V.

SYSTEM RELIABILITY ASSESMENT USING PROBABILITY

DISTRIBUTIONS

System reliability analysis has three basic steps:

1. Determination of component failure rates or component reliability functions.

2. Construction of series-parallel equivalent.
3. Reliability calculation.

Component failure rates or component reliability functions can be determined from past data or
from some accelerated life tests by using parameter estimation or best fitting curve estimation. It
was the task of Chapter III. Several methods were introduced in Chapter IV to determine series-
paralell equivalents of complex connected systems. However, those methods were given for the
components having constant failure probabilities/reliabilities. All of them are also valid for the
components which are modeled by time dependent reliability expressions. Therefore, compex
connected systems comprising the units having time dependent reliabilities will first be
represented by their series-parallel equivalents. Then, system reliabilities and MTTFs can be
calculated by the techniques those will be given in this chapter. We will assume that the series-
parallel equivalent of the system is derived and we will concentrate ourselves on the calculation
of system reliability.

5.1 Serially Connected Units

Reliability and failure probability of a system comprising n-serially connected units can be
expressed in term of unit reliabilities. If unit reliabilities are R1(t),R2(t),...,Rn(t), then;

  
R s ( t )  R1( t ). R 2 ( t ) R1( t ). R 3( t )  R1( t )  R 2 ( t ) .... R n ( t )  R1( t ) ...  R n 1( t ) 
Qs ( t )  1  R s ( t )

If the units operate independently, then;

n
R s ( t )  R1( t ). R 2 ( t ). R 3 ( t ).... R n ( t )  
i 1
R i (t)

System reliability can be expressed in terms of unit failure rates as;

n n
 t 
Rs (t)  
i 1
R i ( t )   exp   i () d
i 1  0 

Mean time to failure of n- serially connected units can be calculated as;

   n
 t 
m  MTTF   t. fs (t) dt   Rs (t) dt   
i 1
exp  i () d dt
 0 
0 0 0
 2

All those reliability, failure probability and MTTF expressions are valid for all types of failure
distribution functions. On the other hand, equations will be more simple for the systems
comprising the units which have constant failure rates.

n n n
 t 
Rs (t )   exp   i d 
 0 
 e
 i .t
e
 e .t
; e   i
i 1 i 1 i 1

ms   Rs (t ) dt  1e   1   21... n  1 1 1
1

 Min m 1, m 2 ,..., m n 
0  ...
m1 m 2 mn
1
s    e   1   2 ... n
ms

This final Rs(t) expression show that, equivalent of serially connected units having constant
failure rates/exponential failure density functions will have a constant failure rate/exponential
failure density function.

In case of units having linearly increasing failure rates (Rayleigh Failure Density Functions),

 i ( t )  Ki t
n n n
 t 
Rs (t)   exp  K i  d 
 0 
 e K i t 2 / 2
e K e t 2 / 2
; Ke   Ki
i 1 i 1 i 1

ms   Rs (t ) dt  
2. K e


2. K i 
0
1
s 
ms

Final Rs(t) expression show that, equivalent of serially connected units having linearly
increasing failure rates/Rayleigh failure density functions will have a linearly increasing failure
rate/Rayleigh failure density function.

Raliabilities and MTTFs of the other distributions can be obtained similarly. However, in general
we do not expect the same failure rate model for the equivalent.

5.2 Parallel Connected Units

Failure probability and reliability of a system comprising n-parallel connected units can be
expressed in term of unit reliabilities. If unit reliabilities are R1(t),R2(t),...,Rn(t), then;
 3

  
Q p ( t )  Q1( t ). Q 2 ( t ) Q1( t ). Q3 ( t )  Q1( t )  Q 2 ( t ) .... Q n ( t )  Q1( t )  Q 2 ( t ) ...  Q n 1( t ) 
R p (t)  1  Qp (t)

If the units operate independently, then;

n
Q p ( t )  Q1( t ). Q 2 ( t ). Q3( t ).... Q n ( t )  
i 1
Qi ( t )

n
R p (t)  1  Qp (t)  1  
i 1
Qi ( t )

System failure probability can be expressed in terms of unit failure rates as;

n n n   t 
Qp (t)  
i 1
Qi ( t )  1  R i (t )  1  exp   i () d
i 1 i 1   0 
n   t 
R p (t)  1  Qp (t)  1   1  exp   i () d
i 1   0 

Mean time to failure of n-parallel connected units can be calculated as;

   
n n   t 

m  R p ( t ) dt  1 

   1  R i ( t ) dt  1 

  1  exp   i () d dt
i 1   0 
0 0  i 1 0


All those reliability, failure probability and MTTF expressions are valid for all types of failure
distribution functions. On the other hand, equations will be more simple for the systems
comprising the units which have constant failure rates.

n
 t  n
  .t
R p (t )  1  1  exp   i d   1   1  e i 
i 1  0  i 1
 
 1   1   
1 1 1
m p   R p (t )dt     ...     ...    ... 
0  1  2   1   2 1  3   1   2  3 
1  1   1  1
         ...  (1)
n 1

i i  i   j  i, j , k  i   j   k 
i, j  i   j  ...   n
1
p 
mp

This final Rp(t) expression show that, equivalent of parallel connected units having constant
failure rates/exponential failure density functions will not anymore have a constant failure
rate/exponential failure density function.
 4

Three or more parallel units are rarely seen in practical applications. In case of 2-parallel
components,

 1.t  2 .t (  1   2 )t
R p (t)  e e e

 
 
m m
m  R p (t) dt  11  12   1 1  2  m 1  m 2  m 1 1 m2 2   Max m 1, m 2
0

Example: Calculate the reliability and the MTTF of two identical units both for series and for
parallel comnncetions. Compare them with of the single unit.

1
Single unit : R(t )  e  t , m 

2
1
2  serially units : Rs (t )   e
 i .t
 e 2 t  R(t )  e  t , ms  m
i 1 2

  2.e
2
2 1
2  parallel connected units : R p (t )  1   1  e
 i .t  t
 e 2 t  R(t )  e  t , m p   m
i 1  2

R(t)
1

Rp(t) R(t) Rs(t)

2/e-1/2e
1/e
1/2e
t
1

It is clear that serially connected units give the worst reliability (Lower bound structure) and
pparallel connected units give the best reliability (upper bound structure).

5.3 Partial Redundant (k-out of n) Systems

Partial redundant systems are between the two extreme cases; namely, series and parallel
connections. They require some of available components operate for system success. Yhey are
also known as majority vote systems.
All the methods described in Chapter 4 are also valid when probability distributions are used
instead of constant reliabilities. System reliability can be calculated as the some of appropriate
terms in [R(t)+Q(t)] n expression.
 5

Example: Consider a system comprising 4 identical components each having a failure rate of 0.1
f/year. Calculate the reliabilities for 0.5 year and for 5 year of missions
R(t )  exp    d   et , Q(t )  1  R(t )  1  et
t

 0 
R(t )  Q(t )  R (t )  4.R3 (t )Q(t )  6.R 2 (t )Q2 (t )  4.R(t )Q3 (t )  Q4 (t )
4 4

R2 / 4 (t )  R 4 (t )  4.R 3 (t )Q(t )  6.R 2 (t )Q 2 (t )  e 4. t  4.e 3. t .(1  e  t )  6.e 2. t .(1  e  t ) 2

R(0.5) = 0.99955 and R(5) = 0.82824

n
If the units are not identical, appropriate terms of  R (t )  Q (t )
i 1
i i product is used for the

reliability of k-out-of-n system.

Example: Calculate the reliability for t=100 h of the following system comprising constant
failure rate units. Calculate the approximate value of the MTTF.

 2/3 system
F/year
4 F/year
2
5 F/year
1
IN 3 OUT F/year
6
F/year

R23 (t )  R2 (t )  R3 (t )  R2 (t ) R3 (t )
R456 (t )  R4 (t ) R5 (t ) R6 (t )  R4 (t ) R5 (t )Q6 (t )  R4 (t )Q5 (t ) R6 (t )  Q4 (t ) R5 (t ) R6 (t )
R8 (t )  R1 (t ) R23 (t ) R456 (t )
R(t )  R7 (t )  R8 (t )  R7 (t ) R8 (t )
R(t  100 h)  0.99943
m2 m3 
m23  m2  m3   6333 .3 h
m2  m3  1
 m8  1 1 1
 1618 .4 h
1 3 3   
m456     2777 .7 h
34 24 34  m1 m23 m456
m7 m8
m  m7  m8   2723 .9 h
m7  m8
 6

5.4 Standby Redundant System

A
S

Figure 5.1 Standby Redundant System

Time-independent failure probability of a Standby Redundant System was previously given as,

Q  QA  QA PS (1  QB )

Standby redundant system is preferred since they alleviate some problems of parallel redundant
systems where the components are simultaneously operating mode. Here, standby unit operates
whenever the main unit fails. They are more advantageous if the failure rate of the standby unit is
low while iits idle phase.
The fundamental problem in standby redundant system is that a failure sensing and switching
device is required to bring the standby unit into service. These additional units may affect the
system reliability. It is therefore impossible to generalize whether standby redundant systems are
more reliable or less reliable than the paralle redundant systems. For the sake of simplicity we
will only deal with the units having constant failure rates.
Before deriving the expressions, we will better derive the probability of a constant failure rate-
unit failing m times in a specified period.

Pm (t) : The probability of m failures in 0 , t period,

Pm a , b : The probability of m failures in a ,b period,
.dt : The probability of failing t , t+dt given that the unit is surviving at time t,

Pm (t  dt)  Pm (t ).P0 t, t  dt  Pm1 (t ).P1t, t  dt  Pm2 (t ).P2 t, t  dt  ...

If dt is small enough so that the probability of 2 or more failure occurring is zero, then,

Pm (t  dt)  Pm (t ).P0 t, t  dt  Pm1 (t ).P1 t, t  dt  Pm (t ).1  dt Pm1 (t ).dt 

Pm (t  dt)  Pm (t ) dP (t )
  Pm (t ).  Pm1 (t ).  m  Pm (t )  Pm1 (t )
dt dt

dt

d t

e Pm (t )  e
 t dPm (t )
dt
t  t  dP (t )  t
 e Pm (t )  e  m   Pm (t )  e  Pm1 (t )
 dt 
 7

 d e 
t t
t t 0 t t
Pm (t )  e Pm (t )  e .Pm (0)  e Pm (t ) e  Pm1 (t )dt

0 0 0
t
 t t
Pm (t )   e e
0
Pm1 (t )dt

If the final equation is used as a recursize algorithm, then,

t t t
 t t t t t t t
P1 (t )  e e
0
P
0 (t ) dt  e e
0
e dt  e  dt  te
0
R (t )

(t ) 2 t
t t t
t t t t t t
P2 (t )  e  e P1 (t ) dt e
0
 e t.e
0
dt  e  t dt 
0
2
e

(t ) t (t ) 2 (t )3 t

t t 2 t
t t t t
P3 (t )  e  e P2 (t )dt e
0
e
0
2
.e dt  e t 
0
2
dt 
2.3
e

…
(t ) m t
t
t t
Pm (t )  e
0
 e Pm1 (t ) dt
m !
e

That is, the probability of a unit which has a constant failure rate of  failing m times during
[0,t] interval can be represented by a Poisson distribution with an average of a = t.

Let’s try to derive the reliability and the MTTF of a standby system where Error! Bookmark
not defined.BError! Bookmark not defined.A and switch S is perfect. We can either use
two dimensional random variable and a joint probability density function or a single variable
represented by a single probability density function.

1. By using joint probability density function

Assume that  represents the failure time of the main unit (the first random variable) and t
represents the failure time of the standby unit/system (the second variable). Let ft(,t) be the
joint probability density function of  and t.

f, t (, t ) , 0    t

If conditional probability density functions are used, then,

f ,t ( , t )  f t (t  ). f ( )
A
f ( )   A .e , 0    t     (t  )
B (t  )   f ,t ( , t )  A .e A B .e B 0    t 
f t (t  )  B .e 
 8

 

 AB   B t   A t
t
f t (t )   f , t ( , t ).d   f , t ( , t ).d 
 A  B
e e
 0

A
 
t
R(t )  1   f t ( ) d  e   At  e   At  e   B t
0
B   A

1 1
m  MTTF   R(t ) dt  
0
A B

2. By using single random variable

Fig. 5.2 shows the failure mechanism of the standby redundant system.

0 A operating d B Operating
 t
A Failed
Fig. 5.2 Probable operating/failure sequence in standby redundant system

- A .t
P0 (t) = e : The survival probability of unit A in 0 , t period,
P1 (t) : The probability of failing A at <t and operating B up to t

R(t )  P0 (t )  P1(t )
 A t
dP1 ( )  e .( A .d ).e   B (t  ) 
A
 
t t
P1 (t )   dP1 (t )   e   A t . A .e   B (t  ) d  e  A t  eB t
0 0
B   A

R(t )  P0 (t )  P1 (t )  e   A t 
A
B   A

e A t  e B t 

1 1
m  MTTF   R(t ) dt  
0
A B

The first term in the reliability expression corresponds to the probability of unit A has not failed
up to t (reliability of A) and the second term corresponds to the probability of unit B has not
failed up to t whenever Afails at <t (reliability of B). It is clear that the second term includes a
switching phenomena. Therefore, if the switching wEğeas not perfect then this term should be
multiplied by a perfect switching probability Ps.
 9

R(t )  e   At  Ps
A
B   A

e   At  e   B t 

1 Ps
m  MTTF   R(t ) dt  
0
A B

Let’s assume the case where there are two standby units B and C for an operating basic unit A.
Let the failure rates of the units be different than eachother. The probable operating phases are
given in Fig. 5.3.

A is operating or A has failed and the first

C is operating
standby unit B is operating.
0 d
ıl Birim Arızalı Birinci Yedek Sağlam 
t
B fails
Figure 5.3 Probable operating phases of a syetm wehere there are two standby units.

 A t
P0 (t )  e
t
  A (t  )  A   A t  B t 
P1 (t )   P0 ( )( Ad )e  e e
0
 B   A  
t
  C (t  )
P2 (t )   P1 ( )( B d )e
0
R(t )  P0 (t )  Ps P1 (t )  P2 (t )

m  MTTF   R(t )dt
0
Similarly, reliability of a standby redundant system comprising n standby units can be derived.
Let’s assume that the failure rate of the basic unit is 0 and failure rates of the standby units are
1 ,2 ,...,n.

P0 (t )  e 0 t
t
Pi (t )   Pi 1 ( )(i 1d )e  i (t  ) , i  1,2,..., n
0
n
R(t )  P0 (t )  Ps  Pi (t )
i 1

m  MTTF   R(t ) dt
0

So far, we have assumed that the failure probability during the idle phases of the standby units
were zero. If the standby units can fail during their idle phases, we have take into account its
probability. Assume a standby redundant system comprising a single standby unit. Let the failure
 10

rate of the main unit be A and the failure rates of the standby unit are B and By for the
operating and idle phases, respectively. The status of the standby unit is not important while
main unit is operating, and it should be taken into account when the basic unit has failed.

A operating and B has not B operaing

failed during its iddle phase d
0 t

A fails

Figure 5.4 Operating phases of the standby system when there is possibility of failing the standby
unit during its idle phase.

P0 (t )  et
dP1 ( )  P0 ( ).Py ( )( A.d ).eB (t  ) 
t
  By . A eB t  e( A  By )t 
P1 (t )   e A .e ( Ad )eB (t  )   
0
 A   B y  B
A e B t  e( A  By )t 
R(t )  P0 (t )  P1 (t )  eB t   
 A  B y   B

Failure rates of the basic unit and the standby unit can be the same in some applications. We can
use the Poisson distributşon for those cases. Assume that the basic unit and n standby units have
the same failure rate of  ,
P0 (t )  e t
t t
  (t  )  t  t
P1 (t )   P0 ( )(d )e  .e  d  t.e
0 0
t t
(t ) 2   t
P2 (t )   P1 ( )(d )e   (t  )  2e  t   .d  .e
0 0
2
t t
( ) 2   (t ) 3   t
P3 (t )   P2 ( )(d )e   (t  )   .e ..e   (t  ) d  .e
0 0
2 2.3
.......
( ) n 1  
t t
  (t  ) (t ) n   t
Pn (t )   Pn 1 ( )(d )e  .e ..e   (t  ) d  .e
0 0
( n  1)! n!

n  n
(t )i   t
R(t )  P0  Ps  Pi (t )  1  Ps  e
i 1 
 i 1 i! 
 n
(t )i   t 1  nPs
m  MTTF   1  Ps  e dt 
0  i 1 i!   
 11

In case of a single standby unit,

R(t )  P0  Ps P1 (t )  1  Ps t e t

1  nPs
m  MTTF   1  Ps t e  t dt 
0

The same expressions can also be obtained by taking the limit of the previous cases where the
failure rates were different.

R(t )  Lim
  A t
e  Ps
A
B   A

e A t  e B t   1  P t e
s
 t
 B  A    

1  Ps
m  MTTF   R(t )dt 
0


Example: A standby redundant system comprises one basic unit and one standby unit. Failure
rates of the units are  = 0.02 f/h. Calculate the reliability for t = 10 h and the MTTF of
the system if the perfect switching probability is 100%. Calculate the same values for
the parallel redundant system comprising the same units.
2
Standby redundant system: R(t )  (1  t ).e  t  R(10)  0.9825 , m   100 h

Parallel redundant system: R(t )  1  QA (t ).QB (t )  1  1  RA (t )  2e t  e 2t
2

R(10)  0.96714
2 1
m   75 h
 2
Actually, this result may be misleading since it is affected from the perfect switching probability.

Example: Calculate the same values if the perfect switching probability in the previous example
was Ps=0.98,0.94 and 0.9.

Ps  0.98  R(10)  0.97920 , m MTTF  99 h

Ps  0.94  R(10)  0.97265 , m MTTF  97 h
Ps  0.90  R(10)  0.96610 , m MTTF  95 h

The results show that the reliability is sensitive to perfect switching probability but MTTF is
slightly affected from it. In practical applications, there are n standby units for N operating
(basic) ones. If all the basic and the standby units are identical and have a constant failure rate of
, then,

n  n
(et )i   e t  n
( Nt )i   Nt
R(t )  P0 (t )  Ps  Pi (t )  1  Ps  e  1  Ps  e ; e  N .
i 1  i 1 i!    i 1 i! 
1  nPs
MTTF 
N
 12

Example: A system comprises 50 idetical units each having a constant failure rate of 0.001 f/h
and its success requires the operation of all units. Calculate the reliability of the system
for t=20 h and the MTTF for 0,1,2,3,4,5,6 equivalent standby redundant components.

For 0 standby units R(20 h) =0.367900 , MTTF = 20 hours

For 1 standby unit R(20 h) =0.735760 , MTTF = 40 hours
For 2 standby units R(20 h) =0.919700 , MTTF = 60 hours
For 3 standby units R(20 h) =0.981000 , MTTF = 80 hours
For 4 standby units R(20 h) =0.996340 , MTTF = 100 hours
For 5 standby units R(20 h) =0.999400 , MTTF = 120 hours
For 6 standby units R(20 h) =0.999917 , MTTF = 140 hours

It is clear that even a single standby unit doubles the reliability at t=20 h. Each additional
unit increases the reliability but the increasing rate decreases with increasing number of
standby units. However, MTTF is a linear function of the number of standby units. The
important question then “How many standby redundant units?”. The answer of this
question depends on the application. If reliability is very important regardless of its cost
then the number of standby units can be determined wrt desired reliability. However, if
MTTF is the basic criteria, the number of standby units can be chosen wrt the desired
MTTF values.

Example: Consider the following system where units 4,5 and 6 operate as a 2 out of 3 system.
Calculate the reliability for a mission of 100 hours and the MTTF of the system.

4 1  12.10 5 failure / h
2
5 2  20.10 5 failure / h
3
1 S 6 3  7  10 5 failure / h
4
4  5  10 4 failure / h
2
3 7 6  5.10 5 failure / h
Ps  0.98, Rs  1
 13

Let the standby system, parallel combination and 2/3 systems be represented by equivalent
units 8, 9 and 10 respectively.
1t 1 e 1t  e 2t 
R8 (t )  e  Ps
9 10 2  1  

8 4 R9 (t )  1  Q2 (t )Q3 (t )
 t  t (  )t
 e 2 e 3 e 2 3
3 7
R10 (t )  R4 (t ) R5 (t ) R6 (t )  R4 (t ) R5 (t )Q6 (t )
 R4 (t )Q5 (t ) R6 (t )  Q4 (t ) R5 (t ) R6 (t )

Minimal Paths : 8
910 , 8
37 , 8
947 , 83410

T1 T2 T3 T4
Minimal Cut Sets : 8 , 3
- 9 , 7
-
10 , 4-7
- 9 , 3
-
4 -
10

C1 C2 C3 C4 C5

4 3
Equivalent network 3 7
8 7 4
9 10
9 10

Q(t )  Q(C1  C2  C3  C4  C5 )
 Q(C1 )  Q(C2 )  Q(C3 )  Q(C4 )  Q(C5 )
 Q8 (t )  Q3 (t )Q9 (t )  Q7 (t )Q10 (t )  Q4 (t )Q7 (t )Q9 (t )  Q3 (t )Q4 (t )Q10 (t )
Q(100 h)  0.000355  (1.978  19.717  0.01968  0.1962 )  3.5513810  4
R(100 h)  0.999465

The following MTTF calculatşion is an approximation, because of several transformations.

1 Ps
m8    13233 .3 h
1 2
m2 m3
m9  m23  m2  m3   100238 .1h
m2  m3

1 2 1
m10  m456   R10 (t ) dt   2  10333 .3 h
0
4  5 4  6 4  5  6
 14

mC1  m8  13233 .3 h
m9 m3
mC2  m9  m3   150178 h
m9  m3
m m
mC3  m10  m7  10 7  100968 h
m10  m7
1  1 1 1 1 1  1
mC 4         150329 h
9 4 7  9  4 9  7 4  7  9  4  7
1 1 1  1 1 1  1
mC5         101631 h
3 4 10  3  4 3  10 4  10  3  4  10
1
m  9206 hours
1 1 1 1 1
   
mC1 mC 2 mC3 mC 4 mC5

5.5 Wearout and Component reliability

We have assumed that the failure rate was constant (unit was operating during its useful life) so
far. However, in practice the units may fail long before their MTTF values due to ageing.
Therefore, calculations with a MTTF value which was determined for a constant failure rate may
not be realistic.
Failure density function at wear-out region can be represented by a normal distribution with an
average value of M. The shapes of the failure density function, failure/survivor probability and
the failure rate are given in Figure 5.5

( tM )2
1 
f (t )  e 2 2
, M : Mean wear - out time,  : Standart deviation
2 
 15

(a) (b)
0.399 f(t) 1.000
 0.841 Q(t)

0.500

0.159
R(t)
M t M M M t

(c)
t

f(t)
(t) = R(t)
0.798/ 

M t

Figure 5.5 Reliability functions at wear-our region

a) Failure density function
b) Falure distribution (failure probability) and reliability functions
c) Failure rate

If the unit is operating at its wear-out region, ageing failures must be taken into account in
addition to its chance failures. The reliability of a unit operating at its wear-out region and has
not failed up to T is,

T t  T t 

 fW (t )dt  fW (t )dt   fW (t )dt  fW (t )dt RW (T  t )

RW (t T )  1  QW (t T )  1  T  T T  T t 
   RW (T )
 fW (t )dt  fW (t )dt  fW (t )dt
T T T
 16

If the chance failures are also taken into account,

RW (T  t ) t
RB (t )  RW (t T ).RC (t )  e
RW (T )
For a particular case if T=0; that is if wearout region starts from initial time, then,

RW (T  t )  t
RB (t )  Lim e  RW (t )et
T 0 RW (T )

The reliability in the wearout region depends on the relative values of m and M. Fig. 5.6 shows
the reliability for two cases; namely for m>M and for m<M. It is clear that chance failures
dominate the reliability of the system op to T and wearout failures are not effective. Wearout of
the unit should be taken into account after T. If m<M, chance failures are effective up to t=m and
the weraout is effective beyond this time. However, wearout process starts much before m if
m>M. Therefore, an approporiate maintenance is required to increase the average maintenance
time M.

1.0 Rw(tT) 1.0

Rw(tT)
Rc(t)
RB(t)=Rw(tT).Rc(t) Rc(t)
0.5 0.5 RB(t)=Rw(tT).Rc(t)
0.368 0.368

t t
0 T M m m T M

Fig. 5.6 Reliability of a unit including the wearout region

Example: Failure rate of a pump during its useful life period is 100 failure per 106 hours. Mean
wearout time of the pump is 10000 h with a standard deviation of 2000 hr. Evaulate
the followings, assuming that the wearout failure density function is Gaussian.
a) A 100 h mission starting at 9900 h period of its life cycle, and
b) A similar mission starting 1000 h later.

1 100 f
m  MTTF    10  4 f / h
 6
10 h
 17

a) Rc (t )  e t  Rc (100 h)  e 100*0.0001  0.99005

9900  10000
M  10000 ,   2000 , T  9900 h   0.05  RW (T )  0.5199
2000 Pr ob. table
10000  10000
T  t  10000 h   0  RW (T  t )  0.5
2000 Pr ob. table
RW (T  t )
RW (t T )   0.961723  RB (t )  RW (t T ).Rc (t )  0.952154
RW (T )

b) Rc (t )  e t  Rc (100 h)  e 100*0.0001  0.99005

10900  10000
M  10000 ,   2000 , T  10900 h   0.45 RW (T )  0.3264
2000 table
10100  10000
T  t  10100 h   0.5  RW (T  t )  0.3085
2000 table
R (T  t )
RW (t T )  W  0.94516  RB (t )  RW (t T ).Rc (t )  0.935755
RW (T )