STATISTICAL PROCESS

CONTROL (S P C)

Presented By:

Nadeem Alam
(Consulting Engineer-PIQC)

1
 STATISTICAL PROCESS
CONTROL (S P C)
 Have 10 years of diversified experience including world-class
Canadian Company
 Master’s in Mechanical Engineering from Concordia University,
Canada

 Published Papers in International Journals

 Lead Assessor ISO 9000:2000, QMS

 ASQ Certified Quality Engineer

 Six Sigma Black Belt Candidate

 Provided Consultancy and Training within and outside the country

2
 IN GOD WE TRUST, ALL OTHERS
BRING DATA
• GDP growth rate • % of Rejection (PPM)
• GNP • No of late deliveries
• Inflation rate • No of complaints
• Development projects • No of breakdowns
• Export • Defects prioritization
• Import • Process variation
• Trade deficit/surplus • Vendor performance
• Stock exchange • No of suggestions
• Shares trend • Employee performance
• Budget • Rework, returns.
3
What is SPC?
SPC stands for

Statistical Collection, analyzing and interpreting data

Process An activity which transforms input into

output by utilizing resources

Control Measuring and monitoring performance

4
 What is SPC?

Statistics helps us Statistical Analysis is a

make sense out of critical skill in quality
data, and transform Improvement and
crude numbers into problem-solving.
useful information
and knowledge.
5
 The Misuse of Statistics

There are three kinds of lies: lies, damned lies, and statistics.
(Benjamin
Disraeli)

Figures don’t lie, but liars will figure.

(Charles H Grosvenor)

6
 Could we just replace all
Engineers with Statisticians?
Statistics is not a substitute for Engineering judgment!

Process Knowledge and Experience are still vital to

problem-solving.

Strong Good Excellent

Engineering PLUS Statistical LEADS Problem
knowledge Skills TO Solver

7
 Benefits of SPC

Factual decision

Waste reduction
PERFORMANCE

IMPROVEMENT
Increased monitoring

Operator involvement

COPQ reduction

Customer satisfaction
8
 Why SPC

External Pressure

Internal Improvements

Survival

9
PROCESS CONTROL SYSTEM
MODEL WITH FEED BACK
VOICE OF THE PROCESS

STATISTICAL
STATISTICAL
METHODS
METHODS

THE
THEWAY
WAY PRODUCTS
PEOPLE
WE WORK/
WE WORK/ PRODUCTS
OR CUSTOMERS
EQUIPMENT OR CUSTOMERS
MATERIALS
BLENDING
BLENDINGOF
OF
SERVICES
SERVICES
METHODS RESOURCES
RESOURCES IDENTIFYIGN
ENVIRONMENT IDENTIFYIGN
CHANGING NEEDS
CHANGING
AND NEEDS
EXPECTATIONS
INPUTS PROCESS/SYSTEM OUTPUTS AND EXPECTATIONS

VOICE OF CUSTOMERS

10
 THREE BASIC REASONS OF
NON-CONFORMITY

1. The work activity standard not defined or

not understood by the concerned

2. The Process selected or followed having

inadequate capability.

3. The process is not properly controlled

11
 SPC – A Continuous Improvement Tool
PLAN – Implementation of SPC

• Does the data exist?

• If so, is it readily available in the current format?
• Where is the data to be collected from?
• How many staff are available and who is going to collect?
• What should be the format of data collection?
• When to be collected and at what frequency?
• What will be the possible outcome of SPC?
• Concerned Personnel are trained for conducting and
analyzing data?
• Who will be responsible for the collection, analysis and
interpretation of data?
12
 SPC – A Continuous Improvement Tool
DO – Data Collection

• Appropriate process is identified

• Proper format has been given
• Operator / Process has been trained enough
• Equipment if needed is properly maintained
• Factual data (biased - free)

13
SPC – A Continuous Improvement Tool
CHECK – Data Analysis

• Data is represented and analyzed by using an

appropriate graph, chart or SPC tool.

• Interpretation of data

14
 SPC – A Continuous Improvement Tool
ACT – Recommendations for improvements

• Based on data analysis & interpretation,

suggestions are made for continuous
improvement.

15
 ATTRIBUTE

VARIABLE

DATA
DATA and its Types

LOCATIONAL
16
 DATA and its Types

ATTRIBUTE DATA
Counted data or attribute data answers to the questions of “how
many” or “how often.”

Examples are:
• How many of the final products are defective?
• How often are the machines repaired?
• How many people are absent each day?
• How many days did it rain last month?

17
 DATA and its Types

VARIABLE DATA
Measured data (variable data) answers to the questions like “how
long,” “what volume,” “how much time” and “how far.” This data
is generally measured with some instrument or device.

Examples are:
• How long is each item?
• What is the diameter of the product?
• How long did it take to complete the task?
• What is the weight of the product?

18
 DATA and its Types
Variable Attribute

Characteristics measurable countable

continuous discrete units or
occurrences good/bad

Types of Data length no. of defects

volume no. of defectives
time no. of scrap items

Examples width of door gap audit points lost

lug nut torque paint chips per unit
fan belt tension defective lamps

Data Examples 1.7 inches 10 scratches

32.06 psi 6 rejected parts
10.542 seconds 25 paint runs
19
 DATA and its Types

Is the following data continuous or discrete?

. A station wagon weighs 3478.6 Ibs

. On the last SSBB exam, 800 people failed

. Out of 300 steel rods, 12 proved to be defective

. A heat treatment lasted 6 hours and 18 minutes

20
 QUALITY TOOLS

Quality Control Tools

(Mostly dealing with)
Numerical data
Seven Basic QC Tools
 Graphs
 Check Sheets
 Pareto Diagrams
 Cause & Effect Diagrams
 Scatter Diagram
 Histograms
 Control Chart

21
 What are 7-QC Tools

Graphs Scatter Diagram

Pareto diagram Cause & Effect

Diagram

Histograms Control Chart

Check Sheets

22
 Graphs
Examples: Line graph, Bar graph, Pie chart

Purpose:
To represent data with lines and shapes for
Their better understanding and analysis.

“APPLICATION”
For representation of rate of defectives, rejections, rework, quality losses, vendor
performances, yearly sales vs expenses, rate of Absentees, productivity & efficiency.

23
 Check Sheet

Switch Assembly op236 Plastic Footer Operator________

Chart began July 12, 1995

Week 1 Totals
Burns III IIII
Misrun III II
Bad Finish IIII
Porosity I
Flash IIII IIII I IIII I
Color

Purpose:
Forms/Formats specially prepared to enable data to be collected simply by check marks.
Application:
Used by inspectors during checking of the process.
e.g. check sheets for process Distribution
check sheet for defective items
check sheet for defect location

24
 Pareto Diagram

Purpose:
To graphically represent quality problems or their associated
cost in order of their significance. And facilitate decision
making priority wise.
Application:
Commonly used to show types of defects, rejects, reworks or
complaints in products, processes and systems.
25
 Cause & Effect Diagram
Purpose:
A graphical tool to find out the reasons for problems and carryout their root cause analysis through brain storming.
Application:
Commonly used for analyzing the causes of problems which occur in products & processes or systems eg. reason for poor
paint, reasons for high absenteeism, reason for customers’ complaints, reasons for low productivity.
Outputs:
Identification of most contributing causes with consensus.

People Methods
Fail to clean Don’t measure
Make too strong
Bad
coffee
Old coffee Dirty coffeepot

Materials Machinery
26
 Scatter Diagram

Purpose:
To study and evaluate the interrelationship between two characteristics.
Application:
Commonly used to investigate the effect of one characteristic on other e.g.
temp. effect on the strength of metal; moisture effect on elongation, ink
viscosity effect on the registration in the printing.
Outputs:
Positive correlation, No. correlation
Negative correlation.

27
 Histogram

Purpose:
To display the spread (shape) of the data
Application:
Used to display key quality characteristics of variable data,
e. g. length, hardness, thickness, tensile strength.
Outputs of a histogram
Average (x)
Standard Deviation (s)
Normal Distribution curve

28
 HISTOGRAM

1. A stable process is characterized by a ‘Histogram’

exhibiting unimodal or bell-shaped curve
2. A stable process is predictable
3. An unstable process is often characterized by a
histogram that does not exhibit a bell-shaped curve
4. Frequency column graphs display an static picture of
process behavior
5. Histogram requires a minimum of 50-100 data points
6. In a bell-shaped curve, most of the points fall near the
centerline or average
29
 HISTOGRAM

7. The centerline divides the curve into two symmetrical

halves
8. Some of the points approach the minimum and
maximum values
9. Very few points are outside the curve
10. One sigma limits contain 68.26 percentage of data
11. Two sigma limits contain 95.46 percentage of data
12. Three sigma limits contain 99.73 percentage of data

30
Normal Curve – Bell Shaped Curve

2σ
(68%)

µ
4σ (95.5%)
6σ (99.7%)
31
 Control Charts
Purpose:
To display variation by a line graph in a time ordered fashion. A control line and control limits (based on +
3 standard deviations) are placed on the graph to help analyze the patterns in data.
Application:
To show whether or not a process is in control. It is used in monitoring key Quality Characteristics during
continuous processes.
Common Types:
For variable data For Attribute data
- XR-chart - p- chart, np chart,
- individual X moving range chart - c-chart, u-chart

32
Theory of Variation

“Variation exists in Nature”

33
 Control Charts

• The Control Chart will tell you whether your process is in

statistical control – W. Edwards Deming
• Control Charts are the most powerful tools to analyze
variation in most processes either manufacturing or
administrative.
• Control Charts are line graph that display a dynamic picture
of process behavior
• Control Charts require 100 data values to calculate upper
and lower control limits
• But, it requires only periodic small subgroups to continue to
monitor the process

34
 Types of Shewhart Control Charts

Control Charts for Variables Data

X and R charts: for sample averages and ranges.

X and s charts: for sample means and standard deviations.

Md and R charts: for sample medians and ranges.

X charts: for individual measures; uses moving ranges.

Control Charts for Attributes Data

p charts: % age Defectives

np charts: Defectives

c charts: Number of Defects

u charts: Average number of defects

35
 Process Control Chart Factors

UCL Factor LCL Factor Factor for

Control Limit for Ranges for Ranges
Sample Factor for Estimating
(Range (Range Sigma
(Subgroup) Averages
(Mean Charts) Charts) Charts) ( = R/d2)
Size (A2) (D4) (D3)
(n) (d2)

2 1.880 3.267 0 1.128

3 1.023 2.575 0 1.693

4 0.729 2.282 0 2.059

5 0.577 2.115 0 2.326

6 0.483 2.004 0 2.534

7 0.419 1.924 0.076 2.704

8 0.373 1.864 0.136 2.847

9 0.337 1.816 0.184 2.970

10 0.308 1.777 0.223 3.078

36
 CONTROL CHARTS
TERMINOLOGY

• CENTER LINE (PROCESS AVERAGE)

• UPPER CONTROL LIMIT (UCL):
HIGHEST POINT BEFORE PROCESS IS OUT OF
CONTROL
• LOWER CONTROL LIMIT (LCL):
LOWEST POINT BEFORE PROCESS IS OUT OF
CONTROL

37
 CONTROL CHARTS
THE BASICS

CONTROL CHART

Upper Control
Limit
Y (results)

X (Grand Average
or
(Expected Result)
Lower Control
Limit

X (observations)

38
 CONTROL CHARTS
VARIATION

 CONTROL CHARTS DISTINGUISHES BETWEEN:

 NATURAL VARIATION (COMMON CAUSE)
 UNNATURAL VARIATION (SPECIAL CAUSE)

UNNATURAL VARIATION
UCL

Average NATURAL VARIATION

LCL
UNNATURAL VARIATION

39
 Drift

 A gradual change of a mean or average of a

population.

40
 Shift

 A sudden or Step change in the mean or

average of a population

41
 CONTROL CHARTS
XBAR - R CHART STEPS

 DETERMINE SAMPLE SIZE (n=2-6)

 DETERMINE FREQUENCY OF SAMPLING
 COLLECT 20-25 DATA SETS (Around 100 data points)
 AVERAGE EACH SAMPLE (X-bar)
 RANGE FOR EACH SAMPLE (R)
 AVERAGE OF SAMPLE AVERAGES =
 X-double bar (defines center line)
 AVERAGE SAMPLE RANGES =
 R-bar (defines center line)

42
 CONTROL CHARTS
XBAR - R CHART STEPS

 XBAR CONTROL LIMITS: - UCL = XDBAR +

(A2)(RBAR ) - LCL = XDBAR - (A2)(RBAR )

 R CONTROL LIMITS: - UCL = (D4)(RBAR ) -

LCL = (D3)(RBAR )

43
 Process (Out-of-Control)

 A process is said to be ‘Out-of-Control’ if any point falls

outside the control limits in average or range chart
 When a process is ‘Out-of-Control’, then special causes of
variation are present in average or range chart
 Control limits are the boundaries set by the process which
alert us to process stability and variability

44
 Process (Out-of-Control)

Average Out-of-Control
Range In-Control

Average In-Control Process

Range Out-of-Control Out-of-Control

Average Out-of-Control
Range Out-of-Control

Average In-Control Process

Range In-Control In-Control

45
 X bar and R Chart

Sample Number
1 2 3 4 25

Values
4 7 6 7
Select 25 small samples
(in this case, n=4) 6 3 9 6
Find X and R of each 5 8 8 6
sample.
5 6 9 5
The X chart is used to Sum 20 24 32 24 28 Total
control the process mean.
X 5 6 8 6 7 150
The R chart is used to R
control process variation. 2 5 3 2 3 75

46
 X and R Charts

Sample Number
n A2 D4 D3 d2 1 2 3 4 25
4 7 6 7

Values
2 1.880 3.267 0 1.128 6 3 9 6
3 1.023 2.575 0 1.693 5 8 8 6
4 0.729 2.282 0 2.059 5 6 9 5
Sum 20 24 32 24 28 Total
X 5 6 8 6 7 150
R 2 5 3 2 3 75

47
 X and R Charts
n A2 D4 D3 d2

2 1.880 3.267 0 1.128

3 1.023 2.575 0 1.693
4 0.729 2.282 0 2.059
Sample Number
X = 150 / 25 = 6 1 2 3 4 25
4 7 6 7
R
–
= 75 / 25 = 3
– 6 3 9 6

Values
A2R = 0.729(3) = 2.2 5 8 8 6
– 5 6 9 5
UCL
– X
= X + A2R = 6 + 2.2 = 8.2 20 24 32 24 28 Total
Sum 5 6 8 6 7 150
LCLX = X
– - A2R– = 6 - 2.2 = 3.8
– 2 5 3 2 3 75
X
UCLR = ––D4R =–2.282(3) = 6.8 R

– = 0(3) = 0
LCLR = D3R
–
σ = R /d2 = 3/2.059
48
 Sample Number
X and R Charts 1 2 3 4 25
n A2 D4 D3 d2
4 7 6 7

Values
6 3 9 6
2 1.880 3.267 0 1.128 5 8 8 6
3 1.023 2.575 0 1.693 5 6 9 5
4 0.729 2.282 0 2.059 Sum 20 24 32 24 28 Total
X 5 6 8 6 7 150
X = 150 / 25 = 6 R 2 5 3 2 3 75
–
– = 75 / 25 = 3
R UCL X– = 8.2
– R = 0.729(3) = 2.2 –
–

Mean
A 2 X = 6.0
UCL – = X + A R = 6 + 2.2 = 8.2
X
–
2 LCL X– = 3.8
– –
LCLX = X - A2R = 6 - 2.2 = 3.8 UCL R = 6.8
–
– –
Range

UCLR = D4R = 2.282(3) = 6.8 –

– R = 3.0
LCLR = D3R– = 0(3) = 0 LCL R = 0
49
Trend Analysis

Examples of Trends

A Stable Process
50
 Trend Analysis

Examples of Trends
51
 TEST FOR
CONTROL CHARTS

1. One or more points outside control limits.

2. Two or three consecutive points outside 2 sigma
limits but still inside control limits. (same side)
3. Four or five consecutive points beyond 1 sigma
limits. (same side)
4. A run of 9 consecutive points on one side of the
central line.
5. Six points in a row steadily increasing or decreasing,

52
 TEST FOR
CONTROL CHARTS

6. Fifteen points in a row in zone “C” (up & down).

7. Fourteen points in a row alternating up and down.
8. Eight points in a row on both sides of the center line
with none in zone “C”.

53
 RECOMMENDED ACTIONS

CAUSES CORRECTIVE ACTION

Tool Wear Repair Machine

Tired Operator Use Alternate Machine

Change in Material Composition Change, Repair or Sharpen Tool

Mixed Material from Various Suppliers Investigate Material

New Operator or Machine Keep Consistent Material Supply

Modification of Production Method Check Out Machine Capability

Change in Inspection Device Examine Operator Methods and

Instructions
Regular Rotation of Operator
Check Out Calibration of Measuring
Environment (Temperature, Humidity) Devices

Schedule Maintenance Adjust Environment, if controllable 54

 Process Capability

 Differs Fundamentally from Control

Charting
 Focuses on improvement, not control
 Variables, not attributes, data involved
 Normal Distribution
 Inherent capability (Process Width) is
compared to specifications (Design
Tolerance)
 Requires Process First to be In Control

55
 Process Capability

 A process is capable when its spread on

the normal curve is narrower than the
tolerance range

 Process variation is compared to

engineering specifications

56
 Process Capability - Applications

 Evaluation of new equipment

 Reviewing tolerances based on inherent
variability of a process
 Assigning equipment to product
 Routine process performance audits
 The effects of adjustments during
processing

57
 Process Capability

Process Capability (PC) is the range in which "all" output

can be produced.

Definition:
PC = 6σ

µ
6σ (99.7%)
58
 Process Capability Chart

Process output
distribution

Output Output
out of spec out of spec

5.010

4.90 4.95 5.05 5.10 5.15

5.00
cm
X
Tolerance band
LSL USL
Inherent capability (6 ) σ

59
 Process Capability

This process is
CAPABLE of
producing all good
output.
Lower
Spec ➤ Control the process.
Upper
Limit Spec
Limit
× This process is
NOT CAPABLE.
CAPABLE

➤ INSPECT - Sort out

the defectives
60
 Process Capability

Process Capability Formula:

Cp = Design Spec Width / Process Width
Cp = (USL-LSL) / 6σ (σ = R-bar / d2 )

Thumb Rule:
Cp > 1.33 Capable
Cp = 1.00 Capable with tight control
Cp < 1.00 Incapable

61
 Process Capability

Process Capability Index (account for Mean Shifts):

Cpk = Min (Cpl, Cpu)

Cpl = (Xbar - LSL) / 3σ
Cpu = (USL - Xbar) / 3σ

62
 Process Capability

Cpk Meaning
Negative. Process Mean outside Spec Limits
0 - 1.0 Portion of process spread falls Outside Specs
> 1.0 Process spread falls within Spec Limits

For Six Sigma, an approximate value of Cpk = 1.5

63
 (a) (b) (c)
Process Capability: C pk Variations

Cpk = 1.0 Cpk = 1.33 Cpk = 3.0

LSL USL LSL USL LSL USL

(d) (e) (f)
Cpk = 1.0 Cpk = 0.60 Cpk = 0.80

LSL USL LSL USL LSL USL

64
 Process Capability

(Desired Performance) / (Actual Performance)

Note that average performance is
not centered between the spec
limits
This curve is the
distribution of data from The shaded areas represent
the process the percentage of off-spec
production

Voice of Customer
Voice of Process
65
 Process Capability Analysis

• Do Nothing
• Change the Specifications
• Center the Process
• Reduce Variability

66
 Cpk vs. PPM

Cpk PPM

0.33 317,500
1.00 2700
1.10 967
1.20 318
1.30 96
1.40 27
1.50 6.8
67
Link between SPC and SIX SIGMA

OPTIMIZATION 68
 EXERCISES
EXERCISES

Case Study 1: X-bar and R Chart

Case Study 2: Histogram

Case Study 3: Process Capability

MINITAB EXERCISES
69