STATISTICAL PROCESS

CONTROL

OUTLINE

Introduction.
Summarising Data.
Variation.
Some Fundamentals.
Control Charts.
Stability ,Capability & Continuous
improvement.
Demings 14 points.

Introduction

Management is responsible for

95% of Quality problems.
E.Deming

What is Quality?
How to improve Quality?
What is value?

Summarizing Data

Types of data: Variables & Attributes.

Purpose of Data gathering: Decision making

Can we make it right?

Are we making it right?
Have we made it right?
Can we make it better?

Reasons for collection of Data

To understand an actual situation
For analysis
Data for process control and acceptance and
rejection.

Summarizing Data

Measures of Location
Average
E.g. 3,10,16,1,5,14,11,6
Median

Measures of Dispersions
Range
RMSP
Standard Deviation

Histograms

Run Chart.

Variation
If I had to reduce my message for
management to just a few words, Id
say it all had to do with reducing
variation.
E. Deming.

Variation

There is variation among the products

from any manufacturing process.

We would like to completely get rid of

variation but we know that cant be done.

The goal is to be able to measure/identify

variation and than work to minimize it.

Variation
Sources of Variation

Equipment

Direct Materials

Indirect Materials

People

Methods

Orders

Facilities

Variation
Indirect Material
Tools
Coolant
Bushings
Lubricants
Chemicals

Equipment

Machines
Fixtures
Gauges/Masters
Spindles/Collects

Direct Material

Training
Skill
Motivation
Communication
Attitude
Trust

Product
Quality

Hardness
Machiniability
Pick up points

Orders

Clarity
Timeliness
Adequacy
Product mix
Quantity

People

Methods
Facilities

Temperature
Cleanliness
Humidity
Noise
Central coolant system

Specification
Speeds
Feeds
Process Documentation

Variation
Types of variation

Controlled variation (Common causes):

Is stable and consistent pattern of variation
over time.

Uncontrolled variation (Assignable

cause):
Is a pattern of variation that changes over
time.

Variation
Frequency for each size

Controlled Variation

Frequency for each size

Time

Prod

uct

size

Time

Prod
uct

Uncontrolled Variation
size

Variation
Difference between Common and Assignable cause
Common Cause
1) A large number are in effect
2)
3)

4)
5)

at any time.
Each has an individual effect
that is too small to mention.
Only a change in the system
will reduce that part of the
variability.
Only management has the
ability to make changes.
Remain constant over time.

Assignable Cause
1) Very few in effect at any

time.
2) The effect is measurable.
3) They can be found and
eliminated.
4) The machine operator is
best able to discover and
make changes.
5)They occur infrequently in an
unpredictable manner.

Variation
Approaches to Variation

Fabricate

Pack

Inspect

Quality Audit

Inspect
Rework

Ship

Scrap
Report to Management

Detection based system

Variation
T test - Exercise

t Test demonstrating 100% inspection.

Instructions: When the instructor says
go, please count the number of ts in
the paragraph. You will have one minute.
At the end of that time, be able to state
how many ts are in the paragraph.

Variation
Approaches to Variation

Production,
assembly,
inspection, test
Suppliers of
raw materials

Test of
raw
material

Consumers and
service to customer
Distribution

Final
inspection

Prevention based system

Some Fundamentals

50%

50%

Balanced
Normal Distribution

Some Fundamentals

68%
95%
99.73%

1s

1s

1s

1s

1s

1s

Normal Distribution.

Some Fundamentals

Process has Lack of

Control

Process displays
control

Four possibilities for any Process

Threshold State
Process in control
Some non confirming Product
Must Either
Change process, or
Change specifications
Sorting is temporary fix
Control Charts
Maintain process in control
Evaluate efforts at improvement

Ideal State
Process in Control
100% conforming products
Control Charts
Maintain process in control
Evaluate efforts at improvement
Chaos managers

State of Chaos
Brink of Chaos
Process out of control
Process out of control
100% conforming product
Some non conforming product
Assignable causes still dominate
All may seem okay, but
Random fluctuations due to
Assignable causes determine what
assignable causes will eventually
is produced by the process!
Quality and conformance can
frustrate efforts at process
change in a moment.
improvement.
The only way out of chaos is to first
eliminate the assignable caused
Entropy slide
Some nonconforming products
100% conforming products
produced
produced

Some Fundamentals
We are being ruined by best efforts.
Deming
The Funnel Experiment:
This experiment can be carried out with a relatively simple
apparatus:
1. A funnel, such as found in kitchen.
2. A holder for funnel, e.g. a desk lamp.
3. A marble, small enough to pass through the stem.
4. A table covered with washable table cloth.
5. A felt tip pen with non permanent ink. And
6. A ruler and a protractor.

Some Fundamentals
Funnel Experiment
Rule One :
Dont shift the funnel, irrespective of where the
marble comes to rest.

Lets do something. Dont just sit there.

Do something about it Move the
Funnel.

Some Fundamentals
Funnel Experiment
Rule two :
Move the funnel in attempt to compensate for the
amount by which the marble misses the target.
Suppose marble finishes six inch east of the target
funnel is moved six inch west of the target.
Example
Firing rifle bullet on the target.

100 marbles dropped using rule 2 of the

funnel

Some Fundamentals
Funnel Experiment
Rule three :
Move the funnel in attempt to compensate for the
amount by which the marble misses the target.
Suppose marble finishes six inch east of the target
funnel is moved six inch west
of relative current
Example
position.
Over control

100 marbles dropped using rule 3 of the

funnel

Some Fundamentals
Funnel Experiment

Rule four :
Move the funnel directly over the position where the
marble has just landed.
E.g.
1. Colour matching where each
batch of paint is matched with previous
batch.
2. Operator that tries to achieve
consistency by making each piece as
previous one.
3. Worker training worker.

100 marbles dropped using rule 4 of the

funnel

Control charts
Purpose of control chart
The function of a control chart is to
minimize the net economic losses from
over adjustment and under adjustment.

Control charts
Preparatory steps for Control Chart

Establish an environment suitable for action.

Define the process.

Determine characteristics to be charted.

Define the measurement system.

Minimize unnecessary variation.

Control Chart

Uses sub groups to monitor process

Control Chart
Type of control charts
Variables
Based on sample size:
x

and R

and s

and moving range

Attribute
Based on type of data:
Is the interest in non conforming
unit i.e. percent defective
p chart
np chart (sample size constant)
Is the interest in non conformities
i.e. discrepancies/part?
c chart (sample size constant)
u chart

Control chart
Foundations of Control Charts

Always use control limits which are set at 3 sigma limits

on either side of central line.

In computing three sigma limit one must always use an

average dispersion statistic.

Use rational sampling and rational subgrouping.

Control Chart are effective only to the extent that the

organization can use, in a effective manner, the
knowledge gained from the charts.

Logic of control chart

Assume process
displays controlled
variation

Predict Behaviour of Averages and

ranges (Calculate Control Limit)

Compare observed or
Future Average and
range values with
control limit

If observations are
consistent with
predictions then
process may be
stable

Continued operation of
process within limits is
the only proof of
stability

If observations are
inconsistent with
predictions then
process is definitely
unstable

Take action to identify

and remove assignable
causes

Control Chart
Control Limits

Compute average & range for each sub group of k subgroups

Compute the grand average, X, by averaging each of the k sub
group averages.
Compute average range R by averaging each of k subgroup
ranges.
The certral line for X- chart is X
The center line for R chart is R.
Find the values for A2, D3, D4, which correspond to the sub
group size n.
Multiply R by A2 = A2R
Add and subtract A2R from X to get UCL and LCL for X-chart.
Multiply R by D4 to get UCL for R chart
Multiply R by D3 to get LCL for R chart.

Control Chart
Control Charts for Subgroup size one

Used in several situation where the logical subgroup

size is n = 1. This might happen when each
measurement represents on batch or the measurement
are widely spaced over time.

UNPLx = X + 2.66mR
CLx = X
LNPLx = X - 2.660 mR
UCLr = 3.268 mR
CLr = mR

Control Chart
Rational Subgrouping

Never subgroup unlike thing together.

Minimize the variation within each subgroup.
Maximize the opportunity for variation
between subgroups.
Average across noise not across signal
Treat the chart in accordance with the data.

Control Charts
Analyzing Control Charts

Choice of scale for control charts: Scale should be twice as

the control limit spread.

What is Reasonable Degree of statistical Control?

Process is said to be in statistical control if 100 consecutive
observations fail to indicate lack of control i.e 25 consecutive sub
groups with size 4 or 20 consecutive with subgroup size 5.

Myths about control Chart.

Data should be normally distributed before they can be placed on
control chart.
It has been said that the data must be in control before one can plot
them on a control chart.
Only useful in shopfloor.

Control Chart
Four rules for defining Lack of Control

A lack of control is indicated whenever a single point falls out

side the control limit

A lack of control is indicated whenever at least two out of

three successive values fall on the same side of and more
than two sigma units away from the central line

A lack of control is indicated whenever at least four out of

five successive values fall on the same side of, and more
than one sigma units away from the central line

A lack of control is indicated whenever at least seven

successive values fall on same side of central line

Stability, Capability &

Continuous Improvement
The first step in examination of data is accordingly to question
the state of statistical control that produced the data
Deming
Process Capability
Capability refers to the quantification of common cause
variation. The presence of special or assignable cause make
the prediction of capability impossible and meaning of
capability unclear.
Cpk: What is it?
A prediction of the level of quality (or measure of the inherent
uniformity) a process can produce.

Stability, Capability &

Continuous Improvement
Process capability indices
LSL

USL

6s
Total Tol.

Cp or Pp = Total Tol
6s

LSL

USL

Avg.

3s
Zu

3s
Zl

Cpk or Ppk = Min of zu & zl

3s

Note : Pp/Ppk is calculated for short run of a process &

Cp/Cpk is calculated for long run of process

Stability, Capability &

Continuous Improvement
Limits for Individual values
For X-R chart
Find the value d2 which corresponds to the sub group
size n
Divide the Average Range by d2 to get Sigma(X) =
R/d2.
Multiply this quantity by 6 to get 6 sigma(X) = 6* R/d2.
For X- mR Chart
Using the grand average as the central line add and
subtract the above value to get UNPLx & LNPLx
respectively.

Stability, Capability &

Continuous Improvement
Capability Analysis

Short term capability study.

Will only reveal part of story as assignable causes may not have
an opportunity to arrive.
Uses anywhere between 25 - 50 values.

Steps.
Plot individual vales observe any indication of lack of statistical
control.
Plot XmR chart
Calculate control limits. Difference in UNL and LNL will give 6
sigma value. Use the same in computing Pp and Ppk.

Stability, Capability &

Continuous Improvement

Cost per good unit of product

Cpk logic applied to Jurans Optimal Quality

Cost model

100%
defective

T0

T0

Total
quality
costs
Failure
costs

Cost of
Appraisal plus
prevention
100%
good

Quality of Conformance
Cpk 1

1.33

Stability, Capability &

Continuous Improvement
SPC Implementation
CONDUCT GR&R
GAUGE CAPABLE
CONDUCT PROCESS
POTENTIAL STUDY
INCAPABLE
100% CONTROL UNTIL
IMPROVEMENT MADE
CONTROL CHART FOR
PROBLEM SOLVING

CAPABLE

VERY CAPABLE

TARGETING CRITICAL MUST REACT QUICKLY

TO OUT OF CONTROL
CONDITION

CONTROL CHARTS TO
IDENTIFY SOURCE OF
VARIATION

CONTROL CHART TO
REDUCE VARIABILITY
(SPC ON GOING
CONTROL)

MARGINAL
CAPABILITY
100% OR SPC
IMMEDIATE
ACTION ON OUT
OF CONTROL
CONDITIONS
REQUIRED

CONTINUE CHARTING
UNTIL DESIRED LONG TERM
CAPABILITY IS OBTAINED
CAPABLE Cpk
1.33 - 2.50
CONTINUE X
AND R
CHART

EXTREMELY
CAPABLE Cpk
2.5 - 4.0
CONTINUE X
AND R
REDUCE
FRQUENCY

Cpk>
4.0
N=2
POSSIBLE

Stability, Capability &

Continuous Improvement
World Class Quality
World class quality is Quality which is on
target with minimal variation.

Loss

measurement

LSL

USL

Loss Function Representing the

conformance to Specification
Interpretation of Quality

Stability, Capability &

Continuous Improvement
World Class Quality

Loss

K = Targeting error

measurement

A quadratic loss function with a product distribution

Taguchis Loss function states that there exists a target, or ideal value,
for any characteristic and a continuous loss function which indicates that
the further a given product lies away from the target the higher the total
economic loss to society

Stability, Capability &

Continuous Improvement

Control Chart is effective tool to reduce process

dispersion and lead to continual improvement.
Control charts used just for monitoring is waste of
valuable resources.
Refer Handout for example on continuous
improvement.
PDCA cycle
Act

Specification

Plan

Production

Check
Inspection

Quality Control process

shown as circle

Do

Quality Control process

shown as circle

Stability, Capability &

Continuous Improvement
Why Continuous Improvement?

Knowledge gets the process into good shape, and

makes them so well understood, that the chance of any
thing going seriously wrong with them is negligible which itself constitutes huge saving.
Operation becomes hassle free.
It contributes to innovation.
Enhanced morale and pride in workmanship.
Less or no warranty.
Meeting specification are not good enough.

Demings Fourteen Points

Constancy of Purpose.
The new philosophy.
Cease dependence on mass inspection.
End lowest tender contract.
Improve every process.
Institute training.
Institute Leadership of people.
Drive out fear.
Break down barriers.
Eliminate Exhortations.
Eliminate Arbitrary numerical targets.
Permit pride in workmanship.
Encourage education.
Top management commitment and action.