UNIT 15

BASIC FUNCTIONS OF
RELIABILITY

Structure
15.1 Introduction Relationship between Functions
R(t), F(t), f(t) and λ(t)
Excepted Learning Outcomes
Mean Time to Failure and Median of
15.2 Definition of Reliability
the Random Variable T
15.3 Basic Functions in
15.4 Bathtub Curve
Reliability and their
Relationship 15.5 Estimation of Reliability
Functions from Failure Data
Reliability Function
15.6 Summary
Cumulative Failure Distribution
Function 15.7 Terminal Questions
Failure Density Function 15.8 Solutions/Answers
Hazard Rate

15.1 INTRODUCTION
Due to the rapid advancement of science and technology and to meet the
ever-increasing needs of society, industries are increasingly introducing
automation for producing goods ranging from the simplest to highly complex
systems. However, all processes and products are prone to failure. For The science which
example, a television remote control may stop functioning or a malfunction deals with the
may occur in any other household appliance; an automobile engine starter concepts used in
the insurance
may fail, an aeroplane may crash due to the failure of some of its industry is known
component(s), credit card transactions may fail, and so on. You can think of as actuarial
science. That is
many more similar examples from your daily life. So we would like to why many concepts
understand failures, present them in a different manner and see how we can of reliability and
actuarial science
try to prevent interruption due to failure! Reliability is nothing but a science of are the same
better presentation of failure. For example, the statement that 20% of students except the
terminology used
of the class failed the test can be better presented as 80% of students of the there.
class passed the test!
The mathematical foundations of reliability were laid down when queueing
theory and renewal theory were developed in the early 1930s and 1940s.
However, reliability as a subject gained importance after World War II because 151
* Dr. Rajesh, School of Sciences, IGNOU, New Delhi
Block 4 Reliability Theory

a high failure rate of electronic equipment was observed during the war. The
first journal on the subject, Institute of Electrical and Electronics Engineers
(IEEE) –Transactions on Reliability, came out in 1963. A number of textbooks
on the subject were published in the 1960s.
We give a formal definition of reliability in Sec. 15.2. Four basic functions in
reliability, namely, the reliability function, cumulative failure distribution
function, failure density function and the hazard rate and their relations are
described in Sec. 15.3. We also explain the terms mean time to failure and
median of time to failure in Sec. 15.3. When a population of identical
components is tested, the curve of the relative failure rate of the entire
population of the components without replacement is generally of the shape of
a bathtub. This curve, called the bathtub curve, is discussed in Sec. 15.4. We
explain how the basic reliability functions can be estimated from failure data in
Sec. 15.5.
In the next unit, we shall discuss the reliability evaluation of simple systems.

Expected Learning Outcomes

After studying this unit, you should be able to:
 define reliability;
 explain the basic functions in reliability, namely, reliability function,
cumulative failure distribution function, failure density function and
hazard rate;
 explain three phases of the bathtub curve; and
 estimate basic reliability functions from complete failure data.

15.2 DEFINITION OF RELIABILITY

In articles on reliability, many authors suggest that the word ‘reliability’ was
first coined by the poet Samuel Taylor Coleridge when he wrote a piece in
praise of his friend, the poet Robert Southey in 1816. However, reliability as a
subject of study was taken up after World War II when a high failure rate of
electronic equipment used in the war was observed. Today it has taken a new
shape by blending itself in all phases of the product life cycle from proposal
and design to manufacturing. So in the literature on reliability, we find that the
definitions of reliability vary from one practitioner or researcher to another.
Still, a comprehensive and widely accepted definition of reliability may be
stated as follows:

Definition of Reliability: Reliability of a component/device/equipment/unit/

system is the probability that it performs its intended function adequately
for a specified period of time under the given operating conditions.
To understand what this definition actually means, we have to clearly
understand the meaning of the above-mentioned bold words, which are
keywords in this definition. Let us explain what we mean by these keywords:

Probability: When a number of identical components operate under similar

conditions, the failure times of components generally vary from component to
152 component and we cannot predict their failure times exactly in advance. But
Unit 15 Basic Functions of Reliability

we can describe the phenomenon of failure in probabilistic terms. For

example, suppose we put N0 identical components in operation at time t = 0
under identical conditions. Suppose after t units of time, we have
Ns (t) = number of components that are operating at time t

Nf (t) = number of components that have failed at time t

Then reliability R(t) at time t is defined as

Ns (t)
R(t) = … (1)
N0

Thus, the reliability of a component/system at time t is the probability that it

performs its function without failure. Being a probability, its value lies between
0 and 1. From equation (1), we can see that
if Ns (t) = 0, R(t) = 0 and

if Ns (t) = N0 , R(t) = 1.

By definition Ns (t) satisfies the following relation:

0 ≤ Ns (t) ≤ N0 for any time interval [0,t]

0 Ns (t) N0
⇒
N0
≤
N0
≤
N0
[on dividing by N0 ]

⇒ 0 ≤ R(t) ≤ 1

What does this mean? Suppose that the reliability of a refrigerator of a

particular company at 10 years is R(t) = 0.9452. We can interpret this
statement as follows: Out of 10000 refrigerators of this particular company,
approximately 9452 refrigerators worked without failure for 10 years.

Intended Function: The intended function of a product must be described in

unambiguous terms early on in the design process so that the expected
requirements of the customers can be ensured/addressed. Thus, the intended
function of a product is the function/work/job that it is expected to perform
when we put it in operation under stated conditions. For example, if a pump is
designed to deliver at least 300 gallons of water per minute, the intended
function of the pump is to deliver 300 gallons or more water per minute.

Adequately: A product is said to perform adequately if it is performing its

intended function as expected. For example, if the intended function of a water
pump is to deliver 300 gallons or more water and it is delivering 300 gallons or
more water per minute, it means that it is performing adequately. But if the
pump delivers less than 300 gallons of water per minute we say that it has
failed or is not performing adequately.

Time: You should always keep in mind that reliability can be meaningful only if
it is related to a time interval or period of time. For example, we can say that
the reliability of a component is 0.98 for a mission time of 100 hours. But the
statement ‘reliability of a component is 0.95’ is meaningless since the time
interval is unknown. From equation (1), reliability at time ‘t’ can be defined as
follows: 153
Block 4 Reliability Theory
Number of components performing intended function at time ‘t’
R(t) = ... ( 2 )
Number of component at start (i.e., when ' t ' = 0)

Number of components performing the intended function at a time ‘t’

In the above definition, time ‘t’ means the interval [0,t].
In equation (2), R(0) = 1 since no component fails at t = 0, and R(t → ∞ ) =0
since no component can survive forever. This point will become clearer to you
in the next section when we discuss the wear-out period of a component.
Given Operating Condition: By given operating condition, we mean that the
performance of the product should be observed under normal stated
conditions in which it is expected to perform. The environmental conditions
(such as temperature, humidity, shock, vibration, altitude, etc.), design loads
(such as voltages, pressure, etc.) and operating conditions (such as
maintenance, storage, etc.) affect the reliability of a product. So a product will
perform better in the field if its design takes into account and is representative
of how it is actually used by the customers. For example, a car which is
designed for smooth roads will not perform well if a customer uses it on rough
roads.
You may like to try the following Self Assessment Questions to check your
understanding of the definition of reliability.

SAQ 1
Which of the following is not a part of the definition of reliability?
(a) Probability (b) Intended function
(c) Time (d) Certainty

15.3 BASIC FUNCTIONS IN RELIABILITY AND

THEIR RELATIONSHIP
The goal of this section is to define the following basic functions and explain
the relationship between them:
• Reliability function,
• Cumulative failure distribution function,
• Failure density function, and
• Hazard rate.
In this section, we also obtain expressions for the mean and median of the
random variable called the time to failure or failure time.
Let us explain these functions one at a time.
15.3.1 Reliability Function
In Sec. 15.2, you have learnt that reliability is a probability. If reliability is a
probability, then a random variable has to be associated with it because
probability is always associated with a random variable. You are familiar with
the concept of a random variable as it has been discussed in detail in the
154 course MST-012. In MST-012, you have also studied some discrete as well as
 Unit 15 Basic Functions of Reliability

continuous probability distributions. Now, can you guess what the random
variable in the case of reliability is? Think for a while.
We hope that you have given this a thought and arrived at the following idea:
Generally, the random variable in the case of reliability is time – the time in
which a component/system fails. Let this random variable, called time to
failure or failure time of the component/system, be denoted by T. Then, by
definition of reliability [R(t)], the component/system at time t is given by
R(t) = Probability that the time to failure of the component/system is > t

Probability of any event A 

= P [T > t ] ≥ 0 always satisfies 0 ≤ P(A) ≤ 1 … (3)
 
The reliability of a component/system when expressed in terms of time t is
known as reliability function and is denoted by R(t).
Let us consider two particular cases of the reliability function:
When t 0 and when t → ∞.
=
 we assume that at t =0, the component/ 
R(0)= P [ T > 0]= 1 system is performingits intended function. 
 
no component / system canperform 
∞ ) P [ T > t(t → ∞
R(t → = = )] 0  
itsintended function forever. 
Therefore, we can say that the probability of survival of a component
decreases as the lifetime of the component increases. Ultimately this
probability will approach zero since no component can perform its intended
function forever. A typical shape of the reliability function is shown in Fig. 15.1.
From Fig. 15.1, you can see that the reliability function is a non-increasing
function in time (t).

Fig. 15.1: Typical shape of reliability function R(t).

15.3.2 Cumulative Failure Distribution Function

Let us define the function F(t) by
F(t) = 1 – R(t) … (4)

∴ F(0) =
1 − R(0) = 1 – 1 = 0
155
Block 4 Reliability Theory
and F(t → ∞ ) = 1 − R(t → ∞ ) = 1 − 0 = 1

Using equation (3), we can write equation (4) as

F(t) =
1 − R(t) =
1 − P[X > t]

 You know that events P[X > t] and [X ≤ t] 

= P[X ≤ t] are complementary to each other. 
 
Hence, P[X > t] + P[X ≤ t] =
1. 

or F(t)
= P[T ≤ t] … (5)
Also, as shown above F(0)
= 0, F(∞
= ) 1 … (6)

Study equations (5) and (6). Can you identify a probability distribution function
given in course MST-012, which has similar characteristics? Recall that the
cumulative distribution function has these characteristics. But in reliability
terminology, this distribution function is known as cumulative failure
distribution function and gives unreliability of the component up to time t.
So if we want to calculate the probability of failure of a component at time t
(known as unreliability of the component), then we have to simply obtain the
value of the function F(t). Therefore, if we denote the reliability and unreliability
of a component by R and Q, respectively, then they satisfy the relation:
R+Q=1 … (7)
Note that the probability of failure of a component increases after its useful life
period is over. Ultimately this probability will approach 1. A typical shape of
cumulative failure distribution is shown in Fig. 15.2.

Fig. 15.2: Typical shape of cumulative failure distribution.

15.3.3 Failure Density Function

You have learnt in course MST-012 that the derivative of the cumulative
distribution function of a random variable is known as the probability density
function (pdf) of the random variable. But in the terminology of reliability, the
probability density function is known as failure density function (fdf). As
explained in Sec. 15.3.2, the cumulative distribution function in reliability
terminology is the cumulative failure distribution function. So if f(t) denotes
156 failure density function (lifetime density function), then
Unit 15 Basic Functions of Reliability
d
f(t) =
dt
[F(t)] … (8)

From course MST-012, recall the following relationship exists between f(t) and
F(t):
t
F(t) = ∫ f(t)dt
−∞
… (9)

Since t represents time therefore t ≥ 0, we can write equation (9) as

t
F(t) = ∫ f(t)dt
0

From equation (4), we have

t
R(t) = 1 − ∫ f(t)dt
1 − F(t) = … (10)
0

∞
Since f(t) is the pdf of the random variable T, therefore ∫ f(t) dt = 1, i.e., the sum
0

of all probabilities of a random variable is always equal to 1. So we can

∞
replace 1 in equation (10) by the integral ∫ f(t) dt and write
0

∞ t
R (t)
= ∫ f(t) dt − ∫ f(t) dt
0 0

t ∞
 t  We know that 
R(t) =  ∫ f(t)dt + ∫ f(t)dt  − ∫ f(t)dt b c b

0 t  0  ∫ f(x)dx
= ∫ f(x)dx + ∫ f(x)dx,a < c < b 
a a c 
∞
or R(t) = ∫ f(t) dt
t

Thus, we have
t
F(t) = ∫ f(t) dt … (11)
0

and
∞
R(t) = ∫ f(t) dt … (12)
t

If the use of t within the integral sign as well as in the limits of the integral
creates confusion for you, then the variable t inside the integral sign can be
changed to any other dummy variable, say u. Then we can write F(t) and R(t)
as:
t
F(t) = ∫ f(u) du … (13)
0

and

∞
R(t) = ∫ f(u) du … (14)
t
157
Block 4 Reliability Theory

These functions may be interpreted geometrically as areas under the curve of

f(t) as shown in Fig 1.3.

Fig. 15.3: Typical shape of failure density function and geometrical interpretation of F(t)
and R(t).

15.3.4 Hazard Rate

The instantaneous rate of failure at time t is known as the hazard rate and is
generally denoted by λ(t) .

To define it mathematically, let T denotes the time to failure of the

component/system. Let us consider the time interval [t,t + Δt]. Then, we have

 P [a ≤ X ≤ b]= F(b) − F(a) as you

P [ t ≤ T ≤ t + Δt ]= F(t + Δt) − F(t)  
have learnt in Unit 3 of MST-012 
= [1 − R(t + Δt)] − [1 − R(t)] [R(t) + F(t) = 1]
= R(t) − R(t + Δt)
The conditional probability of failure in the time interval [t,t + Δt] given that the
component/system has performed its intended function till time t is given by
P [ t ≤ T ≤ t + Δt ]
P  t ≤ T ≤ t + Δt T > t  =
P [T > t ]

 P(A ∩ B) 
 P ( A B ) = P(B) . If you want torefresh your 
 
memory, you may refer to Unit1 of MST-012. 
Using equation (12) and R(t)
= P[T > t], we get
R(t) − R(t + Δt)
P  t ≤ T ≤ t + Δt T > t  =
R (t)

Then the conditional probability of failure per unit of time is given by

R(t) − R(t + Δt)

,
R(t)Δt
and instantaneous rate of failure at time t (i.e., hazard rate λ(t) ) is given by

R(t) − R(t + Δt) 1 R(t + Δt) − R(t)

λ(t) = lim = − lim
158 Δt →0 R(t)Δt R(t) Δt → 0 Δt
Unit 15 Basic Functions of Reliability
1 d
= −
R(t) dt
[R(t)]
d d d
Since f(t) =
dt
[F(t)] =−
dt
[1 R(t)] =
− [R(t)] ,
dt
d
−
dt
[R(t)] f(t)
λ (t) = = … (15)
R(t) R(t)

15.3.5 Relationship between Functions R(t), F(t), f(t)

and λ(t)
So far we have defined the functions R(t), F(t), f(t) and λ(t). In this section,
you will learn how these functions can be obtained from the failure data
collected by testing a fixed number of identical components. And finally, we
form a table showing the relationship between them, which will be very useful
when we solve numerical problems.
Suppose that initially a fixed number (N0 ) of identical components are put to
test. Let
Ns (t) = number of components that are performing the intended function
adequately at time t
Nf (t) = number of components that have failed at time t

where Ns (t) + Nf (t) =

N0

Now, from equation (1), we know that reliability at time t is given by

Ns (t) N0 − Nf (t)
R(t) = = [ Ns (t) + Nf (t) =
N0 ] … (16)
N0 N0
Or
Nf (t)
R(t)= 1 −
N0

Differentiating R(t) with respect to t, we get

d 1 d
dt
[R(t)] = −
N0 dt
[Nf (t)] … (17)

Also, from equations (8) and (4), we know that

d d d
f(t) =
dt
[F(t)
= ] dt
[1 − R(t)] = − [R(t)]
dt
… (18)

 1 d 
=−  − [Nf (t)]  [Using equation (17)]
 N0 dt 
or
1 d
f(t) =
N0 dt
[Nf (t)] … (19)

By definition, hazard rate is given by

number of failuresper unit time at time t
λ (t) =
number of components which are exposed to failure at time t
159
Block 4 Reliability Theory
dy
Since derivative denotes the rate of change of y with respect to x,
dx
therefore, we can write the numerator term of the above expression as follows:
d
dt
[Nf (t)]
λ(t) = … (20)
Ns (t)
or
N0 1 d
λ (t) =
Ns (t) N0 dt
[Nf (t)] [on multiplying and dividing by N0]
 Ns (t) 
=
1
f(t) R(t) = N from equation 
R(t)  0 
(1)andusingequation(18) 
or
f(t)
λ (t) = … (21)
R(t)
1 d
= −
R(t) dt
[R(t)] [using equation (18)]

1 d
∴ λ(t) =
−
R(t) dt
[R(t)] … (22)

Integrating λ ( t ) with respect to t on both sides within the limits t = 0 to t = t,

we get
d
t
1 d
t t
dt
[R(t)]
∫0 λ(t)dt =
−∫
R(t) dt
[R(t)] dt =
−∫
R(t)
dt
0 0

 d 
t  b dx [ f(x)] b
= − ln R(t)  0  ∫ dx = ln f(x)  a 
 a f(x) 
 

− ln R(t) − ln R(0) 

=

− ln R(t) − ln1
= [R(0) = 1]

 ln(1) 0 andR(t) ≥ 0 ∀ t,

=
− ln [R(t)] − 0
=  
∴ R(t) =R(t) 
= − lnR(t)
t
⇒ − ∫ λ(t)dt =lnR(t)
0

or
 t 
= exp  − ∫ λ(t)dt 
R(t)  ln a = m ⇒ a = em  … (23)
 0 
Note that if λ ( t ) is constant and λ ( t ) = λ, then equation (23) gives

160
R(t) = e− λt
Unit 15 Basic Functions of Reliability
We now put together the relations between R(t), F(t), f(t) and λ(t) in
Table 15.1 as follows:
Table 15.1: Relationship between the Functions R(t), F(t), f(t) and λ(t)

R(t) F(t) f(t) λ(t)

∞
 t 
R(t) – 1 – F(t) ∫ f(t)dt
t
exp  − ∫ λ(t)dt 
 0 
t
 t 
F(t) 1 – R(t) – ∫ f(t)dt
0
1 − exp  − ∫ λ(t)dt 
 0 

d d  t 
f(t) − [R(t)] [F(t)] – λ(t) exp  − ∫ λ(t)dt 
dt dt  0 

d f(t)
λ(t)
d
− [lnR(t)] dt
[F(t)] ∞
–
dt
1 − F(t) ∫ f(t)dt
t

After studying the relations between R(t), F(t), f(t) and λ(t) , let us discuss next
function of the reliability.

15.3.6 Mean Time to Failure and Median of the Random

Variable T

In reliability terminology, the mean of the random variable T in the absence of

repair and replacement is known as the mean time to failure (MTTF).
Usually, the users of a product/component are interested in knowing the
average life of the product/component rather than the complete failure details.
MTTF is a measure which simply gives a number (in units of time in which the
life of the product/component is measured) that tells us on average how long
the product/component performs its intended function successfully. It is
calculated by taking the mean of the lifetimes obtained on the basis of the
results of a sample of such identical products/components tested under stated
conditions. For example, suppose we put 10 identical components to test
under the stated conditions. Let t i ,(i = 1,2,...,10) denote the time for which the
ith component performs its intended function successfully. The results of the
sample are shown in Fig. 15.4. Thus, the mean time to failure (MTTF) of such
components on the basis of the results of this sample is given by

1 10 1
MTTF
= ∑
10 i=1
=ti
10
(5 + 2 + 11 + 6 + 9 + 1 + 5 + 4 + 6 + 1)

50
= = 5hr
10
In general, if n components are put to the test and t i ,(i = 1,2,3,...,n) denote the
time for which the ith component performs its intended function, then the mean
time to failure (MTTF) of such components is given by

1 n
MTTF = ∑ ti
n i=1
161
Block 4 Reliability Theory

Fig. 15.4: Times for which each of the 10 components operates successfully.

However, if we are given the basic functions of a component then by definition,

MTTF is given by
∞
MTTF
= E(T)
= ∫ t f(t) dt
0
[refer to Unit 5 of MST-012] ...(24)

∞
 d   d 
∫0 t  − dt R(t)  dt
=

− R(t) from equation(16a)
 f(t) =dt 
I II

Integrating by parts, we get

∞ ∞
t [ −R(t)] 0 − ∫ (1) [ −R(t) ] dt tR(t) t →∞ − 0 + ∫ R(t)dt
∞
MTTF = =
0 0

∫
− λ(t )dt
n −x
Since=
R(t) e 0
and lim x=e 0 for n > −1,
x →∞

∞
MTTF = ∫ R(t)dt … (25)
0

To calculate MTTF, you should use equation (25) instead of equation (24),
because, in general, dealing with equation (25) is easier.
Note 1: If the failure rate is constant, i.e., λ(t) = λ, we have
t

∫
− λdt
R
= ( t ) e=
0
e− λt

∞ ∞ ∞
 e− λt  1 1
∫ R(t)dt =∫ e
− λt dt
∴ MTTF = =  =− (0 − 1) =
0 0  −λ  0 λ λ

1 1
i.e. MTTF = or Constant failure rate =
162
constant failure rate MTTF
Unit 15 Basic Functions of Reliability

Now, to obtain an expression for median of the random variable T, we have to

recall the definition of median. The median is the value of the variate which
divides the distribution of the variate into two equal parts. Therefore, if tmd
denotes the median of the random variable T, then
R(t md ) = P [ T > t md ] = 0.5
… (26)
Let us now consider a few examples to apply the concepts discussed in this
section to numerical problems.
Example 1: The Failure density function of the random variable T is given by
0.012 e −0.012t, t≥0
f(t) = 
0, otherwise

Calculate
(i) Reliability of the component.
(ii) Reliability of the component for a 100 hour mission time.
(iii) Mean time to failure (MTTF).
(iv) Median of the random variable T.
(v) What is the life of the component if the reliability of 0.96 is desired?
Solution: Failure density function of the random variable T is given as
0.012 e −0.012t, t≥0
f(t) = 
0, otherwise

(i) Since f(t) is given so we can compute the reliability using equation (12)
as
∞ ∞

∫= ∫ 0.012e
−0.012t
R(t)
= f(t)dt dt
t t

 0.012e −0.012t 
∞
 b  ekt  
b

=   ∫ e dt   
kt

 −0.012  t  a  k  a 

= − 0 − e0.012t   e

−0.012t
→ 0 as t → ∞ 

= e−0.012t … (27)
(ii) The reliability of the component for a 100 hour mission time is obtained
by replacing t by 100 in equation (27):
= e−0.012=
R(100) ×100 −1.2
e= 0.3012

(iii) We know from equation (22) that

∞ ∞ ∞
 e−0.012t 
MTTF ∫=
= R(t)dt ∫ e =dt  −0.012t

0 0  −0.012  0

1 1
=− [0 − 1] = =83.3333.
0.012 0.012

(iv) Let t md denote the median time to failure. Then from equation (26), we
have 163
Block 4 Reliability Theory
R(t md ) = P [ T > t md ] = 0.5
∞
⇒
tmd
∫ f(t)dt =
0.5

⇒ e −0.012tmd =
0.5 [replacing t by t md in (27)]

Taking natural logarithm on both sides

−0.012t md =
ln(0.5) ⇒ −0.012t md =
−0.6931

0.6931
⇒ t=
md = 57.7583hr
0.012
(v) The life of the component for a reliability of 0.96 can be obtained by
solving (27) for t and replacing R(t) by 0.96 as follows:
0.96 = e−0.012t
Taking natural logarithm on both sides, we get
0.0408
ln(0.96) =
−0.012t ⇒ −0.0408 =
−0.012t ⇒ t = = 3.4hr
0.012
Example 2: The hazard rate of a component is given by λ (=
t ) 0.6 × 10−6 t,
where t is in years.
Calculate the reliability of the component for the first 2 years.
Solution: It is given that the hazard rate of a component is λ(t)
= 0.6 × 10 −6 t .

We know from equation (22) that

 1   t    −0.6 × 10 −6 t 2  t 
R(t) = exp  − ∫ λ(t)dt  = exp  − ∫ 0.6 × 10 t dt  = exp  
−6
 
 0   0   2 0 


( )
= exp −0.3 × 10 −6 t 2 − 0 = exp( −0.3 × 10 −6 t 2 )

Now reliability of the component for the first two years is given by just
replacing t by 2. Therefore.
−6 2 −6
−0.3×10 ×2
R(2) e=
= e−1.2×10

Example 3: Show that if the hazard rate of a component is constant, say λ,

then the failure distribution of the component follows an exponential
distribution. Also, show that MTTF = 1/ λ .
Solution: We are given that the hazard rate of a component is constant = λ .
From equation (23), we have
t t

∫ − ∫ λdt
− λ(t )dt
R(t) e=
= 0
e 0 [ λ(t)= λ= constant]

t
−[ λ t ]
λ(t − 0)
= e=0
e−= e− λt

Using equation (4), we have

F(t) = 1 − e− λt
1 − R(t) =

164 which is nothing but the cumulative distribution function of an exponential

Unit 15 Basic Functions of Reliability

distribution. Hence, the failure distribution of the component follows an

exponential distribution.
Now we calculate mean time to failure (MTTF) of the component. From
equation (25),
∞ ∞
 e− λt 
∞
 b  ekt  
b

= ∫=
MTTF R(t)dt ∫=e dt  
− λt
∫ e dt  k  
 = kt

0 0  −λ  0  a a

1
=−
λ
[0 − 1] as t → ∞,e − λt → 0 

1
=
λ
1 1
Hence MTTF= =
λ constant hazard rate

Note 2: Thus we see that if hazard rate = constant = λ = then

1 1
MTTF
= = or λ
λ MTTF
Now, you may like to solve the following Self Assessment Question to obtain
MTTF.

SAQ 2
Calculate MTTF using the information given in Example 2.

15.4 BATHTUB CURVE

In Sec. 15.5.4, we have discussed the hazard rate of a component. Consider a
population of new identical components and suppose that all components are
operative at some point in time, say, t = 0. In general, the curve of the relative
failure rate of the entire population of the components without replacement has
the typical shape of a bathtub. Therefore, it is known as the bathtub curve as
shown in Fig. 15.5. The bathtub curve is nothing but a graphical representation
of the failure rate of a population of identical components versus time.
The shape of this curve has three different phases, which are explained as
follows:
Early Failure or Infant Mortality
When a new identical component starts to perform its intended function, a high
failure rate is observed in the initial stages due to many factors such as
manufacturing errors, poor quality or substandard items, incorrect adjustment
or positioning, bad assembly, human factors, improper design, etc. This phase
of the bathtub curve is known as the period of early failure or infant
mortality. It is also known as the burn-in or decreasing failure rate or
debugging period. Most failures in this period are due to manufacturing errors,
design problems or poor quality for which the manufacturer may be
responsible. So this period is generally covered by a warranty period by the
manufacturer. The duration of this period varies from component to
component. However, it is typically for the period of time till the failure rate
decreases.
The hazard rate in this phase can be reduced by increasing quality control at 165
Block 4 Reliability Theory

the production level (you have studied quality control in detail in the first eight
units of this course). However, even increased quality control at the production
stage cannot completely eliminate infant mortality. Therefore, components
should be tested at the factory before delivering them to the customers. That is
why good companies generally test the components before supplying them to
users.
Normal Life, Useful Life or Chance Failure Period
Most components having manufacturing errors, improper design, bad
assembly, etc. have failed in the early failure period. So for a long period of
time (length of the period may vary from component to component), fewer
failures are reported and the failure rate remains almost at a constant level.
This phase of the life of the component is known as the normal life, useful
life or chance failure period as shown in Fig. 15.5.
Usually, failures in this period are due to stress-related causes, random
fluctuations, etc. But these failures cannot be predicted exactly and happen
randomly. That is why this phase is known as the chance failure period. The
appropriate distribution used in this phase is the exponential distribution as
shown in Example 3. We know that in the case of a constant failure rate, the
failure distribution of the component follows an exponential distribution.
Wear-out Failures
Since no component is perfect, it cannot last forever. So after the span of
useful life, the failure rate of components starts increasing due to the ageing of
the components. This phase of the bathtub curve when the components begin
to deteriorate is known as the period of wear-out failures as shown in
Fig.15.5. In the wear-out phase, the failure rate increases with time. So if we
wish to minimise this interruption in the smooth working of a real-life system,
one of the strategies which can be helpful is the replacement of the
component when it reaches mean time to failure of its life. The probability
distributions generally used to study the failure characteristics in this phase
are normal (due to the shape of the failure curve in this phase shown in Fig.
15.5), Weibull and Gamma (due to the presence of shape parameter in these
distributions).

166 Fig. 15.5: Bathtub curve.

Unit 15 Basic Functions of Reliability

Now, you can try the following Self Assessment Questions based on the
concept of the bathtub curve.

SAQ 3
In different phases of the bathtub curve, which one is the phase of increasing
failure rate? Choose the most appropriate alternative.
(a) Infant mortality (b) Useful life (c) Wear out (d) Early failures

15.5 ESTIMATION OF RELIABILITY FUNCTIONS

FROM FAILURE DATA
In Sec. 15.3, we have discussed the basic reliability functions, namely,
reliability, cumulative failure distribution, failure density and hazard rate. It is a
real challenge for the people working in the field of reliability to estimate the
basic reliability functions from the failure data of a sample of identical
components obtained either from test generated failures or collected from the
actual field of operation of these components.
The following two approaches are generally used for this purpose:
(i) Fitting a theoretical distribution, which is appropriate to the failure data.
(ii) Deriving a function for hazard rate directly from the failure data and
consequently other reliability functions using the relations mentioned in
Table 15.1.
Here we follow the second approach as the first approach is beyond the scope
of the course. Further, the choice of the method in the second approach also
varies depending upon whether we have complete failure data or censored
(incomplete) failure data. Let us first explain what we mean by complete and
censored failure data.
Complete and Censored Failure Data: Suppose a sample of n identical
components numbered from 1 to n is put to the test. We record the lifetime of
the components that fail. If the test is allowed to run until the failure of all n
components, we shall have a record of the lifetimes of all n components. The
data thus obtained is known as complete failure data. But if the test is not
allowed to run until the failure of all n components due to some reason, we
shall not have a record of the lifetimes of all n components. The data thus
obtained is known as censored failure data.
Here we discuss only the case of complete failure data as censored failure
data is beyond the scope of the course. Let us explain the procedure of
estimating the basic reliability functions with the help of the following example.
The data considered in this example is purely hypothetical.
Example 4: The failure data for 1000 electronic components is shown in
Table 15.2.
Table 15.2: Failure Data of 1000 Electronic Components

Operating Time (in hours) 0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80
Number of Failures 167 116 98 76 65 56 44 36
Operating Time (in hours) 80-90 90-100 100-110 110-120 120-130 130-140 140-150
Number of Failures 48 64 77 61 46 27 19 167
Block 4 Reliability Theory

Estimate and plot the reliability, cumulative failure distribution, failure density
and failure rate functions.
Solution: In this example, the number of components in the sample is large.
So instead of recording the time to failure of each component individually, it is
a common practice to record the number of failures in some suitable intervals
of time. In this example, the interval of time is taken as 10 hours. The estimate
of reliability, cumulative failure distribution, failure density and failure/hazard
rate functions are given in columns 5, 6, 7 and 8 of Table 15.3. The
calculations for different columns are explained below:
Calculations of Columns 1 and 2
In column 1, we simply enter time in hours starting from the lower bound of the
first-class interval up to the upper bound of the last-class interval for the data
given in Table 15.2. Entries in the second column represent the number of
failures observed during the test in each interval. The number of failures
(frequencies) listed against each class interval of time in Table 15.2 is the sum
of the numbers of all components that fail during that time interval. In the
second column of the table, this frequency is entered against the upper bound
of the class interval during which it is observed.
Calculation of Column 3
Entries in this column simply represent the cumulative frequencies of the
frequencies written in column 2. You are familiar with the concept of
cumulative frequencies as it has been explained and used by you a number of
times in other courses of this programme. If Nf (t) denotes the number of
components that have failed by time t, then the first three entries of the column
are calculated as follows:
First entry Nf (0)= 0= same as the first entry of column 2
Second entry Nf (10) =+
0 167 =
167 =sum of the first two entries of column 2
Third entry Nf (20) = 167 + 116 = 283 = sum of the first three entries of column
2
You can calculate the remaining cumulative frequencies as explained above.
Calculation of Column 4
Entries of column 4 represent the number of components that are performing
their intended function adequately at time t. If Ns (t) denotes the number of
components that are performing their intended function at time t, then
initially we started with 1000 
Ns (0) = 1000 components and it is assumed that 
 
no component fails at time t = 0 
167 components have failed within the 
Ns (10) = 1000 − 167 = 833  first 10 hours of their operational life 
 
out of theremaining 833 components,
Ns (20) = 833 − 116 = 717 116 have failed within the next 
 
10 hours of their operational life 

168 The other entries in this column are calculated in the same way.
Unit 15 Basic Functions of Reliability

Calculations of Column 5
Entries in column 5 represent values of reliabilities at different points in time. If
R(t) denotes the reliability at time t, then by definition [equation (16)]
 where N0 = the size of the sample that 
N (t)  
R(t) = s  we start with and Ns (t)is the same as 
N0
explained in calculations of column 4 

The first three entries of this column are calculated as follows:

Ns (0) 1000 Ns (10) 833
R(0)
= = = 1, R(10)
= = = 0.833
N0 1000 N0 1000

Ns (20) 717
R(20)
= = = 0.717
N0 1000

In fact, entries in this column can simply be calculated by dividing the

corresponding entries of column 4 by 1000 (the total number of components in
the beginning).
Calculations of Column 6
Entries in column 6 represent values of cumulative failure distribution at
different points in time. Therefore, if F(t) denotes cumulative failure distribution
at time t, then
Nf (t)  whereNf (t) denotes the number of 
F(t) = ,  
N0 components that have failed by time t 

The first three entries of this column are calculated as follows:

Nf (0) 0 Nf (10) 167
F(0)
= = = 0, F(10)
= = = 0.167
N0 1000 N0 1000

Nf (20) 283
F(20)
= = = 0.283
N0 1000

In fact, entries in this column can simply be calculated by dividing the

corresponding entries of column 3 by 1000 (the total number of components in
the beginning) or by column 5 as F(t) = 1 – R(t).
Calculations of Column 7
Entries of this column represent values of failure density function at different
points in time. If f(t) denotes the value of the failure density function at time t,
then from equation (19), we have
d
dt
[Nf (t)] N (t + Δt) − Nf (t)
f(t) = = lim f [by definition of derivative]
N0 Δt →∞ N0 Δt

Thus, approximate values of f(t) can be obtained from the expression

Nf (t + Δt) − Nf (t)
f(t) = , where Δt(class interval) = 10 in this example
N0 Δt

Nf (t + 10) − Nf (t)
= = , where t 0,10,20,30,...,
= and N0 1000
10 × 1000 169
Block 4 Reliability Theory

The first three entries of this column are calculated as follows:

Nf (10) − Nf (0) 167 − 0
f(0)
= = = 0.0167
10000 10000
Nf (20) − Nf (10) 283 − 167 116
f(10)
= = = = 0.0116
10000 10000 10000
Nf (30) − Nf (20) 98
f(20)
= = = 0.0098
10000 10000
In fact, entries in this column can simply be calculated by dividing the entries
of column 2 by 10000.
Calculations of Column 8
Entries of this column represent values of failure rate or hazard rate at
different points in time. If λ(t) denotes the value of failure rate at time t, then
from equation (17), we have
d
dt
[Nf (t)]
λ(t) =
Ns (t)

Nf (t + Δt) − Nf (t)
= lim [by definition of derivative]
Δt →0 Ns (t) Δt
Thus, approximate values of λ(t) can be obtained from the expression
Nf (t + Δt) − Nf (t)
λ(t) = , where Δt = 10 in this example
Ns (t) Δt

Nf (t + 10) − Nf (t)
= , where t = 0, 10, 20, 30, …,
10Ns (t)

The first three entries of this column are calculated as follows:

Nf (10) − Nf (0)
λ(0) =
10Ns (0)

167 − 0
=
10 × 1000
167
= = 0.0167
10000
Nf (20) − Nf (10)
λ(10) =
10Ns (10)

283 − 167 116

= =  0.0139
10 × 833 8330
Nf (30) − Nf (20)
λ(20) =
10Ns (20)

98 98
= =  0.0137
10 × 717 7170

In fact, entries of this column can be calculated by dividing the entries of

column 2 by the preceding entries of column 4 and further dividing the
170 resulting value by 10.
Unit 15 Basic Functions of Reliability

Table 15.3: Estimation of R(t), F(t), f(t) and λ(t)

Number Cumulative Failure

Cumulative Number of Hazard
Time of Reliability Failure Density
Failures Survivors Rate
Failures Distribution Function

t f Nf(t) Ns(t) R(t) F(t) f(t) λ(t)

(1) (2) (3) (4) (5) (6) (7) (8)

0 0 0 1000 1 0 0.0167 0.0167

10 167 167 833 0.833 0.167 0.0116 0.0139

20 116 283 717 0.717 0.283 0.0098 0.0137

30 98 381 619 0.619 0.381 0.0076 0.0123

40 76 457 543 0.543 0.457 0.0065 0.0120

50 65 522 478 0.478 0.522 0.0056 0.0117

60 56 578 422 0.422 0.578 0.0044 0.0104

70 44 622 378 0.378 0.622 0.0036 0.0095

80 36 658 342 0.342 0.658 0.0048 0.0140

90 48 706 294 0.294 0.706 0.0064 0.0218

100 64 770 230 0.230 0.770 0.0077 0.0335

110 77 847 153 0.153 0.847 0.0061 0.0399

120 61 908 92 0.092 0.908 0.0046 0.0500

130 46 954 46 0.046 0.954 0.0027 0.0587

140 27 981 19 0.019 0.981 0.0019 0.1000

150 19 1000 0 0 1 − −
The plot of the reliability, cumulative failure distribution, failure density and
hazard rate functions are shown in Figs. 15.6 to 15.9, respectively.

0.9

0.8

0.7

0.6
R(t)

0.5

0.4

0.3

0.2

0.1

0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Time

Fig. 15.6: Reliability function.

171
Block 4 Reliability Theory

1
0.9
0.8
0.7
0.6

F(t)
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Time

Fig. 15.7: Cumulative failure distribution function.

0.02
0.018
0.016
0.014
0.012
f(t)

0.01
0.008
0.006
0.004
0.002
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
Time

Fig.15.8: Failure density function

0.06

0.05

0.04
λ(t)

0.03

0.02

0.01

0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
Time

172 Fig.15.9: Failure rate.

Unit 15 Basic Functions of Reliability

Now, you can try the following Self Assessment Questions to obtain all these
functions.

SAQ 4
The failure data of 10 electronic components are shown in Table 15.4 given
below. Estimate the reliability, cumulative failure distribution, failure density
and failure rate functions.
Table 15.4: Failure Data

Failure Number 1 2 3 4 5 6 7 8 9 10

Operating Time (in hours) 2 6 31 51 76 116 140 182 250 302

Let us summarise the main points that we have covered in this unit.

15.6 SUMMARY
• Definition of Reliability: Reliability of a component/device/equipment/
unit/system is the probability that it performs its intended function
adequately for a specified period of time under the given operating
conditions.
• Reliability of a component/system when expressed in terms of time t is
known as reliability function and is denoted by R(t).
• In reliability terminology, distribution function of the random variable,
time to failure (T), is known as cumulative failure distribution
function.
• In reliability terminology, probability density function of the random
variable, time to failure (T), is known as failure density function.
• Instantaneous rate of failure at time t is known as the hazard rate at
time t and is generally denoted by λ(t).
• Some commonly used relations between R(t), F(t), f(t) and λ(t) are given
as follows:
∞
 t 
R(t) = ∫t
1 − F(t) = f(t) dt exp
=  − ∫ λ(t) dt 
 0 
t
 t 
F(t) = ∫0
1 − R(t) = f(t) dt 1
= − exp  − ∫ λ(t) dt 
 0 

d d  t 
− [R(t)] =
f(t) = [F(t)] =
λ(t) × exp  − ∫ λ(t) dt 
dt dt  0 
d
d dt
[F(t)] f(t)
− [lnR(t)] =
λ(t) = =
dt 1 − F(t) ∞
∫ f(t)dtt

• The bathtub curve has three phases known as Early Failure, Normal
Life and Wear out Failure. 173
Block 4 Reliability Theory

15.7 TERMINAL QUESTIONS

1. The maximum possible value of reliability for a given period of time is
A) 1 B) 0 C) 0.1 D) 100

2. If hazard rate of a component is 0.2 × 10 −4 , then obtain MTTF.

3. In the phase of the useful life of a component, which distribution is the

best suited to study the failure phenomenon? Choose the most
appropriate alternative.
A) Exponential B) Normal C) Weibull D) Gamma
4. Check whether the following statement is true or false. Give reasons in
support of your answer.
A sample of n identical components numbered from 1 to n is put to test.
The test is allowed to run until all n components fail. The data of records
of the lifetimes of all n components thus obtained is an example of
censored failure data.

15.8 SOLUTIONS/ANSWERS
Self Assessment Questions (SAQs)
1. The definition of reliability is:
Reliability of a component/device/equipment/unit/system/ is the
probability that it performs its intended function adequately for a
specified period of time under the given operating conditions.
From the above definition, it is clear that the correct option is D.
2. In Example 2, we have calculated
−6 2
R(t) = e−0.3×10 t

∞ ∞

∫ R(t) dt= ∫e
−0.3×10−6 t 2
∴ MTTF
= dt
0 0

Putting 0.3 × 10 −6 t 2 =
x, we get

x
t=
0.3 × 10−3
103 1
or dt = dx
0.3 2 x
Also, when=t 0,=
x 0
and when t → ∞, x → ∞
∞ ∞
103 103
∫0 2 0.3 ∫
−x −1/2
∴ MTTF = e x dx = x1/2−1 e− x dx
2 0.3 0

103 1  ∞

∫x
n −1 − x
 = n e dx,n > 0 
1.0954 2  0 
174
Unit 15 Basic Functions of Reliability

107  1 
= π  = π 
10954  2 

3. When a population of identical components is put to test, the failure rate

of the components starts increasing due to the ageing of the
components. In the bathtub curve, this phase is known as wear out.
Hence the correct option is (C).
4. Here the number of components under test is small in size. So we can
record the failure time of each component easily, which is given in Table
15.5. Calculations of the entries of different columns have already been
explained in Example 4 with necessary clarifications. You can follow
similar steps and calculate the entries of different columns of Table 15.5.
It is a small exercise given to you for practice.
Table 15.5: Estimation of R(t), F(t), f(t) and λ (t)

Time Reliability Cumulative Failure Density Hazard Rate

Cumulative
Number of

Number of
Survivors

Failure Function
Failures
Failures

Distribution

t Nf(t) Ns(t) R(t) F(t) f(t) λ(t)

(1) (2) (3) (4) (5) (6) (7) (8)
10 1 1
0 0 0 10 =1 0 = 0.05 = 0.05
10 10 × (2 − 0) 10(2 − 0)

9 1 1 1
2 1 1 9 = 0.9 = 0.1 = 0.025  0.0278
10 10 10(6 − 2) 9(6 − 2)

8 2 1 1
6 1 2 8 = 0.8 = 0.2 = 0.004 = 0.005
10 10 10(31 − 6) 8(31 − 6)

7 3 1 1
31 1 3 7 = 0.7 = 0.3 = 0.005  0.007
10 10 10(51 − 31) 7(51 − 31)

6 4 1 1
51 1 4 6 = 0.6 = 0.4 = 0.004  0.0067
10 10 10(76 − 51) 6(76 − 51)

5 5 1 1
76 1 5 5 = 0.5 = 0.5 = 0.0025 = 0.005
10 10 10(116 − 76) 5(116 − 76)

4 6 1 1
116 1 6 4 = 0.4 = 0.6  0.0042  0.0104
10 10 10(140 − 116) 4(140 − 116)

3 7 1 1
140 1 7 3 = 0.3 = 0.7  0.0024  0.0079
10 10 10(182 − 140) 3(182 − 140)

2 8 1 1
182 1 8 2 = 0.2 = 0.8  0.0015  0.0074
10 10 10(250 − 182) 2(250 − 182)

1 9 1 1
250 1 9 1 = 0.1 = 0.9  0.0019  0.0192
10 10 10(302 − 250) 1 × (302 − 250)

0 10
302 1 10 =0 =1 − −
10 10

Terminal Questions (TQs)

1. Reliability of a component/device/equipment/unit/system is the
probability that it performs its intended function. Being a probability, its
value will vary from 0 to 1. Hence, A is the correct option. 175
Block 4 Reliability Theory

2. We know that if hazard rate of a component = constant = λ, then

1
MTTF = .
λ
Here=
λ 0.2 × 10 −4 , which is a constant.

1 104 105
∴ MTTF = = = = 50000
0.2 × 10 −4 0.2 2
3. We know that in the phase of the useful life of a population of identical
components, hazard rate = constant and constant hazard rate implies
that the distribution of failures follows the exponential distribution.
Hence, (A) is the correct alternative.
4. This statement is false because we know that if we have failure data of
each component of the sample under test then data thus obtained are
known as complete failure data, while the failure data set is said to be
censored if the test is not allowed to run until all n components fail due to
one reason or another.

176