MEC 541T–Probabilistic Engineering Design

Module 0: Introduction

Siva Prasad AVS,

219 – D, Laboratory Building.
avs@iiitdm.ac.in
Why Probabilistic Approach In Design?
• Uncertainty in Science and Engineering is inevitable because a majority of laws and hypotheses are proposed
based on approximations and assumptions.
• The increasing complexity of modern engineering systems adds to this uncertainty.
• Several phenomena in nature exhibit a high amount of randomness.
• In order to be able to take into account the uncertainty and unpredictability of the systems, both natural as well
as man–made, we need to take a probabilistic route of designing and building the systems.
• Examples are: variation in material properties; variation in estimated loads; variation in market demand for a
product and so on.
Course Content

Module 0 Introduction : Need for taking uncertainty into account to assess the reliability
Module 1 Review of Basic Probability Theory : Set theory, definition of probability, conditional probability, total
probability theorem, Baye’s rule; Random Variables and Probability Distributions
Module 2 Joint and Marginal Distributions
Module 3 Functions of Random Variables : Distributions for functions of random variables; moments of a function
of several random variables; moment–generating function; functions of several random variables
Module 4 Transient Reliability of Components : Reliability and hazard functions; Mean time to failure (MTTF);
Analysis of various systems; Reliability enhancement
Module 5 Reliability concepts in Mechanical Design : Geometry, Material Strength, Loads; Strength based
reliability
Module 6 Further topics : Examples in Design of Mechanical Components and Systems; Monte–Carlo Simulation;
Textbooks & References

• D C Montgomery and G C Runger, Applied Statistics and

Probability for Engineers, 3rd edition, John Wiley & Sons,
2003

• K. C. Kapur and L. R. Lamberson, Reliability in Engineering

Design, Wiley India Pvt Ltd., 2009.

• Singiresu S Rao, Reliability Engineering, Pearson

Publishers, 2016

• E. E. Lewis, An Introduction to Reliability Engineering, Wiley

Publishers, 1996.
Evaluation Policy

Evaluation Description Weightage

Component
• Two quizzes of 15 marks each will be conducted.
Midsem 30
• The tentative dates of the two quizzes are: 08 Sep and 08 Oct
• Students have to form groups of utmost 3 students by August 15,
2021.
• Each group has to choose a “stochastic design problem” by
Seminar 40
September 15, 2021.
• Each group is required to give a seminar in the first week of
November, 2021. The actual dates will be intimated in October.
This will be conducted at the end of the semester as per the schedule
Final
given by the academics office. The duration of this exam will range 30
Assessment
from 90 minutes to 2 hours.
Uncertainty
Uncertainty

Stochastic or aleatory or Subjective or reducible or

irreducible uncertainty epistemic uncertainty
• System itself may display • Lack of complete
random behaviour. knowledge about the
• Inherent randomness in the system. For example,
inputs of the system may starting a new process, a
cause the output also to be new design.
random. • This can be reduced over
• Uncertainty in modelling time by acquiring more and
more knowledge about the
system.
Reliability – Need and importance
• Therefore, given the uncertainty in a system, the question that arises is what is the reliability of the system? How
reliable is its function for the intended purpose?
• For example, you found a flaw of size 60 μm in a component during inspection. Is this flaw big enough to discard
the component and replace it with a new one? Or can it be allowed to continue for some more time in service?
This question can be answered through probabilistic considerations.
• Reliability is the probability of a device performing its function over a specified period of time and under specified
operating conditions.
• A component or system level reliability needs to be assessed.
Bathtub curve
• A bathtub curve is a representation of the failure regimes of a
component or a system. It is the variation of a product’s

Failure Rate
failure rate with time. The design considerations of a
particular product are formulated based on this curve.

Infant
• An ideal bathtub curve (the failure rate vs time) will exhibit a
mortality Wearout
Useful life
drastic decrease of failure rate during its early (infant) stages
and a gradual increase during its wearout stages. Time
Factor of Safety and Reliability
Allowable stress
Factor of safety =
Working stress
• Allowable stress can be as yield strength, ultimate strength, fatigue strength etc, which are material properties.
They are obtained from experiments conducted on plain specimens and the results are usually taken as average of
several repetitions.
• Working stress is the maximum expected stress magnitude within the component that is being designed.
• Thus, in a real – life situation, both the applied load and the strength of the material are prone to variations. It is
therefore imperative to define factory of safety as,

expected strength of the material

Factor of safety =
expected load during service
Factor of Safety and Reliability
expected strength of the material
Factor of safety =
expected load during service

Example to understand the consequence of this definition:

Let there be a component which is subjected to a stress
magnitude of 100 MPa on an average. And, let the material
has an ultimate strength of 160 MPa, on an average. The
factor safety as calculated from the above is 1.6. The
distributions of the applied stress and strength are as shown in
the adjacent plot. The blue curve corresponds to the applied
stress and the red curve corresponds to the strength of the
material. You can see that there is still a finite probability of
failure of the component, even though the above calculated
factory of safety is used.
Factor of Safety and Reliability
Exercise:
• Investigate what happens to the failure probability when both the average load and the average strength are scaled
by same amount so that the factor of safety remains same.
• Investigate what happens to the failure probability when the standard deviations of the distribution of the load and
the strength are scaled by same factor.
• What are your inferences?
Reliability
• After performing the exercise, you might have realized the necessity behind the use of “Reliability” principles to
relate the functioning of a component satisfactorily.
• The steps involved in a reliability analysis are:
1. Identify the parameters which exhibit random behaviour in a design
2. The quality characteristic or the performance criteria of the component/system
3. Collection of data
4. Histogram of random parameters
5. Fit a probability density or distribution function
6. Estimate the reliability of the component/system
7. Risk and cost associated with the computed reliability
8. Modify the design to reduce the risk and cost by repeating steps 1 to 8.
Some Examples of Structural Failures in the History
Hyatt Regency Walkway Collapse, 17 July 1981
Some Examples of Structural Failures in the History
Tacoma Narrows Bridge Collapse, 07 November 1940
Some Examples of Structural Failures in the History
Space Shuttle Columbia Disaster, 01 February 2003
Some Examples of Structural Failures in the History
Crash of El Al Boeing 747–200, 04 October 1992
Some Examples of Structural Failures in the History
Space Shuttle Challenger disaster, 28 January 1986
Some Examples of Structural Failures in the History
Laboratory for Hypersonic and Shockwave Research
(LHSR), IISc Bangalore, 05 December 2018

https://www.thehindu.com/news/cities/bangalore/blast-turns-spotlight-on-safety-at-iisc/article25805461.ece
 MEC 541T–Probabilistic Engineering Design
Module 1: Basic Probability Theory

Siva Prasad AVS,

219 – D, Laboratory Building.
avs@iiitdm.ac.in
Experiment, Outcome and Event
• Experiment: Tossing a coin; Measuring a property or performance etc.,

• Outcomes: All the possible results of the experiment. While measuring the density of steel, the possible outcomes
are 7000 kg/m3 to 8000 kg/m3. The outcomes can be discrete or continuous, depending upon the type of the
variable.

• Event: An event is looking for a particular outcome or a group of outcomes. So, in measuring density, looking for
all the set of measured values between 7500 kg/m3 and 7800 kg/m3 is an event.
Mutually Exclusive
If the occurrence of an event precludes the occurrence of other events in a given experiment, the events are called
mutually exclusive. So, this means the events cannot occur simultaneously. The event of tail and the event of head, for
example, cannot occur simultaneously and hence are mutually exclusive. “Pass/Fail”, “defective/non – defective” etc.,
are some more examples.

NOTE: “Mutually exclusive events” is NOT same as “independent events”.

Set Theory
• Likelihood of an event is
better understood using the
language of set theory.
• All the possible outcomes of
an experiment is universal
set.
• All the concepts such as
union, intersection,
complement etc., are
summarized.
Set Theory
• Likelihood of an event is
better understood using the A B A B
language of set theory.
• All the possible outcomes of
C C
an experiment is universal
set.
AՈBՈC (A Ս C) Ո B
• All the concepts such as
union, intersection,
complement etc., are
summarized.

Exercise: Draw the step–wise Venn diagrams to prove the distributive law.
Sample points and sample space
Example: Suppose you are given the following table, which contains the densities of five common alloys used for
engineering applications. Note that the density values are given as a range. Now, you are given a specimen of one of
these alloy materials. You are expected to measure its density and find out which alloy it is. If the measured value of
density is chosen as a random variable, the sample space for this experiment is set of all the possible values ranging
from the 2.64 g/cc to 8.7 g/cc.
S = {ρ| 2.64 ≤ ρ ≤ 8.7}
Every element of this set is called a sample point. Note that the size of the sample space is infinite.

S. No. Alloy Density (g/cc)

1 Ti6Al4V 4.429 – 4.512
2 AlSi 2.64 – 2.7
3 SS316 7.87 – 8.07
4 INCONEL 8.12 – 8.44
5 Brass–casting 8.4 – 8.7
https://www.azom.com/
https://www.engineeringtoolbox.com/metal-alloys-densities-d_50.html
Sample points and sample space
Example: Suppose you are given the following table, which contains the densities of four common alloys used for
engineering applications. Note that the density values are given as a range. Now, you are given a specimen of one of
these alloy materials. You are expected to measure its density and find out which alloy it is. If the type of alloy is
chosen as a random variable, the sample space for this experiment is just the set of all the alloys.
S = {Ti6Al4V, AlSi, SS316, INCONEL}
So, there are four sample points in the sample space S. Note that the size of the sample space is finite (=4).

S. No. Alloy Density (g/cc)

1 Ti6Al4V 4.429 – 4.512
2 AlSi 2.64 – 2.7
3 SS316 7.87 – 8.07
4 INCONEL 8.12 – 8.44

https://www.azom.com/
https://www.engineeringtoolbox.com/metal-alloys-densities-d_50.html
Sample points and sample space
Example: Four specimens made of either Ti6Al4V or IN718 alloys are given, but it is not known which is what. It is
required to determine which alloy each specimen is, by conducting some standard tests (such as EDAX, XRD etc.,).
The sample space with all the possible outcomes is,
S = {TTTT, TTTI, TTIT, TITT, ITTT, TTII, TITI, TIIT, ITIT, ITTI, IITT, TIII, ITII, IITI, IIIT, IIII}
So, there are 16 sample points in the sample space S.

Event Description of the event Sample points of the event

E1 All the four specimens are Ti6Al4V {TTTT}
E2 At least three specimens are Ti6Al4V {TTTT, TTTI, TTIT, TITT, ITTT}
E3 At most one specimen is IN718 {TTTT, TTTI, TTIT, TITT, ITTT}
E4 At least two specimens are IN718 {TTII, TITI, TIIT, ITIT, ITTI, IITT,
TIII, ITII, IITI, IIIT, IIII}
E5 At most four specimens are Ti6Al4V S

Exercise: State any two events which are mutually exclusive to occur.
Definition of Probability
Two Approaches

• Relative frequency (Statistical) Definition

𝑛
𝑃 𝐸 = lim
𝑁→∞ 𝑁

where, n is the number of occurrences of event E and N is the total number of trials or repetitions of the experiment.
0≤𝑃 𝐸 ≤1

• Axiomatic Definition
1. 𝑃 𝐸 ≥ 0
2. 𝑃 𝐸 = 1
3. If E1 and E2 are mutually exclusive events, then 𝑃 𝐸1 ∪ 𝐸2 = 𝑃 𝐸1 + 𝑃 𝐸2
Counting Techniques – Permutations and Combinations
Permutations – Order of arrangement matters
Given S = {a, b, c}. Permutation is the ordered sequence of these elements. In this case it is 3 X 2 X 1 = 3! = 6 ways.

There are n (n – 1) (n – 2) (n – 3)…1 = n! ways to form ordered sequences of n different elements.

There are n (n – 1) (n – 2) (n – 3)…(n – r + 1) ways to form ordered sequences of r elements, drawn from n different
𝑛!
elements. This can be written as which is denoted as 𝑛𝑃𝑟 .
𝑛−𝑟 !

That is, there are 𝑛𝑃𝑟 ways of forming ordered sequences of r elements drawn from n different elements.

Example: A printed circuit board has 8 different locations in which a component can be placed. If four different
components are to be place on the board, how many different designs are possible?
Counting Techniques – Permutations and Combinations
Permutations – Order of arrangement matters
There are 𝑛𝑃𝑟 ways of forming ordered sequences of r elements drawn from n different elements.
Suppose, out of n elements, there are n1 of type 1, n2 of type 2, n3 of type 3 ……, nr of type r then, number of
permutations formed using n elements is

𝑛!
𝑛1 ! 𝑛2 ! 𝑛3 ! … 𝑛𝑟 !

Example: A piece of sheet metal needs two identical diameter holes drilled and two identical size notches cut. We
denote a drilling operation as ‘D’ and notching operation as ‘N’. In determining a schedule for a machine shop, we
might be interested in the number of possible sequences for two drilling operations and two notching operations.
What is this number?
Counting Techniques – Permutations and Combinations
Combinations–Order of arrangement does not matter (A combination lock should have been called a
permutation lock according to this definition)
Recall: Suppose, out of n elements, there are n1 of type 1, n2 of type 2, n3 of type 3 ……, nr of type r then, number
of permutations formed using n elements is

𝑛!
𝑛1 ! 𝑛2 ! 𝑛3 ! … 𝑛𝑟 !

Now, number of combinations of r elements from n different elements means number of subsets of r elements that
can be formed from n different elements, without any particular order. So, out of n different elements, r elements can
be thought of as type 1, and the remaining (n – r) can be thought of as type 2. Then using the above formula, number
of combinations of r elements from n different elements is

𝑛!
= 𝑛𝐶𝑟
𝑟! 𝑛 − 𝑟 !
Laws of Probability
Union of Events:
𝑃 𝐸1 ∪ 𝐸2 = 𝑃 𝐸1 + 𝑃 𝐸2 − 𝑃 𝐸1 ∩ 𝐸2
𝑃 𝐸1 ∪ 𝐸2 ∪ 𝐸3 = 𝑃 𝐸1 + 𝑃 𝐸2 + 𝑃 𝐸3 − 𝑃 𝐸1 ∩ 𝐸2 − 𝑃 𝐸1 ∩ 𝐸3 − 𝑃 𝐸2 ∩ 𝐸3 + 𝑃 𝐸1 ∩ 𝐸2 ∩ 𝐸3

Union of Mutually Exclusive Events:

𝑃 𝐸1 ∪ 𝐸2 = 𝑃 𝐸1 + 𝑃 𝐸2
i.e., 𝑃 𝐸1 ∩ 𝐸2 = 0, if E1 and E2 are mutually exclusive.

For any event E, the events E and E’ (the complement of E) are mutually exclusive.
Conditional Probability
• The occurrence of the event E2 when it is know that E1 has occurred is known as the conditional event and is
denoted as E2|E1. The conditional probability of this event is,

𝑃 𝐸1 ∩ 𝐸2
𝑃 E2|E1 =
𝑃 E1

Example: A class consists of 35 boys and 15 girls. Of the 35 boys, 15 are science majors and 20 are engineering
majors. Of the girls, 5 major in science and 10 major in engineering.
Define two events G → a student is a girl and S → a student majoring in Science. Find all the probabilities.
Conditional Probability
Example: In a manufacturing process, 10% of the parts contain visible surface flaws and 25% of the parts with
surface flaws are functionally defective. However, only 5% of parts without surface flaws are defective. Define events
D → a part is functionally defective; S → a part has surface flaws. Find all the probabilities.
Conditional Probability
Example: If p(A|B) = 1, must A = B? Reason out using Venn diagram.
Conditional Probability
Example: Suppose A and B are mutually exclusive events. Construct a Venn diagram that contains the three events A,
B, and C such that P(A|C) = 1 and P(B|C) = 0.
Statistically independent Vs Mutually exclusive events
• Two events are independent if their occurrences are not interdependent.

𝑃 𝐸1 ∩ 𝐸2 = 𝑃 𝐸1 𝑃 𝐸2

• Two events are mutually exclusive if occurrence of one precludes the other.

𝑃 𝐸1 ∩ 𝐸2 = 0
Multiplication Rule
𝑃 𝐸1 ∩ 𝐸2 = 𝑃 𝐸1 𝐸2 𝑃 𝐸2 = 𝑃 𝐸2 𝐸1 𝑃 𝐸1

As noted before, if two events are independent,

𝑃 𝐸1 ∩ 𝐸2 = 𝑃 𝐸1 𝑃 𝐸2

Exercise: The probability that an automobile battery subject to high engine compartment temperature suffers low
charging current is 0.7. The probability that a battery is subject to high engine compartment temperature is 0.05. What
is the probability of high engine compartment temperature and low charging current?
Statistically independent Vs Mutually exclusive events
Exercise: The functioning of a system depends upon two independent components A and B, whose probabilities of
failure are 0.2 and 0.1, respectively. What is the probability of failure of the system in this case?
If the system is redesigned so that its functioning depends only on any one component, what is the probability of
failure of the system?
Multiplication Rule
Exercise: A batch of 500 containers for frozen orange juice contains 5 that are defective. Two are selected, at random,
without replacement from the batch. What is the probability that the second one selected is defective given that the
first one was defective? What is the probability that both are defective? What is the probability that both are
acceptable?
Total Probability Theorem
𝑃 𝐸1 = 𝑃 𝐸1 ∩ 𝐸2 + 𝑃 𝐸1 ∩ 𝐸2′ = 𝑃 𝐸1 𝐸2 𝑃 𝐸2 + 𝑃 𝐸1 𝐸2′ 𝑃 𝐸2′
Generalizing this,

If E1, E2, E3, ….En are n mutually exclusive events, then the probability of occurrence of another event A can be
written as,
𝑛

𝑃 𝐴 = ෍ 𝑃 𝐴 𝐸𝑖 𝑃 𝐸𝑖
𝑖=1

𝑃 𝐸1 𝑃 𝐸1 ∩ 𝐸2 𝑃 𝐸1 ∩ 𝐸2′

= +
E1 E2 E1 E2 E1 E2

Exercise: In the problem in the previous slide, what is the probability that the second one selected is defective?
Total Probability Theorem
Exercise: A firm manufactures 70% cotton rolls and 30% nylon rolls. 2% cotton rolls are defective and 3% nylon
rolls are defective. Find the probability that a fabric roll selected at random is defective.
Total Probability Theorem
Exercise: In a chip manufacturing company, 70% of the chips are of high purity, 20% are of medium purity and 10%
are of very low purity. Their corresponding failure probabilities are 0.001, 0.01 and 0.1. What is the probability of
failure of a chip drawn at random?
Total Probability Theorem
Exercise: In the manufacturing of a chemical adhesive, 3% of batches have raw materials drawn from two different
lots. This occurs when the holding tanks are replenished and the remaining portion of a lot is insufficient to fill the
tanks. Only 5% of batches with raw material from a single lot require reprocessing. However, 40% of the batches with
raw materials drawn from two or more lots require reprocessing. If event A denotes a batch is formed from two or
more lots and B denotes a batch requiring processing, find (i) p(A) (ii) p(B|A) (iii) p(B|A’) (iv) p(A∩B) (v) p(A∩B’)
(vi) p(B)
Total Probability Theorem
Exercise: Computer keyboard failures are due to faulty electrical connects (12%) or mechanical defects (88%).
Mechanical defects are related to loose keys (27%) or improper assembly (73%). Electrical connect defects are caused
by defective wires (35%), improper connections (13%), or poorly welded wires (52%). Find the probability that a
failure is due to loose keys. Find the probability that a failure is due to improperly connected or poorly welded wires.
Independence
Exercise: The following circuit operates only if there is a path of functional devices from left to right. The probability
that each device functions is shown on the graph. Assume that devices fail independently. What is the probability that
the circuit operates?
Independence
Exercise: Disks of polycarbonate plastic from a supplier are analyzed for scratch and shock resistance. The results
from 100 disks are summarized in the table below. Let A denote the event that a disk has high shock resistance, and
let B denote the event that a disk has high scratch resistance. Are events A and B independent?
Bayes’ Theorem
Recall the multiplication rule of probabilities:

𝑃 𝐴∩𝐵 = 𝑃 𝐴 𝐵 𝑃 𝐵 = 𝑃 𝐵 𝐴 𝑃 𝐴

𝑃 𝐴𝐵 𝑃 𝐵
⇨𝑃 𝐵𝐴 =
𝑃 𝐴
Now, we can use total probability theorem.

If B1, B2, B3, ….Bn are n mutually exclusive events, then the probability of occurrence of another event E2 can be
written as,
𝑛

𝑃 𝐴 = ෍ 𝑃 𝐴 𝐵𝑖 𝑃 𝐵𝑖
𝑖=1
Using this in the above gives the Bayes’ theorem.

𝑃 𝐴 𝐵𝑗 𝑃 𝐵𝑗
𝑃 𝐵𝑗 𝐴 =
σ𝑛𝑖=1 𝑃 𝐴 𝐵𝑖 𝑃 𝐵𝑖
Bayes’ Theorem
Exercise: A new medical procedure for diagnosis of illness is being tested on a population. The probability that test
correctly identifies someone with illness as positive is 0.99 and the probability that the test correctly identifies
someone without illness as negative is 0.95. Incidence of illness in the general population is 0.0001. A person takes
the test and got positive. What is the probability that he/she has illness?
Bayes’ Theorem
Exercise: Semiconductor lasers used in optical storage products require higher power levels for write operations than
for read operations. High-power-level operations lower the useful life of the laser. Lasers in products used for backup
of higher speed magnetic disks primarily write, and the probability that the useful life exceeds five years is 0.95.
Lasers that are in products that are used for main storage spend approximately an equal amount of time reading and
writing, and the probability that the useful life exceeds five years is 0.995. Now, 25% of the products from a
manufacturer are used for backup and 75% of the products are used for main storage.

Let A denote the event that a laser’s useful life exceeds five years, and let B denote the event that a laser is in a
product that is used for backup. Find (i) p(B) (ii) p(A|B) (iii) p(A|B’) (iv) p(A∩B) (v) p(A∩B’) (vi) p(A) and (vii)
What is the probability that a laser that failed before five years came from a product used for backup?
Bayes’ Theorem
 MEC 541T–Probabilistic Engineering Design
Module 2: Probability Distributions

Siva Prasad AVS,

219 – D, Laboratory Building.
avs@iiitdm.ac.in
Definition of a random variable
X → a random variable which can be a measured value of a quality characteristic
x → a particular value taken by the random variable X

Discrete
X
Continuous
Probability Mass Function – Discrete variables
Cumulative Distribution Function – Discrete variables
Cumulative Distribution Function – Discrete variables
Example: 250 castings of an engine are manufactured in day, of which 22 have blowholes. A sample of two castings
are drawn at random for inspection. No. of defective castings in the sample is chosen as a random variable. Generate
probability mass and distribution functions.
Cumulative Distribution Function – Discrete variables
Example: Given the cumulative distribution function below, find the probability mass function

0, 𝑥 < −2
0.2, −2 ≤ 𝑥 < 0
𝑃 𝑋≤𝑥 =𝐹 𝑥 =
0.7, 0≤𝑥<2
1, 2≤𝑥
Probability Density Function – Continuous variables
(Cumulative) Probability Distribution Function – Continuous variables
Probability Distribution Function – Continuous variables
Example: Given the probability density function of the hardness number of a material below, find the probability
distribution function. Find the probability that the hardness number exceeds 80 units given that it is greater than 20
units.

3𝑥
𝑓 𝑥 = 100 − 𝑥 , 0 ≤ 𝑥 ≤ 100
500000
Measures of Central Tendency – Mean, Median and Mode
Measures of Central Tendency – Mean, Median and Mode
Measures of Central Tendency – Mean, Median and Mode
Example: For the probability density function given below, find the mean, median and the mode.
2
𝑓 𝑥 = 2𝑥𝑒 −𝑥 , 𝑥 ∈ 0, ∞ .
Measures of Central Tendency – Mean, Median and Mode
Moments of Random Variables – About Origin, Central moment
Moments of Random Variables – Second Central moment
Moments of Random Variables – Third and Fourth Central moments
Thresholds on central moments as a design approach
Chebyshev Inequality
Example: Given the Chebyshev’s inequality

1
𝑃 𝑋 − 𝜇𝑋 ≥ 𝑡𝜎𝑋 ≤ 2
𝑡
Design a circular bar such that the stress does not exceed a permissible value S0. The maximum probability of
exceeding the permissible value (failure) should be 0.0025. The load L fluctuates about μL with a standard
deviation of σL. Take the area of the cross–section, A, to be deterministic.
Chebyshev Inequality
Probability Distributions for Discrete Random Variables
Uniform Distribution

Binomial Distribution
Probability Distributions for Discrete Random Variables
Example: 10 questions of multiple choice type are there in a test, each question having four options, of which only
one is correct. What is the probability that 2 questions are incorrect if all the questions are answered by guess?
Probability Distributions for Discrete Random Variables
Geometric Distribution: If in a random experiment the random variable is chosen as the number of trials required till
first success, it follows Geometric Distribution.
Example: 10 questions, 4 options of which only one is correct. Find the distribution for the number of trials till the
question attempted is incorrect (i.e, the first success).
Probability Distributions for Discrete Random Variables
Negative Binomial Distribution:
X → No. of Bernoulli trials required to obtain r successes.
Probability Distributions for Discrete Random Variables
Hypergeometric Distribution:
Similarity with binomial distribution, but with no approximations about replacement.
Probability Distributions for Discrete Random Variables
Poisson Distribution:
Probability Distributions for Continuous Random Variables
Uniform Distribution:
Probability Distributions for Continuous Random Variables
Normal Distribution:
Probability Distributions for Continuous Random Variables
Exponential Distribution:
Probability Distributions for Continuous Random Variables
Lognormal Distribution:
Probability Distributions for Continuous Random Variables
Weibull Distribution:
Problem Solving
Example: Determine the cumulative distribution function of a binomial random variable with n = 3 and p = 1/2.
Problem Solving
Example: The phone lines to an airline reservation system are occupied 40% of the time. Assume that the events that
the lines are occupied on successive calls are independent. Assume that 10 calls are placed to the airline. Generate the
probabilities for X → number of calls in which lines are occupied, Y → number of calls in which lines are not
occupied. What is the probability that for at least one call the lines are not occupied?
Problem Solving
Example: A computer system uses passwords that are exactly six characters and each character is one of the 26 letters
(a–z) or 10 integers (0–9). Suppose there are 10,000 users of the system with unique passwords. A hacker randomly
selects (with replacement) one billion passwords from the potential set, and a match to a user’s password is called a
hit.
(a) What is the distribution of the number of hits?
(b) What is the probability of no hits?
(c) What are the mean and variance of the number of hits?
Problem Solving
Example: Mean customer arrival at a single window cash–counter in a bank is 25/hr and can be described as a
Poisson process. What is the probability that there are no customers in an interval of 6 minutes?
Problem Solving
Example: The probability density function of the length of a metal rod is f(x) = 2 for 2.3 ≤ x ≤ 2.8 m. Specifications
desired by a customer for length of the rod are 2.25 m to 2.75 m. What proportion of the rods will fail to meet these
specifications?
Problem Solving
Example: Current measurements in a strip of wire are assumed to follow normal distribution, with μX = 10 mA and
𝜎𝑋2 = 4 𝑚𝐴 2 . What is the threshold value a measurement will lie below with a probability of 0.98?
Problem Solving
Example: The time taken for a cell to undergo mitosis (cell division) is normally distributed with an average of 1
hour and a standard deviation of 5 minutes. What time does it take for 99% of all the cells to undergo mitosis?
A few observations on Binomial, Poisson and Normal distribution
Binomial distribution:
𝑝 𝑘 = 𝑛𝑐𝑘 𝑝ҧ 𝑘 1 − 𝑝ҧ 𝑛−𝑘
Poisson Distribution:
𝑒 −λ λ𝑘
𝑝 𝑘 =
𝑘!
Where, 𝑝ҧ is the average fraction defective,
λ = n𝑝ҧ is the mean of Poisson distribution.

n = 500
𝑝ҧ = 0.08
Problem Solving
Example: No. of bits received in error follows binomial distribution with the probability that a bit transmitted is in
error being 10–5. If 16 million bits are transmitted, what is the probability of receiving 150 errors or more?
 MEC 541T –Probabilistic Engineering Design
Module 3: Joint and Marginal Distributions

Siva Prasad AVS,

219 – D, Laboratory Building.
avs@iiitdm.ac.in
Concept of Joint Distributions
Example: In the development of a new receiver for the transmission of digital information, each received bit is rated
as acceptable, suspect, or unacceptable, depending on the quality of the received signal, with probabilities 0.9, 0.08,
and 0.02, respectively. Assume that the ratings of each bit are independent. In the first four bits transmitted, let
X1 → number of acceptable bits and X2 → number of suspect bits

Analysis: Separately these two random variables follow binomial distribution with n = 4, p = 0.9 and n = 4, p = 0.08,
respectively. That is, with the given information, P(X1 = x1) for x1 = 0, 1, 2, 3, 4 and P(X2 = x2) for x2 = 0, 1, 2, 3, 4
can be computed.
Can we now derive the probability mass function for P(X1 = x1 and X2 = x2)?

Solution:
P(X1 = 0 and X2 = 0) = P(All are unacceptable) = (0.02)4 = 16 X 10–8
4!
P(X1 = 1 and X2 = 0) = P(X1 = 1 and 3 unacceptable) = 0.9 (0.02)3 = 2.88 X 10–5
3!

4!
P(X1 = 2 and X2 = 0) = P(X1 = 2 and 2 unacceptable) = (0.9)2(0.02)2 = 1.944 X 10–3
2!2!
Concept of Joint Distributions
Example: In the development of a new receiver for the transmission of digital information, each received bit is rated
as acceptable, suspect, or unacceptable, depending on the quality of the received signal, with probabilities 0.9, 0.08,
and 0.02, respectively. Assume that the ratings of each bit are independent. In the first four bits transmitted, let
X1 → number of acceptable bits and X2 → number of suspect bits

Solution:
4!
P(X1 = 3 and X2 = 0) = P(X1 = 3 and 1 unacceptable) = (0.9)3 (0.02) = 0.05832
3!

P(X1 = 4 and X2 = 0) = P(X1 = 4) = (0.9)4 = 0.6561

4!
P(X1 = 0 and X2 = 1) = P(X2 = 1 and 3 unacceptable) = 0.08(0.02)3 = 2.56 X 10–6
3!

4!
P(X1 = 1 and X2 = 1) = P(X1 = 1 and X2 = 1 and 2 unacceptable) = (0.9) (0.08)(0.02)2 = 3.456 X 10–4
2!

P(X1 = 2 and X2 = 1) = P(X1 = 2 and X2 = 1 and 1 unacceptable) = 0.015552

P(X1 = 3 and X2 = 1) = P(X1 = 3 and X2 = 1) = 0.23328
Concept of Joint Distributions
Example: In the development of a new receiver for the transmission of digital information, each received bit is rated
as acceptable, suspect, or unacceptable, depending on the quality of the received signal, with probabilities 0.9, 0.08,
and 0.02, respectively. Assume that the ratings of each bit are independent. In the first four bits transmitted, let
X1 → number of acceptable bits and X2 → number of suspect bits

Solution:
P(X1 = 0 and X2 = 2) = P(X2 = 2 and 2 unacceptable) = 1.536 X 10–5
P(X1 = 1 and X2 = 2) = P(X1 = 1 and X2 = 2 and 1 unacceptable) = 1.3824 X 10–3
P(X1 = 2 and X2 = 2) = P(X1 = 2 and X2 = 2) = 0.031104
P(X1 = 0 and X2 = 3) = P(X2 = 3 and 1 unacceptable) = 4.096 X 10–5
P(X1 = 1 and X2 = 3) = P(X1 = 1 and X2 = 3) = 1.8432 X 10–3
P(X1 = 0 and X2 = 4) = P(X2 = 4) = (0.08)4 = 4.096 X 10–5
Concept of Joint Distributions
Observations:
• Sum of all the probabilities is equal 1
• Summing up along any one random variable keeping the other random variable fixed gives the probability of the
random variable taking the fixed value. i.e.,

𝑃 𝑋1 = 𝑥1 = ෍ 𝑃 𝑋1 = 𝑥1 and 𝑋2 = 𝑥2
𝑥2

𝑃 𝑋2 = 𝑥2 = ෍ 𝑃 𝑋1 = 𝑥1 and 𝑋2 = 𝑥2
𝑥1

• The individual mass functions or distributions P(X1 = x1) and P(X2 = x2) are called Marginal Distributions and
the combined probability mass function or distribution is called Joint Probability Distribution.
Marginal distributions for continuous case
Expectation and Variance from Joint Distributions
Moments of Continuously Distributed Random Variables
Moments of Continuously Distributed Random Variables
Moments of Continuously Distributed Random Variables
Conditional Probability Distributions
Problems
Example: Determine the marginal distributions for X1 and X2 in the first example from the joint distribution and
validate with Binomial distribution for each random variable
We have already computed joint probability mass function. The corresponding values are given below.
Note that the bold subscript is being used to indicate that it is a vector of two random variables – X1 and X2.
fX (0, 0) = 16 X 10–8; fX (1, 0) = 2.88 X 10–5; fX (2, 0) = 1.944 X 10–3; fX (3, 0) = 0.05832; fX (4, 0) = 0.6561
fX (0, 1) = 2.56 X 10–6; fX (1, 1) = 3.456 X 10–4; fX (2, 1) = 0.015552; fX (3, 1) = 0.23328
fX (0, 2) = 1.536 X 10–5; fX (1, 2) = 1.3824 X 10–3; fX (2, 2) = 0.031104
fX (0, 3) = 4.096 X 10–5; fX (1, 3) = 1.8432 X 10–3
fX (0, 4) = 4.096 X 10–5

4−𝑥1 4−𝑥2

𝑃 𝑋1 = 𝑥1 = ෍ fX (𝑥1 ,𝑥2 ) for 𝑥1 = 0,1,2,3,4 and 𝑃 𝑋2 = 𝑥2 = ෍ fX (𝑥1 ,𝑥2 ) for 𝑥2 = 0,1,2,3,4

𝑥2 =0 𝑥1 =0
Problems
Example: Determine the marginal distributions for X1 and X2 in the first example from the joint distribution and
validate with Binomial distribution for each random variable
We have already computed joint probability mass function. The corresponding values are given below.
Note that the bold subscript is being used to indicate that it is a vector of two random variables – X1 and X2.
fX (0, 0) = 16 X 10–8; fX (1, 0) = 2.88 X 10–5; fX (2, 0) = 1.944 X 10–3; fX (3, 0) = 0.05832; fX (4, 0) = 0.6561
fX (0, 1) = 2.56 X 10–6; fX (1, 1) = 3.456 X 10–4; fX (2, 1) = 0.015552; fX (3, 1) = 0.23328
fX (0, 2) = 1.536 X 10–5; fX (1, 2) = 1.3824 X 10–3; fX (2, 2) = 0.031104
fX (0, 3) = 4.096 X 10–5; fX (1, 3) = 1.8432 X 10–3
fX (0, 4) = 4.096 X 10–5

𝑥1 0 1 2 3 4
𝑃 𝑋1 = 𝑥1 0.0001 0.0036 0.0486 0.2916 0.6561
𝑥2 0 1 2 3 4
𝑃 𝑋2 = 𝑥2 0.71639296 0.24918016 0.03249876 0.00188416 4.096 X 10–5
Problems
Example: Determine the expectation of X1 and X2 in the first example.

We have already computed joint probability mass function. The corresponding values are given below.
Note that the bold subscript is being used to indicate that it is a vector of two random variables – X1 and X2.
fX (0, 0) = 16 X 10–8; fX (1, 0) = 2.88 X 10–5; fX (2, 0) = 1.944 X 10–3; fX (3, 0) = 0.05832; fX (4, 0) = 0.6561
fX (0, 1) = 2.56 X 10–6; fX (1, 1) = 3.456 X 10–4; fX (2, 1) = 0.015552; fX (3, 1) = 0.23328
fX (0, 2) = 1.536 X 10–5; fX (1, 2) = 1.3824 X 10–3; fX (2, 2) = 0.031104
fX (0, 3) = 4.096 X 10–5; fX (1, 3) = 1.8432 X 10–3
fX (0, 4) = 4.096 X 10–5

4 4−𝑥1 4 4−𝑥2

𝐸 𝑋1 = ෍ 𝑥1 ෍ fX (𝑥1 ,𝑥2 ) and 𝐸 𝑋2 = ෍ 𝑥2 ෍ fX (𝑥1 ,𝑥2 )

𝑥1 =0 𝑥2 =0 𝑥2 =0 𝑥1 =0

Answers: E[X1] = 3.6 and E[X2] = 0.32

Problems
Example: Given the following data of joint probabilities: Determine
(a) P(X < 2.5, Y < 3) (b) P (X < 2.5) (c) P(Y < 3) (d) P(X > 1.8, Y > 4.7) (e) P(X = x) (f) P(Y = y) (g) E[X] (h) E[Y]
(i) P( Y = y | X = 1.5) (j) P(X = x | Y = 2) (k) Comment about the independence of X and Y
Problems
Example: Given the following data of joint probabilities: Determine
(a) P(X < 2.5, Y < 3) (b) P (X < 2.5) (c) P(Y < 3) (d) P(X > 1.8, Y > 4.7) (e) P(X = x) (f) P(Y = y) (g) E[X] (h) E[Y]
(i) P( Y = y | X = 1.5) (j) P(X = x | Y = 2) (k) Comment about the independence of X and Y
Problems
Example: Determine the value of c such that the function f(x, y) = cxy for 0 < x < 3 and 0 < y < 3 satisfies the
properties of a joint probability density function. Further, determine (i) P(X<2, Y < 3) (ii) P(X < 2.5) (iii) E[X]
(iv) P(X>1.8, 1 < Y < 2.5)
Problems
Example: Given the joint density function f(x,y) = 12 exp(–4x–2y) for 0 ≤ y ≤ x; 0 ≤ x ≤ ∞ and 0 in the remaining 2–
D space. Determine the marginal distributions for the random variables X and Y and compute their expected values.
Problems
Example: Two methods of measuring surface smoothness are used to evaluate a paper product. The measurements
are recorded as deviations from the nominal surface smoothness in coded units. The joint probability distribution of
the two measurements is a uniform distribution over the region 0 < x < 4, 0 < y, and |y–x| < 1. That is, fXY (x, y) = c
for x and y in the region. Determine the value for c such that fXY(x, y) is a joint probability density function.
Problems
Example: Four electronic printers are selected from a large lot of damaged printers. Each printer is inspected and
classified as containing either a major or a minor defect. Let the random variables X and Y denote the number of
printers with major and minor defects, respectively. Determine the range of the joint probability distribution of X and
Y.
Problems
Example: In the transmission of digital information, the probability that a bit has high, moderate, and low distortion
is 0.01, 0.04, and 0.95, respectively. Suppose that three bits are transmitted and that the amount of distortion of each
bit is assumed to be independent. Let X and Y denote the number of bits with high and moderate distortion out of the
three, respectively. Determine (i) fXY(x, y) (ii) fY|X=1(y) (iii) E[Y|X=1]
Problems
Example: A missile is fired to hit a target point O. Due to uncertainties associated with firing, the missile can hit any
point within a radius of a units from the point O with equal probability. What is the joint probability density fXY(x, y)
function of the fired points? Find the marginal density functions, fX(x) and fY(y) for the fired points to lie in |x| < a and
|y| < a, respectively. What is the probability of hitting any point within a radius of a/4 from O? What is the probability
of hitting any point, within a distance of a/4 from the y – axis?
Problems
Joint Normal Distribution
Joint Normal Distribution
 MEC 541T–Probabilistic Engineering Design
Module 4: Functions of Random Variables

Siva Prasad AVS,

219 – D, Laboratory Building.
avs@iiitdm.ac.in
Functions of Random Variables

• Given a functional relationship between two random variables, X and Y as Y = g(X), and if the probability density
and cumulative distribution functions fX(x) and FX(x) are known, can we determine fY(y) and FY(y)?

• Given a scalar valued function of a vector, Y=g(X), where X=(X1,X2,….,XN), and the joint distribution fX(x) of
the random vector variable X, can we determine fY(y) and FY(y)?

• Given a vector valued function of a vector, Y=g(X), where X = (X1, X2, …., XN), g = (g1(X), g2(X), …, gM(X))
and the joint distribution fX(x) of the random vector variable X, can we determine the joint distribution for the
random vector variable Y, i.e., fY(y)?
Function of a single random variable
Function of a single random variable
Function of a single random variable
Function of a single random variable
Function of a single random variable
Function of a single random variable
Example 1, Linear relationship: Given Y = a X + b, a, b > 0. Find the density function for Y, if X follows a density
function given by fX(x).
Function of a single random variable
Example 2: The angular speed (not the velocity, because the direction of rotation is fixed) of a rotating flywheel
follows a normal distribution with mean μ and standard deviation σ. Determine the probability density function for the
energy stored in the flywheel
Function of a single random variable
Example 3: X is a discrete random variable which can take values {-3, -2, -1, 0, 1, 2, 3} with probability given by

𝑥
27 1
𝑃 𝑋=𝑥 =
1093 3

Determine the probability mass function for Y = X2.

Function of two random variables – Discrete Case
Function of two random variables – Continuous case
Function of two random variables – Continuous case
Function of two random variables – Sum of two random variables
Example 4: Find the distribution for sum of two independent random variables following Poisson distributions with
means α and β.
Function of two random variables – Sum of two random variables
Example 5: Find the distribution for sum of two independent random variables following Normal distributions with
means μ1 and μ2, standard deviations σ1 and σ2, respectively.
Function of two random variables – Sum of two random variables
Function of two random variables – Sum of two random variables
Function of two random variables – Product of two random variables
Example 6: Given Y = X1X2, determine the distribution for Y.
Function of two random variables – Quotient of two random variables
Example 7: Given Y = X1/X2, determine the distribution for Y. Using this result, determine the distribution for the
stress induced in a uniform bar subjected to a load L with cross–sectional area A, when both L and A follow
exponential distributions with means 1/α and 1/β. It can be assumed that variation in L does not influence the
variation in A.
Function of two random variables – Quotient of two random variables
Function of several random variables
Moments of a function of several random variables
Example 8: What is the mean and variance of a random variable which is the linear combination of n random
variables.
Moments of a function of several random variables
Moments of a function of several random variables
Homework: Determine the mean and variance of the product of two independent random variables.
Mean and Variance of a Generalized Non–linear Function of Several Random
Variables
Mean and Variance of a Generalized Non–linear Function of Several Random
Variables
Mean and Variance of a Generalized Non–linear Function of Several Random
Variables
Mean and Variance of a Generalized Non–linear Function of Several Random
Variables
Homework: The inertia force developed in a reciprocating engine (F) is given by

𝑟 2 𝜔2
𝐹 = 𝑚𝑟𝜔2 cos 𝜔𝑡 + 𝑚 cos 2𝜔𝑡 ,
𝑙
Where, m → mass of the piston, r → length of the crank, l → length of the connecting rod and ω → angular velocity
ഥ = 1 kg, 𝑟ҧ = 7.5 cm, 𝑙 ҧ = 30 cm and 𝜔
of the crank. The means of these variables are 𝑚 ഥ = 157.079 rad/s, and
standard deviations are σm = 0.1 kg, σr = 0.00375 m, σl = 0.015 m, and σω = 15.708 rad/s. Determine the mean and
standard deviation of the inertia force using first order approximations. Assume all the variables are statistically
independent.

Answers:
𝐹ത = 1833.9515 N
σF = 347.5902 N
Moments generating function for a distribution
Moments generating function for a distribution
Example: Moment generating function for a normal distribution
Moments generating function for a distribution
Example: Moment generating function for a normal distribution
Moments generating function for a distribution
Example: Finding the mean and variance of log – normal distribution using the moment generating function of
normal distribution
Functions of several random variables
Functions of several random variables
Functions of several random variables
Example: Given

1 𝑋1
𝑓𝑿 = 2 2 , 𝑌1 = 𝑋1 𝑋2 , 𝑌2 =
𝑥1 𝑥2 𝑋2
Determine the joint distribution for the variables Y1 and Y2.
Functions of several random variables
 MEC 541T–Probabilistic Engineering Design
Module 5: Time – Dependent Reliability

Siva Prasad AVS,

219 – D, Laboratory Building.
avs@iiitdm.ac.in
Failure Rate
• Study of failure rate to know the residual service life of a system
• Study of the reliability of a system

Failure Rate
• Enhancement of the reliability of a system
Examples: Reliability of the usage of a chemical after it has been
stored for some time; Reliability of a component that undergoes
Infant
repeated/cyclic operation conditions etc.
mortality Wearout
Useful life
Learning Objectives of This Module:
Time
o Deriving reliability& hazard rate functions
o Mean–Time–To–Failure (MTTF)
o Expected Residual Life
o Series, Parallel and Complex systems and their analysis
o Reliability enhancement and allocation
Mathematical definitions of reliability and hazard functions
• Time – dependent Reliability at time t can be defined as the probability that a system’s life exceeds t.
• It can also be defined as the number of components survived at time t. If N identical components are considered
out of which Ns(t) have survived up to time t, then,

𝑁𝑠 𝑡 𝑁 − 𝑁𝑓 𝑡
𝑅 𝑡 = =
𝑁 𝑁
Where, Nf(t) is the number components failed up to time t.

𝑑𝑅 1 𝑑𝑁𝑠 𝑡 1 𝑑𝑁𝑓 𝑡
= =−
𝑑𝑡 𝑁 𝑑𝑡 𝑁 𝑑𝑡
Mathematical definitions of reliability and hazard functions
• Hazard rate or hazard function or instantaneous failure rate or failure rate is defined as the ratio of the rate of
increase in the number of failed components to the number of survived components up to time t.

1𝑑𝑁𝑓 𝑡
ℎ 𝑡 =
𝑁𝑠 𝑡 𝑑𝑡
𝑑𝑅
• From the definition of R(t) and ,
𝑑𝑡

𝑁 1 𝑑𝑁𝑓 𝑡 1 𝑑𝑅 𝑡
ℎ 𝑡 = =−
𝑁𝑠 𝑡 𝑁 𝑑𝑡 𝑅 𝑡 𝑑𝑡

Or

𝑡
𝑅 𝑡 = 𝑒𝑥𝑝 − න ℎ 𝑡 𝑑𝑡
0
Reliability and hazard functions from failure time distributions
Relationship between reliability and hazard functions using conditional probability
Hazard rate or hazard function can also be defined as the rate of the probability of an event that a component fails
during the time interval t to t+Δt, given that it has not failed up to t.
Modelling of Failure Rates

Failure Rate
Infant
mortality Wearout
Useful life

Time
Failure rate from empirical data
Mean–Time–To–Failure (MTTF)
Mean–Time–To–Failure (MTTF)
Example: Given the reliability of forging press:

3
𝑡
𝑅 𝑡 = ൞ 1 − 𝑇𝑚𝑎𝑥 ; 0 ≤ 𝑡 ≤ 𝑇𝑚𝑎𝑥

0; Otherwise
Determine fT(t), FT(t), h(t) and MTTF.
Reliability and Hazard Functions for Exponential Distribution
Reliability and Hazard Functions for Exponential Distribution – Problem
Example: Number of failures per hour of a system is 0.01. (i) Find the probability that the system will fail within 25
hours. (ii) If the system has functioned successfully for 200 hours, find the probability that the system will fail in the
next 25 hours.
Reliability and Hazard Functions for Normal Distribution
Reliability and Hazard Functions for Log–Normal Distribution
Reliability and Hazard Functions for Log–Normal Distribution – Problem
Reliability and Hazard Functions for Weibull distribution
Reliability and Hazard Functions for Weibull distribution
Reliability and Hazard Functions for Weibull distribution
Example: The time to failure in hours of a bearing in a gearbox, T, follows Weibull distribution with parameters η =
6000 hours and β = 0.5. Find (i) MTTF (ii) R(t = 7000) and (iii) the time of service at which the bearing has 90%
reliability.
Reliability and Hazard Functions for Gamma distribution
Reliability and Hazard Functions for Rayleigh distribution
Expected Residual (Remaining) Life
L(t0) → Expected life remaining beyond t0, is E[T–t0|T>t0], where T is the random variable denoting the life of a
component/system
Expected Residual (Remaining) Life
Expected Residual (Remaining) Life
Expected Residual (Remaining) Life
Reliability Analysis of Systems in Series
Reliability Analysis of Systems in Series
Reliability Analysis of Systems in Series
Reliability Analysis of Systems in Parallel
Reliability Analysis of Systems in Parallel
Reliability Analysis of (k, n) systems
Reliability Analysis of (k, n) systems
Reliability Analysis of (k, n) systems
Reliability Analysis of Complex Systems
1. Enumeration method 2. Conditional probability method 3. Cut – set method
Reliability Analysis of Complex Systems
1. Enumeration method 2. Conditional probability method 3. Cut – set method
Reliability Analysis of Complex Systems
1. Enumeration method 2. Conditional probability method 3. Cut – set method
Reliability Analysis of Complex Systems
1. Enumeration method 2. Conditional probability method 3. Cut – set method
Reliability Analysis of Complex Systems
1. Enumeration method 2. Conditional probability method 3. Cut – set method
Reliability Analysis of Complex Systems
1. Enumeration method 2. Conditional probability method 3. Cut – set method
Reliability Analysis of Complex Systems
1. Enumeration method 2. Conditional probability method 3. Cut – set method
Reliability Enhancement – Systems in Series
Reliability Enhancement – Systems in Series
Reliability Enhancement – Systems in Series
Reliability Enhancement – Systems in Parallel
Reliability Enhancement – Systems in Parallel
Reliability Allocation – Agree method
Subsystem Number of Operating period in Probability of System Failure Due to
(i) components (ni) hours (ti) Failue of ith subsystem (wi)
1 5 10 0.15
2 2 25 0.1
3 8 5 0.2
4 6 20 0.05
5 4 15 0.025
Reliability Allocation – Agree method

Desired overall reliability R0 = 0.99

Subsystem Number of Operating Probability of System For a desired overall reliability R0,
(i) components period in Failure Due to Failue desired expected life time in hours
(ni) hours (ti) of ith subsystem (wi) for each subsystem calculated as per
AGREE method (mi)
1 5 10 0.15 746.24
2 2 25 0.1 3109.348
3 8 5 0.2 310.9348
4 6 20 0.05 414.58
5 4 15 0.025 2332.0116
 MEC 541T–Probabilistic Engineering Design
Module 6: Strength Based Reliability

Siva Prasad AVS,

219 – D, Laboratory Building.
avs@iiitdm.ac.in
Reliability of a component based on strength
• The word “Strength” refers to the threshold value above (or below) which the system does not work.
❑ Examples: Capacity of a dam, container, motor; Maximum allowable deflection; Minimum Required Strength

• The word “Load” condition or state of the system under service.

Reliability of a component based on strength
• If S and L, respectively, denote the random variables representing the “Strength” and “Load”, then the reliability
of the system is R = P(S>L).
Reliability of a component based on strength
• If S and L, respectively, denote the random variables representing the “Strength” and “Load”, then the reliability
of the system is R = P(S>L).
Reliability of a component based on strength
Example: If S and L follow normal distribution, what is the reliability? The torque transmitting capacity of a cone
clutch follows normal distribution with an average at 396.33 lb–in and standard deviation 26.4225 lb – in. The torque
required for operation of the clutch is also normally distributed with mean 400 lb – in and standard deviation 40 lb –
in, what is the reliability of the working of the clutch?
Reliability of a component based on strength
Example: If S and L follow normal distribution, what is the reliability? The torque transmitting capacity of a cone
clutch follows normal distribution with an average at 396.33 lb–in and standard deviation 26.4225 lb – in. The torque
required for operation of the clutch is also normally distributed with mean 400 lb – in and standard deviation 40 lb –
in, what is the reliability of the working of the clutch?
Reliability of a component based on strength
Example: A mechanical fuse has a specially designed screw head that is required to be tightened by applying a mean
torque of 400 lb–in. The fuse is permitted to have a maximum of 100 failures per 50000 units. Find the torque at
which the fuse is to be designed to fail if the standard deviation of shear strength of the material of fuse, expressed in
terms of applied torque, is 15 lb–in. The standard deviation of the torque wrenches used to tighten the screw head of
the fuse is 10 lb–in of torque. Both applied torque and shear strength follow normal distribution.
Reliability of a component based on strength
Example: S and L follow log – normal distribution.
Reliability of a component based on strength
Example: S and L follow log – normal distribution.
Reliability of a component based on strength
Example: The bending stress developed in a gear tooth is given by the Lewis equation,

𝐹𝑡 𝑝𝑑
𝑆= ,
𝑤𝑌
Where Y is the form factor, Ft is the tangential force; pd is the diametrical pitch and w is the face width of the gear.
The tangential load is observed to follow lognormal distribution with mean 3720 lb and a coefficient of variation
0.080645. The yield strength of the material is also known to follow lognormal distribution with a mean value of
32000 lb/in2. Find the maximum permissible value of the coefficient of variation of yield strength required in order to
have a reliability of 0.9999 for the gear tooth. Other parameters in the above equation are w = 0.5 in, form factor Y =
0.3 and diametrical pitch = 1.
Reliability from experiments
Factory of Safety For a Given Reliability
Can we related factory of safety to the desired reliability?
Factory of Safety For a Given Reliability
Can we related factory of safety to the desired reliability?
Reliability of systems involving more than two random parameters
Reliability of systems involving more than two random parameters
First order second moment method
Hasofer – Lind Reliability Index With Two Normally Distributed Variables
Hasofer – Lind Reliability Index With Two Normally Distributed Variables
Hasofer – Lind Reliability Index With Several Normally Distributed Variables
Hasofer – Lind Reliability Index With Several Normally Distributed Variables
Hasofer – Lind Reliability Index With Several Normally Distributed Variables
Hasofer – Lind Reliability Index With Several Normally Distributed Variables
Example Problems
Determine the reliability when S and L follow exponential distribution with means μS and μL.
Example Problems
Find the reliability and the corresponding central factor of safety of a component for which the mean strength is
15000 lb/in2, mean load is 10000 lb/in2 and the corresponding standard deviations, 3000 lb/in2 and1000 lb/in2 when
both the strength and load follow normal distribution. What should be the factory safety for a reliability of 0.99?
Example Problems
The split piston ring, shown in the figure below, is mounted on a piston. The tangential deflection of the split ring
under a force F is given by,
3𝜋𝐹𝑅3 𝜋𝐹𝑅 6𝜋𝐹𝑅
𝛿= + + ,
𝐸𝐼 𝐸𝐴 5𝐺𝐴
In the above equation, the three terms are contributions from bending moment, axial load and transverse shear. R, the
radius of the piston ring is 2 in, E = 18000 ksi, b = 0.2 in (width of cross – section), h = 0.3 in (height of cross –
section), G = 7000 ksi. The tangential force is normally distributed with mean 50 lb and standard deviation 10 lb. Find
the probability of the two ends X and Y overlapping if the installed gap is 0.45 in.
Example Problems
4. Solution
Example Problems
The maximum stress induced near the hole in a rectangular plate of width w and thickness t subjected to an axial load
P is given by smax = kts0, where kt is the stress concentration factor and s0 = P/{(w–d)t}is the nominal stress induced in
the cross – section far away from the hole. If both the stress concentration factor and yield strength are normally
distributed with means 2.5 and 30 ksi; standard deviations 0.1 and 1 ksi, respectively, find the reliability of the plate
with dimensions w = 10 in, d = 2 in, t = 0.1 in subjected to a load P = 10000 lb.
Reliability of Structures
Reliability of Structures
Reliability of Structures
Reliability of Structures
Reliability of Structures
Reliability of Structures
Reliability of Structures
L
Reliability of Structures
Reliability of Structures L

9m

12 m
3.75 m
Reliability of Structures
Reliability of Structures
Reliability of Structures
A chain consisting of three identical links carries a load L as shown below. The strengths of the links are uniformly
distributed over the range 1 to 6 units and the load L is also uniformly distributed over the range 0 to 5 units. What is
the probability of failure of a link? What is the probability of failure of the chain?
Reliability of Structures
Reliability of Structures
Reliability of Structures
 MEC 541T–Probabilistic Engineering Design
Module 7: Probabilistic Design

Siva Prasad AVS,

219 – D, Laboratory Building.
avs@iiitdm.ac.in
Probability and Reliability Based Design
• To design for a given probability of failure (or reliability)

• To estimate the probability of failure and the bounds

Probability and Reliability Based Design
A cantilever beam of circular cross–sectional area is subjected to a load W = 1 kN at the free end. The length of the
cantilever is L = 1 m. The yield strength of the material is SY = 250 MPa.
(i) Determine the diameter in a deterministic design approach with a factor of safety n = 1.25.
(ii) If the load W and the yield strength SY are uniformly distributed over the ranges [0.7, 1.3] kN and [200, 300] MPa,
respectively, determine the diameter for a maximum probability of failure of 0.3.
(iii) The maximum and minimum values of the load and yield strength variation given in (ii) are obtained from 5
different lots of data. Determine the probability distributions for minimum strength and maximum load.
Probabilistic Fatigue Design
16 kN • Fillet Radius at C = 1 mm and the geometric stress
C D concentration factor is 1.75.
A B
• Mean ultimate tensile strength is 600 MPa
50 mm 75 mm
• Mean load is 16 kN with a standard deviation of 1.6
0.2 m 0.2 m 0.4 m kN
• Assume all the random variables to be normally
distributed
Probabilistic Analysis of a Gear Train
 MEC 541T–Probabilistic Engineering Design
Module 8: Monte Carlo Simulation

Siva Prasad AVS,

219 – D, Laboratory Building.
avs@iiitdm.ac.in
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers
Generation of Random Numbers