Chapter 15
APPLICATIONS TO RELIABILITY AND LIFE TESTING
15.1 Since this is a series system, we need to find R such that
R8 = .95 so R = (.95)1/8 = .9936.
15.2 Since this is a parallel system, we need to find R such that
1  .96 = (1  R)5.
so
R = 1  (1  .96)1/5 = .475.
15.3 The reliability of the B, C parallel segment is :
RBC = 1  (1  .80)(1  .90) = .980. 587
588
Chapter 15 RELIABILITY AiYD LIFE TESTLYG
The reliability of the E, F parallel segment is :
REF = 1  (1  .90)(1  .85) = .985.
The reliability of the A. B, C combination segment is :
R·wc = RA REC = (.95)(.980) = .931.
The reliability of the D, E, F combination segment is :
RDEF = RD REF = (.99)(.985) = .97515.
Thus. the reliability of the system is:
R 1 (1  RAEC)(I RDEF)
1  (1  .931)(1  .97515) = .9983.
15.4 The diagram of the system is shown in Figure 15.1.
Figure 15.1: A Diagram of the System for Exercise 5.4.
(a) The reliability of the B component is :
589
RB = 1  (1  .9995)(1  .999)(1  .20) = .9999996.
.The reliability of the C component is :
Rc = 1  (1  .95) (1  .8.5) = .9925.
Thus, the reliability of the system is:
R = RA RB Rc = (.9999)(.9999996)(.9925) = .9924.
(b) In this case, the reliability of the B component is
RB = 1  (1  .9995)(1  .999)(1  .99) = 1  5 X 109.
Thus, the reliability of the system is:
R = RA RB Rc = (.9999)(1  5 x 109)(.9925) = .9924.
(c) The reliability of the B component ( before the increase in reliability of B3 )
IS :
RB = 1  (1  .9995)(1  .20) = .9996.
Thus, the reliability of the system is:
R = (.9999)(.9996)(.9925) = .9920.
After the increase, the reliability of the component B3 is :
RB = 1  (1  .9995)(1  .99) = .999995.
590
Chapter 15 RELIABILITY AND LIFE TESTING
so the reliability of the system is
R = (.9999)(.999995)(.9925) = .9924.
The reliability" of the system has increased by .0004 .
(d) The reliability of the C component is now
Rc = 1  (1  .95)(1  .85)(1  .80) = .9985
so the reliability of the system is
R = (.9999) (.9999996) (.9985) = .9984
which is an increase of .006 over the original system,
15.5 (a) We have that
f(t) = Z(t) exp [  fat Z(t) dt].
where
Z(t) = { g(1 t/a)
for 0 < t < a. elsewhere.
Thus
f(t) = { f30 (1  t/a) exp [  J~ f3(1  x/a) dx] for 0 < t < a. elsewhere.
so
f(t) = { f30(1  t/a) exp [f3(t  t2/(2a) )] for 0 < t < a. elsewhere.
The distribution function is
591
F(t) = { f~of(x) dx for 0 < t < a. fo f(:r) dx for t > a.
or
F(t) = { 1 exp [3(t  t2j(2a))] for 0 < t < a.
1  exp [ a3j2] for t > a.
(b) the probability of initial failure is F(a) = 1  ea3/2 as ShO\YIl in the previous part.
15.6 (a)
F(t)
{ 8 + rt IB)Bo dt for a < t < ,8.
1 Ja (3a
1  82 for t = B.
{ 8 + (IB) (i2)(ta) for a < t < p.
_ 1 (3a
1  82 for t = p.
(b)
Z(t)
f(t) (1 81  (h)j({3  a)
=
1  F(t) 1  81  (1  81  82)(t  a)j(3  a)
1  81  82
(c) Figure 5.2 gives the plot of
.2
, 100::; t < 15, 000 14,900(.95) + (.2)(t  100)
15.7 (a) Since the failure rate is constant,
f(t) = (.02)e(·02)t
592
Chapter 15 RELIABILITY AND LIFE TESTING
0 LOi
"'" ,~
,.... ali
J
al

(1j 0 ~j
II C')
al ,.... ali
l... !
:J
. iii J
LL
I
0 LOI
C\J '"
,.... all
,
!
0 5000
10000
15000
Hours
Figure 15.2: Failure Rate. Exercise 5.6
and
F(t) = 1  e(·02)t.
The unit of time is 1000 hours. The probability that the chip will last longer than 20,000 hours is
1  F(20) = e(·02).20 = .6703.
(b) The 5000hour reliability of four such chips in a series is
R = [1  F(5)]4 = e4(.02).5 = .6703.
593
15.8 (a) The probability of failure during the first 365 days of operation is:
1  e(·0005)(365) = .167.
(b) The probability that two independent components will survive the first 365
days of operation is:
e(·0005)(365) . e(·0005)(365) = .694.
15.9 Using the formula for the mean time between failures
1 1 1 1 1 1
fl =(1+++···:)= (1+++···+).
p ex 2 n 9 x 104 2 n
Since the failure rate is the inverse of the mean time between failures, in this case
we require that flp ~ 1/ (4 X 104). Thus we must find n such that
9 1 1
 < (1 +  + ... + ).
4  2 n
A trial and error search yields n = 5 and the failure rate is
9.0 X 104 = 3.94 X 104. 2.2833
15.10 The mean time between failures for a series system is:
lvlTBF
1
_1_+_1_+_1_+_1_
1'1 JL2 JL3 JL4
1 = .11628
l.9 + 3.1 + l.4 + 2.2
or 1,162.8 hours.
594
Chapter 15 RELIABILITY AND LIFE TESTIVe
1.5.11 (a) The probability that a part will last 500 hours is:
1  F(.500) = e500/1000 = .606.5.
(b) The probability that at least one will fail is one minus the probability that
none will fail or
1  e3(lOOO)/lOOO = .9502.
(c) The probability that a particular part fails is:
1  e6oo/1000 = .4.512.
The probability that exactly two parts fail is given by the binomial distribution with p = .4512 or
( ~ ) (.4512)2 (1  .4512)2 = .3679.
15.12 The Weibull density is
f(t) = "'{3tJ11eo:t{3 f (3 t 0
<< or a, , > .
Thus,
F(t) = 1 _ eo:tfJ.
The probability that the component will operate at least 5,000 hours is
1  F(5000) = e(.005)(5000)80 = .0106 for a, {3, t > O.
15.13 (a) The probability of failure on any trial is p and the probability of no failure is 1  p provided failure has not occurred previously. thus, failure on the x'th trial means no failure for xI trials and then a failure. Since the trials are
595
independent
f(x) = (1 p)Xl p for x = l.2.3, ...
(b)
x xl
F(x) 'L(1  pr1p = » 'L(1 p)i
i=l
i=O
l(lp)X ( )X
P =1 1p
p
for x = 1,2,3""
(c) The probability that the switch survives 2,000 cycles is
1  F(2000) = (1  6 x 104)2000 = .30l.
15.14 (a) T; = tit; = 50·760 = 38,000. Since X~75 with 2· r = 16 degrees of freedom is 6.908 and X~025 with 16 degrees of freedom is 28.845, the 95 percent confidence interval for /L is
2 . 38000 2 . 38000
28.845 < /L < 6.908
or
2,635 < /L < 11,002
(b) The null hypothesis is /L = 10,000 and the alternative is /L < 10,000. Since X~95 with 2· r = 16 degrees of freedom is 7.962 , we reject the null hypothesis at the .05 level if
121
t: < 2/LOX.95 = 2(10,000)(7.962) = 39,810.
Since T; = 38,000 , we reject the null hypothesis at the 0.05 level of signifi
cance.
596
Chapter 15 RELIABILITY AND LIFE TESTLVG
15.15 (a)
r
T; :L i, + (n  r )tr
i=l
250 + 380 ""7"" 610 + 980 + 1250 + 30· 1250 = 40,970.
Since X2995 with 10 degrees of freedom is 2.156 and X2005 with 10 degrees of freedom is 25.118, the 99 percent confidence interval is
2·40.970 2·40,970
25.118 < /L < 2.156
or
3.253.1 < /L < 38,005.6
(b) The null hypothesis is /L = 5.000 and the alternative is /L > 5,000. Since X205 with 2· r = 10 degrees of freedom is 18.307, we reject the null hypothesis at the .05 level if
1 2 1
Tr > 2/LOX.05 = 2(5,000)(18.307) = 45,767.5
Since T; = 40,970 , we cannot reject the null hypothesis at the .05 level of significance. vVe cannot be sure the manufacturer's claim is true.
15.16 We calculate the successive values of total time on test
Tl = i, + (n  l)tl = 250 + 34(250) = 8,750
T2 = t, + t2 + (n  2)t2 = 250 + 380 + 33(380) = 13,170 .
Continuing, we use
to find T3 = 20,760, T4 = 32,600 and, n = 40,970. The points
597
t.tt, .214 .321 .507 .796 1
i/r .2 .± .6 .8 1 are graphed as a total time on test plot in Figure 15.3.
~ •
•
.._
I <0
 0
f= •
•
C\J •
0
0.2 0.4 0.6 0.8 1.0
ilr Figure 15.3: Total Time OIl Test Plot. Exercise 5.16
15.17 (a)
i=l
211 + 350 + 384 + 510 + 539 + 620 + 715 = 3,329.
Since X~975 with 14 degrees of freedom is 5.629 and X2025 with 14 degrees of freedom is 26.119, the 95 percent confidence interval is
2 . 3329 2 . 3329
26.119 < JL < 5.629
or
254.9 < JL < 1,182.8
598
Chapter 15 RELIABILITY AND LIFE TESTING
(b) The null hypothesis is 11 = 500 and the alternative is 11 =I 500. Since ,\205 with 14 degrees of freedom is 23.658 and X29;) with 11 degrees of freedom is 6.57l. we reject the null hypothesis at the .10 level if
or if
Since T; = 3,329 , we cannot reject the null hypothesis at the .10 level of significance.
15.18 (a) From Exercise 15.15 we have that T; = 40,970 and X205 = 18.307 for 10 degrees of freedom. Thus, the onesided 95 percent tolerance limit is
') ·40.970· In(.8)
t* = ~. = 998.8.
18.307
(b) From Exercise 15.17 we have that T; = 3,329. Since X201 = 29.141 for 14 degrees of freedom, the onesided 99 percent tolerance limit is
* 2·3329 ·In(.9)
t = = 24.07.
29.141
15.19 Since 2Tr/1l is a X2 random variable with 2r degrees of freedom,
[ 2 2Tr
p X1n/2 <  11
2Tr 2 1
and  < Xn/2 = a,
11
where XLn/2 and X;/2 are the chisquare quantiles for 2r degrees of freedom. Multiplying the first inequality by Il/XLn/2 and the second by Il/X;/2 gives:
[ 2Tr 2Tr 1
p 11 < 2 and 2  < 11 = a.
X1n/2 Xn/2
599
Thus the (1  a) 100 percent confidence interval is:
2Tr 2Tr
,)  < jJ. < 2:
XZ,/2 \ 10:/2
15.20 \\e solve the equation
for n = 100 , r = 10 and the first 10 failure times 7.0, 14.1, 18.9, 31.6, 52.8, 80.0, 164.5, 355.4, 451.0, 795.1. Using a computer, we obtain the solution (3 = .43449 so that
1
a= ~~
~ L~=l tf + (n  r)t~
110 LI~l t;43449 + 90(795.1).43449 = .005818.
1
The failure rate Z(t) = a(3t31 , at t = 1000 , is estimated by
Z(t) = &!Jt(31 = (.005818)(.43449)(1000).434491 = .0508 X 103 hours.
If the exponential model had been used, the estimate of the mean would have
p, = Tr = 73,529.4 = 7,352.94.
r 10
Thus the estimate of the constant failure rate is 1/ p, = .0136 X 103 hours.
15.21 There are r = 3 failures and total time on test is
T3 = 2076 + 3667 + 9102 + 197(9102) = 1,807,939
Since X~05 = 12.592 for 2r = 6 degrees of freedom, the 95 percent lower confidence
600
Chapter 15 RELIABILITY AND LIFE TESTING
bound is
2~, = 2(1. 807, 939) = 287, 156.8
X.05 12.592
15.22 \Ve solve the equation
for n = 60 , r = 9 and the first 9 failure times 3.6. 6.9, 9.5, 15.7, 27.3,41.2, 81.7, 178.3, 227.1 . Using a computer, we obtain the solution S = .5062 so that
1
a= ~
_,1 ~r t'r3 + (n  r) tiJ
r L...z=1 z
1
1 9 062 ( ) 0 2 = .0105.
9 Li=l t;" + 51 227.1 .0 6
The estimate of the mean life is
, 1 1
P, = o:l/Pf(1 + ;c) = (.0105)1j.5062f(1 + ) = 15 863.99
f3 .5062 '
If the exponential model had been used, the estimate of the mean would have
II = Tr = 12, 173.4 = 1.352.6.
,... T 9 '
which is far too low.
15.23 The probability the diaphragm valve will perform at least 150 hours is 1 minus the probability that it fails before 150 hours. Since
1  F(150) = 1  (1  ea(150)i3),
0: = .0105 and S = .5062, we estimate this probability by
e6(150)!3 = e(·0105)(150)·5062 = .8758.
601
15.24 The \Veibull density is
Thus,
1t lot p
F(t) = 1(:1') dx = a3.r81eax dx.
o .0
Using the change of variables u = a~·3, we haw
Since R(t) = 1  F(t), we have
15.25 Suppose T has the Weibull distribution with parameters a and (3. In the text it was shown that
1 1
E(T) = a1/i3f(1 + ~)
Since Var(T) = E(T2)  [E(T)j2, we must find E(T2). The Weibull density is
af3t{31 e cd13
so
Let u = at{3. Then
2 _ 1 r'o 2/ {3 u _ 1 ( 2 )
E(T )  a2/{3 Jo u e du  a2/{3 r 1 + ~
602
Chapter 15 RELIABILITY AND LIFE TESTING
so that
1.5.26 Since this is a parallel system, we need to find R such that
1.90=(1R)7.
so
R = 1 (1  .90)1/7 = .28.
15.27 (a) The probability of failure during the first 250 hours of operation is:
1  e(·0045).250 = .6753.
(b) The probability that two independent components will survive the first 100 hours of operation is:
e(·0045).100 . e(·0045).100 = .4066.
15.28 Using the formula for the mean time between failures
1 1 1 1 1 1
/l =(1+++ ... +)= (1+++···+).
p ex 2 ti 2 x 104 2 n
Since the failure rate is the inverse of the mean time between failures, in this case we require that /lp ~ 1/(5 x 105). Thus we must find ti such that
603
A trial and error search yields n = 31 and the failure rate is
2.0 X 104 
 = 4.966 x 100. 4.027
15.29 The mean time between failures for a series system is:
1
JVfT B F = 1 1 1 1 I 1 , 1
+++,rIII /12 /13 /14 Il5 /16
1
 = .0833 thousand hours l.8 + 2.4 + 2.0 + l.3 + 3.0 + l.5
or 83.3 hours.
15.30 (a)
r
T; = Lti + (71  r)tr = 16.5 + 19.2 + 20.8 + 37.3 + (11)37.3 = 504.l.
i=l
Since X295 with 8 degrees of freedom is 2.733 and X205 with 8 degrees of freedom is 15.507, the 90 percent confidence interval is
2·504.1 2·504.1
15.507 < 11 < 2.733
or
65.0 < 11 < 368.9
(b) A failure rate of .004 failures per hour corresponds to a mean life of 1/.004 = 250 hours. Thus, we test the null hypothesis 11 = 250 against the alternative 11 > 250. Since X~Ol with 2 . r = 8 degrees of freedom is 20.090, we reject the null hypothesis at the .01 level if
t; > ~110X201 = ~(250)(20.090) = 2, 51l.25
604
Chapter 15 RELIABILITY AND LIFE TESTING
Since T; = 504.1 , we cannot reject the null hypothesis at the .01 level of
significance.
15.31 There are T = 4 failures and total time on test is
T; = 3582 + 8482 + 8921 + 16303 + 296\16303) = 4,862,976
Since X205 = 15.507 for 2r = 8 degrees of freedom. the 95 percent lower confidence bound is
2T4 _ 2(4,862,976) _ T
2  6~1.197.5
X.05 15.507
15.32 We solve the equation
for ti = 100 , r = 10 and the first 10 failure times 28, 46, 50, 63, 81, 101, 116, 137, 159, 175.
Using a computer, we obtain the solution j = 1.3665 so that
1 1
& =  1 10 = .0000909.
1 /3 f3 10 Li=l d·3793 +90(175)1.3665
~L;=lti +(nr)tr
The estimate of the mean life is
. 1 1
it = &l/i3r(l + 7) = (.0000909)1/1.3665[(1 + ) = 829.8
f3 1.3665
If the exponential model had been used, the estimate of the mean would have
it = Tr = 12,331 = 1,233.1.
r 10
15.33 The probability the circuit will perform at least 100 hours is 1 minus the probability
605
that it fails before 100 hours. Since
1  F(100) = 1  (1  ee>(lOO)3)
& = .0000909 and /3 = 1.3665, we estimate this probability by
eci(lOO)6 = e(·0000909)(lOO)1.3665 = .9520.
15.34 (a) Calculate the successive values of total time on test
Tl = tl + (n  l)tl = 28 + 74(28) = 2, 100
T2 = i, + t2 + (n  2)t2 = 28 + 46 + 73(46) = 3,432
Continuing, use
to find T3 = 3,724, T4 = 4,660, T5 = 5,938, T6 = 7,338, T7 = 8,373, Ts = 9,801, Tg = 11,275, TlO = 12,331. The points
t.tt: .170 .278 .302 .378 .482 .595 .679 .795 .914 1
i/r .1 .2 .3 .4 .5 .6 .7 .8 .9 1 are graphed in Figure 15.4 as a total time on test plot.
(b) Using the approximation F(ti) = i/(n+1), with n = 75 , calculate the r = 10 pairs (Xi, Yi) where
and
1
Yi = ln ln _,
1  F(ti)
G06
Chapter 15 RELIABILITY AND LIFE TESTING
0 • i
I
I I
i I
• I
co I
0 •
I I
I •
,__
I <0 I
 0 , •
i=
•
~ 
a •
•
•
C\J
a •
0.2 0.4 0.6 0.8 1.0
i/r Figure 15.4: Total Time on Test Plot. Exercise 5.34
F(ti) ti Xi Yi
.0132 28 3.33 4.32
.0263 46 3.83 3.62
.0395 50 3.91 3.21
.0526 63 4.14 2.92
.0658 81 4.39 2.69
.0789 101 4.62 2.50
.0921 116 4.75 2.34
.1053 137 4.92 2.20
.1184 159 5.07 2.07
.1316 175 5.16 1.96
In Figure 15.5, these points are graphed as a Weibull plot. 0 •
N •
I
•
•
•
CJ)
Q) •
~
0 0 •
o
CJ) Ct)
I
:::l •
.0
'05
S •
0
.q:
I
•
3.5 4.0 4.5 5.0
In(t) Figure 15.5: Weibull Plot. Exercise 5.34
15.35 (a) We are given that X has
F(x) = 1  e·01x and f(x) = .01e·Olx
and Y has distribution
G(y) = 1 e·005y and g(y) = .005e·005y.
Consequently,
R pry > Xl = i: F(y)g(y) dy
 i: [1  e·olYl·005e·o05y dy
i: .005e·005Ydy  i: (.005)e(·OO5+.01)y dy
607
608
Chapter 15 RELIABILITY AND LIFE TESTING
.005
1  = .6667 . . 015
(b) \Ve are given that X has
F(x) = 1  e005x and fCr) = .005e005x
and Y has distribution
G(y) = 1  e005y and g(y) = .005e·005y.
Consequently,
R pry > Xl = i: F(y)g(y) dy i: [1  e·005Y](.005)e·005y dy
_ i: (.005)e·005y dy  i: .005e(·005+.005)y dy 1  .005 = .50
.010
(c) Since ln X and ln Yare independent and each has a normal distribution, ln Y In X has a normal distribution with mean /Ly  /Lx = 80  60 = 20 and variance 0"; + 0"; = 52 + 52 = 50. To avoid numerical integration, we note that
R  pry > X] = P[ In Y > In X] = P[ In Y  In X > 0] 20
1  F( v'5Q) = 1  F( 2.83)
where 1  F( 2.83) = F(2.83) = .9977 is obtained from the standard normal distribution.