-
1000
CHAPTER 31RELIABILITY IN THE MECHANICALDESIGN PROCESS
B. S. DhillonDepartment of Mechanical EngineeringUniversity of
OttawaOttawa, Ontario, Canada
1 INTRODUCTION 1000
2 STATISTICAL DISTRIBUTIONSAND HAZARD RATE MODELS 10012.1
Statistical Distributions 10012.2 Hazard Rate Models 1002
3 COMMON RELIABILITYNETWORKS 10033.1 Series Network 10043.2
Parallel Network 10053.3 SeriesParallel Network 10053.4
ParallelSeries Network 10063.5 K-out-of-m-Unit Network 10073.6
Standby System 10083.7 Bridge Network 1009
4 MECHANICAL FAILURE MODESAND CAUSES OF GENERALAND GEAR FAILURES
1010
5 RELIABILITY-BASED DESIGNAND DESIGN-BY-RELIABILITYMETHODOLOGY
1011
6 DESIGN RELIABILITYALLOCATION ANDEVALUATION METHODS 10126.1
Failure Rate Allocation Method 10136.2 Hybrid Reliability
Allocation
Method 1013
6.3 Safety Factor and Safety Margin 10146.4 StressStrength
Interference
Theory Method 10156.5 Failure Modes and Effect
Analysis (FMEA) 10166.6 Fault Tree Analysis (FTA) 1017
7 HUMAN ERROR ANDRELIABILITY CONSIDERATIONIN MECHANICAL DESIGN
1017
8 FAILURE RATE ESTIMATIONMODELS FOR VARIOUSMECHANICAL ITEMS
10188.1 Brake System Failure Rate
Estimation Model 10188.2 Compressor System Failure Rate
Estimation Model 10198.3 Filter Failure Rate
Estimation Model 10198.4 Pump Failure Rate
Estimation Model 1019
9 FAILURE DATA AND FAILUREDATA COLLECTION SOURCES 1019
REFERENCES 1021
BIBLIOGRAPHY 1022
1 INTRODUCTIONThe history of reliability may be traced back to
the early 1930s when probability conceptswere applied to problems
related to electric power systems.17 During World War II,
Germanresearchers applied the basic reliability concepts to improve
reliability of their V1 and V2rockets. During the period 19451950,
the U.S. Department of Defense (DOD) conductedvarious studies that
revealed a denite need to improve equipment reliability.
Consequently,the DOD formed an ad hoc committee on reliability in
1950. In 1952, this committee wastransformed to a permanent body:
Advisory Group on the Rer-liability of Electronic Equip-ment
(AGREE).8 The group released its report in 1957.
Mechanical Engineers Handbook: Materials and Mechanical Design,
Volume 1, Third Edition.Edited by Myer Kutz
Copyright 2006 by John Wiley & Sons, Inc.
-
2 Statistical Distributions and Hazard Rate Models 1001
In 1951, W. Weibull proposed a function to represent time to
failure of various engi-neering items.9 Subsequently, this function
became known as the Weibull distribution and isregarded as the
starting point of mechanical reliability along with the works of A.
M. Freu-denthal.1011
In the early 1960s, the National Aeronautics and Space
Administration (NASA) playedan important role in the development of
mechanical reliability, basically due to the followingthree
factors12:
The loss of Syncom I in space in 1963 due to a bursting
high-pressure gas tank The loss of Mariner III in 1964 due to a
mechanical failure The frequent failure of components such as
valves, regulators, and pyrotechnics in the
Gemini spacecraft systems
Consequently, NASA initiated and completed many projects
concerned with mechanical re-liability. A detailed history of
mechanical reliability is given in Refs. 1315 along with
acomprehensive list of publications on the subject up to 1992.
2 STATISTICAL DISTRIBUTIONS AND HAZARD RATE MODELSVarious types
of statistical distributions and hazard rate models are used in
mechanicalreliability to represent failure times of mechanical
items. This section presents some of thesedistributions and models
considered useful to perform various types of mechanical
reliabilityanalyses.
2.1 Statistical DistributionsThis section presents three
statistical or probability distributions: exponential, Weibull,
andnormal.
Exponential DistributionThis is probably the most widely used
distribution in reliability work to represent the failurebehavior
of various engineering items.16 Moreover, it is relatively easy to
handle in perform-ing reliability analysis in the industrial
sector. Its probability density function is expressedby14,16
t(t) e for 0 t 0 (1)
where (t) probability density function distribution parameter;
in reliability work, it is known as the constant failure
ratet time.
The cumulative distribution function is given by14,16
t t tF(t) (t) dt e dt0 0 (2)t 1 e
where F(t) is the cumulative distribution function.
-
1002 Reliability in the Mechanical Design Process
Weibull DistributionThis distribution was developed by W.
Weibull in the early 1950s and can be used to rep-resent many
different physical phenomena.9 The distribution probability density
function isexpressed by14,16
1t(t) e(t /) for t 0 0 0 (3)
where and are the distribution scale and shape parameters,
respectively. The cumulativedistribution function is given
by14,16
t t1t (t / )F(t) (t) dt e dt
0 0
(t / ) 1 e (4)For 1 and 2, the Weibull distribution becomes the
exponential and Rayleighdistributions, respectively.
Normal DistributionThis is one of the most widely known
distributions. In mechanical reliability, it is often usedto
represent an items stress and strength. The probability density
function of the distributionis expressed by
21 (t )(t) exp t (5) 222where and are the distribution
parameters (i.e., mean and standard deviation, respec-tively). The
cumulative distribution function is given by14,16
t t21 (x )
F(t) (t) dt exp dx (6) 222
2.2 Hazard Rate ModelsIn reliability studies, the term hazard
rate is often used. It simply means the constant ornonconstant
failure rate of an item. Thus, the hazard rate of an item is
expressed by
(t)h(t) (7)1 F(t)where h(t) is the item hazard rate.
This section presents four hazard rate models considered useful
to perform various typesof mechanical reliability studies:
exponential, Weibull, normal, and general.
Exponential DistributionBy substituting Eqs. (1) and (2) into
Eq. (7), we get the following equation for the expo-nential
distribution hazard rate function:
-
3 Common Reliability Networks 1003teh(t)
t1 (1 e ) (8)
As the right-hand side of Eq. (8) is independent of time, is
called the failure rate.
Weibull DistributionBy substituting Eqs. (3) and (4) into Eq.
(7), we get the following equation for the Weibulldistribution
hazard rate function:
1 (t / ) 1[(t / ) e ] th(t) (9)(t / ) 1 [1 e ] For 1 and 2, Eq.
(9) becomes the hazard rate function for the exponential
andRayleigh distributions, respectively.
Normal DistributionBy substituting Eqs. (5) and (6) into Eq.
(7), we get the following equation for the normaldistribution
hazard rate function:
2 21/ (2) exp [(t ) /2 ]h(t) (10)t
2 21 1/(2) exp [ (x ) /2 ] dx
General DistributionThe distribution hazard rate function is
dened by17
m1 m1 th(t) ct (1 c)mt e for 0 c 1 , , m, 0 (11)where , scale
parameters
, m shape parameterst time
The following distribution hazard rate functions are the special
cases of Eq. (11): Bathtub; for m 1, 0.5 Makeham; for m 1, 1
Extreme value; for c 0, m 1 Weibull; for c 1 Rayleigh; for c 1, 2
Exponential; for c 1, 1
3 COMMON RELIABILITY NETWORKSComponents of a mechanical system
can form congurations such as series, parallel, seriesparallel,
parallelseries, k out of m, standby, and bridge. Often, these
congurations arereferred to as the standard congurations. Sometime
during the design process, it might be
-
1004 Reliability in the Mechanical Design Process
1 2 3 m
Figure 1 Series system block diagram.
desirable to determine the reliability or the values of other
related parameters of systemsforming such congurations. All these
congurations or networks are described below.14,16
3.1 Series NetworkThe block diagram of an m-unit series network
or conguration is shown in Fig. 1. Eachblock represents a system
unit or component. If any one of the components fails, the
systemfails; that is, all of the series units must work normally
for the system to succeed. Forindependent units, the reliability of
the system shown in Fig. 1 is
R R R R R (12)S 1 2 3 mwhere RS series system reliability
m number of unitsRi reliability of unit i for i 1, 2, 3, ... ,
m
For constant unit failure rates of the units, Eq. (12)
becomes14m
t t t t1 2 3 mR (t) e e e e exp t (13) S ii1
where RS(t) series system reliability at time ti constant
failure rate of unit i for i 1, 2, 3, ... , m
The system hazard rate is given by14
(t) 1 dR (t)S S (t) S 1 F (t) R (t) dtS Sm
(14) ii1
where S(t) series system hazard rate or total failure rateS(t)
series system probability density functionFS(t) series system
cumulative distribution function
It is to be noted that the system total failure rate given by
Eq. (14) is the sum of the failurerates of all of the units. It
means that whenever the failure rates of units are added, it
isautomatically assumed that the units are acting in series (i.e.,
if one unit fails, the systemfails). This is the worst-case
assumption often practiced in the design of engineering
systems.
The system mean time to failure is given by14
m
MTTF R (t) dt exp t dt S S ii1
0 0
1 (15)
m (1 / )i1 iwhere MTTFS is the series system mean time to
failure.
-
3 Common Reliability Networks 1005
1
2
3
mFigure 2 Parallel system block diagram.
3.2 Parallel NetworkThis type of conguration can be used to
improve a mechanical systems reliability duringthe design phase.
The block diagram of an m-unit parallel network is shown in Fig.
2.Each block in the diagram represents a unit. This conguration
assumes that all of its unitsare active and at least one unit must
work normally for the system to succeed. For indepen-dently failing
units, the reliability of the parallel network shown in Fig. 2 is
expressed by14
m
R 1 (1 R ) (16)p ii1
where Rp reliability of parallel networkRi reliability of unit i
for i 1, 2, 3, ... , m
For constant failure rates of the units, Eq. (16) becomes14m
tiR (t) 1 (1 e ) (17)pi1
where Rp(t) parallel network reliability at time ti constant
failure rate of unit i for i 1, 2, 3, ... , m
For identical units, the network mean time to failure is given
by14
m1 1MTTF R (t) dt (18)p p ii1
0
where MTTFp parallel network mean time to failure unit constant
failure rate
3.3 SeriesParallel NetworkThe network block diagram is shown in
Fig. 3. Each block in the diagram represents a unit.This network
represents a system having m number of subsystems in series. In
turn, eachsubsystem contains n number of active units in parallel.
All subsystems must operate nor-mally for the system to succeed.
For independent units, the reliability of the seriesparallelnetwork
shown in Fig. 3 is given by18,19
-
1006 Reliability in the Mechanical Design Process
1
1
2
n
1
2
n
1
2
n
2 m
Figure 3 Seriesparallel network block diagram.
m n
R 1 F (19) Sp iji1 j1
where RSp seriesparallel network reliabilitym number of
subsystemsn number of units
Fij failure probability of the ith subsystems jth unitFor
constant unit failure rates, Eq. (19) becomes18,19
m n
tijR (t) 1 (1 e ) (20) Spi1 j1
where RSp(t) seriesparallel network reliability at time tij
constant failure rate of unit ij
For identical units, the network mean time to failure is given
by19
MTTF R (t) dtSp Sp0
m in1 1mi1 (1) (21) i ji1 j1where MTTFSp seriesparallel network
mean time to failure
unit failure rate
3.4 ParallelSeries NetworkThis network represents a system
having m number of subsystems in parallel. In turn, eachsubsystem
contains n number of units in series. At least one subsystem must
function nor-mally for the system to succeed. The network block
diagram is shown in Fig. 4. Each blockin the diagram represents a
unit. For independent units, the reliability of the
parallelseriesnetwork shown in Fig. 4 is expressed by18,19
m n
R 1 1 R (22) pS iji1 j1
-
3 Common Reliability Networks 1007
1
1
1
1
2
2
2
2
3
3
3
3
n
n
n
n
1
2
3
m
Figure 4 Parallelseries network block diagram.
where RpS parallelseries network reliabilitym number of
subsystemsn number of units
Rij reliability of the jth unit in the ith subsystemFor constant
unit failure rates, Eq. (23) becomes18,19
m n
tijR (t) 1 1 e (23) pSi1 j1
where RpS(t) parallelseries network reliability at time tij
constant failure rate of unit ij
For identical units, the network mean time to failure is given
by19
m1 1MTTF R (t) dt (24)pS pS n ii1
0
where MTTFpS is the parallelseries network mean time to
failure.
3.5 K-out-of-m-Unit NetworkThis network is sometimes referred to
as a partially redundant network. It is a parallelnetwork with a
condition that at least K units out of the total of m units must
operate normallyfor the system to succeed.
For independent and identical units, the network reliability is
given by14,16m
m i miR R (1 R) (25) K /m iiKwhere
m!m i i!(m i)!
-
1008 Reliability in the Mechanical Design Process
0
1
2
m
Figure 5 Block diagram of an (m 1)-unit standby system.
RK /m is the K-out-of-m-unit network reliability, and R is the
unit reliability. For K 1 andK m, Eq. (25) becomes the reliability
expression for parallel and series networks, respec-tively. More
specically, parallel and series networks are the special cases of
the K-out-of-m-unit network.
For constant failure rates of the units, Eq. (25)
becomes14,16m
m it t miR e (1 e ) (26) K /m iiKwhere RK /m K-out-of-m-unit
network reliability at time t
unit constant failure rate
The network mean time to failure is given by14,16
m1 1MTTF R (t) dt (27)K / m k /m iik
0
where MTTFK /m is the K-out-of-m-unit network mean time to
failure.
3.6 Standby SystemThis is another conguration used to improve
reliability. The block diagram of an (m 1)-unit standby system is
shown in Fig. 5. Each block in the diagram represents a unit. In
thisconguration, one unit operates and m units are kept on standby.
As soon as the operatingunit fails, it is replaced by one of the
standbys. The system fails when all of its units fail(i.e.,
operating plus all standbys). For perfect switching, independent
and identical units, andas-good-as-new standby units, the standby
system reliability is given by14,16
m t i t[ (t) dt] exp( (t) dt)0 0R (t) (28)SS i!i0where RSS(t)
standby system reliability at time t
(t) unit hazard rate or time-dependent failure ratem number of
standbys
For constant failure rates of the units [i.e., (t) ] Eq. (28)
becomesm i t( t) e
R (t) (29)SS i!i0
-
3 Common Reliability Networks 1009
1 2
3
4 5Figure 6 Block diagram of a ve-unit bridge network.
where is the unit constant failure rate. The system mean time to
failure is given by14
MTTF R (t) dtSS SS0
m i t(t) e
dt i!i00
m 1 (30)
where MTTFSS is the standby system mean time to failure.
3.7 Bridge NetworkThe block diagram of a bridge network is shown
in Fig. 6. Each block in the diagramrepresents a unit. Mechanical
components sometimes can form this type of conguration.For
independent units, the reliability of the bridge network shown in
Fig. 6 is20
5 4
R 2 R R R R R R R R R bn i i 1 3 5 1 4 2 5i1 i1
5 4 3 5
R R R R R R R R R R (31) i i 5 i 1 i 1 2 4 5i2 i1 i1 i3
where Rbn bridge network reliabilityRi unit i reliability for i
1, 2, 3, 4, 5
For identical units and constant failure rates of the units, Eq.
(31) becomes145t 4t 3t 2tR (t) 2e 5e 2e 2e (32)bn
where is the unit constant failure rate. The network mean time
to failure is given by
49MTTF R (t) dt (33)bn bn 600
where MTTFbn is the bridge network mean time to failure.
-
1010 Reliability in the Mechanical Design Process
4 MECHANICAL FAILURE MODES AND CAUSES OF GENERALAND GEAR
FAILURES
A mechanical failure may be dened as any change in the shape,
size, or material propertiesof a structure, piece of equipment, or
equipment part that renders it unt to carry out itsspecied mission
adequately.13 Thus, there are many different types of failure modes
asso-ciated with mechanical items. Good design practices can reduce
or eliminate altogether theoccurrence of these failure modes. Some
of these failure modes are as follows2123:
Fatigue failure Material aw failure Bending failure
Metallurgical failure Bearing failure Instability failure Shear
loading failure Compressive failure Creep/rupture failure Tensile
yield strength failure Ultimate tensile strength failure Stress
concentration failure
There are many causes of product failures. Some of these are as
follows24:
Defective design Wear-out Defective manufacturing Wrong
application Incorrect installation Failure of other parts
A study performed over a period of 35 years reported a total of
931 gear failures.24They were classied under four categories:
breakage (61.2%), surface fatigue (20.3%), wear(3.2%), and plastic
ow (5.3%). The causes of these failures were grouped into the
vecategories shown in Fig. 7.
These ve categories were further divided into various elements.
The elements of theservice-related classication were continual
overloading (25%), improper assembly (21.2%),impact loading
(13.9%), incorrect lubrication (11%), foreign material (1.4%),
abusive han-dling (1.2%), bearing failure (0.7%), and operator
errors (0.3%). The elements of the heat-treatment-related
classication were incorrect hardening (5.9%), inadequate case
depth(4.8%), inadequate core hardness (2%), excessive case depth
(1.8%), improper tempering(1%), excessive core hardness (0.5%), and
distortion (0.2%). The elements of the design-related classication
were wrong design (2.8%), specication of suitable heat
treatment(2.5%), and incorrect material selection (1.6%). The
elements of the manufacturing-relatedclassication were grinding
burns (0.7%) and tool marks or notches (0.7%). Finally, the
threeelements of the material-related classication were steel
defects (0.5%), mixed steel or in-correct composition (0.2%), and
forging defects (0.1%).
-
5 Reliability-Based Design and Design-by-Reliability Methodology
1011
Heattreatment related (16.2%)
Design related (6.9%)
Materialrelated (0.8%)
Manufacturingrelated (1.4%)
Service related (74.7%)
Classifications
Figure 7 Classications of gear failure causes.
5 RELIABILITY-BASED DESIGN AND DESIGN-BY-RELIABILITY
METHODOLOGYIt would be unwise to expect a system to perform to a
desired level of reliability unless itis specically designed for
that level. Desired system/equipment /part reliability design
spec-ications due to factors such as well-publicized failures
(e.g., the space shuttle Challengerdisaster and the Chernobyl
nuclear accident) have increased the importance of
reliability-based design. The starting point for reliability-based
design is during the writing of the designspecication. In this
phase, all reliability needs and specications are entrenched into
thedesign specication. Examples of these requirements might include
item mean time to failure(MTTF), mean time to repair (MTTR), test
or demonstration procedures to be used, andapplicable
documents.
Over the years the DOD has developed various reliability
documents for use during thedesign and development of an
engineering item. Many times, such documents are entrenchedinto the
item design specication document. Table 1 presents some of these
documents.Many professional bodies and other organizations have
also developed documents on variousaspects of reliability.15,22,25
References 22 and 26 provide descriptions of documents devel-oped
by the DOD.
Reliability is an important consideration during the design
phase. According to Ref. 27,as many as 60% of failures can be
eliminated through design changes. There are manystrategies the
designer could follow to improve design:
Eliminate failure modes. Focus design for fault tolerance. Focus
design for fail safe. Focus design to include mechanism for early
warnings of failure through fault diag-
nosis.
-
1012 Reliability in the Mechanical Design Process
Table 1 Selected Design-Related Documents Developed by the U.S.
Department of Defense
Document No. Document Title
MIL-STD-721 Denitions of Terms for Reliability and
MaintainabilityMIL-STD-217 Reliability Prediction of Electronic
EquipmentMIL-STD-781 Reliability Design, Qualication, and
Production Acceptance Tests: Exponential DistributionMIL-STD-756
Reliability Modeling and PredictionMIL-STD-785 Reliability Program
for Systems and EquipmentMIL-HDBK-251 Reliability /Design Thermal
ApplicationsMIL-STD-1629 Procedures for Performing a Failure Mode,
Effects, and Criticality AnalysisRADC-TR-75-22 Non-Electronic
Reliability NotebookMIL-STD-965 Parts Control ProgramMIL-STD-2074
Failure Classication for Reliability Testing
During the design phase of a product, various types of
reliability and maintainability analysescan be performed, including
reliability evaluation and modeling, reliability allocation,
main-tainability evaluation, human factors / reliability
evaluation, reliability testing, reliabilitygrowth modeling, and
life-cycle cost. In addition, some of the design improvement
strategiesare zero-failure design, fault-tolerant design, built-in
testing, derating, design for damagedetection, modular design,
design for fault isolation, and maintenance-free design.
Duringdesign reviews, reliability and maintainability-related
actions recommended/ taken are to bethoroughly reviewed from
desirable aspects.
A systematic series of steps are taken to design a reliable
mechanical item. The designby methodology is composed of such
steps14,28,29:
Dene the design problem under consideration. Identify and list
all associated design variables and parameters. Perform failure
mode, effect, and criticality analyses according to
MIL-STD-1629.30
Verify critical design parameter selection. Establish
appropriate relationships between the failure-governing criteria
and the crit-
ical parameters. Determine the failure-governing stress and
strength functions and then the most ap-
propriate failure-governing stress and strength distributions.
Estimate the reliability utilizing the failure-governing stress and
strength distributions
for all critical failure modes. Iterate the design until
reliability goals are achieved. Optimize design with respect to
factors such as safety, reliability, cost, performance,
maintainability, weight, and volume. Repeat the design
optimization process for all critical components. Estimate item
reliability. Iterate the design until item reliability goals are
fully satised.
6 DESIGN RELIABILITY ALLOCATION AND EVALUATION METHODSOver the
years, many reliability allocation and evaluation methods have been
developed foruse during the design phase.14,29 This section
presents some that are considered useful, par-ticularly in
designing mechanical items.
-
6 Design Reliability Allocation and Evaluation Methods 1013
6.1 Failure Rate Allocation MethodThis method is used to
allocate failure rates to system components when the overall
systemrequired failure rate is given. The method is based on the
following three assumptions14:
All system components fail independently. Component failure
rates are constant. System components form a series network.
Thus, the system failure rate using Eq. (14) ism
(34)S ii1
where S system failure ratem total number of system componentsi
failure rate of component i for i 1, 2, 3, ... , m
If the specied failure rate of the system is Sp, then the
component failure rate is allocatedsuch that
m
* (35) i Spi1
where is the failure rate allocated to component i for i 1, 2,
3, ... , m. The following*ithree steps are associated with this
approach:
Estimate failure rates of the system components (i.e., i for i
1, 2, 3, ... , m) usingthe eld data.
Calculate the relative weight i of component i using the
preceding step failure ratedata and the expression
i for i 1, 2, 3, ... , m (36)i m i1 iIt is to be noted that i
represents the relative failure vulnerability of component i
and
m
1 (37) ii1
Allocate the failure rate to part or component i by using the
equation
* for i 1, 2, 3, ... , m (38)i i SpA solved example in Ref. 14
demonstrates the application of this method.
6.2 Hybrid Reliability Allocation MethodThis method combines two
reliability allocation methods: similar familiar systems and
factorsof inuence. The method is more attractive because it
incorporates the benets of bothsystems.14,29
The basis for the similar-familiar-systems approach is the
familiarity of the designerwith similar systems as well as the
utilization of failure data collected on similar systemsfrom
various sources during the allocation process. The principal
disadvantage of the similar-
-
1014 Reliability in the Mechanical Design Process
familiar-systems method is the assumption that the reliability
and life-cycle cost of similarsystems are satisfactory or
adequate.
The factors-of-inuence method is based upon four factors that
are considered to affectitem reliability: failure criticality,
environment, complexity / time, and state of the art.
The failure criticality factor is concerned with how critical is
the failure of an item (e.g.,the failure of some auxiliary
equipment in an aircraft may not be as critical as the failureof an
engine). The environmental factor takes into account the
susceptibility of items toconditions such as vibration, humidity,
and temperature.
The complexity / time factor relates to the number of
item/subsystem components /partsand the operating time of the item
under consideration during the total system operatingperiod.
Finally, the state-of-the-art factor relates to the advancement in
the state of the artfor an item under consideration.
During the reliability allocation process, each item is rated
with respect to each of thesefour factors by assigning a number
from 1 to 10. The assignment of 1 means the item isleast affected
by the factor under consideration and 10 means the item is most
affected bythe same factor. Subsequently, reliability is determined
by weighing these numbers for allfour factors.
6.3 Safety Factor and Safety MarginThe safety factor and safety
margin are arbitrary multipliers used to ensure the reliability
ofmechanical items during the design phase. These indexes can
provide satisfactory design ifthey are established using
considerable past experiences and data.
Safety FactorA safety factor can be dened in many different
ways.13,3135 Two commonly used denitionsfollow.
Denition I. The safety factor is dened by36
mShSF 1 (39)mSS
where SF safety factormSh mean failure governing strengthmSS
mean failure governing stress
This index is a good measure of safety when both stress and
strength are normally distributed.However, when the spread of both
strength and/or stress is large, the index becomes mean-ingless
because of positive failure rate.14
Denition II. The safety factor is dened by3237
USSF (40)WSwhere SF safety factor
WS working stress expressed in pounds per square inch (psi)US
ultimate strength expressed in psi
-
6 Design Reliability Allocation and Evaluation Methods 1015
Safety MarginThe safety margin (SM) is dened as14,31
SM SF 1 (41)The negative value of this measure means that the
item under consideration will fail. Thus,its value must always be
greater than zero.
The safety margin for normally distributed stress and strength
is expressed by14,31
as msSM (42)Sh
where as average strengthms maximum stressSh strength standard
deviation
In turn the maximum stress ms is expressed by
C (43)ms SS SSwhere SS mean stress
SS stress standard deviationC factor between 3 and 6
6.4 StressStrength Interference Theory MethodThis method is used
to determine the reliability of a mechanical item when its
associatedstress and strength probability density functions are
known. The item reliability is denedby13,14,29
R P(y x) P(x y) (44)where R item reliability
P probabilityx strength random variabley stress random
variable
Equation (44) is rewritten in the form13,14,29
R (y) (x) dx dy (45) y
where (x) strength probability density function(y) stress
probability density function
Special Case Model: Exponentially Distributed Stress and
StrengthIn this case, an items stress and strength are dened by
y(y) e 0 y (46)and
-
1016 Reliability in the Mechanical Design Processx(x) e 0 x
(47)
where and are the reciprocals of the mean values of stress and
strength, respectively.Using Eqs. (46) and (47) in Eq. (45)
yields
y xR e e dx dy 0 y (48)
For and Eq. (48) becomes1/y, 1 /x,x
R (49)x y
where x mean strengthy mean stress
Similarly, models for other stress and strength probability
distributions can be developed. Anumber of such models are
presented in Ref. 13.
6.5 Failure Modes and Effect Analysis (FMEA)FMEA is a vital tool
for evaluating system design from the point of view of reliability.
Itwas developed in the early 1950s to evaluate the design of
various ight control systems.38
The difference between the FMEA and the failure mode, effect,
and criticality analysis(FMECA) is that FMEA is a qualitative
technique used to evaluate a design, whereasFMECA is composed of
FMEA and criticality analysis (CA). Criticality analysis is a
quan-titative method used to rank critical failure mode effects by
taking into account their occur-rence probabilities.
As FMEA is a widely used method in industry, there are many
standards/documentswritten about it. In Ref. 30 45 such
publications were collected and evaluated, prepared byorganizations
such as the DOD, NASA, and the Institute of Electrical and
Electronic Engi-neers (IEEE). These documents include39:
DOD: MIL-STD-785A (1969), MIL-STD-1629 (draft) (1980),
MIL-STD-2070 (AS)(1977), MIL-STD-1543 (1974), AMCP-706-196
(1976)
NASA: NHB 5300.4 (1A) (1970), ARAC Proj. 79-7 (1976) IEEE: ANSI
N 41.4 (1976)
Details of the above documents as well as a list of publications
on FMEA are given in Ref.24.
The main steps involved in performing FMEA are as follows29:
Dene carefully all system boundaries and detailed requirements.
List all parts /subsystems in the system under consideration.
Identify and describe each part and list all its associated failure
modes. Assign failure rate /probability to each failure mode. List
effects of each failure mode on subsystem/system/plant.
-
7 Human Error and Reliability Consideration in Mechanical Design
1017
Enter remarks for each failure mode. Review each critical
failure mode and take appropriate measures.
This method is described in detail in Ref. 14.
6.6 Fault Tree Analysis (FTA)This method, so called because it
arranges fault events in a tree-shaped diagram, is one ofthe most
widely used techniques for performing system reliability analysis.
In particular, itis probably the most widely used method in the
nuclear power industry. The technique iswell suited for determining
the combined effects of multiple failures.
The fault tree technique is more costly to use than the FMEA
approach. It was developedin the early 1960s at Bell Telephone
Laboratories to evaluate the reliability of the MinutemanLaunch
Control System. Since that time, hundreds of publications on the
method have ap-peared.15
The FTA begins by identifying an undesirable event, called the
top event, associatedwith a system. Fault events that could cause
the occurrence of the top event are generatedand connected by AND
and OR logic gates. The construction of a fault tree proceeds
bygeneration of fault events (by asking the question How could this
event occur?) in asuccessive manner until the fault events need not
be developed further. These events areknown as primary or
elementary events. In simple terms, the fault tree may be described
asthe logic structure relating the top event to the primary events.
This method is described indetail in Ref. 14.
7 HUMAN ERROR AND RELIABILITY CONSIDERATIONIN MECHANICAL
DESIGN
As in the reliability of any other system, human reliability and
error play an important rolein the reliability of mechanical
systems. Over the years many times mechanical systems/equipment
have failed due to human error rather than hardware failure.
Careful considerationof human error and reliability during the
design of mechanical systems can help to eliminateor reduce the
occurrence of non-hardware-related failures during the operation of
such sys-tems. Human errors may be classied under the following
seven distinct categories4042:
Design errors Operator errors Assembly errors Inspection errors
Maintenance errors Installation errors Handling errors
Each of these categories is described in detail in Ref. 14.There
are numerous causes for the occurrence of human error including
poor equipment
design, complex tasks, poor work layout, poorly written
operating and maintenance proce-dures, poor job environment (e.g.,
poor lighting, high/ low temperature, crowded work space,high noise
level), inadequate work tools, poor skill of involved personnel,
and poor moti-vation of involved personnel.4042
-
1018 Reliability in the Mechanical Design Process
Human reliability of time-continuous tasks such as aircraft
maneuvering, scope moni-toring, and missile countdown can be
calculated by using the equation42,43
t
R (t) exp (t) dt (50)h 0
where Rh(t) human reliability at time t(t) time-dependent human
error rate
For constant human error rate [i.e., (t) ], Eq. (50)
becomest
tR (t) exp dt e (51)h 0
where is the constant human error rate. The subject of human
reliability and error isdiscussed in detail in Ref. 42.
Example. A person is performing a certain time-continuous task.
Assume his or her errorrate is 0.004 per hour. Calculate the
persons probability of performing the task correctlyduring a 5-h
period.
By using the values in Eq. (51), we get(0.004)5R (5) e 0.98h
So, there is approximately a 98% chance that the person will
perform the task correctlyduring the specied period.
8 FAILURE RATE ESTIMATION MODELS FOR VARIOUS MECHANICAL
ITEMSMany mathematical models available in the literature can be
used to estimate failure ratesof items such as bearings, pumps,
brakes, lters, compressors, and seals.4446 This sectionpresents
some of these models.
8.1 Brake System Failure Rate Estimation ModelThe brake system
failure rate is expressed by44
6
(52)brs ii1
where brs brake system failure rate, expressed in failures /106
h1 brake housing failure rate2 total failure rate of actuators3
total failure rate of seals4 total failure rate of bearings5 total
failure rate of springs6 total failure rate of brake friction
materials
The values of i for i 1, 2, ... , 6 are obtained through various
means.44,47
-
9 Failure Data and Failure Data Collection Sources 1019
8.2 Compressor System Failure Rate Estimation ModelThe
compressor system failure rate is expressed by45
6
(53)comp ii1
where comp compressor system failure rate, expressed in failures
/106 h1 failure rate of all compressor bearings2 compressor casing
failure rate3 failure rate due to design conguration4 failure rate
of valve assay (if any)5 failure rate of all compressor seals6
failure rate of all compressor shafts
Procedures for calculating 1, 2, 3, 4, 5, and 6 are presented in
Ref. 45.
8.3 Filter Failure Rate Estimation ModelThe lter failure rate is
expressed by46
6
(54)ft b ii1
where ft lter failure rate, expressed in failures /106 hb lter
base failure ratei ith modifying factor; i 1 is for temperature
effects, i 2 for water contam-
ination effects, i 3 for cyclic ow effects, i 4 for differential
pressureeffects, i 5 for cold-start effects, and i 6 for vibration
effects
Procedures for estimating b, 1, 2, 3, 4, 5, and 6 are given in
Ref. 46.
8.4 Pump Failure Rate Estimation ModelThe pump failure rate is
expressed by46
5
(55)pm ii1
where pm pump failure rate, expressed in failures /106 cycles1
pump uid driver failure rate2 pump casing failure rate3 pump shaft
failure rate4 failure rate of all pump seals5 failure rate of all
pump bearings
Procedures for calculating 1, 2, 3, 4, and 5 are presented in
Ref. 46.
9 FAILURE DATA AND FAILURE DATA COLLECTION SOURCESFailure data
provide invaluable information to reliability engineers, design
engineers, man-agement, and so on, concerning the product
performance. These data are the nal proof of
-
1020 Reliability in the Mechanical Design Process
Table 2 Failure Rates for Selected Mechanical Items
Item Description Failure Rate 106 h
Roller bearing 8.323Bellows (general) 13.317Filter (liquid)
6.00Compressor (general) 33.624Pipe 0.2Hair spring 1.0Pump (vacuum)
10.610Gear (spur) 3.152Seal (O-ring) 0.2Nut or bolt 0.02Brake
(electromechanical) 16.00Knob (general) 2.081Washer (lock)
0.586Washer (at) 0.614Duct (general) 2.902Note: Use environment:
ground xed or general.
Table 3 Selected Failure Data Sources for Mechanical Items
Source Developed By
Ref. 48 Reliability Analysis Center, Rome Air DevelopmentCenter,
Grifs Air Force Base, Rome, New York
Component Reliability Data for Use inProbabilistic Safety
Assessment (1998)
International Atomic Energy Agency, Vienna,Austria
R. G. Arno, Non-Electronic Parts ReliabilityData (Rept. No.
NPRD-2, 1981)
Reliability Analysis Center, Rome Air DevelopmentCenter, Grifs
Air Force Base, Rome, New York
Government Industry Data ExchangeProgram (GIDEP)
GIDEP Operations Center, U.S. Dept. of Navy,Seal Beach, Corona,
California.
Ref. 49 Reliability Analysis Center, Rome Air DevelopmentCenter,
Grifs Air Force Base, Rome, New York
the success or failure of the effort expended during the design
and manufacture of a productused under designed conditions. During
the design phase of a product, past informationconcerning its
failures plays a critical role in the reliability analysis of that
product. Failuredata can be used to estimate item failure rate,
perform effective design reviews, predictreliability and
maintainability of redundant systems, conduct trade-off and
life-cycle coststudies, and perform preventive maintenance and
replacement studies. Table 2 presents failurerates for selected
mechanical items.13,48,49
There are many different ways to collect failure data. For
example, during the equipmentlife cycle, there are eight identiable
data sources: repair facility reports, development testingof the
item, previous experience with similar or identical items,
customers failure-reportingsystems, inspection records generated by
quality control and manufacturing groups, testsconducted during eld
demonstration, environmental qualication approval, and eld
instal-lation, acceptance testing, and warranty claims.50 Table 3
presents some sources for collectingfailure data for use during the
design phase.13
-
References 1021
REFERENCES1. W. J. Layman, Fundamental Considerations in
Preparing a Master System Plan, Electrical World,
101, 778792 (1933).2. S. A. Smith, Spare Capacity Fixed by
Probabilities of Outage, Electrical World, 103, 222225
(1934).3. S. A. Smith, Probability Theory and Spare Equipment,
Edison Electric Inst. Bull., Mar. 1934, pp.
310314.4. S. A. Smith, Service Reliability Measured by
Probabilities of Outage, Electrical World, 103, 371
374 (1934).5. P. E. Benner, The Use of the Theory of Probability
to Determine Spare Capacity, General Electric
Rev., 37, 345348 (1934).6. S. M. Dean, Considerations Involved
in Making System Investments for Improved Service Relia-
bility, Edison Electric Inst. Bull., 6, 491496 (1938).7. B. S.
Dhillon, Power System Reliability, Safety, and Management, Ann
Arbor Science, Ann Arbor,
MI, 1983.8. A. Coppola, Reliability Engineering of Electronic
Equipment: A Historical Perspective, IEEE
Trans. Reliabil., 33, 2935 (1984).9. W. Weibull, A Statistical
Distribution Function of Wide Applicability, J. Appl. Mech., 18,
293
297 (1951).10. A. M. Freudenthal, and E. J. Gumbel, Failure and
Survival in Fatigue, J. Appl. Phys., 25, 110
120 (1954).11. A. M. Freudenthal, Safety and Probability of
Structural Failure, Trans. Am. Soc. Civil Eng., 121,
13371397 (1956).12. W. M. Redler, Mechanical Reliability
Research in the National Aeronautics and Space Adminis-
tration, Proceedings of the Reliability and Maintainability
Conference, 1966, pp. 763768.13. B. S. Dhillon, Mechanical
Reliability: Theory, Models, and Applications, American Institute
of
Aeronautics and Astronautics, Washington, DC, 1988.14. B. S.
Dhillon, Design Reliability: Fundamentals and Applications, CRC,
Boca Raton, FL, 1999.15. B. S. Dhillon, Reliability and Quality
Control: Bibliography on General and Specialized Areas, Beta
Publishers, Gloucester, Ontario, Canada, 1992.16. P. Kales,
Reliability: For Technology, Engineering, and Management,
Prentice-Hall, Upper Saddle
River, NJ, 1998.17. B. S. Dhillon, A Hazard Rate Model, IEEE
Trans. Reliabil., 29, 150151 (1979).18. B. S. Dhillon, Systems
Reliability, Maintainability, and Management, Petrocelli Books, New
York,
1983.19. B. S. Dhillon, Reliability, Quality, and Safety for
Engineers, CRC, Boca Raton, FL, 2005.20. J. P. Lipp, Topology of
Switching Elements Versus Reliability, Trans. IRE Reliabil. Quality
Con-
trol, 7, 2134 (1957).21. J. A. Collins, Failure of Materials in
Mechanical Design, Wiley, New York, 1981.22. W. Grant Ireson, C. F.
Coombs, and R. Y. Moss (eds.), Handbook of Reliability Engineering
and
Management, McGraw-Hill, New York, 1996.23. R. L. Doyle,
Mechanical System Reliability, Tutorial Notes, Annual Reliability
and Maintaina-
bility Symposium, Las Vegas, NV, 1992.24. C. Lipson, Analysis
and Prevention of Mechanical Failures, Course Notes No. 8007,
University
of Michigan, Ann Arbor, MI, June 1980.25. S. S. Rao,
Reliability-Based Design, McGraw-Hill, New York, 1992.26. J. W.
Wilbur, and N. B. Fuqua, A Primer for DOD Reliability,
Maintainability, and Safety, Standards
Document No. PRIM 1, Rome Air Development Center, Grifss Air
Force Base, Rome, NY, 1988.27. D. G. Raheja, Assurances
Technologies, McGraw-Hill, New York, 1991.28. D. Kececioglu,
Reliability Analysis of Mechanical Components and Systems, Nucl.
Eng. Design,
19, 259290 (1972).29. B. S. Dhillon, and C. Singh, Engineering
Reliability: New Techniques and Applications, Wiley, New
York, 1981.
-
1022 Reliability in the Mechanical Design Process
30. Procedures for Performing Failure Mode, Effects, and
Criticality Analysis, MIL-STD-1629, Depart-ment of Defense,
Washington, DC, 1980.
31. D. Kececioglu, and E. B. Haugen, A Unied Look at Design
Safety Factors, Safety Margin, andMeasures of Reliability,
Proceedings of the Annual Reliability and Maintainability
Conference,1968, pp. 522530.
32. G. M. Howell, Factors of Safety, Machine Design, July 12,
1956, pp. 7681.33. R. B. McCalley, Nomogram for Selection of Safety
Factors, Design News, Sept. 1957, pp. 138
141.34. R. Schoof, How Much Safety Factor? Allis-Chalmers Elec.
Rev., 1960, pp. 2124.35. J. E. Shigley, and L. D. Mitchell,
Mechanical Engineering Design, McGraw-Hill, New York, 1983,
pp. 610611.36. J. H. Bompass-Smith, Mechanical Survival: The Use
of Reliability Data, McGraw-Hill, London,
1973.37. V. M. Faires, Design of Machine Elements, Macmillan,
New York, 1955.38. J. S. Countinho, Failure Effect Analysis, Trans.
N.Y. Acad. Sci., 26, 564584 (1964).39. B. S. Dhillon, Failure Modes
and Effects Analysis: Bibliography, Microelectron. Reliabil.,
32,
719732 (1992).40. D. Meister, The Problem of Human-Initiated
Failures, Proceedings of the Eighth National Sym-
posium on Reliability and Quality Control, 1962, pp. 234239.41.
J. I. Cooper, Human-Initiated Failures and Man-Function Reporting,
IRE Trans. Human Factors,
10, 104109 (1961).42. B. S. Dhillon, Human Reliability: With
Human Factors, Pergamon, New York, 1986.43. T. L. Regulinski, and
W. B. Askern, Mathematical Modeling of Human Performance
Reliability,
in Proceedings of the Annual Symposium on Reliability, 1969, pp.
511.44. S. Rhodes, J. J. Nelson, J. D. Raze, and M. Bradley,
Reliability Models for Mechanical Equipment,
Proceedings of the Annual Reliability and Maintainability
Symposium, 1988, pp. 127131.45. J. D. Raze, J. J. Nelson, D. J.
Simard, and M. Bradley, Reliability Models for Mechanical
Equip-
ment, Proceedings of the Annual Reliability and Maintainability
Symposium, 1987, pp. 130134.46. J. J. Nelson, J. D. Raze, J.
Bowman, G. Perkins, and A. Wannamaker, Reliability Models for
Mechanical Equipment, Proceedings of the Annual Reliability and
Maintainability Symposium,1989, pp. 146153.
47. T. D. Boone, Reliability Prediction Analysis for Mechanical
Brake Systems, NAVAIR/SYSCOMReport, Department of Navy, Department
of Defense, Washington, DC, Aug. 1981.
48. M. J. Rossi, Non-Electronic Parts Reliability Data, Report
No. NRPD-3, Reliability Analysis Center,Rome Air Development
Center, Grifss Air Force Base, NY, 1985.
49. R. E. Schafer, J. E. Angus, J. M. Finkelstein, M. Yerasi,
and D. W. Fulton, RADC Non-electronicReliability Notebook, Report
No. RADC-TR-85-194, Reliability Analysis Center, Rome Air
Devel-opment Center, Grifss Air Force Base, NY, 1985.
50. B. S. Dhillon, and H. C. Viswanath, Bibliography of
Literature on Failure Data, Microelectron.Reliabil., 30, 723750
(1990).
BIBLIOGRAPHYBompas-Smith, J. H., Mechanical Survival,
McGraw-Hill, London, 1973.Carter, A. D. S., Mechanical Reliability,
Macmillan Education, London, 1986.Carter, A. D. S., Mechanical
Reliability and Design, Wiley, New York, 1997.Dhillon, B. S., Robot
Reliability and Safety, Springer-Verlag, New York, 1991.Frankel, E.
G., Systems Reliability and Risk Analysis, Martinus Nijhoff, The
Hague, 1984.Haugen, E. B., Probabilistic Mechanical Design, Wiley,
New York, 1980.Kapur, K. C., and L. R. Lamberson, Reliability in
Engineering Design, Wiley, New York, 1977.Kivenson, G., Durability
and Reliability in Engineering Design, Hayden, New York,
1971.Little, A., Reliability of Shell Buckling Predictions, MIT
Press, Cambridge, MA, 1964.
-
Bibliography 1023
Little, R. E., Mechanical Reliability Improvement: Probability
and Statistics for Experimental Testing,Marcel Dekker, New York,
2003.
Mechanical Reliability Concepts, American Society of Mechanical
Engineers, New York, 1965.Middendorf, W. H., Design of Devices and
Systems, Marcel Dekker, New York, 1990.Milestone, W. D. (ed.),
Reliability, Stress Analysis and Failure Prevention Methods in
Mechanical Design,
American Society of Mechanical Engineers, New York,
1980.Shooman, M. L., Probabilistic Reliability: An Engineering
Approach, Krieger, Melbourne, FL, 1990.Siddell, J. N.,
Probabilistic Engineering Design, Marcel Dekker, New York,
1983.
Additional publications on mechanical design reliability may be
found in Refs. 13 and 15.