Reliability

(Ref. Ch. 11, QCI, Amitava Mitra)

Prof. Sayak Roychowdhury

 Reliability
• Definition: Reliability is the probability of a product
performing its intended function for a stated period of
time under certain specified condition.
❑ Numerical value of reliability is between 0-1
❑ Product function should be specified, e.g. breaking
strength of nylon cord is 1000 Kg.
❑ Time period for satisfactory function of the product
needs to be specified. E.g. Light bulb life 1000 hours
❑ Operating or environmental conditions are specified.
e.g. Dry conditions, temperature range.
Life Cycle Curve
• Often called ‘Bathtub curve’, failure rate 𝜆 is plotted as a
function of time.

Fundamentals of Quality Control and Improvement, Amitava Mitra

Probability Distribution for Failure Rate
• In the chance failure phase (useful lifetime of product),
failure rate 𝜆 is constant.
• Exponential distribution can be used to provide time of
failure in this phase
𝑓 𝑡 = 𝜆𝑒 −𝜆𝑡 , 𝑡 ≥ 0
• The reliability 𝑹(𝒕) at time 𝑡, is the probability that the
product will last upto time 𝑡:
𝑅 𝑡 = 1 − 𝐹 𝑡 = 1 − 1 − 𝑒 −𝜆𝑡 = 𝑒 −𝜆𝑡
1
• Mean time to failure is given as MTTF =
𝜆
𝑓 𝑡 𝜆𝑒 −𝜆𝑡
• Failure rate function: 𝑟 𝑡 = = =𝜆
𝑅 𝑡 𝑒 −𝜆𝑡
Exponentially Distributed Failure Time

Fundamentals of Quality Control and Improvement, Amitava Mitra

Examples
1. An amplifier has an exponential time-to-failure distribution with a failure
rate of 8% per 1000 hours. What is the reliability of the amplifier at 5000hours? Find the
mean time to failure.
Weibull Distribution for Failure Time
• Weibull distribution can be used to model failure time of products with varying
failure rate. It can be used to model the debugging phase or the wear-out phase.

𝛽 𝑡−𝛾 𝛽−1 𝑡−𝛾 𝛽

𝑓 𝑡 = exp − ,𝑡 ≥ 𝛾
𝛼 𝛼 𝛼

−∞ ≤ 𝛾 ≤ +∞ location parameter, 𝛼 > 0 scale parameter,

𝛽 > 0 shape parameter. For reliability modelling 𝛾 = 0
• The reliability 𝑹(𝒕) at time 𝑡, is the probability that the product will last upto
time 𝑡:
𝑡 𝛽
𝑅 𝑡 = exp −
𝛼
1
Mean time to failure is given as MTTF =𝛼Γ( + 1)
𝛽
𝑓 𝑡 𝛽𝑡 𝛽−1
• Failure rate function: 𝑟 𝑡 = =
𝑅 𝑡 𝛼𝛽
Weibull Distribution for Failure Time
• Debugging phase: Example 𝛼 = 1, 𝛽 = 0.5
• Wear-out phase: Example 𝛼 = 1, 𝛽 = 3.5 (approximates
normal distribution)
• Failure rate fixed: 𝛽 = 1

Fundamentals of Quality Control and Improvement, Amitava Mitra

Example
• Capacitors in an electrical circuit have a time-to-failure
distribution that can be modeled by the Weibull
distribution with a scale parameter of 400 hours and a
shape parameter of 0.2. What is the reliability of the
capacitor after 600 hours of operation? Find the mean
time to failure. Is the failure rate increasing or decreasing
with time?
Availability
• Availability of a system at time 𝑡 is the probability that the
system will be up and running at time 𝑡.
𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑡𝑖𝑚𝑒
𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑖𝑙𝑖𝑡𝑦 =
𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑡𝑖𝑚𝑒 + 𝑑𝑜𝑤𝑛𝑡𝑖𝑚𝑒

𝑑𝑜𝑤𝑛𝑡𝑖𝑚𝑒 = 𝑟𝑒𝑝𝑎𝑖𝑟 𝑡𝑖𝑚𝑒 + 𝑎𝑑𝑚𝑖𝑛𝑖𝑠𝑡𝑟𝑎𝑡𝑖𝑣𝑒 𝑡𝑖𝑚𝑒 + 𝑙𝑜𝑔𝑖𝑠𝑡𝑖𝑐 𝑡𝑖𝑚𝑒

At steady-state:
𝑀𝑇𝑇𝐹
𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑖𝑙𝑖𝑡𝑦 =
𝑀𝑇𝑇𝐹 + 𝑀𝑇𝑇𝑅
System Reliability
• When components are in series:

A B C
• 𝑅𝑠 = 𝑅1 × 𝑅2 … × 𝑅𝑛 System reliability is the product of
component reliabilities
• When components are exponentially distributed with
failure rates 𝜆1 , … 𝜆𝑛 , R s = exp − σ𝑛𝑖=1 𝜆𝑖 𝑡
1
• 𝑀𝑇𝑇𝐹 = σ𝑛
𝑖=1 𝜆𝑖
System Reliability
• When components are in parallel:

C
• Considering independence, probability of system failure is
𝑛
𝐹𝑠 = 1 − 𝑅1 1 − 𝑅2 … 1 − 𝑅𝑛 = ෑ (1 − 𝑅𝑖 )
𝑖=1
System Reliability (Parallel)
• 𝑅𝑠 = 1 − 𝐹𝑠 = 1 − ς𝑛𝑖=1 1 − 𝑅𝑖
• When failure time is exponentially distributed:
• 𝑅𝑠 = 1 − ς𝑛𝑖=1 1 − 𝑒 −𝜆𝑖 𝑡
• When failure rates of all components are same 𝜆,
−𝜆𝑡 𝑛
• 𝑅𝑠 = 1 − 1 −𝑒
1 1 1 1
• MTTF = 1 + + +⋯+
𝜆 2 3 𝑛
Example
1) Find the reliability of the system shown in Figure 11-5 with three
components (A, B, and C) in parallel. The reliabilities of A, B, and C are
0.95, 0.92, and 0.90, respectively.
Example
Consider the seven-component system shown in Figure The
reliabilities of the components are as follows: Ra= 0.96, Rb = 0.92, Rc
= 0.94, Rd= 0.89, Re = 0.95, Rf = 0.88, Rg = 0.90. Find the reliability of
the system. If you had a choice of improving system reliability by
modifying any two components, how would you proceed?
OC-Curve
• Parameters of life testing plan are
test time 𝑇
failure rate 𝜆
1
mean life 𝜃 =
𝜆
sample size = 𝑛
acceptance number = 𝑐
The number of failures of the item is Poisson distributed.
Assume if any item fails, it is replaced immediately
X- axis : Vary 𝜆 and determine 𝑛𝑇𝜆
Y- axis: Probability of acceptance using Poisson
𝑃𝑎 = 𝑃 𝑋 ≤ 𝑐 where 𝑋 denotes number of failures.
OC-Curve
Standby systems
• In a standby configuration, one or more parallel components wait to take
over operation upon failure of the currently operating component. Here,
it is assumed that only one component in the parallel configuration is
operating at any given time.

• Because of this, the system reliability is higher than for comparable

systems with components in parallel.
Standby Systems
• Exponential Model: If the time to failure of the
components is assumed to be exponential with failure
rate 𝜆, the number of failures in a certain time 𝑡 adheres
to a Poisson distribution with parameter 𝜆𝑡.
• Probability of 𝑥 failures in time 𝑡 is given by:
𝑒 −𝜆𝑡 𝜆𝑡 𝑥
𝑃 𝑥 =
𝑥!
In general, if there are n components on standby along with
the basic component (for a total of n + 1 components in the
system), the system reliability is given by
𝑛
𝑒 −𝜆𝑡 𝜆𝑡 𝑥
𝑅𝑠 = ෍
𝑥!
𝑥=0
Reliability and Life Testing Plans
• Failure Terminated Test: The tests are terminated when a
preassigned number of failures occurs in a chosen sample. Lot
acceptance is based on the accumulated test time of the items
when the test is terminated.
• Time Terminated Test: This type of test is terminated when a
preassigned time 𝑇 is elapsed. Acceptance of lot is based on
the observed number of failures during the test time.
• Sequential Reliability Test: Accumulated test results on test
time and number of failures are used to decide whether to
accept or reject the lot. Acceptance and rejection lines are
found using mean life, consumer’s and producer’s risks. (for a
similar degree of protection, the expected test time or the expected number of
failures required to reach a decision is less than that for a time- or a failure-
terminated plan.)
• Standards: US DoD- MIL-STD-690B and Handbook H-108
Sequential Testing

Fundamentals of Quality Control and Improvement, Amitava Mitra

Failure Terminated Test (Exponential Dist)
• Suppose 𝑟 failures occur at the following times, in rank
order: 𝑡1 < 𝑡2 < … < 𝑡𝑟 for sample size 𝑛.
• The accumulated life until the 𝑟 𝑡ℎ failure (𝑇𝑟 ), assuming
failed items are not replaced,
𝑇𝑟 = σ𝑟𝑖=1 𝑡𝑖 + 𝑛 − 𝑟 𝑡𝑟 or
𝑇𝑟 = 𝑛𝑡𝑟 (if replaced)
𝑇𝑟
መ
• Estimated mean life 𝜃 =
𝑟
• Confidence interval
2𝑇𝑟 2𝑇𝑟
<𝜃<
𝜒2 𝛼 𝜒𝛼2
1− 2 ,2𝑟 2 ,2𝑟
Example
• Life testing is conducted for a sample of 15 transistors.
The time to failure for each is exponentially distributed.
The test is terminated after four failures, with no
replacement of the failed items. The failure times (in
hours) of the four transistors are 400,480,610, and 660.
Estimate the mean life of the transistors and the failure
rate. Find a 95% confidence interval for the mean life.
• Assume that each item that fails is replaced by an
identical item. The failure times of the four transistors are
420,490,550, and 580 hours. Estimate the mean life of the
transistors and the failure rate. Find a 95% confidence
interval for the mean life.
Time Terminated Test (Exponential Dist)
• Suppose 𝑇 is the pre-assigned time to terminate the test for
sample size 𝑛.
• Let 𝑡𝑖 be the time of failure of 𝑖 𝑡ℎ item.
• Observed number of failure 𝑥 is the random variable.
• The accumulated life for the test items
𝑇𝑥 = σ𝑥𝑖=1 𝑡𝑖 + 𝑛 − 𝑥 𝑇 or
𝑇𝑥 = 𝑛𝑇(if replaced)
𝑇𝑥
መ
• Estimated mean life 𝜃 =
𝑥
• Confidence interval
2𝑇𝑥 2𝑇𝑥
<𝜃<
𝜒2 𝛼 𝜒𝛼2
1− 2 ,2(𝑥+1)
2 ,2(𝑥+1)
Example
• A sample of 12 electronic components is tested for 1000
hours with no replacement of failed components. The
time to failure is exponentially distributed. Three
components failed within the prescribed test time, the
failure times being 650,680, and 720 hours. Estimate the
mean time to failure and the failure rate. Find a 90%
confidence interval for the mean time to failure.
Standard Life Testing Plans Handbook H-108
• The Quality Control and Reliability Handbook H-108 (U.S. Department of
Defense, 1960) was developed by the Bureau of Naval Weapons, U.S.
Department of the Navy.
• Life testing plans in the handbook are based on a time-to-failure
distribution that is exponential, and all three types of plans (failure-
terminated, time-terminated, and sequential life testing) are included.
• Failure Terminated: A sample of n items is selected from the lot and
tested until the occurrence of the rth failure. If the estimated mean life
𝜃෠ ≥ 𝐶 (a criterion value given by the plan), the lot is accepted. The
producer's risk 𝛼 is the probability of rejecting a lot with a satisfactory
mean life 𝜃0 .
• Time Terminated: The preassigned test time is denoted by T. An
acceptable mean lot life is denoted by 𝜃0 , where the probability of
rejection is 𝛼, the producer's risk, and a minimum mean life is denoted by
𝜃1 , where the probability of acceptance is 𝛽, the consumer's risk. The
rejection criterion number 𝒓 is obtained from the tables. If the observed
number of failures within the preassigned test time is greater than or
equal to r, the lot is rejected; otherwise, the lot is accepted
Examples
• A life testing plan is to be terminated upon occurrence of
the sixth failure. The plan should accept a lot having an
acceptable mean life of 1200 hours with a probability
0.95. Twenty items are placed on test. The six failures
occur at the following times (in hours): 480, 530, 560,
600, 640, and 670. Failed items are not replaced. Using
Handbook H-108, determine whether the lot should be
accepted.
Example (Time Terminated)
• Find a time-terminated life testing plan (using Handbook
H-108) that rejects lots with a mean life of 1250 hours
with a probability of 0.05 and accepts lots with a mean
life of 400 hours with a probability of 0.10. Items that fail
during the test are not replaced.