SUPPLEMENT TO

CHAPTER 4
Reliability

SUPPLEMENT OUTLINE Solved Problems Learning

Defining Reliability and Stating Discussion and Review Objectives
Two Ways of Using It Questions
Internet Exercises After completing this supple-
Finding Probability of Functioning ment, you should be able to:
When Activated Problems
Finding Probability of Functioning Mini-Case: Engineer Tank LO1 Define reliability and state
for a Given Length of Time Mini-Case: Sonar System two ways of using it.
Key Terms LO2 Find probability of
functioning when
activated, and explain the
purpose of redundancy in
a system.
LO3 Find probability of
functioning for a given
length of time, and define
failure rate per hour, mean
time to failure, and
availability.
4S-2 PART 3 System Design

LO1 Defining Reliability and Stating Two Ways of Using It

reliability The ability of a Reliability is a measure of the ability of a product or part to perform its intended function under
product or part to perform its a prescribed set of conditions. In effect, reliability is a probability.
intended function under a Suppose that an item has a reliability of .90. This means that it has a 90 percent probability of
prescribed set of conditions. functioning as intended. The probability that it will fail, i.e., its failure rate, is 1 - .90 = .10, or 10
percent. Hence, it is expected that, on average, 1 out of every 10 such items will fail or, equiva-
lently, that the item will fail, on average, once in every 10 trials. Similarly, a reliability of .985
implies 15 failures per 1,000 parts or trials.
Reliability of a product or part is used in two ways.
1. Reliability when activated.
2. Reliability for a given length of time.
The first of these focuses on one point in time and is often used when a product or part must
operate for one time, such as a missile or an air bag in a car. The second of these focuses on the
length of service, such as most other products e.g., a car. The distinction will become more apparent
as each of these approaches is described in more detail.
Reliability is an important dimension of product quality. Reliability management involves
establishing, achieving, and maintaining reliability objectives for products, e.g., the expected life of
a particular make of light bulb may be specified to be 5,000 hours. Achieving reliability usually
falls on the shoulder of reliability engineers who use a variety of techniques to build reliability
into products (e.g., by using reliable key components), test their performance, and estimate their
reliability. If the reliability is inadequate, the types of failure and their effect on the product should
be determined, their root cause(s) identified, and potential failure prevented.
We will mainly focus on reliability measurement, which involves statistics and probability
theory. The average reliability of a part is measured by testing several units over time until some
or all fail. However, this time may be very long (several years). To accelerate this, the items are
stressed by using extreme environmental conditions such as high temperature, temperature
cycles (e.g., hot–cold), high humidity, high vibration, high voltage, surges in power, etc. The result-
ing life estimate is then adjusted appropriately. Reliability of a product is determined from the
reliability of its parts.

LO2 Finding Probability of Functioning When Activated

The probability that a part or product will operate as planned is an important concept in product
design. Determining that probability when the product consists of a number of independent com-
ponents requires the use of rules of probability for independent events. Independent events have
no relation to the occurrence or nonoccurrence of each other. What follows are examples illustrat-
ing the use of two probability rules to determine whether a given product will operate successfully.
Let Pi = probability that event i occurs, i = 1, 2, 3, . . .
Rule 1. If two or more events are independent and “success” is defined as the occurrence of all of
the events, then the probability of success Ps is equal to the product of the probabilities of the
events occurring, i.e., Ps = P1 × P2 × . . .
Example. Suppose a room has two lamps, but to have adequate light, both lamps must work
(success) when turned on. Here the product is the lighting system that has two component lamps.
One lamp has a probability of working of .90, and the other has a probability of working of .80. The
probability that both will work is .90 × .80 = .72.
This lighting system can be represented by the following diagram where the two components
are connected in series:

Lamp 1 Lamp 2

.90 .80
 SUPPLEMENT TO CHAPTER 4 Reliability 4S-3

CR 0.9 CR 0.98 CR 0.99 Figure 4S-1

1.0 Relating product and component
.9 reliabilities.
.8
Product reliability

.7
.6
.5
.4
.3
.2
.1
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Number of components in product

Even though the individual components of a series system (product) might have high reliabil-
ity, the series system (product) as a whole can have considerably less reliability because all its
components must function (i.e., the system is dependent on each of its components). As the num-
ber of components in a series system (product) increases, the system (product) reliability decreases.
For example, a series system (product) that has eight components, each with a reliability of .99, has
a reliability of only .998 = .923. See Figure 4S-1 for plots of product reliability as a function of num-
ber of its components for selected component reliability, CR.
Many products have a large number of component parts that must all operate, and some way to
increase overall reliability is needed. One approach is to overdesign, i.e., enhance the design to
avoid a particular type of failure. For example, use a more durable and higher quality (but more
expensive) material in a product. Another is design simplification, i.e., reduce the number of com-
ponents in the product. The third approach is to use redundancy in the design. This involves redundancy The use of
providing backup components. backup components to
increase reliability.
Rule 2. If two or more events are independent and “success” is defined as occurrence of at least
one of the events, then the probability of success Ps is equal to 1 - probability that none of the
events will occur, i.e., 1 - (1 - P1)(1 - P2)(1 - P3). . . . Simplifying, Ps = P1 + (1 - P1) P2 + (1 - P1)
(1 - P2) P3 + · · ·
Example. There are two lamps in a room. When turned on, one has probability of working of .90
and the other has probability of working of .80. Only a single lamp is needed to light the room for
success (note that the threshold for success is different in this example). Then, probability of suc-
cess Ps = 1 - (1 - .90)(1 - .80) = .98.
Conceptually, we can think of this system as a lamp with a backup. If the first lamp fails to light
when turned on, the backup lamp is turned on. The probability of success Ps is probability that
the first lamp operates plus probability that the first lamp fails and the backup lamp operates, i.e.,
.90 + (1 - .90) × .80 = .98.
This backup system can be represented by the following diagram.

.80
Lamp 2 (backup)

.90
Lamp 1

Example. Three lamps have probabilities of .90, .80, and .70 of lighting when turned on. Only one
lighted lamp is needed for success. Then, probability of success Ps = 1 - (1 - .90)(1 - .80)(1 -
.70) = .994.
4S-4 PART 3 System Design

Conceptually, we can think of this system as a lamp with a backup which in turn has a
backup. If the first lamp fails to light when turned on, the second lamp is turned on. If the sec-
ond lamp also fails to light when turned on, the third lamp is turned on. The probability of suc-
cess Ps is probability that the first lamp operates plus probability that the first lamp fails and the
second lamp operates plus probability that the first and second lamps fail and the third lamp
operates, i.e.:
3#1 operates4 + 3#1 fails and #2 operates4 + 3#1 fails and #2 fails and #3 operates4
.90 + 11 − .902 × .80 + 3 11 − .902 × 11 − .802 × .70 4 = .994
This double backup system can be represented by the following diagram:

.70
Lamp 3 (backup for Lamp 2)

.80
Lamp 2 (backup for Lamp 1)

.90
Lamp 1

In general, a product (system) may be composed of some parallel components and some series
components. The product’s reliability is calculated in two stages: (1) first calculate the reliability of
the parallel component(s) and then (2) use these to calculate the reliability of the resulting series
system.

Example 1 Determine the reliability of the system shown below.

.90 .92

.98 .90 .95

SOLUTION (a) The system can be reduced to a series of three components:

.98 .90 ⫹ .90(1 ⫺ .90) .95 ⫹ (1 – .95).92

(b) The system reliability is, then, the product of these component reliabilities:
.98 × .99 × .996 = .966

LO3 Finding Probability of Functioning for a Given Length of Time

The second way of looking at reliability considers a use factor, usually the time dimension: prob-
abilities are determined relative to a specified length of time. This approach is most common.
Product warranties, e.g., one year free repair, should be based on this definition of reliability.
In this case, failure rate per hour is defined as the number of failures divided by total operating
hours.
 SUPPLEMENT TO CHAPTER 4 Reliability 4S-5

Two hundred units of a particular component were subjected to accelerated life testing equivalent Example 2
to 2,500 hours of normal use. One unit failed after 1,000 hours and another after 2,000 hours. All
other units were still working at the conclusion of the test. What is the failure rate per hour?

The failure rate per hour = 2/[198 (2,500) + 1,000 + 2,000] = 0.000004016 per hour SOLUTION
Note that this formula assumes constant failure rate over time.

If failure rate is constant over time, time-quantity transposition is applicable, i.e., one can reduce
the test time but use more test items. For example, in the above example, instead of 200 items
tested for 2,500 hours, the reliability engineer could have used 400 items tested for 1,250 hours.
A typical profile of failure rate over time is shown in Figure 4S-2. Because of its shape, it is
referred to as the bathtub curve. Usually, a number of products or parts fail shortly after they are put
into service, because they are defective to begin with. Examples include electronics components
such as capacitors. The rate of failure decreases rapidly as the defective items are weeded out. Dur-
ing the second phase, random failures occur. In many cases, this phase covers a relatively long
period of time (several years). In the third phase, failures occur because the items are worn out, and
the failure rate increases.

Figure 4S-2
Failure rate is generally a function
of time and follows the bathtub
Failure rate

curve.

0
Steady state (random)
Burn-in Wear-out
failures
Time, T

The following example illustrates the bathtub curve. In a reliability testing study, 1,000 light
bulbs were lighted until they failed. Each failure time was recorded. The number of light bulbs
remaining (survivors) over time can be seen in Figure 4S-3. Note that initially and close to the end,
there are sharp drops in the number of light bulbs remaining (survivors), reflecting the burn-in and
wear-out phases, respectively.

1,000 Figure 4S-3

Number of light bulbs remaining
Number remaining

over time.

0 Time

The inverse of failure rate per hour is mean time to failure (MTTF), the average length of mean time to failure (MTTF)
time (in hours) before failure. For data in Example 2, The average length of time
before failure of a product or
MTTF = 1/Failure rate per hour = 1/0.000004016 = 249,000 hours
component.
Note that this formula assumes that failure rate is constant.
4S-6 PART 3 System Design

mean time between failures For repairable items, a similar term, mean time between failures (MTBF), is usually used.
(MTBF) The average time MTBF is the average time from the uptime after the repair following a failure to the next failure.
from the uptime after the
repair following a failure to
the next failure. uptime (after repair) failure

Up

between failures
Down

The time to failure of a nonrepairable item during the steady-state phase can often be modelled
by the Exponential distribution with an average equal to the MTTF (see Figure 4S-4). Similar
results hold for repairable items. The probability that the item put into service at time 0 will fail
before some specified time, T, is equal to the area under the curve between 0 and T. Reliability of
the item is the probability that it will last at least until time T; therefore, reliability is equal to the
area under the curve beyond T. (Note that the total area under the curve is 100 percent.) Observe
that as the specified length of service increases, the area under the curve to the right of that point
(i.e., the reliability of the item) decreases.

Figure 4S-4 f (T)

An Exponential distribution.

⫺T/MTTF
Reliability = e
1 ⫺ e ⫺T /MTTF

0
T
Time

The Exponential distribution is completely described using a single parameter, its average, in
this case the mean time to failure (or between failures). Using the symbol T to represent length of
service, the reliability or probability that failure will not occur before time T (i.e., the area in the
right tail) is easily determined by:
Reliability = P1no failure before T 2 = e −T/MTTF
where
e ≈ 2.7183
T = Length of service before failure
MTTF = Mean time to failure
The probability that failure will occur before time T is 1 minus reliability:
P1failure before T 2 = 1 − e −T/MTTF
Selected values of e–T/MTTF (i.e., reliability), given values for T/MTTF are listed in Table 4S-1.
 SUPPLEMENT TO CHAPTER 4 Reliability 4S-7

T/MTTF e⁻T/MTTF T/MTTF e⁻T/MTTF T/MTTF e⁻T/MTTF Table 4S-1

.10 .9048 2.60 .0743 5.10 .0061 Values of reliability = e–T/MTTF
.20 .8187 2.70 .0672 5.20 .0055
.30 .7408 2.80 .0608 5.30 .0050
.40 .6703 2.90 .0550 5.40 .0045
.50 .6065 3.00 .0498 5.50 .0041
.60 .5488 3.10 .0450 5.60 .0037
.70 .4966 3.20 .0408 5.70 .0033
.80 .4493 3.30 .0369 5.80 .0030
.90 .4066 3.40 .0334 5.90 .0027
1.00 .3679 3.50 .0302 6.00 .0025
1.10 .3329 3.60 .0273 6.10 .0022
1.20 .3012 3.70 .0247 6.20 .0020
1.30 .2725 3.80 .0224 6.30 .0018
1.50 .2231 4.00 .0183 6.50 .0015
1.60 .2019 4.10 .0166 6.60 .0014
1.70 .1827 4.20 .0150 6.70 .0012
1.80 .1653 4.30 .0136 6.80 .0011
1.90 .1496 4.40 .0123 6.90 .0010
2.00 .1353 4.50 .0111 7.00 .0009
2.10 .1255 4.60 .0101
2.20 .1108 4.70 .0091
2.30 .1003 4.80 .0082
2.40 .0907 4.90 .0074
2.50 .0821 5.00 .0067

By means of extensive testing and data collection, a manufacturer has determined that a particular Example 3
model of its vacuum cleaners has an expected life that is Exponential with a mean of four years
and insignificant burn-in phase. Find the probability that one of these vacuum cleaners will have
a life that ends:
a. After the initial four years of service.
b. Before four years of service are completed.
c. Not before six years of service.

MTTF = 4 years SOLUTION

a. T = 4 years:
4 years
T/MTTF = = 1.00
4 years
From Table 4S-1, e–1.00 = .3679.
b. The probability of failure before T = 4 years is 1 - e–1.00, or 1 - .3679 = .6321.
c. T = 6 years:
6 years
T/MTTF = = 1.50
4 years
From Table 4S-1, e–1.50 = .2231.
4S-8 PART 3 System Design

Figure 4S-5
A Normal curve.

Reliability

0 z
z scale

Mean life T
Years

Mechanical items such as ball bearings, valves, and springs tend to have insignificant
burn-in and steady-state phases, and start to wear out right away. Item failure due to wear-out
can sometimes be modelled by a Normal distribution. Obtaining Normal probabilities
involves the use of the standard Normal table (see Appendix B, Table B). The table provides
areas under a Normal curve up to a specified point z, where z is a standardized value calculated
using the formula:

T − Mean wear-out time

z=
Standard deviation of wear-out time

This area is the probability that service life will not exceed some value T. To find the reliability,
subtract this probability from 1. See Figure 4S-5.
To obtain the value of T that will provide a given probability, work in reverse, i.e., locate the
nearest probability in Appendix B, Table B, and pick up the associated z value. Then, insert the z
value in the above formula and solve for T.

Example 4 The mean life of a certain ball bearing can be modelled using a Normal distribution with a mean
of six years and a standard deviation of one year. Determine each of the following:
a. The probability that a ball bearing will fail before seven years of service.
b. The probability that a ball bearing will fail after seven years of service (i.e., find its reliability).
c. The service life that will provide a failure probability of 10 percent.

SOLUTION Wear-out mean = 6 years

Wear-out standard deviation = 1 year
Wear-out is Normally distributed
a. Calculate z using the above formula:

7−6 .8413
z= = +1.00
1

0 + 1.00
z scale

6 7
Years
 SUPPLEMENT TO CHAPTER 4 Reliability 4S-9

Use the calculated z to obtain the required probability from Appendix B, Table B. Thus, P (T < 7) =
P (z < 1) = .8413 (see the graph on the previous page).
b. Subtract the probability determined in part a from 1.00 (see the graph below).

1.00 − .8413 = .1587

.1587
0 + 1.00
z scale

6 7
Years

c. Use the standard Normal table in reverse, i.e., find the value of z that corresponds to an area
under the curve (starting from the left side) of .10. Thus, z = -1.28 from Appendix B, Table B.
Now, insert this in the z formula above:

90%
T−6
z = 1.28 =
1
10%
z = ⫺1.28 0
z scale

4.72 6
Years

Solving for T, we find T = 4.72 years (see the graph above).

A more general distribution than Exponential is the Weibull distribution. The probability density
function of a Weibull random variable x is:

k x k−1 −1x/λ2k
a b e x ≥ 0,
f 1x; λ, k2 = • λ λ
0 x > 0,

where k > 0 is the shape parameter, λ > 0 is the scale parameter of the distribution, and x repre-
sents time t. If k = 1 the failure rate is constant over time, and Weibull is identical to the Exponen-
tial distribution. If k < 1 failure rate decreases over time, and if k > 1 failure rate increases over
time. Because of its flexibility, Weibull distribution is commonly used to model the time to failure
during the burn-in (k < 1) and wear-out (k > 1) phases.
4S-10 PART 3 System Design

OM in ACTION

PTC®-Windchill Quality Solutions1

PTC®-Windchill Quality Solutions’ reliability software assists in
designing products and parts with reduced chance of failure,
collecting reliability data, estimating failure rate and mean time
between failures, finding the cause of failure, and many other
reliability activities. See the screenshot of the bill of material of
the processing unit of a computer, with component failure rate
(per million hours) provided from the library of the software,
based on quality and expected operating conditions such as
temperature.

Source: http://www.ptc.com/products/windchill/quality.

Source: Image courtesy PTC® Windchill Quality Solutions™, http://

www.ptc.com.

Please note that the probability rules of Section 2 above for series and parallel systems can also
be used to determine reliability of a system for a given length of time based on the reliability of its
components over the same length of time.

Availability
For repairable items, the measure of importance to customers, and hence to designers, is
availability The fraction of availability. It measures the fraction of time a piece of equipment or a repairable product is
time a piece of equipment or expected to be available for operation (as opposed to being down for repair). Availability can
a repairable product is range from zero (never available) to 1.00 (always available). Companies that can offer equipment
expected to be available for with high availability have a competitive advantage over companies that offer equipment with
operation. lower availability. Availability is a function of both the mean time between failures and the mean
time to repair (the average length of time to repair a failed item). We assume that there is little
mean time to repair The delay before a failed item begins to be repaired. The availability factor can be calculated using the
average length of time to following formula:
repair a failed item.
MTBF
Availability =
MTBF + MTTR
where
MTBF = Mean time between failures
MTTR = Mean time to repair

Example 5 A copier is expected to operate for 200 hours after repair, and the mean repair time is expected to
be two hours. Determine the availability of the copier.
MTBF = 200 hours, and MTTR = 2 hours
Availability = MTBF/1MTBF + MTTR2 = 200/1200 + 22 = .99

To increase availability, designers increase MTBF but also decrease MTTR. Laser printers, for
example, are designed with print cartridges that can be easily replaced, thus requiring a small
MTTR.

1
http://www.ptc.com/products/windchill/quality. Image courtesy PTC®-Windchill Quality Solutions.
 SUPPLEMENT TO CHAPTER 4 Reliability 4S-11

Key Terms
availability mean time to failure (MTTF) redundancy
mean time between failures (MTBF) mean time to repair (MTTR) reliability

Solved Problems
Problem 1
A product designer must decide if a redundant component is cost-justified in a product. The product
in question has a critical component with a probability of .98 of operating. Product failure would
involve a cost of $20,000. For a cost of $100, a switch and backup component could be added that
would automatically transfer the control to the backup component in the event of a failure. Should
the backup component be added if its operating probability is also .98?

Solution
Because no probability is given for the switch, we will assume that its probability of operating when
needed is 1.00. The expected cost of failure (i.e., without the backup) is $20,000 (1 - .98) = $400.
With the backup, the probability of not failing would be:

.98 + .021.982 = .9996

Hence, the probability of failure would be 1 - .9996 = .0004. The expected cost of failure with the
backup would be the added cost of the backup component plus the failure cost:

$100 + $20,0001.00042 = $108

Because this ($108) is less than the expected cost without the backup ($400), adding the backup
component is definitely cost-justified.

Problem 2
Due to the extreme cost of interrupting production, a manufacturer has two standby machines
available in case a particular machine breaks down. The machine in use has a reliability of .94, and
the backups have reliabilities of .90 and .80. In the event of a failure, a backup machine is brought
into service. If this machine also fails, the other backup is used. Calculate the system reliability.

Solution
R1 = 0.94, R2 = .90, and R3 = .80

The system can be depicted in this way:

.80

.90

.94

Rsystem = R1 + R2 11 − R1 2 + R3 11 − R2 211 − R1 2
= .94 + .9011 − .942 + .8011 − .90211 − .942 = .9988
Problem 3
A hospital has three independent fire alarm systems, with reliabilities of .95, .97, and .99. In the event
of a fire, what is the probability that a warning would be given?
4S-12 PART 3 System Design

Solution
A warning would not be given if all three alarms failed. The probability that at least one alarm would
operate is 1 - P (none operate):

P1none operate2 = 11 − .95211 − .97211 − .992 = .000015

P1warning2 = 1 − .000015 = .999985

Alternatively, P (warning) = .95 + .97(1 - .95) + .99(1 - .95) (1 - .97) = .999985

Problem 4
A weather satellite has expected life of 10 years from the time it is placed into Earth’s orbit.
Determine its reliability for each of the following lengths of service (assume that Exponential
distribution is appropriate.)
a. 5 years
b. 12 years
c. 20 years
d. 30 years
Solution
MTTF = 10 years
Calculate the ratio T/MTTF for T = 5, 12, 20, and 30, and obtain the values of e–T/MTTF from Table 4S-1.
These are the solutions (= reliabilities).
T MTTF T/MTTF e⁻T/MTTF

a. 5 10 .50 .6065
b. 12 10 1.20 .3012
c. 20 10 2.00 .1353
d. 30 10 3.00 .0498

Problem 5
What is the probability that the satellite described in Solved Problem 4 will fail between 5 and 12
years after being placed into Earth’s orbit?
Solution

P15 years < failure < 12 years2 = P1failure after 5 years2

− P1failure after 12 years2

Using the probabilities shown in the previous solution, we obtain:

P (failure after 5 years) = .6065

- P (failure after 12 years) = .3012
.3053
See the following chart:

P(5 < Failure < 12)

f(t )

.3053

0 5 12
Years
 SUPPLEMENT TO CHAPTER 4 Reliability 4S-13

Problem 6
One line of specialty tires has a wear-out life that can be modelled using Normal distribution with a
mean of 25,000 km and a standard deviation of 2,000 km. Determine each of the following:
a. The percentage of tires that can be expected to wear out within ± 2,000 km of the average
(i.e., between 23,000 km and 27,000 km).
b. The percentage of tires that can be expected to fail between 26,000 km and 29,000 km.
c. For what tire life would you expect 4 percent of the tires to have worn out?

Solution
Note: Kilometres are analogous to time and are handled in exactly the same way.
a . The phrase “within ± 2,000 km of the average” translates to within one standard deviation
of the mean because the standard deviation equals 2,000 km. Therefore, the range of z is z =
-1.00 to z = +1.00, and the area under the curve between those points is found as the differ-
ence between P (z < +1.00) and P (z < -1.00), using values obtained from Appendix B, Table B.

P1z < +1.002 = .8413

−P1z < −1.002 = .1587

P (−1.00 < z < +1.00) = .6826, which means 68.26% of tires will wear out between 23,000 km
and 27,000 km (see the following chart):

.8413

.1587

.6826

– 1.00 0 + 1.00

b. Wear-out mean = 25,000 km

Wear-out standard deviation = 2,000 km

P126,000 < Wear-out < 29,0002 = P1z < z29,000 2 − P1z < z26,000 2
29,000 − 25,000
z29,000 = = +2.00 S .9772 (from Appendix B, Table B)
2,000
26,000 − 25,000
z26,000 = = +.50 S .6915 (from Appendix B, Table B)
2,000
The difference is .9772 - .6915 = .2857, which means 28.57 percent of tires will wear out
between 26,000 km and 29,000 km (see the following chart).

.9772
.6915

.2857

0 + 0.50 + 2.00

c. Use Appendix B, Table B to find z for 4 percent: z = -1.75

Find tire life using μ + zσ: 25,000 - 1.75(2,000) = 21,500 km.
4S-14 PART 3 System Design

Discussion and Review Questions

Note: An asterisk indicates that a question or problem may be more challenging.

1. Define the term reliability and give an example. LO1 5. What is the significance of the bathtub curve in reliabil-
2. Explain why a product might have an overall reliability ity? Give an example of an item with failure rate in each
that is low even though its components have fairly high phase. LO3
reliabilities. LO2 6. How is reliability determined if the distribution of time
3. What is redundancy and how can it improve product to failure is Exponential? LO3
reliability? LO2 7. What is availability and how can it be increased? LO3
4. How is failure rate per hour calculated? MTTF? Give an
example of both. LO3

Internet Exercises
1. Visit either http://www.smrpjobboard.com or http:// 3. Visit http://www.itl.nist.gov/div898/handbook/apr/
www.sre.org/current/current.htm, pick a reliability job section1/apr18.htm, find the information about the
announcement, and briefly summarize the duties Standby System and parallel system, and briefly explain
involved. LO1 how they differ. LO3
2. Read http://asq.orgcertification/reliability-engineer/bok. 4. Visit http://www.sqconline.com/resistor-mtbf-mil-hdbk-
html, and briefly summarize the knowledge and skills 217-rev-f-notice-2, find the failure rate and MTBF of a resis-
required of a Certified Reliability Engineer. LO1 tor, and identify the relationship between the two. LO3

Problems
1. Consider the following system: LO2 4. A product engineer has developed the following equa-
tion for the cost of a component: C = (10P)2, where C is
the cost in dollars and P is the probability that the com-
.90 .90
ponent will operate as expected. The product is com-
posed of two of these components, both of which must
Determine the probability that the system will operate operate for the product to operate. The engineer can
under each of these conditions: spend a total of $173 for the two components. To the
a. The system as shown. nearest two decimal places, what is the largest compo-
b. Each system component has a backup with a nent reliability that can be achieved? LO2
reliability of .90 and a switch that is 100 percent 5. The guidance system of a ship is controlled by a com-
reliable. puter that has three major modules. In order for the
c. Each system component has a backup with .90 reli- computer to function properly, all three modules must
ability and a switch that is 99 percent reliable. function. Two of the modules have reliability of .97, and
2. A product is composed of four parts. In order for the the other has reliability of .99. LO2
product to function properly, each of the parts must a. What is the reliability of the computer?
function. Two of the parts each have .96 probability of b. A backup computer identical to the one being used
functioning, and the other two each have probability of can be installed to improve overall reliability.
.99. What is the overall probability that the product will Assuming that the new computer can automatically
function properly? LO2 function if the first computer fails, determine the
3. A system consists of three identical components. In resulting reliability.
order for the system to perform as intended, all of the c. If the backup computer must be activated by a
components must perform. Each component has the switch in the event that the first computer fails, and
same probability of performance. If the system is to the switch has a reliability of .98, what is the overall
have .92 probability of performing, what probability of reliability of the system? (Both the switch and the
performance is needed by each of the individual backup computer must function in order for the
components? LO2 backup system to function.)
 SUPPLEMENT TO CHAPTER 4 Reliability 4S-15

6. One of the industrial robots designed by a leading pro- 10. Repeat Problem 9 under the condition that one of the
ducer has four major components. Components’ reliabil- components will have a backup with reliability equal to
ity are .98, .95, .94, and .90. All of the components must that of any one of the other components. LO2
function in order for the robot to operate effectively. LO2 11. Hoping to increase the chances of reaching a perfor-
a. Calculate the reliability of the robot. mance goal, the director of a research project has
b. Designers want to improve the reliability of the robot assigned the same task to three separate research teams.
by adding a backup component. Due to space limita- The director estimates that the team probabilities for
tions, only one backup can be added. The backup for successfully completing the task in the allotted time are
any component will have the same reliability as the .9, .8, and .7. Assuming that the teams work indepen-
unit for which it is the backup. Which component dently, what is the probability that the project will not
should get the backup in order to achieve the be completed in time? LO2
highest reliability of the robot? 12. An electronic chess game has a useful life that is Expo-
c. If one backup with a reliability of .92 can be added nential with a mean of 30 months. Determine each of
to any one of the main components, which compo- the following: LO3
nent should get it to achieve the highest overall
a. The probability that any given unit will operate for at
reliability?
least (1) 39 months, (2) 48 months, (3) 60 months.
7. A production line has three machines A, B, and C, with b. The probability that any given unit will fail sooner
reliabilities of .99, .96, and .93, respectively. The than (1) 33 months, (2) 15 months, (3) 6 months.
machines are arranged so that if one breaks down, the c. The length of service time after which the percentage
others must shut down. Engineers are weighing two of failed units will approximately equal (1) 50 percent,
alternative designs for increasing the line’s reliability. (2) 85 percent, (3) 95 percent, (4) 99 percent.
Plan 1 involves adding an identical backup line (i.e., a
13. A manufacturer of programmable calculators is attempt-
series backup), and plan 2 involves providing a backup
ing to determine a reasonable warranty period for a
for each machine (i.e., a parallel backup). In either case,
model it will introduce shortly. The manager of product
three additional machines (A, B, and C) would be used
testing has indicated that the calculators have an
with reliabilities equal to the original three. LO2
expected life of 30 months. Assume product life can be
a. Which plan will provide higher reliability? described by Exponential distribution. LO3
b. Explain why the two reliabilities are not the same.
a. If warranties are offered for the expected life of the
c. What other factors might enter into the decision of
calculators, what percentage of those sold would be
which plan to adopt?
expected to fail during the warranty period?
8. Refer to the previous problem. LO2 b. What warranty period would result in a failure
a. Assume that a single switch is used in plan 1 to trans- chance of approximately 10 percent?
fer production to the backup line if the first line failed, 14. A type of light bulb has a life that is Exponentially dis-
and this switch is 98 percent reliable, while reliabili- tributed with a mean of 5,000 hours. Determine the
ties of the machines remain the same. Recalculate the probability that one of these light bulbs will last: LO3
reliability of plan 1. Compare this reliability with the
a. At least 6,000 hours.
reliability of plan 1 calculated in solving the original
b. No longer than 1,000 hours.
problem. How much did reliability of plan 1 decrease
c. Between 1,000 hours and 6,000 hours.
as a result of a 98 percent-reliable switch?
b. Assume that three switches are used in plan 2 to 15. According to its designers, a satellite will have an
transfer production to the backup machines if the expected life of six years. Assume that Exponential
original machines failed, and these switches are all distribution applies. Determine the probability that
98 percent reliable, while reliabilities of the machines it will function for each of the following time periods:
remain the same. Recalculate the reliability of plan 2. LO3
Compare the reliability of this plan with the reliabil- a. More than 9 years.
ity of plan 2 calculated in solving the original prob- b. Less than 12 years.
lem. How much did reliability of plan 2 decrease? c. More than 9 years but less than 12 years.
9. A Web server has five major components that must all d. At least 21 years.
function in order for it to operate as intended. Assuming 16. An office manager has received a report from a
that each component of the system has the same consultant on equipment replacement. The report
reliability, what is the reliability each one must have indicates that the scanners have a service life that is
in order for the overall system to have a reliability Normally distributed with mean of 41 months and
of .98? LO2 standard deviation of 4 months. On the basis of this
4S-16 PART 3 System Design

information, determine the percentage of scanners that premium battery, and offering a 54-month warranty
can be expected to fail in the following time periods: LO3 on it. What percentage of “premium” batteries would
a. Before 38 months of service. you expect to fail before the warranty period expires?
b. Between 40 and 45 months of service. 24. In practice, for a series system the failure rate is esti-
c. Within 2 months of the mean life. mated by adding the failure rate of its components. For a
17. A copier manufacturer has determined that its major system made of n identical components in series, each
product has a service life that can be modelled by Nor- having a probability of failure = Pf , probability of sys-
mal distribution with mean of six years and standard tem failure is approximately n(Pf ) provided that Pf is suf-
deviation of half year. LO3 ficiently small. Choose a value of n > 1 and Pf < .05, and
show the above result. LO2
a. What probability can you assign to service lives of
(1) at least five years? (2) at least six years? (3) at most 25. The MTTF of the central processing unit (CPU) of a
seven and a half years? single board computer is estimated to be 150,000 hours.
b. If the manufacturer offers a service warranty of four You can assume Exponential distribution for operating
years on these copiers, what percentage can be time of this component until failure. What is the
expected to fail during the warranty period? probability that this component will operate without
failure for: LO3
18. Refer to Problem 17. What warranty period would result
in percentage failure of: LO3 a. 2.5 years?
b. 5 years?
a. 2 percent?
c. 10 years?
b. 5 percent?
(Hint: Use a calculator, instead of Table 4S-1, to obtain
19. Determine the availability for each of these cases: LO3
more accurate probabilities.)
a. MTBF = 40 days, MTTR = 53 days
*26. A study was performed to determine the reliability of
b. MTBF = 300 hours, MTTR = 6 hours
components of personal computers (PCs) used by Rolls-
20. A machine can operate for an average of 50 days before Royce staff.2 The operating lives of the components of
it needs to be overhauled, a process that takes two days. 341 PCs were measured over a 22-month period.
Calculate the availability of this machine. LO3 The study fit the Weibull distribution to the sample
21. A manager must decide between two machines. operating lives of each component and, based on
Machine A has an average operating time of 142 hours the estimated parameters k (shape) and λ (scale), calcu-
and an average repair time of 7 hours. Times for lated its average life or MTTF. LO3
machine B are an average operating time of 65 hours
and an average repair time of 2 hours. What is the avail- k x k−1 − 1x/λ2k
ability of each machine? LO3 a b e x ≥ 0,
f 1x; λ, k2 = µ λ λ
22. A designer estimates that she can (1) increase the
average time to failure of a product by 5 percent at a 0 x < 0.
cost of $450, or (2) reduce the average repair time by
In the above formula, x represents time. The mean of the
10 percent at a cost of $200. Which option would be
Weibull distribution is λ Γ(1 + 1/k) where Γ (y) can be
more cost-effective? Currently, the average time to
computed in Excel using “ = EXP(GAMMALN(y))”.
failure is 100 hours and the average repair time is
While some components had a larger failure rate at the
4 hours. LO3
beginning of their life (e.g., hard disks with k = .51),
23. A battery’s life is Normally distributed with mean of others had almost constant failure rate over time (e.g.,
4.7 years and standard deviation of .3 year. The batteries motherboards with k = .99). Mice had k = .86, λ =
are warrantied to operate for a minimum of four years. 22,440, and average life of 24,000 hours; keyboards had
If a battery fails within the warranty period, it will be k = .76, λ = 41,919, and average life of 49,000 hours;
replaced with a new battery at no charge. LO3 hard disks had k = .51, λ = 136,752, and average life of
a. What percentage of batteries would you expect to 264,000 hours; and monitors had k = .76, λ = 58,395,
fail before the warranty period expires? and average life of 69,000 hours.
b. The manager is toying with the idea of using the a. For motherboards, k = .99 and λ = 49,171. Estimate
same battery with a different exterior, labelling it as a the average life of a motherboard.

2
M. Bradley and R. Dawson, “The Cost of Unreliability: A Case Study,” Journal of Quality in Maintenance
Engineering, 4(3), 1998, pp. 212–218.
 SUPPLEMENT TO CHAPTER 4 Reliability 4S-17

b. Suppose 90 motherboards (out of 341) failed during Hint: The distribution of time to failure for the car is also
the study period (22 months), and their lives added Exponential.
to 540,000 hours. Estimate the average life of a *32. A computer has two independent and identical Central
motherboard (in hours) directly (i.e., not using Processing Units (CPUs) that are both used when the
Weibull distribution). computer is on. Suppose that the failure rate of a CPU is
27. A car has four independent and identical tires. The reli- .0001 failures per hour (with time to failure being Expo-
ability of a tire is .99. If any tire is flat, the car cannot be nential), and the computer will operate if at least one
driven. Calculate the reliability of a car with respect to CPU operates. Calculate the mean time to CPU failure of
its tires. LO2 the computer. Hint: It can be shown that
28. A computer has two independent and identical Central 1 m 1
MTTFp = a
Processing Units (CPUs). The computer will operate if at λ i=1 i
least one CPU operates. The reliability of a CPU is .99. where λ is the failure rate of one CPU and m is the num-
Calculate the reliability of the computer with respect to ber of parallel CPUs (in this case m = 2). LO3
its CPUs. LO2 *33. A standby system consists of two independent and
*29. A communication network between two cities, A and B, identical units. When the first unit fails, the second unit
consists of five independent and identical relay units (standby) kicks in. Note that unlike the parallel system of
forming a bridge network as shown below. For the net- the previous problem, the standby component is
work to work, a least one path between the two cities brought into operation only when needed. A unit’s fail-
should work. If the reliability of each relay is .99, what is ure rate is .0001 failures per hour (with time to failure
the reliability of the network RN? Hint: RN = 1 - sum of being Exponential). Calculate the system mean time to
probability that any minimal two or three relays that failure (i.e., both units failing). Hint: It can be shown that
together cut off the network will fail simultaneously. m
MTTF =
Specifically, any of sets of relays 1&4, 2&5, 1&3&5, and λ
2&3&4 cut off the two cities. LO2 where λ is the failure rate of a unit and m is the number
of units (in this case m = 2). LO3
1 2 *34. The following data shows the time to failure of 100 units
of a generator that were run until failure.3 LO3
City A 3 City B a. Compute the failure rate during each month as a pro-
portion of numbers that survived until the beginning
of that month. Draw these failure rates against
4 5 months. Determine if the failure rate is decreasing,
constant, or increasing.
b. If we assume that failure rate is constant, determine
30. A plane has two independent and identical engines. At
the average failure rate (per month) and MTTF (in
least one engine should operate for the plane not to
months).
crash. The reliability of an engine is .999. Calculate the
reliability of the plane with respect to its engines. LO2 Month 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
31. Suppose that the failure rate of a tire is .0001 per hour of No. failed 1 0 2 1 4 10 18 22 14 8 10 5 3 1 0 1
driving (with time to failure being Exponential), and that
the failure rates of the four tires are independent and 35. Five units of an electronic component were subjected to
identical. LO3 stress testing until they failed.4 The observed failure times
a. What is the failure rate of the car with respect to its (in days) were 550, 680, 820, 910, and 1,110. Calculate
tires? Hint: It is the sum of the failure rate of the four the MTTF and the failure rate for this component. LO3
tires because any tire’s failure would result in the 36. Twenty units of a component with constant failure rate
car’s failure. were subjected to high-stress testing. After 25 hours,
b. Calculate the mean time to a tire failure for the car. seven failed at times (in hours) 2.1, 8.3, 10.9, 15.2, 16.3,
c. Calculate the reliability of the car with respect to its 20.5, 23.8.5 Calculate the MTTF and the failure rate for
tires if a 25-hour journey is going to take place. this component. LO3

3
R. D. Leitch, Basic Reliability Engineering Analysis, London: Butterworths, 1988, p. 21.
4
Ibid., p. 33.
5
Ibid., 1988, p. 34.
4S-18 PART 3 System Design

*37. The following data shows the number of failures of 38. A piece of equipment contains 300 integrated circuits,
1,000 units of an electronic component in time intervals 25 amplifiers, 150 transistors, 500 resistors, and 450
of 100 hours.6 LO3 capacitors.7 Failure of any component causes a system
a. Compute the failure rate during each time interval as failure. Using the estimated failure (hazard) rates below,
a proportion of numbers that survived until the calculate the failure (hazard) rate of the equipment. LO3
beginning of that interval. What can you conclude Hazard Rate
about the failure rates? Component (failures per billion hour)
b. Calculate the MTTF for this component. Integrated circuit 10
Time Interval No. of Failures Amplifier 30
Transistor 20
0–100 95
Resistor 1
100–200 86
Capacitor 1
200–300 78
300–400 70
400–500 64
500–600 58
600–700 52
700–800 47
800–900 42
900–1000 39

MINI-CASE

Engineer Tank8 set to 420 hours. (The target for MTBF for loss of propulsion is
3,500 hours). During reliability testing for 70 engineer tanks, the
Engineer tank is an armoured vehicle used on the battlefield to need for unscheduled maintenance arose after the following
perform engineering tasks such as opening routes, digging, and
number of hours: LO3
bulldozing. During design, the mean time between the need for
unscheduled maintenance (other than loss of propulsion) was

No. Failed hr No. Failed hr No. Failed hr No. Failed hr No. Failed hr No. Failed hr No. Failed hr

1 9 11 52 21 105 31 178 41 232 51 284 61 416

2 18 12 64 22 114 32 194 42 232 52 291 62 425
3 22 13 68 23 122 33 195 43 239 53 292 63 441
4 23 14 75 24 126 34 198 44 241 54 294 64 451
5 27 15 88 25 126 35 202 45 244 55 315 65 454
6 33 16 89 26 130 36 210 46 247 56 323 66 456
7 36 17 91 27 151 37 215 47 247 57 325 67 557
8 42 18 94 28 153 38 216 48 252 58 350 68 560
9 46 19 95 29 165 39 223 49 273 59 360 69 947
10 50 20 99 30 168 40 223 50 275 60 378 70 997

a. Make a histogram of these times. Does the distribu- b. Compute the mean. Is the target MTBF met?
tion look like Exponential?

6
D. J. Klinger et al., Editors, AT&T Reliability Manual, New York: Van Nostrand Reinhold, 1990, p. 7.
7
Ibid., p. 94.
8
U. D. Kumar et al., Reliability and Six Sigma, New York: Springer, 2006, p. 102.
 SUPPLEMENT TO CHAPTER 4 Reliability 4S-19

MINI-CASE

Sonar System9
The components of an underwater sonar system are given in the block diagram below:

Transmitter/
Power
Transducer Heat Exchanger Receiver
Supply
Switch

System
Controller
Power Front-end
Pre Amplifier
Amplifier Processor
System
Controller

Display
Monitor

SDLC Display
Display
BUS Processor
Monitor
Signal Audio
Processor Processor
SDLC Display
BUS Monitor
Display
Processor
Display
Monitor

The failure rate of one unit of each component is estimated below using available data tables:
Subsystem Failure rate
λ/106 hours

Transducer 1,000
Power Supply 20
Heat Exchanger 20
Transmitter/Receiver Switch 6
Power Amplifier 120
Pre Amplifier 32
Front-end Processor 400
System Controller (parallel) 10
SDLC Bus (parallel) 15
Signal Processor 450
Display Processor (parallel) 150
Display Monitor (parallel) 50
Audio Processor 20

Calculate the reliability of the sonar system to work 100 hours failure-free. Hint: For two parallel components, the overall failure
rate of the two is the failure rate of one divided by 1.5 (see Problem 32 above).

9
Ibid., pp. 139–141.