Statistical Quality Control

2009 Prentice-Hall, Inc.

17 1

Learning Objectives
After completing this chapter, students will be able to:
 Define the quality of a product or service  Determine the benefits derived from SQC  Review statistical background  Refresh topics- mean, median std.deviation, ave.

deviation  Develop four types of control charts: x, R, p, and c

2009 Prentice-Hall, Inc.

17 2

Chapter Outline
17.1 17.2 17.3 17.4 17.5 Introduction Defining Quality and Total Quality Mgt Statistical Process Control Control Charts for Variables Control Charts for Attributes

2009 Prentice-Hall, Inc.

17 3

Introduction
Quality is often the major issue in a purchase decision Poor quality can be expensive for both the producing firm and the customer Quality management, or quality control (QC is QC), critical throughout the organization Quality is important for manufacturing and services We will be dealing with the most important statistical methodology, statistical process control (SPC SPC)
2009 Prentice-Hall, Inc. 17 4

Defining Quality and TQM

Quality of a product or service is the degree to which the product or service meets specifications Increasingly, definitions of quality include an added emphasis on meeting the customer s needs Total quality management (TQM refers to a quality TQM) emphasis that encompasses the entire organization from supplier to customer Meeting the customer s expectations requires an emphasis on TQM if the firm is to complete as a leader in world markets
2009 Prentice-Hall, Inc. 17 5

Defining Quality and TQM

Several definitions of quality Quality is the degree to which a specific product conforms to a design or specification. (Gilmore, 1974) Quality is the totality of features and characteristics of a product or service that bears on its ability to satisfy stated or implied needs. (Johnson and Winchell, 1989) Quality is fitness for use. (Juran, 1974) Quality is defined by the customer; customers want products and services that, throughout their lives, meet customers needs and expectations at a cost that represents value. (Ford, 1991) Table Even though quality cannot be defined, you know 17.1 it what 2009 Prentice-Hall, Inc. 17 6 is. (Pirsig, 1974)

Statistical Process Control

Statistical process control involves establishing and monitoring standards, making measurements, and taking corrective action as a product or service is being produced Samples of process output are examined If they fall outside certain specific ranges, the process is stopped and the assignable cause is located and removed A control chart is a graphical presentation of data over time and shows upper and lower limits of the process we want to control

2009 Prentice-Hall, Inc.

17 7

Mean, Median, Deviation, Range

2009 Prentice-Hall, Inc.

17 8

Mean
It is the average of a series of results. The mean of a finite number of measurements, xl, x2, x3, x4,........, xn, is commonly represented as x. It may be calculated by taking the average of individual results as shown below :

2009 Prentice-Hall, Inc.

17 9

Median
The median of an even number of results is nothing but the average of the two middle values provided the results are listed in order ; whereas for an odd number of results the median is the middle value itself. However, the mean and the median are exactly identical in the case of a truly symmetrical distribution. In short, median is an useful measure specifically when dealing with very small samples.

2009 Prentice-Hall, Inc.

17 10

Average Deviation (or Mean Deviation)

It is the average of the differences between the individual results and the mean. It is regarded as a measure of variability. In the case of a small number of observations the average deviation is found to be not quite significant statistically. (d)
2009 Prentice-Hall, Inc. 17 11

Relative average deviation

Found by dividing the average deviation by the mean. It is usually expressed in percentage or parts per thousand

2009 Prentice-Hall, Inc.

17 12

Standard Deviation
It is the distance from the mean to the point of inflexion of the normal distribution curve. In comparison to the average deviation the standard deviation is usually considered to be much more useful and meaningful statistically. For a finite number of values it is normally symbolised as S, and may be expressed as follows :

2009 Prentice-Hall, Inc.

17 13

Coefficient of Variation ( )/ relative standard deviation

The coefficient of variation ( ) is simply the standard deviation(s) expressed as a percentage of the mean ( x ) as stated below :

2009 Prentice-Hall, Inc.

17 14

Range
The difference between the largest and smallest result in a series of measurements

2009 Prentice-Hall, Inc.

17 15

Examples:
The normality of a solution of sodium hydroxide as determined by an analyst by FOUR different titrations are found to be 0.5038 ; 0.5049 ; 0.5042 ; and 0.5039. Calculate the mean, median, average deviation, relative average deviation, standard deviation and coefficient of variation.

2009 Prentice-Hall, Inc.

17 16

Statistical Process Control

Patterns to look for in control charts
Figure 17.1

Upper control limit One plot out above. Investigate for cause.

Target

Lower control limit Normal behavior

One plot out below. Investigate for cause.

2009 Prentice-Hall, Inc. 17 17

Statistical Process Control

Patterns to look for in control charts
Upper control limit Target Lower control limit Upper control limit Target Lower control limit

Figure 17.1

Two plots near upper control Investigate for cause.

Two plots near lower control. Investigate for cause.

Run of 5 above central line. Investigate for cause.

Run of 5 below central line. Investigate for cause.

2009 Prentice-Hall, Inc. 17 18

Statistical Process Control

Patterns to look for in control charts
Figure 17.1

Upper control limit

Target

Lower control limit Trends in either direction 5 plots. Investigate for cause of progressive change.

Erratic behavior. Investigate.

2009 Prentice-Hall, Inc.

17 19

Statistical Process Control Building control charts

Control charts are built using averages of small samples The purpose of control charts is to distinguish between natural variations and variations due to assignable causes

2009 Prentice-Hall, Inc.

17 20

Statistical Process Control

Natural variations
Natural variations affect almost every production process and are to be expected, even when the process is in statistical control They are random and uncontrollable When the distribution of this variation is normal it will have two parameters Mean, Q (the measure of central tendency of the average) Standard deviation, W (the amount by which smaller values differ from the larger ones) As long as the distribution remains within specified 2009 Prentice-Hall, Inc. limits it is said to be in control

17 21

Statistical Process Control

Assignable variations
When a process is not in control, we must detect and eliminate special (assignable causes of assignable) variation The variations are not random and can be controlled Control charts help pinpoint where a problem may lie The objective of a process control system is to provide a statistical signal when assignable causes of variation are present
2009 Prentice-Hall, Inc. 17 22

Control Charts for Variables

 The x-chart (mean) and R-chart (range) are

the control charts used for processes that are measured in continuous units  The x-chart is a data analysis technique for determining if a measurement process has gone out of statistical control.  The R-chart tells us when there has been a change in the uniformity of the process  Both charts must be used when monitoring variables
2009 Prentice-Hall, Inc. 17 23

Five Steps to Follow in Using x and R-Charts

1. Collect 20 to 25 samples of n = 4 or n = 5 from a stable process and compute the mean and range of each. 2. Compute the overall means ( x and R ), set appropriate control limits, usually at 99.7% level and calculate the preliminary upper and lower control limits. 99.7% of the time the sample averages will
fall between s3W if the process has only random variations

3. Graph the sample means and ranges on their respective control charts and determine whether they fall outside the acceptable limits

2009 Prentice-Hall, Inc.

17 24

Five Steps to Follow in Using x and R-Charts

4. Investigate points or patterns that indicate the process is out of control. Try to assign causes for the variation and then resume the process. 5. Collect additional samples and, if necessary, revalidate the control limits using the new data

x chart

R-chart
2009 Prentice-Hall, Inc. 17 25

The Central Limit Theorem

The central limit theorem is the foundation for xcharts The central limit theorem says that the distribution of sample means will follow a normal distribution as the sample size grows large Even with small sample sizes the distribution is nearly normal

2009 Prentice-Hall, Inc.

17 26

The Central Limit Theorem

The central limit theorem says 1. The mean of the distribution will equal the population mean 2. The standard deviation of the sampling distribution will equal the population standard deviation divided by the square root of the sample size Wx Qx ! and Wx ! n
 We often estimate Q x and

with the average of all

sample means ( x )
2009 Prentice-Hall, Inc. 17 27

The Central Limit Theorem

 Figure 17.2 shows three possible population

distributions, each with their own mean (Q) and standard deviation (W x)  If a series of random samples ( x 1 , x 2 , x 3 , x 4 , and so on) each of size n is drawn from any of these, the resulting distribution of the x s will appear as i in the bottom graph in the figure  Because this is a normal distribution
1. 99.7% of the time the sample averages will fall between s3W if the process has only random variations 2. 95.5% of the time the sample averages will fall between s2W if the process has only random variations

 If a point falls outside the s3W control limit, we are

99.7% sure the process has changed

2009 Prentice-Hall, Inc.

17 28

The Central Limit Theorem

Population and sampling distributions
Normal Beta Uniform Q = (mean) Wx = S.D. Q = (mean) Wx = S.D. Q = (mean) Wx = S.D.

Sampling Distribution of Sample Means (Always Normal)

99.7% of all x fall within s3Wx 95.5% of all x fall within s2Wx

| 3Wx

| 2Wx

| 1Wx

| Qx = Q (mean)

| +1Wx
Wx n

| +2Wx

| +3Wx

Figure 17.2

Standard error ! W x !

2009 Prentice-Hall, Inc.

17 29

Setting the x-Chart Limits

If we know the standard deviation of the process, we can set the control limits using Upper control limit (UCL) ! x  zW x Lower control limit (UCL) ! x  zW x where
x = mean of the sample means z = number of normal standard deviations = standard deviation of the sampling W W x distribution of the sample means = x n
2009 Prentice-Hall, Inc. 17 30

= number of normal standard deviations

If chosen 99.7% of the time the sample averages will fall between 3W if the process has only random variations
If chosen 95.5% of the time the sample averages will fall between 2W if the process has only random variations

2009 Prentice-Hall, Inc.

17 31

Box Filling Example

 A large production lot of boxes of cornflakes is

sampled every hour  To set control limits that include 99.7% of the sample, 36 boxes are randomly selected and weighed  The standard deviation is estimated to be 2 ounces and the average mean of all the samples taken is 16 ounces  So x ! 16,W x ! 2, n ! 36, z ! 3 and the control limits are 2 UCLx ! x  zW x ! 16  3 ! 16  1 ! 17 ounces 36 2 LCLx ! x  zW x ! 16  3 ! 16  1 ! 15 ounces 36
2009 Prentice-Hall, Inc. 17 32

Box Filling Example

 If the process standard deviation is not available

or difficult to compute (a common situation) the previous equations are impractical  In practice the calculation of the control limits is based on the average range rather than the standard deviation UCLx ! x  A2 R LCLx ! x  A2 R where R = average of the samples A2 = value found in Table 17.2 x = mean of the sample means
2009 Prentice-Hall, Inc. 17 33

Factors for Computing Control Chart Limits

SAMPLE SIZE, n 2 3 4 5 6 7 8 9 10 12 14 16 18 20 25 MEAN FACTOR, A2 1.880 1.023 0.729 0.577 0.483 0.419 0.373 0.337 0.308 0.266 0.235 0.212 0.194 0.180 0.153 UPPER RANGE, D4 3.268 2.574 2.282 2.115 2.004 1.924 1.864 1.816 1.777 1.716 1.671 1.636 1.608 1.586 1.541 LOWER RANGE, D3 0 0 0 0 0 0.076 0.136 0.184 0.223 0.284 0.329 0.364 0.392 0.414 0.459

Table 17.2
2009 Prentice-Hall, Inc. 17 34

Super Cola Example

Super Cola bottles are labeled net weight 16 ounces The overall process mean is 16.01 ounces and the average range is 0.25 ounces What are the upper and lower control limits for this process? UCLx ! x  A2 R LCLx ! x  A2 R ! 16.01 + (0.577)(0.25) ! 16.01 + 0.144 ! 16.154 ! 16.01 (0.577)(0.25) ! 16.01 0.144 ! 15.866

2009 Prentice-Hall, Inc.

17 35

Setting Range Chart Limits

 We have determined upper and lower control    

limits for the process average We are also interested in the dispersion or variability of the process Averages can remain the same even if variability changes A control chart for ranges is commonly used to monitor process variability Limits are set at s3W for the average range R

2009 Prentice-Hall, Inc.

17 36

Setting Range Chart Limits

 We can set the upper and lower controls using

UCLR ! D4 R LCLR ! D3 R where UCLR = upper control chart limit for the range LCLR = lower control chart limit for the range D4 and D3 = values from Table 17.2

2009 Prentice-Hall, Inc.

17 37

Range Example
A process has an average range of 53 pounds If the sample size is 5, what are the upper and lower control limits? From Table 17.2, D4 = 2.114 and D3 = 0 UCLR ! D4 R ! (2.114)(53 pounds) ! 112.042 pounds LCLR ! D3 R ! (0)(53 pounds) !0
2009 Prentice-Hall, Inc. 17 38

Factors for Computing Control Chart Limits

SAMPLE SIZE, n 2 3 4 5 6 7 8 9 10 12 14 16 18 20 25 MEAN FACTOR, A2 1.880 1.023 0.729 0.577 0.483 0.419 0.373 0.337 0.308 0.266 0.235 0.212 0.194 0.180 0.153 UPPER RANGE, D4 3.268 2.574 2.282 2.115 2.004 1.924 1.864 1.816 1.777 1.716 1.671 1.636 1.608 1.586 1.541 LOWER RANGE, D3 0 0 0 0 0 0.076 0.136 0.184 0.223 0.284 0.329 0.364 0.392 0.414 0.459

Table 17.2
2009 Prentice-Hall, Inc. 17 39

Problems:
No. Insp. A 1 2 3 4 10.5 10.4 10.1 10.4 5 sample size of Volume of vials every 30 minutes B 10.3 10.3 10.3 10.1 C 10.6 10.1 10.5 10.3 D 10.7 10.3 10.7 10.7 E 10.2 10.5 10.6 10.2 mean R

a. Calculate the mean, average mean, range, average range b. Calculate the UCL and LCL of both mean and range 2009 Prentice-Hall, Inc. using 99.7% in normal distribution

17 40

Control Charts for Attributes

We need a different type of chart to measure attributes These attributes are often classified as defective or nondefective There are two kinds of attribute control charts
1. Charts that measure the percent defective in a sample are called pcharts 2. Charts that count the number of defects in a sample are called c-charts
2009 Prentice-Hall, Inc. 17 41

p-Charts
Attributes that are good or bad typically follow the binomial distribution If the sample size is large enough a normal distribution can be used to calculate the control limits UCL p ! p  zW p
where

LCL p ! p  zW p

p = mean proportion or fraction defective in the sample z = number of standard deviations W p = standard deviation of the sampling distribution which is estimated by W p where n is the size of each sample Wp ! p(1  p ) n
2009 Prentice-Hall, Inc. 17 42

ARCO p-Chart Example

Performance of data-entry clerks at ARCO (n = 100)
SAMPLE NUMBER 1 2 3 4 5 6 7 8 9 10 NUMBER OF ERRORS 6 5 0 1 4 2 5 3 3 2 FRACTION DEFECTIVE 0.06 0.05 0.00 0.01 0.04 0.02 0.05 0.03 0.03 0.02 SAMPLE NUMBER 11 12 13 14 15 16 17 18 19 20 NUMBER OF ERRORS 6 1 8 7 5 4 11 3 0 4 80 FRACTION DEFECTIVE 0.06 0.01 0.08 0.07 0.05 0.04 0.11 0.03 0.00 0.04

2009 Prentice-Hall, Inc.

17 43

ARCO p-Chart Example

We want to set the control limits at 99.7% of the random variation present when the process is in control so z = 3
p! Wp ! Total number of errors 80 ! ! 0.04 Total number of records examined (100 )(20) (0.04 )(1  0.04 ) ! 0.02 100

UCL p ! p  zW p ! 0.04  3(0.02 ) ! 0.10 LCL p ! p  zW p ! 0.04  3(0.02 ) ! 0 Percentage cant be negative
2009 Prentice-Hall, Inc. 17 44

ARCO p-Chart Example

p-chart for data entry
0.12 0.11 0.10 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0.00

Out of Control
UCLp = 0.10

Fraction Defective

p ! 0.04

LCLp = 0.00

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Sample Number
2009 Prentice-Hall, Inc. 17 45

Figure 17.3

ARCO p-Chart Example

Excel QMs p-chart program applied to the ARCO data showing input data and formulas

Program 17.1A
2009 Prentice-Hall, Inc. 17 46

ARCO p-Chart Example

Output from Excel QMs p-chart analysis of the ARCO data

Program 17.1B
2009 Prentice-Hall, Inc. 17 47

c-Charts
 In the previous example we counted the number of     

defective records entered in the database But records may contain more than one defect We use c-charts to control the number of defects per unit of output c-charts are based on the Poisson distribution which has its variance equal to its mean The mean is c and the standard deviation is equal to c To compute the control limits we use

cs3 c
2009 Prentice-Hall, Inc. 17 48

Red Top Cab Company c-Chart Example

The company receives several complaints each day about the behavior of its drivers Over a nine-day period the owner received 3, 0, 8, 9, 6, 7, 4, 9, 8 calls from irate passengers for a total of 54 complaints To compute the control limits 54 c! ! 6 complaints per day 9
 Thus

UCLc ! c  3 c ! 6  3 6 ! 6  3(2.45 ) ! 13.35 LCLc ! c  3 c ! 6  3 6 ! 6  3(2.45 ) ! 0

2009 Prentice-Hall, Inc. 17 49