You are on page 1of 24

Text Book: Statistical Quality Control; a modern introduction

by Douglas C Montgomery

If a product is to meet customer expectations, it should be produced by a stable


process (low variability)

Statistical Process Control (SPC) is a powerful collection of tools, useful in


achieving process stability and improving process capability through reducing
variability

Seven powerful statistical tools (the Magnificent seven)


-Histogram or stem-and-leaf plot

-Check sheet
-Pareto chart
-Cause and effect diagram
-Defect concentration diagrams
-Scatter diagrams

-Control charts (Shewhart control charts/ most technically sophisticated tool)


2

Chance causes and assignable causes of quality variations


Chance causes (Background noise)

In any process, a certain amount of natural variability exists.

This is the cumulative effect of many small, unavoidable causes

Assignable causes

Mainly due to improperly controlled machines, operator errors or defective


raw material

The objective of Statistical Process Control is to quickly detect the assignable


causes of process shifts so that investigation of process and corrective actions
can be taken before many nonconforming items are produced.
(The major objective of SPC is to eliminate variability in the process)

Basic Principles

General model for a control chart

When L = 3, then they are called three-sigma control limits

Example: Consider a process with a quality characteristic whose mean is 1.5


and sample standard deviation is 0.0671. Determine 3-sigma control limits
5

Choice of control limits


Type I error a point falling beyond the control limits when there is no assignable
cause present (False alarm)
Type II error a point falling between the control limits when the process is
really out of control

Two limits on control charts

Three sigma limits for the outer limits/ actual limits

Two sigma limits for the inner limits/ warning limits

Warning limits increase the sensitivity of the control charts

Example: Consider a process with a


quality characteristic whose mean is 1.5
and sample standard deviation is 0.0671.
Determine warning limits

Variable Control Charts


For quality characteristics that can be numerically measured
Control of process average quality level is done with x control charts

Variability of the process is monitored by control charts for standard deviation (s


control charts) or control charts for range (R control charts)

Suppose samples of size m are taken randomly,


Then sample mean

x1 x2 .....xm
m

If x is normally distributed with mean and standard deviation , then

normally distributed with mean and standard deviation


x

is

Process standard deviation can be estimated by

R
d2

Where R= x max- x min and R R1 R2 ..... Rm


m

10

Example 6.1 of Text book


(Page 231)
Set up the x and R control charts for
the data given

11

Control limits for R charts


LCL = 0

CL= 0.32521
UCL= 0.68749

Control limits for x charts


LCL = 1.31795
CL= 1.5056

UCL= 1.69325

These are called trial control limits


12

Trial control limits are used to determine whether the process has been
in control during the time in which the data were collected. This can be
done by plotting the sample means and range in the developed control
charts

If all the points lie inside the limits, then the process has been in control in
the past and trial control limits (now become actual limits) can be used for
the phase II applications where the future production is monitored

If for any point lies outside the limits, should be investigated for assignable
causes. If any assignable cause is found, it should be eliminated and the limits
should be re-calculated by excluding the corresponding points. This should be
repeated until reliable control limits are obtained.

Then new data should be collected and compared with these revised limits .
This should be repeated until reliable control limits are obtained.

13

Exercise 6.23 of text book Page 276


Given the following data of 20 samples, each with size 4.

14

Control Charts for attributes

Instead of numerical representation, product classification is done as either


defective or non-defective (Conforming and non-conforming)

These are called attributes

Some examples: number of malfunctioning semi conductor chips, number of


errors made in filling a particular form etc

Four types of control charts ;


- p charts (for fraction nonconforming)
- np charts (for number of defects/ nonconforming items)
- c charts (for nonconformities)

charts (for nonconformities per unit)

15

Fraction nonconforming charts (p charts)

The fraction nonconforming is defined as the ratio of the number of


nonconforming items to the total number of items in that population

In some cases, the true fraction non conforming p in the production process is
known or a standard value is given

If it is not known/ standard value is not given, then it should be estimated as


follows

Where,

n is the size of each sample

m is the number of samples and it should be at least 20-25

Then,

is the fraction nonconforming in the ith sample

16

17

Example 7.1 Page 292


30 samples were selected
and sample size is 50

18

np charts for number of nonconforming


items

Use the same data as in example 7.1 (Page 292) and develop
the np control charts
19

Control charts for nonconformities c charts

Sometimes it is possible for some items to have number of defects but still
classified as a conforming item

Ex: Number of functional defects in an electronics device, Number of errors


on a document

Control charts are developed for the total number of nonconformities in a


unit or average number of nonconformities per unit

In some cases, a standard value for number of nonconformities c is given or


otherwise, it can be estimated as the observed average number of non
conformities in a preliminary sample

20

21

Example 7.3 (Page 310- text book)

Determine the trial control limits for number of nonconformities and revise the
limits if necessary, assuming that any out of control point has an assignable cause
22

Control charts for average number of


nonconformities per unit u charts

n is the sample size


23

Example 7.4 (Page 315- text book)

Determine the trial control limits for


number of nonconformities and revise the
limits if necessary, assuming that any out
of control point has an assignable cause

24

You might also like