Quality Control

Assurance and
Reliability
Dr. C.Phaneendra Kiran
Ph.D
BITS Pilani Department of Mechanical Engineering
BITS Pilani Hyderabad Campus
Pilani Campus
 Chapter 4: (Part 2)
Fundamentals of statistical concepts and techniques in
BITS Pilani quality control and improvement
Pilani Campus
Measures of Association

BITS Pilani, Pilani Campus

Probability Distribution

BITS Pilani, Pilani Campus

CDF

BITS Pilani, Pilani Campus

Hyper geometric distibution

BITS Pilani, Pilani Campus

Binomial distribution

BITS Pilani, Pilani Campus

Poisson distribution

BITS Pilani, Pilani Campus

Normal distribution

BITS Pilani, Pilani Campus

Normal distribution

BITS Pilani, Pilani Campus

Exponential distribution

BITS Pilani, Pilani Campus

Weibull distribution

BITS Pilani, Pilani Campus

Example 26 (normal distribution)

The length of a machined part is known to

have a normal distribution with a mean of
100mm and a standard deviation of 2mm

a) What proportion of the part will be above

103.3mm?

BITS Pilani, Pilani Campus

Example 26 - a Solution:

BITS Pilani, Pilani Campus

Example 26 - b
What proportion of the out put will be
between 98.5 and 102.0mm?

BITS Pilani, Pilani Campus

Example 26 - b solution

BITS Pilani, Pilani Campus

Example 26 - c
What proportion of the parts will be shorter than
96.5 mm?

BITS Pilani, Pilani Campus

Example 26 - c Solution

BITS Pilani, Pilani Campus

Example 26 - d

It is important that not many of the parts

exceed the desired length. If a manager
stipulates that no more than 5% of the parts
should be oversized, what specification limit
should be recommended?

BITS Pilani, Pilani Campus

Example 26 - d Solution:

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus
Example 27 (Exponential distribution)
It I known that a battery for a video game has an
average life of 500 hours (h). The failures of
batteries are known to be random and
independent and may be described by an
exponential distribution.

a) Find the probability that a battery will last at

least 600 h.

BITS Pilani, Pilani Campus

Example 27 – a Solution

BITS Pilani, Pilani Campus

Example 27 – b
Find the probability of a battery failing within
200 h.

BITS Pilani, Pilani Campus

Example 27 –b Solution

BITS Pilani, Pilani Campus

Example 27- c
Find the probability of battery lasting
between 300 and 600 h.

BITS Pilani, Pilani Campus

Example 27- c Solution

BITS Pilani, Pilani Campus

Example 27- d
Find the standard deviation of the life of a battery.

BITS Pilani, Pilani Campus

Example 27- d Solution

BITS Pilani, Pilani Campus

Example 27- e
If it is known that a battery has lasted 300 h,
what is the probability that it will last at least
500 h?

BITS Pilani, Pilani Campus

Example 27- e Solution

BITS Pilani, Pilani Campus

Example 28 (weibull distribution )

BITS Pilani, Pilani Campus

Example 28 – a Solution :

BITS Pilani, Pilani Campus

Example 28 – b
What is the probability of a tube operating
for at least 800 hr

BITS Pilani, Pilani Campus

Example 28 – b Solution

BITS Pilani, Pilani Campus

sample distribution

BITS Pilani, Pilani Campus

Example 29 (sample distribution)
The tuft bind strength of synthetic material
used to make carpet is known to have a
mean of 100lb and a standard and a
standard deviation of 20lb. if sample of
size 40 is randomly selected, what is the
probability that the sample mean will be
less than 105 lb?

BITS Pilani, Pilani Campus

Solution

BITS Pilani, Pilani Campus

Interval estimation

BITS Pilani, Pilani Campus

Example 30 (interval estimation)
The output voltage of a power source is
known to have a standard deviation of
10V. Fifty reading are randomly selected,
yielding an average of 118V. Find a 95%
confidence interval for the population
mean voltage

BITS Pilani, Pilani Campus

Solution

BITS Pilani, Pilani Campus

Confidence interval

BITS Pilani, Pilani Campus

Confidence interval

BITS Pilani, Pilani Campus

Example 31 (t- ditribution)
A new process has been developed that transforms ordinary
iron into a kind of super iron called metallic glass. This new
product is stronger than steel alloys and is much more
corrosion-resistant than steel. However it has a tendency to
become brittle at high temperature. It is desired to estimate
the mean temperature at which it become brittle. A random
sample of 20 pieces of metallic glass is selected. The
temperature at which brittleness is first detected is recorded
for each piece. The summary results give a sample mean X̄
of 600oF and a sample standard deviation s of 15oF. find
90% confidence interval for the mean temperature at which
metallic glass become brittle.

BITS Pilani, Pilani Campus

Solution

BITS Pilani, Pilani Campus

Confidence interval for the
difference between two means

1.Variance known

First population n1 , X1 ,X1 , 1 ( known),  12 (unknown)

Second population n 2 , X 2 , X 2 ,  2 ( known),  22 (unkown)
The confidence interval for differnce between two means is given by

BITS Pilani, Pilani Campus

2.Variance unknown

Unknown variances are assumed to be equal

The pooled estimate of the common variance obtained by

Population variances are not equal

BITS Pilani, Pilani Campus

3. Paired samples

• Till now we have assumed samples are independent

• In some cases we can pair sample to reduce impact of
extraneous factors
• Ex: cholesterol level of a patient before and after using
drug
• X1i and X2i are n-paired observations
• The difference is di=X1i-X2i
• If mean and SD are represented by d, sd the confidence
interval is given by

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus
BITS Pilani, Pilani Campus
Confidence interval for
proportion
^

• the sample proportion of nonconforming items which is

p

equal to x/n
• Where x is nonconforming items and n is no. of trials
• Two-sided confidence interval is given by

• IF n is small, binomial table should be referred for

confidence interval
• If n is large and p is small np<5 Poisson approximation
• If [np≥5, n(1-p) ≥5] normal distribution serves as good
approximation

BITS Pilani, Pilani Campus

CI for difference between
two different proportions
• If sample size n1 and p1, n2 and p2 are selected from a
binomial population and sample sizes are large
• The confidence interval p1-p2 can be expressed as

BITS Pilani, Pilani Campus

Problem

BITS Pilani, Pilani Campus

sol

BITS Pilani, Pilani Campus

Confidence interval for the
variance
• Consider random variable X from a ND with mean µ and
variance σ2 (unknown)
• s2 is the estimator of σ2 and this sampling distribution is
known as chi-squared distribution with (n-1) dof

• Generally Chi-square distribution is skewed to right and

it is dependent on the number of DOF
• Two sided confidence interval for the population is

BITS Pilani, Pilani Campus

Problem

BITS Pilani, Pilani Campus

Solution

BITS Pilani, Pilani Campus

Example

BITS Pilani, Pilani Campus

CI for the ratio of two
variances
F-distribution
Radom variables X1 and X2 from two populations with
normal distributions with means and variances
The ratio of sample variance to population variances is
known as F distribution

Two sided confidence interval is given by

The lower tail F-value can be obtained by

BITS Pilani, Pilani Campus

F-distribution

BITS Pilani, Pilani Campus

Problem

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus
BITS Pilani, Pilani Campus
Hypothesis Testing

Determine whether claims on product or process parameters

are valid
Test statistic: a sample statistic used Ex: sample mean length
Ex: For the month of Jan we want to determine mean length
differs from 30 mm
Null hypothesis (Ho): status quo, or circumstance being tested
(which is not rejected unless proven incorrect)
Alternate hypothesis (Ha): what we wish to prove or
established
Two tailed test: departure of parameter is tested in both
directions

BITS Pilani, Pilani Campus

One-tailed test: alternative hypothesis detects departure of a
parameter from a specified value in only one direction
Ex: whether average length exceeds or less than 30mm

Rejection region: If the given parameter falls below critical

value we will reject null hyphothesis

BITS Pilani, Pilani Campus

Level of significance: The rejection region is chosen such
that if the null hypothesis is true, the probability of test
static falling in that region is small (say, 0.01 or 0.05)
Errors in hypothesis:
Type I error: the null hypothesis is rejected when it is
actually true
Probability of Type I error is indicated by α=P(Type I error)=
P(rejecting Ho| Ho is true)
Type II error: the null hypothesis is not rejected even
though it is false
Probability of Type II error is denoted by β

BITS Pilani, Pilani Campus

Steps in hypothesis testing

Step 1: Formulate the null and alternative hypotheses

Step 2: Determine the test statistic
Step 3: Determine the rejection region of null hypothesis
based on a chosen level of significance α
Step 4: Make the decision. If the test statistic lies in the
rejection region, reject the null hypothesis. Otherwise do
not reject the null hypotheses

BITS Pilani, Pilani Campus

Hypotheses testing of mean

Variance known

BITS Pilani, Pilani Campus

Variance unknown

BITS Pilani, Pilani Campus

Problem

BITS Pilani, Pilani Campus

Solution

BITS Pilani, Pilani Campus

Hypotheses testing of
correlation coefficient
• Bivariate normal data on two variables X and Y and a sample of n
observations
• The population correlation coefficient between X and Y be denoted by ρ,
estimator of the sample correlation coefficient r

BITS Pilani, Pilani Campus

Hypothesis testing for
difference between two means
Variances known

BITS Pilani, Pilani Campus

Problem

BITS Pilani, Pilani Campus

Solution

BITS Pilani, Pilani Campus

Variance unknown

BITS Pilani, Pilani Campus

Variance unknown

BITS Pilani, Pilani Campus

Problem

BITS Pilani, Pilani Campus

Solution

BITS Pilani, Pilani Campus

Paired samples

BITS Pilani, Pilani Campus

Hypothesis testing for a
proportion
Assumption: number of trials in binomial experiment is
large

BITS Pilani, Pilani Campus

Problem

BITS Pilani, Pilani Campus

Difference between two
binomial proportions

BITS Pilani, Pilani Campus

Problem

BITS Pilani, Pilani Campus

Hypothesis testing for
variance

BITS Pilani, Pilani Campus

Problem

BITS Pilani, Pilani Campus

Hypothesis testing for ratio
of variances

BITS Pilani, Pilani Campus

Principle

BITS Pilani, Pilani Campus

BITS Pilani, Pilani Campus