develop-ment testing, such as the use of non-destructive testing (NDT) internal or integrated diagnostic systems, also referred to as built-in-test (BIT) or built-in-test-equipment

(BITE), and applied to pilot systems as well as to the engineered installation The

use of maintainability evaluation indices enables engineering designers to evaluate

system and/or equipment characteristics as well as logistics and maintenance prac-tices more precisely during the detail design phase

a) Maintainability Evaluation Indices

Mean time to repair (MTTR) As noted previously, the maintainability measure of

mean time to repair (MTTR) is the total corrective maintenance downtime

accumu-lated during a specific period, divided by the total number of corrective maintenance actions completed during the same period MTTR is commonly used as a general equipment maintainability measure, although it can be applied to each maintenance level individually MTTR considers active corrective maintenance time only Be-cause the frequency of corrective maintenance actions and the number of man-hours expended are not considered, this index does not provide a good measure of the maintenance burden

Maximum time to repair (MaxTTR) MaxTTR is the maximum corrective

main-tenance downtime within which either 90 or 95% (as specified) of all corrective maintenance actions can be accomplished A MaxTTR requirement is useful in those special cases in which there is a tolerable downtime for the system

An absolute maximum is ideal but impractical because there will be failures that require exceptionally long repair times A 95th percentile MaxTTR specification requires that no more than 5% of all corrective maintenance actions take longer than MaxTTR

Maintenance ratio (MR) MR is the cumulative number of man-hours of

main-tenance to be expended in direct labour over a given period of time, divided by the expected cumulative number of end-item operating hours Both corrective and preventive maintenance are included Man-hours for off-system repair of replaced components, and man-hours for daily operational checks are included for some sys-tems Particular care must be taken that the operating hour base be clearly defined, such as in the case of power-generating systems, when either system operating hours

or power delivery hours can be used MR is a useful measure to determine the rela-tive maintenance burden associated with a system It provides a means of comparing systems and is useful in determining the compatibility of a system with the required size of the maintenance organisation

Mean time between maintenance actions (MTBMA) MTBMA is the mean of

the distribution of the time intervals between either corrective maintenance actions, preventive maintenance actions or all maintenance actions This index is frequently used in availability calculations and in statistically oriented maintenance analyses

Average number of maintenance man-hours required The average number of

maintenance man-hours required at each maintenance level provides a quantitative means of expressing the personnel requirements of the overall maintenance concept This index also provides a conversion factor from active downtime to labour hours

Maintainability cost indices Maintainability is a significant factor in the cost of

equipment An increase in maintainability results in a reduction of logistic support costs of engineered installations A more maintainable system inevitably reduces maintenance times and operating costs, and a more efficient maintenance turnaround reduces downtime There are many factors of maintainability that contribute to the investment costs of engineered installations These include a direct effect on sys-tem and equipment hardware costs, support equipment, built-in testing, and contract spares

Off-system maintainability indices The indices MTTR, MaxTTR and MR all

specifically exclude off-system maintenance actions Off-system measures are par-ticularly important if a system’s maintenance strategy involves extensive use of modular removal and replacement for workshop repair/overhaul, since this type of concept transfers the maintenance burden to off-system maintenance As a main-tainability evaluation tool for engineered installations, off-system mainmain-tainability measures are essential Without these, it is not possible to evaluate the ability of