IEM 4103 Quality Control & Reliability Analysis

IEM 5103 Breakthrough Quality & Reliability

Modeling Process Quality

Instructor: Dr. Chenang Liu

Email: Chenang.Liu@okstate.edu
 Outline

 The DMAIC Process

 Describing Variation
 Histogram
 Numerical Summary of Data
 Probability Distributions
 Important Probability Distributions
 Discrete Distribution
 Continuous Distribution
 Some Useful Approximations
 Textbook: Chapter 2,3
1/20/2021

2
 The DMAIC Process

 DMAIC is a structured problem-solving procedure widely used in quality and

process improvement. (a very general procedure)
But not necessarily!
 It is often associated with Six Sigma activities, and almost all implementations of
Six Sigma use the DMAIC process for project management and completion.
 The letters DMAIC form an acronym for the five steps:
 Define: identify the project opportunity.
 Measure: evaluate and understand the current state of the process.
 Analyze: determine the cause-and-effect relationships and understand the variability.
 Improve: creative thinking about the specific changes to have the desired impact.
 Control: ensure that the gains are of help in the process.
1/20/2021 Control charts are an important statistical
3 tool used in the Control step of DMAIC
 The DMAIC Process (cont’d)

1/20/2021

4
 2.1 Describing Variation

 Method 1: Statistical Plot - Histogram

 Method 2: Numerical Summary
 Method 3: Probability Distribution

1/20/2021

5
 Population & Sample

 A population is the entire collection of units/individuals/outcomes in which we

are interested. It is usually very large (and sometimes infinite), so to find out
what is going on in the population we observe a sample - a representative
subset.
 The key word here is representative. A sample should be 'the population in
miniature'. Then by examining a sample we can draw conclusions about the
population. Such conclusions, however, CANNOT be made with 100% certainty
and are stated in terms of probabilities.

Probability
Population Sample
Statistical inference
1/20/2021

6
 2.1.1 Histogram

 Example
 Measured variable: thickness of metal layer on silicon wafer
 Samples: 100 observations

1/20/2021

7
 To construct a histogram

 Find the range of the data

 Min: 413 Max: 487
 Divide the rage of data into equal-width intervals (bins)
 In practice, number of bins≈ 𝑁𝑁 = 100 = 10
487−413
 Width of bins= = 7.4, so use width= 8
10

 Determine the lower and upper bounds of each bin

 Bin 1: [410, 417] 1
 Bin 2: [418, 425] 3
 … …
 Bin 10: [482, 489] 1
 Count the frequency of each bin
1/20/2021

8
 Interpretation based on the Histogram

 Histogram with different bins

 8 bins (top) and 15 bins (bottom)

 Visual display of three properties of sample data

 Shape of distribution
 Roughly symmetric and unimodal
 Central tendency
 Data tend to cluster near 450
 Scatter or variability
 Variability is relatively high (min=413; max=487)

1/20/2021

9
 2.1.2 Numerical Summary of Data

 Central Tendency: sample average/mean

𝑥𝑥1 + 𝑥𝑥2 + ⋯ + 𝑥𝑥𝑛𝑛 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑖𝑖
𝑥𝑥̅ = =
𝑛𝑛 𝑛𝑛
 Scatter/variability: sample variance or sample standard deviation
∑ 𝑛𝑛 2
𝑖𝑖=1 (𝑥𝑥𝑖𝑖 − 𝑥𝑥)
̅
𝑠𝑠 2 = = 𝜎𝜎̂ 2 Not reflect the magnitude
𝑛𝑛 − 1 of data, only the scatter
∑ 𝑛𝑛 2 about the average.
(𝑥𝑥
𝑖𝑖=1 𝑖𝑖 −𝑥𝑥)
̅
𝑠𝑠 = 𝑠𝑠 2 =
𝑛𝑛 − 1
 Median: a value such that at least 50% of the data values are at or below this value
and at least 50% of the data values are at or above this value.
1/20/2021

10
 Example

 Calculate the sample mean, median, variance, and standard deviation of a

sample of observations: 𝑥𝑥1 = 1, 𝑥𝑥2 = 3, 𝑥𝑥3 = 5

 If 𝑥𝑥3 is 500 instead of 5, what is the sample mean and median of the sample?

 If 𝑥𝑥1 = 101, 𝑥𝑥2 = 103, 𝑥𝑥3 = 105, is the sample variance different from the first
sample?

1/20/2021

11
 Numerical Summary of Data - More

 Skewness
 If the distribution of the data is not symmetrical it is called asymmetrical or skewed
 Skewness characterizes the degree of asymmetry of a distribution around its mean
𝑛𝑛
𝑛𝑛 𝑥𝑥𝑖𝑖 − 𝑥𝑥̅ 3
𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = �( )
(𝑛𝑛 − 1)(𝑛𝑛 − 2) 𝑠𝑠
𝑖𝑖=1

 Examples

1/20/2021

12
 Numerical Summary of Data - More

 Kurtosis
 Kurtosis characterizes the relative peakedness or flatness of a distribution
compared with the bell-shaped distribution (normal distribution)

 Examples

1/20/2021

13
 2.1.3 Probability Distributions

 Definition: a probability distribution is a mathematical model that relates the value

of the variable with the probability of occurrence of that value in the population.
 Two types of distributions:
 Continuous: if the variable being measured is expressed on a continuous scale
 Discrete: if the parameter being measured can only take on certain values, e.g. 1,2,3…

 Examples:
 Discrete case (left)
 Continuous case (right)

1/20/2021

14
 Review of probability distribution calculation

Continuous Distribution Discrete Distribution

𝑏𝑏
Probability 𝑃𝑃{𝑎𝑎 ≤ 𝑥𝑥 ≤ 𝑏𝑏} = � 𝑓𝑓 𝑥𝑥 𝑑𝑑𝑑𝑑 𝑃𝑃 𝑥𝑥 = 𝑥𝑥𝑖𝑖 = 𝑝𝑝(𝑥𝑥𝑖𝑖 )
𝑎𝑎

∞ ∞
Population
𝐸𝐸 𝑥𝑥 = 𝜇𝜇 = � 𝑥𝑥𝑥𝑥 𝑥𝑥 𝑑𝑑𝑑𝑑 𝐸𝐸 𝑥𝑥 = 𝜇𝜇 = � 𝑥𝑥𝑖𝑖 𝑝𝑝(𝑥𝑥𝑖𝑖 )
Mean/Expectation −∞
𝑖𝑖=1
∞ ∞

Population Variance 𝑉𝑉𝑉𝑉𝑉𝑉 𝑥𝑥 = 𝜎𝜎 2 = � (𝑥𝑥 − 𝜇𝜇)2 𝑓𝑓 𝑥𝑥 𝑑𝑑𝑑𝑑 𝑉𝑉𝑉𝑉𝑉𝑉 𝑥𝑥 = 𝜎𝜎 2 = �(𝑥𝑥𝑖𝑖 −𝜇𝜇)2 𝑝𝑝(𝑥𝑥𝑖𝑖 )
−∞
𝑖𝑖=1
𝑥𝑥1 + 𝑥𝑥2 + ⋯ + 𝑥𝑥𝑛𝑛 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑖𝑖
Sample mean 𝑥𝑥̅ = =
𝑛𝑛 𝑛𝑛
𝑛𝑛
∑ ̅ 2
𝑖𝑖=1(𝑥𝑥𝑖𝑖 −𝑥𝑥)
Sample variance 2
𝑠𝑠 = = 𝜎𝜎̂ 2
1/20/2021 𝑛𝑛 − 1
15
 Useful results on mean and variance

 If 𝑥𝑥 is a random variable and 𝑎𝑎 is a constant, then

 𝐸𝐸 𝑥𝑥 + 𝑎𝑎 = 𝐸𝐸 𝑥𝑥 + 𝑎𝑎
 𝐸𝐸 𝑎𝑎𝑥𝑥 = 𝑎𝑎𝐸𝐸 𝑥𝑥
 𝑉𝑉𝑉𝑉𝑉𝑉 𝑥𝑥 + 𝑎𝑎 = 𝑉𝑉𝑉𝑉𝑉𝑉 𝑥𝑥
 𝑉𝑉𝑉𝑉𝑉𝑉 𝑎𝑎𝑥𝑥 = 𝑎𝑎 2 𝑉𝑉𝑉𝑉𝑉𝑉 𝑥𝑥

 If 𝑥𝑥1 , 𝑥𝑥2 , ⋯ , 𝑥𝑥𝑛𝑛 are random variables, then

 𝐸𝐸 𝑥𝑥1 + 𝑥𝑥2 + ⋯ + 𝑥𝑥𝑛𝑛 = 𝐸𝐸(𝑥𝑥1 ) + 𝐸𝐸(𝑥𝑥2 ) + ⋯ + 𝐸𝐸(𝑥𝑥𝑛𝑛 )

 If they are mutually independent, and 𝑎𝑎1 , 𝑎𝑎2 , ⋯ , 𝑎𝑎𝑛𝑛 are constant, then
 𝑉𝑉𝑉𝑉𝑉𝑉 𝑎𝑎1 𝑥𝑥1 + 𝑎𝑎2 𝑥𝑥2 + ⋯ + 𝑎𝑎𝑛𝑛 𝑥𝑥𝑛𝑛 = 𝑎𝑎12 𝑉𝑉𝑉𝑉𝑉𝑉(𝑥𝑥1 ) + 𝑎𝑎22 𝑉𝑉𝑉𝑉𝑉𝑉(𝑥𝑥2 ) + ⋯ + 𝑎𝑎𝑛𝑛2 𝑉𝑉𝑉𝑉𝑉𝑉(𝑥𝑥𝑛𝑛 )

1/20/2021

16
 2.2 Important Probability Distributions

 Discrete Probability Distribution

 Hypergeometric distribution
 Binomial distribution
 Poisson distribution

 Continuous Probability Distribution

 Normal distribution
 Chi-square distribution
 𝑡𝑡 distribution
 𝐹𝐹 distribution

1/20/2021

17
 Hypergeometric Distribution

 Example: Suppose that there are 𝑁𝑁 balls in total, including 𝐷𝐷 red balls. 𝑛𝑛 balls
are randomly selected without replacement, and the number of red balls – say
𝑥𝑥 is observed. The probability distribution of 𝑥𝑥 is Hypergeometric.
 Definition:
𝐷𝐷 𝑁𝑁 − 𝐷𝐷
𝑝𝑝 𝑥𝑥 = 𝑥𝑥 𝑛𝑛 − 𝑥𝑥 𝑥𝑥 = 0,1,2, ⋯ , min(𝑛𝑛, 𝐷𝐷)
𝑁𝑁
𝑛𝑛
 The mean and variance of the distribution are
𝑛𝑛𝑛𝑛 𝑛𝑛𝑛𝑛 𝐷𝐷 𝑁𝑁 − 𝑛𝑛
𝜇𝜇 = and 𝜎𝜎 2 = (1 − )( )
𝑁𝑁 𝑁𝑁 𝑁𝑁 𝑁𝑁 − 1

1/20/2021

18
 Binomial Distribution

 Example: A quarter coin is tossed for 𝑛𝑛 times (𝑛𝑛 independent trials). For each
toss, suppose the probability to get the obverse is 𝑝𝑝. The number of “obverse”
in these 𝑛𝑛 trials—say, 𝑥𝑥—is observed. Then the probability distribution of 𝑥𝑥 is
Binomial.
 Definition:
𝑛𝑛 𝑥𝑥
𝑝𝑝 𝑥𝑥 = 𝑝𝑝 (1 − 𝑝𝑝)𝑛𝑛−𝑥𝑥 𝑥𝑥 = 0,1,2, ⋯ , 𝑛𝑛
𝑥𝑥
 The mean and variance of the distribution are
𝜇𝜇 = 𝑛𝑛𝑛𝑛 and 𝜎𝜎 2 = 𝑛𝑛𝑛𝑛(1 − 𝑝𝑝)

1/20/2021

19
 Binomial Distribution (cont’d)

 Assumption:
 Constant probability of success 𝑝𝑝
 Two mutually exclusive outcomes
 All trials statistically independent
 Number of trials 𝑛𝑛 is known and constant

 Application:
 It can be used as a model when sampling from an infinitely large population. The
constant 𝑝𝑝 normally represents the fraction of defective or nonconforming items in
the population.

1/20/2021

20
 Example

 A firm claims that 99% of their products meet specifications. To support this claim, an
inspector draws a random sample of 20 items and ships the lot if the entire sample is
in conformance. Find the probability of committing both of the following errors:
 (1) Refusing to ship a lot even though 99% of the items are in conformance.
 (2) Shipping a lot even though only 95% of the items are conforming.

1/20/2021

21
 Poisson Distribution

 Example: Suppose that the number of wire-bonding defects per unit that occur
in a semiconductor device (i.e., 𝑥𝑥) is Poisson distributed with parameter 𝜆𝜆 = 4
 Definition: the number of random events occurring during a specific “time”
period with the average occurrence rate 𝜆𝜆 known:
𝑒𝑒 −𝜆𝜆 𝜆𝜆𝑥𝑥
𝑝𝑝 𝑥𝑥 = 𝑥𝑥 = 0,1,2, ⋯
𝑥𝑥!
 The mean and variance of the distribution are
𝜇𝜇 = 𝜆𝜆 and 𝜎𝜎 2 = 𝜆𝜆

1/20/2021

22
 Poisson Distribution (cont’d)

 Assumptions:
 The average occurrence rate 𝜆𝜆 (per unit) is a known constant
 Occurrences are equally likely to occur within any unit of time/area
 Occurrences are statistically independent

Poisson probability distributions

for selected values of 𝜆𝜆

1/20/2021

23
 Exercises for discrete distribution

 What is the distribution of x in the following scenarios?

 60% of pulleys are produced using Lathe #1, 40% are produced using Lathe #2. A
random sample of four production parts containing 𝑥𝑥 parts coming from Lathe #1.
 Accidents in a building are assumed to occur randomly with an average rate of 36
per year. There will be 𝑥𝑥 accidents in the coming April.
 The probability that a basketball player will make a free throw is 0.7. Let 𝑥𝑥 denote the
number of free throws he will make in a game of 7 free throw attempts.
 A production process operates with 2% nonconforming output. Every hour a sample
of 50 units of product is taken, and the number of nonconforming units counted as 𝑥𝑥.
 A book of 200 pages with 2 error pages. There are 𝑥𝑥 error pages in a random
selection of 10 pages.

1/20/2021

24
 Normal Distribution (Continuous)

 The normal distribution is probably the most important distribution in both the
theory and application of statistics.
 Definition: if 𝑥𝑥 is a normal random variable, then the probability distribution
of 𝑥𝑥 is defined as follows,
1 1 𝑥𝑥−𝜇𝜇 2
𝑓𝑓 𝑥𝑥 = − (
𝑒𝑒 2 𝜎𝜎 ) − ∞ < 𝑥𝑥 < ∞
𝜎𝜎 2𝜋𝜋
 The mean of the normal distribution is 𝜇𝜇 and the variance is 𝜎𝜎 2 > 0. Also,
2
𝑥𝑥 − 𝜇𝜇
𝑥𝑥~𝑁𝑁 𝜇𝜇, 𝜎𝜎 ⇒ 𝑧𝑧 = ~𝑁𝑁(0,1)
𝜎𝜎 Standard normal distribution
𝑎𝑎 − 𝜇𝜇 𝑎𝑎 − 𝜇𝜇
𝑃𝑃 𝑥𝑥 ≤ 𝑎𝑎 = 𝑃𝑃 𝑧𝑧 ≤ = Φ( )
𝜎𝜎 𝜎𝜎
1/20/2021

25
 Normal Distribution (cont’d)

 If 𝑥𝑥1 and 𝑥𝑥2 are independently normally distributed random variables, then
𝑦𝑦 = 𝑎𝑎1 𝑥𝑥1 + 𝑎𝑎2 𝑥𝑥2 also follows the normal distribution, i.e.,
𝑦𝑦~𝑁𝑁(𝑎𝑎1 𝜇𝜇1 + 𝑎𝑎2 𝜇𝜇2 , 𝑎𝑎12 𝜎𝜎12 + 𝑎𝑎22 𝜎𝜎22 )
 The visual appearance of the normal distribution is a symmetric, unimodal or
bell-shaped curve

𝑃𝑃 𝜇𝜇 − 𝜎𝜎 ≤ 𝑥𝑥 ≤ 𝜇𝜇 + 𝜎𝜎 = 68.26%
𝑃𝑃 𝜇𝜇 − 2𝜎𝜎 ≤ 𝑥𝑥 ≤ 𝜇𝜇 + 2𝜎𝜎 = 95.46%
𝑃𝑃 𝜇𝜇 − 3𝜎𝜎 ≤ 𝑥𝑥 ≤ 𝜇𝜇 + 3𝜎𝜎 = 99.73%
1/20/2021

26
 Normal Distribution (cont’d)

 Example for calculation: 𝑥𝑥~𝑁𝑁(40, 52 )

42.1 − 40
𝑝𝑝 𝑥𝑥 ≥ 42.1 = 1 − 𝑝𝑝 𝑥𝑥 ≤ 42.1 = 1 − Φ = 1 − Φ(0.42)
5

Check the table in Appendix II, the

table will be provided in exams.

1/20/2021

27
 Example

 Three shafts are made and assembled in a linkage. The length of each shaft,
in centimeters, is distributed as follows:
 Shaft 1: 𝑁𝑁(75,0.09)
 Shaft 2: 𝑁𝑁(75,0.16)
 Shaft 3: 𝑁𝑁(75,0.25)
 Assume the shafts’ length are independent to each other:
 (a) What is the distribution of the linkage?
 (b) What is the probability that the linkage will be longer than 160.5 cm?

1/20/2021

28
 Central limit theorem

 In practice, the normal distribution is often assumed as the appropriate

probability model for a random variable. Why?
 The central limit theorem is often a justification of approximate normality.
 The distribution of an average tends to be Normal, even when the distribution from
which the average is computed is decidedly non-Normal.

 Thus, the Central Limit Theorem is the foundation for many statistical
procedures, including Quality Control Charts, because the distribution of the
phenomenon under study does not have to be Normal because its average will
be.

1/20/2021

29
 Central limit theorem (cont’d)

 Mathematical representation:
 If 𝑥𝑥1 , 𝑥𝑥2 , ⋯ , 𝑥𝑥𝑛𝑛 are independent random variables with mean 𝜇𝜇𝑖𝑖 and variance 𝜎𝜎𝑖𝑖2 ,
and if 𝑦𝑦 = 𝑥𝑥1 + 𝑥𝑥2 + ⋯ + 𝑥𝑥𝑛𝑛 , then the distribution of
𝑦𝑦 − ∑𝑛𝑛𝑖𝑖=1 𝜇𝜇𝑖𝑖

∑𝑛𝑛𝑖𝑖=1 𝜎𝜎𝑖𝑖2

approaches the 𝑁𝑁(0,1) distribution as 𝑛𝑛 approaches infinity.

1/20/2021

30
 Chi-square 𝜒𝜒 2 Distribution

 The Chi-square distribution is associated with Normal random variables

𝑦𝑦 = 𝑥𝑥12 + 𝑥𝑥22 + ⋯ + 𝑥𝑥𝑛𝑛2
 𝑦𝑦 follows 𝜒𝜒 2 (𝑛𝑛) if 𝑥𝑥12 , 𝑥𝑥22 , ⋯ , 𝑥𝑥𝑛𝑛2 are normally and independently distributed
with mean 0 and variance 1.
 𝑛𝑛 is the degree of freedom.
 The most popular use of this distribution is for testing hypotheses about
variances of samples from normal distributions.
 The percentage points of the 𝜒𝜒 2 distribution is listed in Appendix III.

1/20/2021

31
 t Distribution

 If 𝑥𝑥 and 𝑦𝑦 are independent standard normal and chi-square random variable

with 𝑘𝑘 degrees of freedom respectively, then
𝑥𝑥
𝑡𝑡 =
𝑦𝑦/𝑛𝑛
is distributed as 𝑡𝑡 with 𝑘𝑘 degrees of freedom.
 This distribution is used for testing hypotheses about two population means.
 As 𝑘𝑘 → ∞ the 𝑡𝑡 distribution approximate to standard normal distribution.
 The percentage points of the 𝑡𝑡 distribution is listed in Appendix IV.

1/20/2021

32
 F Distribution

 If 𝑤𝑤 and 𝑦𝑦 are two independent chi-square random variables with 𝑢𝑢 and 𝑣𝑣

degrees of freedom, respectively, then the ratio
𝑤𝑤/𝑢𝑢
𝐹𝐹𝑢𝑢,𝑣𝑣 =
𝑦𝑦/𝑣𝑣
 is distributed as 𝐹𝐹 with 𝑢𝑢 numerator degrees of freedom and 𝑣𝑣 denominator
degrees of freedom.
 This distribution is used for testing hypotheses about two population variances.
 The percentage points of the 𝐹𝐹 distribution is listed in Appendix V (textbook,
page 715)

1/20/2021

33
 Example

Chi-square distribution for selected values of 𝑛𝑛 The 𝑡𝑡 distribution for selected values of 𝑘𝑘

1/20/2021

34 The 𝐹𝐹 distribution for selected values of 𝑢𝑢 (numerator degrees of freedom)

and 𝑣𝑣 (denominator degrees of freedom).
 2.3 Some Useful Approximation

 The Binomial Approximation to the Hypergeometric

𝑛𝑛 𝐷𝐷
 If < 0.1, use 𝑝𝑝 =
𝑁𝑁 𝑁𝑁

 The Poisson Approximation to the Binomial

 If 𝑛𝑛 is large enough and 𝑝𝑝 is small enough, use 𝜆𝜆 = 𝑛𝑛𝑛𝑛

 The Normal Approximation to the Binomial

 If 𝑛𝑛𝑛𝑛 > 10 and 0.1 ≤ 𝑝𝑝 ≤ 0.9, use 𝜇𝜇 = 𝑛𝑛𝑛𝑛 and 𝜎𝜎 2 = 𝑛𝑛𝑛𝑛(1 − 𝑝𝑝)

1/20/2021

35
 Approximations to distributions

1/20/2021

36
 Thank you!

Any Questions?

1/20/2021

37
 Appendix I

Summary of common probability distributions often used in statistical quality control

1/20/2021

38
 Appendix I (cont’d)

Summary of common probability distributions often used in statistical quality control

1/20/2021

39
 Appendix II

Cumulative Standard Normal Distribution

1/20/2021

40
 Appendix II (cont’d)

Cumulative Standard Normal Distribution

1/20/2021

41
 Appendix III

Percentage Points of the 𝜒𝜒 2 Distribution

1/20/2021

42
 Appendix IV

Percentage Points of the 𝑡𝑡 Distribution

1/20/2021

43
 Appendix V
Percentage Points of the 𝐹𝐹 Distribution

1/20/2021

44
 Appendix V (cont’d)
Percentage Points of the 𝐹𝐹 Distribution (cont’d)

1/20/2021

45
 Appendix V (cont’d)
Percentage Points of the 𝐹𝐹 Distribution (cont’d)

1/20/2021

46