Happiness is taking the

reliability final exam.

Reliability Engineering I
ENM/MSC 565
Review for the Final Exam
 Vital Statistics
• What – R&M concepts covered in the course
• When –
– Monday April 29 from 4:30 – 6:00 pm
• How
– open book
– 50 questions (true-false and multiple choice)
• Distant Learning - Register on-line

Reliability engineering students

enjoying another challenging exam.
 Chap 2.1 - 2.6
• the reliability function, f(t), F(t)
• MTTF, median, mode, std dev
• design life
• hazard rate function, λ(t)
– bathtub curve
– Cumulative and average failure rate
• conditional reliability
– residual MTTF
 Chap 3.1, 3.2, 3.4, 3.5, 3.6
• CFR model
– memoryless property
• failure modes
• two-parameter exponential
• Poisson process
– gamma (Erlang) distribution
• two-component CFR redundant system
 Chap 4.1, 4.2 , 4.3, 4.4
• Weibull
– properties
• mean, median, mode, R(t|T), etc.
– failure modes
– 3-parameter
– 2 component redundant
• Minimum Extreme Value
• Normal
– properties
– central limit theorem
• Lognormal
– Relationship to the normal
– properties
The discriminating student will
• Gamma distribution recognize f(t), F(t), R(t), and λ(t)
for each of these distributions.
 Chap 5.1 - 5.3
• serial configuration
• parallel configuration
• combined series-parallel
– high-level vs. low level
– k out-of-n redundancy
• reliability block diagrams
• complex configurations A reliable block diagram
– decomposition
– Enumeration
 Chap 6.1 - 6.4
• Markov analysis
• load-sharing system
• standby systems
– identical standby units
– standby with switching failures load-sharing

• Degraded systems
 Chap 7.1 - 7.4
• covariate models
• static models
– 3 cases - random stress and strength
– Theoretical distributions
• dynamic models A static model

– periodic loads
– random loads
• physics-of-failure models
 Chap 8.1 - 8.5
• reliability design
• reliability specifications
– system effectiveness
– life-cycle costs
• reliability allocation
– ARINC method
– AGREE method
• design methods
– parts & material selection
– derating
– stress-strength analysis
– redundancy
• Failure mode and effects analysis (FMEA)
• Fault Tree Analysis (FTA)
 Chap 9.1 - 9.6, omit 9.3.3
• repair distribution
• stochastic point processes
– renewal process
– minimal repair
• system MTTR
Pointing at a stochastic process
• reliability under PM
• state-dependent systems with repair
– active redundant
– standby system
 Chap 10.1 - 10.3
• Maintainability design
– specifications
– maintenance concepts (inherent and secondary)
– life cycle cost models
• Design methods
– fault isolation
– standardization & interchangeability
– modularization
– accessibility
– repair vs. replace ability to enter
– replacement model change
– PM model
• Human Factors
 Chap 11.1 - 11.4
• availability
– point
– interval
• steady-state availability An available point
– inherent
– achieved
– operational
• exponential model
• system availability
– steady-state rate diagrams
• inspect and repair model
 Chap 12.1 - 12.2
• empirical analysis of complete ungrouped data
– plotting positions
• empirical analysis of complete grouped data
• empirical analysis of censored ungrouped data
– singly vs. multiply censoring
• type I vs. type II testing
– product limit estimator
– Kaplan-Meier
– rank adjustment

An uncensored group
 Chap 13.1 - 13.4, 13.6
• reliability life testing
• test time calculations
– length of test
– number failures Reliability Test
• burn-in test
– specification model
– cost model
• accelerated life testing
– accelerated cycles
– constant stress models Burn’n test
– Arrhenius and Eyring model
– degradation models
– cumulative damage models
 Chap 14.1, 14.3, 14.4

• reliability growth process

• Duane model
• AMSAA model

A reliable growth process

 Chap 15.1 - 15.3
• Identifying candidate distributions
– use of descriptive stats
– probability plots and least-squares
• parameter estimation
– maximum likelihood estimation

Probably a lily plot

 Chap 16.1 - 16.6
• Chi-square goodness-of-fit test
• Bartlett’s test for exponential
• Mann’s test for Weibull
• K-S test for normal and lognormal
• Trend and GOF for power law process
model
A model of
the powerful
process of
the law
Practicing with the Practice Final Exam

Students resting before the

big reliability test.

I heard that
Recall that a the reliability
renewal process test is easy.
consists of…
 Practice Exam Q1

Two wide bulls in parallel having identical shape parameters

• A system composed
• Ans. False, the
of two Weibulls in
parallel will also be redundant system will
Weibull only if both not be Weibull
components have regardless of the
identical shape parameters of the two
parameters. True or Weibull components.
False?
 Practice Exam Q2
The Pr{T > T0 + t | T >T0} is
equal to which one of the
following expressions:

a. R(t) / R(T0)
b. R(T0) / R(t+T0)
c. R(t) / R(t +T0)
d. R(t+T0) / R(t) • Ans. e. (equation
e. R(t+T0) / R(T0) 2.17)
f. None of the above • Pr{T > T0 + t | T >T0}
= R(t|T0)
g. =
 Practice Exam Q3
Which one of the following
relationships is incorrect? • Ans. c.

a. F(t) = 1 – R(t) dF (t )
f (t ) =
dt
⎡ t ⎤
b. : R (t ) = exp ⎢ − ∫ λ ( x)dx ⎥
⎣ 0 ⎦
− dF (t )
c. : f (t ) =
dt
f (t )
d. : λ (t ) =
1 − F (t ) f.
e. They are all correct
 Practice Exam Q4
• The hazard rate function for • Ans. True, see
the normal distribution is fig. 4.3(c)
always an increasing
function. True or False?

An increasing hazard rate

 Practice Exam Q5
• In a series configuration, • Ans. False,
the system reliability will see equation
always be equal to or greater 5.1
than the least reliable
component. True or False?
 Practice Exam Q6
Which of the following is not • Ans. c.
considered a reliability design
method: Accessibility is a
maintainability
a. Parts selection design method
b. Derating Don Dave Dee
c. Accessibility
d. Choice of Technology
e. Use of redundancy
f. None of the above
Dee rating
 Practice Exam Q7
• A renewal process is a
stochastic point process in which • Ans. True,
the random variables see Section
representing the time between 9.3.1
failures are independent and
identically distributed. True or
False?
I am a
failure.

time between failures

 Practice Exam Q8
A system composed of two active • Ans. d. see
redundant components each
having an availability equal to equation
“A” has an availability equal to: 11.16

a. A2
b. 1-A2
c. 1-(1-A2)
d. 1-(1-A)2
e. None of the above
Two parallel and active
available components
 Practice Exam Q9
• Which of the following
transformations would be used in a
• Ans. d. see
least-squares fit of the lognormal equation 15.6
distribution:

a. xi = ti and yi = ln ln [ 1/(1-F(ti)]
b. xi = ln ti and yi = ln ln [ 1/(1-F(ti)]
c. xi = ti and yi = Φ-1[F(ti)]
d. xi = ln ti and yi = Φ-1[F(ti)]
e. None of the above Distributing normal logs
 Practice Exam Q10
• Since the objective in • Ans. False, the
goodness-of-fit testing is to objective is not
reject the null hypothesis, the to reject the null
level of significance should be hypothesis.
kept small in order to control
the probability of making a
Type I error. True or False?

A goodness-of-fit test
 Practice Exam Q11
Maximum Likelihood • Ans. False,
Estimates are preferred to MLE’s may be
least-square estimates biased.
because they are always
unbiased. True or False?

Very likely a
Maximum hood
 Practice Exam Q12
• The Chi-square goodness- • Ans. True, the test is
of-fit test cannot be used not designed for
when multiply censored multiply censored
data is present. True or data. There is no
False? way to determine
which “failure time”
cells to place the
45¢ x 60¢ = 2700 ¢ censored units.
multiply cents or red data
 Practice Exam Q13

• The parameters of the • Ans. True, the

AMSAA model are Duane model is an
estimated using maximum empirical model
likelihood estimators while while the AMSAA
the parameters of the Duane model is a
model are estimated using theoretical model.
least-squares. True or
False?
a least square
 Practice Exam Q14
Life tables are used to estimate • Ans. b. see
the reliability of: Section 12.2.4

a. Ungrouped complete data

b. Grouped censored data
c. Ungrouped censored data
d. A minimal repair
component
e. Linked networks
Life around the table
 Practice Exam Q15
• The repair versus discard • Ans. True, see
model compares the cost of Section 10.2.4
the two alternatives based
upon an estimate of the
number of failures over the
life of the component. True
or False?

versus
 Practice Exam Q16
The primary objective of
burn-in testing is to identify
the failure modes of
production units in order to
improve product design.
True or False?
burnin ‘n testing
Answer: False, the primary
objective is to increase the mean
residual time.
 Practice Exam Q17
Which of the following would not be considered a
step in conducting a FMECA?
a. Product or process definition
b. Allocating system reliability to components
c. Identification of failure modes
d. Assessment of the effect
The critical effect
e. Classification of severity
of a failure mode
f. They are all steps in a FMECA

Answer: b. Allocating system reliability to components is part of the

reliability design process but not part of a FMECA.
 Practice Exam Q18
Let Ei = the event, the ith component fails where Ec is the
complementary event. Then the event “at least one
component from among 4 components fails” can be
expressed as:
answer: e. The intersection of the
a. E1 ∩ E2 ∩ E3 ∩ E4 complementary events is the event
that none fail. The complement to
b. E1c ∪ E2c ∪ E3c ∪ E4c none failing is at least one failing.
c. 1 − E1 ∩ E2 ∩ E3 ∩ E4
d . E1c ∩ E2c ∩ E3c ∩ E4c
e. ( E ∩ E ∩ E ∩ E
c
1
c
2
c
3
c c
4 )
the intersection of failure events
 Practice Exam Q19
If stress is a constant s and strength is a random variable Y
having PDF, fy(y), then the static reliability, R, can be
expressed as:
s answer: e. one minus the probability that
a. R = ∫ f y ( y )dy the strength is less than the stress level gives
0 the reliability.
∞ Give me the
b. R = ∫f
0
y ( y )dy strength.

∞
c. R = 1 −
∫f
s
y ( y )dy

y( )
d. R = F s where F ( y ) is the CDF Strength less than stress
y

e. R = 1 − Fy ( s ) where Fy ( y ) is the CDF

 Practice Exam Q20
In goodness-of-fit testing, a more powerful test will reject an
incorrect probability distribution with greater probability
than a less powerful test. True or False?

answer: True, the power of a test is its probability to correctly

reject the null hypothesis.

Goodness, a powerful test

giving student’s a fit.
 Bonus Question #1
• Which one of the following is not a measure of
maintainability
a) MTTR
b) H(tp) ≤ p where H(t) is the repair time distribution
c) MTR = MTTR + SDT +MDT
d) Rm(t) , the reliability under Preventive Maintenance
e) Maintenance hours per operating hour
f) Maintenance cost per operating hour
 Bonus Question #2
A low level redundant system introduces
additional failure modes over a high level
redundant system and therefore has a lower
reliability. True or False?
 Bonus Question #3
• Which of the following is a valid reason for fitting a
theoretical distribution to a sample of failure times rather
than fitting an empirical distribution:
a) Management says all failures are Weibull
b) The empirical distribution did not give the desired results
c) Failures must always obey one of theoretical distributions
d) The sample passed a goodness-of-it test for the theoretical
distribution
e) None of the above
 Bonus Question #4
• For a k out n redundant system with identical component
reliabilities R, the probability of a system failure is given by:
a. n ⎛ n ⎞ i
∑ ⎜ i ⎟ R (1 − R)
n −i

i =k ⎝ ⎠

b. k
⎛n⎞ i
∑ ⎜ ⎟
i =0 ⎝ i ⎠
R (1 − R ) n −i

c. 1 − ∑ ⎛ n ⎞ Ri (1 − R)n−i
k

⎜ ⎟
i =0 ⎝i ⎠
n
⎛n⎞
d. 1 − ∑ ⎜ ⎟ R i (1 − R) n −i
i =k ⎝ i ⎠
 Bonus Question #5
• Which one of the following is not a property of maximum
likelihood estimators (MLE)?
a) Consistent
b) Invariant
c) Best asymptotically normal
d) Unbiased
e) Able to accommodate censored data
 The Last Bonus Question
Derive a new reliability model which
will insure failure free operation
of any component over its economic
life. Prove mathematically that it is
superior to anything available today.
Show that your model will handle all
types of failure modes and can be
easily modified to handle reliability
under repair, preventive maintenance,
standby and load-sharing conditions.
Time’s up!
Name your model after your instructor
who made it all possible.
 Are You Ready for The Exam?
The MLE of the beta value
clearly showed a DFR.
Only 5 minutes Therefore, the prince and the
left and I am on princess…
question 12. I
wish I had
studied more. This student
studied too long.
Grandpa, can you tell me the story again
about your days as a reliability engineer
when you maximized the Corporation’s
profits using covariate models,
stochastic point processes, and
reliability growth tests based upon
maximum likelihood estimates?