Statistical Process Control

Manual

Mine Safety Appliances Company P.O. Box 428 Pittsburgh, PA 15230-0428

Telephone: (724) 733-9100

 Table of Contents

I Introduction and Structure Pages 1 – 5

II Definition Pages 6 – 18
• Guidelines
• Continuous Improvement

III Problem Solving Pages 19 – 72

• Seven Steps
• Brainstorming
• Pareto Chart
• Flow Chart
• Cause and Effect Diagram
• Problem Solving Methods condensed
• Statistical Method

IV Analyzing the Process Pages 73 – 87

• Collecting Data
• Histogram
• Normal Curve

V Process Capability Pages 88 – 99

• Capability Ratio
• CPK

VI Control Charts Pages 100 – 172

a. Selecting Type
b. Subgroup Sizes
c. Variables charts
• Average or Mean
• Range
• X and R Chart
• Moving Range and Individual Charts
• Chart Interpretation
d. Attribute Chart
• Symbols and Terminology
• P Chart
• NP Chart
• C Chart
• U Chart
• Chart Interpretation
• Process Capability – P Charts
VII Measurement System Analysis Pages 173 – 178
• Gage Repeatability and Reproducibility Study
• Interpretation of R & R Study

VIII Appendix Pages 179 – 194

• Mathematical Symbols
• Definition of S.P.C. Symbols
• Table of Constants
• Formulas
• Glossary of Terms
 INTRODUCTION

Institutional Statistical Process Control (SPC) is a long term program and is much more involved than simply having
charts on machines to monitor variation.

The key to the successful implementation of SPC is the combined effort of everyone working towards a reduction in
and the eventual elimination of those factors causing inconsistency.

SPC provides a method to:

A. Identify and analyze problems

B. Determine process capabilities

C. Assist in setting design specification

D. Reduce rework and scrap

E. Reduce part-to-part and lot-to-lot variation

F. Increase productivity

Our goal is to reduce product variation resulting in the highest possible quality safety products and accomplish this to
the least possible manufacturing cost.

To assure we meet this goal, management has made the commitment to implement SPC and to make all personnel
knowledgeable with its uses. SPC provides all employees with a technique to ensure continued improvement of quality
and productivity throughout our company. In today’s world market, we will use these techniques to compete.
Everyone is responsible for quality.

1
 MURRYSVILLE SPC MANAGEMENT STRUCTURE

SPC CORPORATE MURRYSVILLE PLANT

MANAGER SPC STEERING TEAM

FACTORY MANAGERS** FACTORY MANAGERS** FACTORY MANAGERS**

SPC TEAM SPC TEAM SPC TEAM

DEPT** DEPT** DEPT**

SPC TEAMS SPC TEAMS SPC TEAMS

DEVELOP AND GUIDE SET SPECIFIC GOALS WORK ON PROBLEMS

S. P. C. PRPGRAM

* TO INCLUDE: ** TO INCLUDE: *** TO INCLUDE:

MANAGER QA FACTORY MANAGERS SUPERVISORS

FACTORY MANAGERS SUPERVISORS INDUSTRIAL ENGINEERS
SPC COORDINATOR INDUSTRIAL ENGINEERS SET-UP MEN
MANUFACRUTING ENGINEERS INSPECTORS
Q. C. ANALYST FLOOR PERSONNEL
Q. C. ENGINEER Q. C. ANALYST
S. P. C. COORDINATOR S. P. C. FACILITATOR

2
 S. P. C. TEAM FUNCTIONS

S. P. C. STEERING TEAM

TO DEFINE AND IMPELEMENT THE S.P.C, PROGRAM FOR THE MURRYSVILLE PLANT.
THE CORPORATE S.P.C. MANAGER WILL ACT AS AN ADVISOR.

FACTORY MANAGERS S. P. C. TEAM

SET SPECIFIC GOALS FOR THEIR DEPARTMENTS, PROVIDE TOOLS, AND TRAINING
FOR MEASURING AND EVALUATING PROCESSES. HAVE ROUTINGS CHANGED TO
REQUIRE S.P.C. ON CRITICAL DIMENSIONS. WORK WITH DESIGN ENGINEERING
TO ESTABLISH MEANIGFUL TOLERANCES AND WORK TO SOLVE PROBLEMS
DISCOVERED BY, BUT NOT CORRECTABLE BY THE DEPARTMENTAL TEAMS.

DEPARTMENTAL S. P. C. TEAMS

TO SOLVE PROBLEMS AND ADVISE NEXT LEVEL OF MANAGEMENT OF PROGRESS

AND NEEDS. TO PROVIDE AN OPEN AND FREE SPIRITED FORUM FOR DISCUSSION
OF ALL QUALITY PROBLEMS WITH AN EMPHASIS ON STATISTICAL THINKING. TO
WORK WITH OTHER DEPARTMENT TEAMS TO CORRECT COMMON PROBLEMS.

3
 STATISTICAL PROCESS CONTROL
(OVERVIEW)

The goal of this manual is to explain your role in maintaining a successful Statistical Process
Control System. This goal will be achieved by giving you a better understanding of the benefits
of using process control charts, and how this information is used to improve quality and
productivity. It contains the basic information we will use to start our program and will be
amended as required by our experience.

Statistical Process Control (SPC) is a technique that is used to monitor, control, evaluate, and
analyze a manufacturing process. It is a system of using charts and statistical methods to detect
defects that need remedied in processes. Since operating characteristics of the process are what
determines quality and productivity, improvement depends on the ability to improve the process.
Once these defects are remedied, the process becomes stable and the true capabilities of process
can be determined. This system can then be used to monitor and continually improve the
process.

In no sense does a statistical procedure ever substitute for human knowledge and ingenuity.
People still have to make decisions as to quality, production, or compatibility of specifications
based upon information they obtain. The statistical method is simply a mathematical device or
tool, which obtains better information than any other method known to date to assist people in
making their decisions.

The points plotted and control limits derived are based on averages, not individuals. The process
average and range are used to determine the stability of the process and are not to be used for
acceptance or rejection of the individual parts. The individual parts will continue to be inspected
to the drawing tolerances. The SPC information will be used to adjust and solve problems in the
manufacturing process. It must be emphasized that the control limits have nothing to do with the
drawing tolerances except in the calculation of process capability. When the process has been
proven to be both capable and in statistical control, we may decide to reduce the level of end
product inspection.

The role of the SPC facilitator in our SPC Program is to check the charts for statistical control
and write reports to the supervisor when they find bad (out of specification) dimensions or out of
control conditions. Also, their product knowledge should be used to correct problems found by
the SPC charts. They will provide statistical knowledge to the department, as well as one-on-one
training of individuals, while working with other facilitators in the plant on problem detection
and solution. In areas where the operators have had simplified training, the facilitator will
calculate limits and provide the first chart, as well as choose the characteristics to be charted.

4
The role of the operator or person in control of the actual manufacturing process shall be to make
measurements or observations, make necessary calculations, and plot points on the appropriate
type SPC chart. They will interpret the charts and make adjustments to the process when
necessary or notify someone who can make the adjustments. They must note on the chart any
changes that are make which may affect the process. As occasional members of the Department
SPC Team, they will help to solve problems found both in their department and by another team
or customer. They will file in-process and completed charts in the QA part number folder, and
make new copies as required.

The Departmental Supervisor is responsible for the process and anything needed to perform the
operator inspections as well as checking the charts periodically for statistical control. Also, they
are to organize and promote group discussions for problem solving, and a continuous
improvement policy.

5
Statistical
Process
Control

6
THE QUALITY / PRODUCTIVITY
KEY:

INTERNAL

PROCESS CONTROL

NOT

EXTERNAL

QUAILTY CONTROL

7
 DETECTION METHOD

(Quality Through Inspection)

ADJUST PROCESS

SCRAP
MACHINES OR
REWORK
MATERIALS

PEOPLE PROCESS
P
METHODS R MASS
O INSPECTION
ENVIRONMENT D
OK
U
C
T

1. Inefficient – Physical and mental fatigue decreases the efficiency of inspection. Experts
estimate 100% inspection is only 80% accurate at best.
2. Costly – Separating good from bad parts is not cost effective. Scrap, rework and repair
all add to the cost of the final product.
3. Customer Dissatisfaction – The inspector of last resort is the customer. Reliance on
inspection to assure quality will result in the customer being a vital ingredient in your
quality control program.
4. Confused Responsibility – The Quality Control Department is not responsible for quality.
Yet, most organizations believe that poor quality means the product has not been
inspected enough.
5. Symptom Oriented – The detection method emphasizes production over quality.
Therefore, adjustments to the process will be geared toward increasing production, not
quality. Problems in the system are not removed but contained,
6. Neglected Improvements – Too busy “fighting fires” to work on process improvements.

8
 PREVENTION METHOD
(Monitoring Process as Each Part is Produced)

MONITOR/ADJUST

POTENTIAL
REDUCTION
OR
ELIMINATION

MACHINES

MATERIALS STATISTICAL
PROCESS PROCESS PRODUCT
PEOPLE CONTROL INSPECTION
METHODS OK

ENVRIONMENT SELECTIVE
MEASUREMENT
* PRODUCT
* PROCESS

1. Change in Philosophy – Away from emphasis on production to emphasis on quality.

Away from short-term profits toward increased profitability through increased
productivity.
2. Change in Focus – Solving problems at the point they occur rather than prior to shipment.
3. Use of Statistics – Reducing problems in the system through the use of statistical
techniques that indicate how consistent, predictable and repeatable the process is.
4. Change in Responsibility – The prevention method brings a deeper sense of responsibility
for quality to the production line.
5. Change in Attitude – Rather than acting as separate units, representatives of design,
engineering, manufacturing, assembly, purchasing, accounting, etc. – act together to
discover problem causes, not just the symptoms or the results of the problems.
6. On-going Improvements – Process orientation and the commensurate increases in process
knowledge will lead to continuous improvement in process capability.

9
 STATISTICAL PROCESS CONTROL

WHAT IS SPC?

- Controlling of a process through the use of a statistical or

mathematical approach.

- Organized method for improving the quality and productivity of a

process.

10
 Statistical Process Control

A guideline for applying SPC tools and related problem solving concepts:

1. Select a candidate for study (try to find data having persuasive power).
a) Pareto Analysis
b) Cause & Effect
c) General Listing

2. Define the process (scope of study).

a) Cause & Effect
b) Flow Chart of Process
c) Brainstorming

3. Procure resources for the study.

a) Develop forms to collect data
b) Designate characteristics to be monitored
c) Who is to collect data? Will process run undisturbed? How will
project progress be reported?

4. Determine adequacy of measurement system

a) Measurement System Analysis
b) Calibration Program

5. Confirm Control Systems

a) Be sure system provides means for identifying units in a manor
allowing you to determine which variables have changed.
(Flow chart, etc.)
b) Assure adequate training is provided
c) Identify as many assignable causes in advance.
(Cause & Effect, Brainstorming)

6. Select methods of analysis

a) Frequency Plot
b) Run/Time Plots
c) Multi-Vari Chart Study
d) Control Charts (X, R, P, NP)
e) Moving Range
f) Scatter Plot

11
7. Gather & Analyze Data
a) Frequency Plot
b) Run/Time Plots
c) Multi-Vari Chart Study
d) Control Charts (X, R, P, NP)
e) Moving Range
f) Scatter Plot

8. Remove special causes of variation until process is stabilized (in statistical

control)

9. Estimate process capability

a) Compute S
b) Compute X
C) Compare 6S to Spec. Spread (capability ratio)

10. Establish quality control plan for process.

a) Standardize what was learned in steps 1 – 9
b) Audit established process
c) Control charts in key process parameters on product characteristics
d) Action to be taken when assignable cause is detected

12
 CONTINUOUS IMPROVEMENT

KEY COMPNENTS

• TOP MANAGEMENT COMMITMENT AND LEADERSHIP

• QAULITY IMPROVEMENT BUILD INTO THE BUSINESS PLAN

• TRAINING AND PROJECTS

• PHILOSOPHY AND CONTINUOUS INPROVEMENT – ALL DISCIPLINES – PROJECT

BY PROJECT

• COST REDUCTION PROJECTS

• MEASURABLE PROGRESS

• COMMITMENT TO HUMAN RESOURCES DEVELOPMENT

• ORGANIZATION AND COMMUNICATION TO SUPPRT CONTINUOUS QUALITY

IMPROVEMENT

• EMPHASIS ON CULTURE CHANGE

• STATISTICAL PROCESS CONTROLL

• REWARD STRUCTURE TO ENCOURAGE QUALITY IMPROVEMENT

• INTERNILIZE QUALITY IMPROVEMENT PROCESS THAT BECOMES A WAY OF

DOING BUSINESS

13
 A PLAN FOR IMPLEMENTING CONTINUOUS IMPROVEMENT

1. To improve the control of processes / operations:

• Determine key product / service variables for each customer / supplier relationship.
(e.g. temperatures, chemistries, equipment availability, purchased material
characteristics, timeliness)

• Determine key process variables that are adjusted to obtain the key product / service
variable (speeds, temperatures, purchasing criteria, frequency of maintenance)

• Establish which process variables are in control. Hold regular review meetings to
discuss status of this effort

• Review or establish specifications for the key product / service and process variables

• As the processes are brought under control, determine their capability.

• Analyze the process to improve capability. Hold regular review meetings of operating
people to discuss success stories, strategies and problems.

• Develop standard operating practices.

• Train staff to follow the standard practices.

• Implement a program of process monitoring and review of capability with an aim

toward organized adjustments in the standardized procedures for the improvement
capability.

• Establish quality improvement teams to attack selected processes and their related
waste. These teams should contain operating people who understand and believe in
continuous improvement.

• Each team should have adequate technical resources, someone knowledgeable in

statistical methods for problem-solving.

• The team should be in an operation whose management is fully committed to the use of
the tools of SPC and SPS.

• Provide a communication system to disperse the results of the teams efforts to other
areas. Avoid having to re-invent the wheel with each department.

14
2. Address Organizational Needs:

• Designate a company coordinator with adequate staff for providing assistance.

• The coordinator must be at a high enough level and report to senior management.

• Identify staff members in each department who will be charged with applying SPC and
SPS techniques to control and improve variability. Ensure that they have enough time
to do this. Do not add this responsibility on top of everything else!

• Every operating are should have at least one person skilled in statistical thinking.

3. Plan for Training Needs:

• Training in statistical fundamentals

• Training at all levels: management, operators and support staff

• Techniques for sampling, determining maintenance programs and statistics for

scheduling.

• Additional resource people with more advanced skills.

• Trainers for additional courses

• Training hourly workers to, at least, plot points and look for out of control conditions.

4. Work Toward Eliminating Waste:

• Long-term change to reward use of SPC and SPS techniques to correct problems when
they occur.

• A system to move away from “productivity” and towards productively following

standard practices, and monitoring and improving the process.

5. Develop a “Cost of Quality” System to Track and Report on the Hidden Cost of Poor Quality:

• Plan for a regular review of the costs of poor quality with emphasis on constant
reduction.

• This must be done carefully to avoid gamesmanship. The data must be used by the
responsible manager to evaluate efforts and not people.

15
6. Obtain the Involvement of Support Areas:

• The purchase department must ensure that suppliers are using statistical process
control.

• The maintenance department should be trained in statistical thinking to improve the

quality of repair efforts.

• The scheduling department can improve its capability through understanding of

statistical principles.

7. How to “Find Time”:

• Examine existing reports and eliminate those which do not help to manage effectively.
We are drowning in enormous efforts to obtain disseminate useless numbers.

• Eliminate the collection and reporting of statistics which do not reflect the natural
variation caused by process changes.

• Examine the decision-making process of management and decide when a phone call is
likely to be ghost-chasing, wasting effort.

16
 MAKING STATISTICAL TRAINING EFFECTIVE

• Customize Training (Site Specific)

• Train Teams as Teams

• Train in Chunks (Use What You Learn)

• Train in the Real World (Real-Life Applications)

• Focus on Results, not Methods

• Install a Problem Solving Tool, not a system

• Use Statistical Problem Solving (SPS) Along with SPC

17
 BENEFITS

1. IMPROVED SALES – People who understand how effective SPC is at controlling

quality and improving productivity are much more likely to buy products produced
under this system.

2. IMPROVED PRODUCT QUALITY – and therefore increase productivity

3. REDUCE END ITEM INSPECTION – Suppliers who successfully use these

techniques during production are candidates for reduced inspection.

4. HELP REDUCE – scrap, rework, repair and unnecessary adjustments to the process.

5. Determine first of all whether a PROCESS IS CAPABLE of producing parts of

specification.

6. Help Product Engineers to ESTABLISH REALISTIC TOLERANCES on future

work or change existing tolerances to coincide with existing process capabilities.

7. HELP DIAGNOSE AND CORRECT PRODUCTION PROBLEMS. Decisions for

corrective action are based on facts rather than intuition or memory.

8. Provide a COMMON LANGUAGE for effective communication between the

machine, operators and management.

18
 Seven Steps of Problem Solving

1. Define Problem / Set Priorities.

This can be done in several ways. Brainstorming will allow you to generate a list
of problems. After the initial list of problems is generated, edit the list to similar
categories and make a Pareto Chart. When collecting data for the Pareto Chart, be
sure to spend enough time to actually represent the entire/complete process.

Examples:

• One job may run a high rate of “defect c.” If data is only collected during
that job run, you may have misleading information.

• One day of order entry may have unusually high kick-outs. Collecting data
for only that day may be misleading.

Accurate data collection may take weeks or months! Experienced team members
may be tempted to make the problem selection based on gut feeling and bypass
data collection but, factual information is important for several reasons:

• Data collected will show:

How big a problem we have.

How much it is costing us in production, waste, dollars, time, etc.

These facts will be invaluable for the Steering Team later in this process.
Cost may be analyzed in all of the 4M’s.

• You may find from the data collected, that the problem is not as great in
magnitude as once suspected.

• When a team arrives at Step Seven and evaluates the action taken, there is
historical data for comparison.

19
When a problem is selected it should be defined by a problem statement or short
paragraph. DO NOT list any causes in this statement. By listing causes you limit
yourself and exclude possibilities.

Often, several Paretos will be made in sequence. The information gained from one
will be helpful for the next Pareto Chart.

Example:

The Team collected data in the injection molding and the types of
defects. They found that contamination is the biggest type of waste.

At this point it may be appropriate to make a Pareto Chart on the types of

contamination discovered. The same sequencing can be applied in the office area.

2. Analyze How The Process Behaves:

This step is often omitted by most teams. It is crucial to understand how this is
affecting your process. Several tools will help:

Histograms: To show where your process is and the distribution.

Control Charts: Will help understand how this problem affects the
process.
Flow Chart: Will help break down all steps in the process.

3. Identify All The Causes:

Many teams already know what is causing the problem because they have lived
with the problem for years. Remember, the team is made up of persons from all
parts of the process so each will see the problem from a different angle. If the
previous two steps were followed properly, new light will likely be shed on the
process.

A cause and Effect Diagram is an excellent tool here because it organizes all the
potential causes to the problem and often opens doors that were not previously
considered. Using the tool properly will take several hours. Upon completion, the
team should go back through each area and if necessary collect data on several key
causes.

20
Flow Charts may also be used at this point is the process. Look at the detailed,
completed chart and find specifically which steps are the snags in the process.

When this step is complete, the team knows the problem (from step one), and the
cause (from step three).

4. Identify All The Possible Solutions:

This is an excellent opportunity for brainstorming. Solutions may be generated by

considering the 4M’s categories from the Cause and Effect Diagram; material,
machines, method, manpower. Solutions may include a change in:

Material: Deciding if one supplier of a material is capable of

delivering consistently high quality at a reasonable price.

Machines: Parts, rebuilding, or new machinery

Method: The way you do things or a change in the process.

Manpower: Training of the operator.

5. Select The Best Solution:

The team’s best approach to this is a Consensus decision. Ask clear questions:

How will this solution correct the problem?

Why is this the best solution?

The decision is one in which, after a structured discussion, everyone will support
as if it were their own. It is always wise to choose an alternate solution.

Keep in mind the value of your solution. If the problem costs the company
$350.00 a month, can your solution of a $1000,000 machine justify itself in a
reasonable amount of time?

21
6. Implement Solution:

On a “common cause” type problem, the solution is presented to the Steering

Team. Have your Key Facilitator (supervisor) look over the proposal. Invite the
Coordinator to your meeting. The homework to be done to present to the Steering
Team will be minimal.

The work has already been done in the previous five steps of Problem Solving.
Bring all data, charts, etc. to the Steering Team meeting. When a decision has
been reached, decide who will implement the solution.

7. Evaluate Action Taken:

This is a very important step to ensure that the solution implemented has worked.
Many methods apply to this step. In essence you are repeating the first step. If a
Pareto Chart was made on different types of contamination, collect data under the
same conditions and categories at last time, and determine if the original biggest
problem has changed to a different category.

Summary

Always think of problems as opportunities for improvement! Opportunities exist

in waste reduction, problem solving, and most importantly process improvement.
The emphasis is on improving productivity and when you improve productivity,
production will automatically increase.

22
 CONSENSUS DECISIONS

Consensus has been reached when:

• All group members agree to support the decision though it may not be
everyone’s first choice.

• Everyone is committed to it as if it were the first choice of all group

members.

• Everyone agrees that he or she has had sufficient opportunity to

influence the decision.

Any member of the group can block a decision. This is precisely why consensus
decisions are both more difficult and more effective than other group decision
methods, such as voting. It forces the group to consider all aspects of the problem
and objections to possible courses of action. Treat differences of opinion as a way
of (1) gathering additional information; (2) clarifying issues and (3) forcing the
group to seek better alternatives.

GUIDELINES

1. Try to get underlying assumptions regarding the situation out into the open
where they can be discussed.

2. Listen and pay attention to what others have to say. This is the most
distinguishing characteristic of successful work groups.

3. Be cautious of early, quick, easy arguments and compromises. They are often
based on erroneous assumptions that need to be challenged.

4. Avoid competing and arguing. In this situation either the group wins or no one
wins. What is “right” is the best collective judgment of the group as a whole.

23
5. Do Not Vote. It will split the group into “winners” and “losers”, it encourages
“either……..or” thinking when there may be other ways, and it fosters
argument rather than rational discussion.

6. Encourage others, particularly the quieter ones, to offer their ideas so that
everyone is included in what is decided.

7. Tension reducing behaviors can be useful so long as meaningful differences are

not “smoothed over” prematurely.

8. The best results flow from a fusion of information, logic and emotion.

24
 THE FIVE WAYS

The solution to some problems is readily apparent if the ROOT CAUSE can be
determined. It is sometimes possible to determine the ROOT CAUSE by asking
the question WHY about FIVE times.

EXAPMLE:

Q. 1. Why are we late in shipping complete units?

A. Final Assy. Is behind in completing gear boxes

Q. 2. Why is Final Assy. Behind in completing gear boxes?

A. Because of late delivery of finished shafts for the gear boxes from
machining.

Q. 3. Why is machining behind in finishing shafts?

A. Because of an increased rate of breaking cutting tools.

Q. 4. Why are the cutting tools breaking more often?

A. The incoming shaft stock is harder.

Q. 5. Why is the incoming shaft stock harder?

A. The supplier sent the wrong grade of steel.

Some thought must be given to asking the right WHY’S. It is not uncommon to
see solutions in less than an hour, when this technique is applicable and used
correctly.

Extreme caution must be used in applying the WHY approach. It can be very
intimidating if used to find blame. Particularly is the questions are used from the
top down and/or with emotion.

25
 FIVE WHYS EXAMPLE

Q. 1. Why did the crust breaker at the point line fail?

A. Because of an air leak

Q. 2. Why was there an air leak?

A. A hole developed in the hose (XYZ)

Q. 3. Why did this hose develop a hole?

A. The hose rubbed thin around the pivot pin.

Q. 4. Why did the hose wear thin at that location?

A. The hose was either to short or the radius of the pin was too
small.

Q. 5. Why did either of these reasons occur?

A. A design problem which could have resulted from lack of
feedback to the design engineer.

26
 BRAINSTORMING

Lesson Objectives….

To help you make the most effective use of this manual and your time in learning
the concepts described, we have listed below your study objectives for this lesson.
These objectives tell you where you are going in this lesson and what you can
expect to learn.

Please read these objectives carefully and keep them in mind as you read and study
this lesson. You should be able to accomplish each of these objectives upon
completion of this lesson.

• Describe the Brainstorming process.

• Identify / List the steps in effective

Brainstorming.

• Describe how / why Brainstorming is an

effective tool.

• Identify possible uses for Brainstorming in your

area.

• Participate effectively in Brainstorming in your

team.

27
After Brainstorming:

Refine and screen the list of ideas. In this step, duplicate ideas are
removed from the list, similar ideas are combined, ideas are put into
categories, ect.

Present the “final” idea list (after refining and screening) to the group.
Get agreement on the list.

Be sure to express appreciation to the group for their participation.

Congratulate them for the idea list achieved.

Results and Benefits:

This process usually results in a good supply of ideas, many of them

new and different.

“Creativity strikes again.”

Persons are more committed to implementing ideas they had a part in

generating.

28
Specific Steps:

1. Introduce Brainstorming and get agreement to use the method.

2. Identify Topic

3. Select and instruct Recorder(s).

4. Set a time period.

5. Establish ground rules:

• all ideas accepted

• no criticism

• no discussion

• everyone participates

• no evaluation

6. Begin idea generation:

• as many ideas as possible

• quantity not quality

• popcorn effect

7. Allow for occasional “collection” times.

8. Stay with the process until all ideas are exhausted, or allotted
time is used.

29
 9. Review list of ideas

10. Refine and screen ideas

11. Allow any additional ideas

12. Present list as refined and screened

Used To Generate / Identify:

• New concepts / Ideas

• Existing problems

• Causes of problems

• New methods for doing things

• Goals and objectives

• Many, many ideas

(Why not brainstorm all the ways that brainstorming can be used?)

30
Summary:

Most group meetings involve both the generation and the immediate evaluation of
members’ ideas. It is critical to keep the generation separate from the evaluation.
The consequences of evaluating ideas as they are being generated is that group
members will be less likely to continue generating ideas if past contributions are
ridiculed or labeled impractical or unrealistic. The result of this, over time, will be
that fewer and fewer ideas will be generated. Moreover, the ideas will tend to be
primarily generated in order to gain group acceptance or respect.

Sooner or later, individuals internalize the idea evaluation process and “pre-edit”
their ideas before expressing them to avoid ridicule by others. This slows down or
eliminates the spontaneity needed for the creative process. Consequently, novel or
outstanding ideas are rarely conceived. When this happens, groups tend to go over
“old ideas” and then lose interest in and respect for the Brainstorming process.

The Brainstorming process attempt to avoid idea slow-down by separating the

generating from the evaluating of ideas. The theory is that increased quantity of
ideas (even though many may be “bad”) will result in increased quality.
Additionally, ideas, which by themselves may not contribute to the problem
solving, may in combination with other ideas, result in effective solutions or
approaches to solutions.

31
Review:

Brainstorming is a method of idea generation used to ____________________ a

topic with Brain Power. This is usually done in __________________________.
This is more beneficial than discussion because ___________________________.
Initially, always go for ______________________ not ______________________.
A very desirable condition of the Brainstorming session is the ________________,
which occurs when one idea sparks another from other team members. The
recorder’s job is very important and he/she must be careful about _________ and
be sure to ___________. After the initial Brainstorming is complete the next step
is to ________________________________. Brainstorm three examples of how
this problem solving method will help in your process.

1._________________________________________________________________

2._________________________________________________________________

3._________________________________________________________________

32
 Pareto Chart

Lesson Objectives...
To help you make the most effective use of this manual and your time in learning the concepts
described, we have listed below your study objectives for this lesson. These objectives tell you
where you are going in this lesson and what you can expect to learn.

Please read these objectives carefully and keep them in mind as you read and study this lesson.
You should be able to accomplish each of these objectives upon completion of this lesson.

• Construct a Pareto Chart using data.

• Construct a Pareto Chart in conjunction with Brainstorming.

• Define the purpose of the Pareto Chart.

• Describe how to collect data for the Pareto Chart.

• Identify how the Pareto Chart is used.

• Describe application of the Pareto Chart in your work area or

team.

Definition:

The Pareto Chart is a bar graph which is used to prioritize items in their order of importance.

The Pareto Chart has a vertical axis which lists the number of times an event occurred and a
horizontal axis which defines the various events which have occurred.

Frequency
(Number of times
the event occurred)

Events(Categories)

33
Specific Steps:

1. Classify data by identifying categories (such as defects) to be

considered.
2. Decide how long data will be collected.
3. Design data collection sheet/check sheet.
4. Collect data by classification such as number of times it occurs,
how much it weighs, how many items, or how much it costs.
5. Show totals from each category.
6. Draw a graph
7. Show each event in its order of importance. Draw the bars on the
graph.
8. Tell what the graph is, and make a legend.

Used To:

The Pareto Chart is used to:

• Set Priorities
• Rank people’s perceptions when used with Brainstorming.

Application:

Apply the Pareto Chart to the process. Follow the steps for constructing a Pareto
Chart. The priorities set by the Pareto Chart will show you were to begin your
problem solving process in order to improve your process(es).

Example:

A completed Pareto Chart will look like the example below. Yours will have the
classifications, numbers and information from your data collection and study.
No. Of Complaints (124)

50
40
30
20
10
0
BACK LATE WRONG ITEM WRONG ITEM OTHER
ORDERS SHIPMENTS PRICING SHORTAGE SHIPPED
Problem Type

34
Vilfredo Pareto (1848 – 1923) was as Italian economist who made contributions
through his use of scientific methods and theory as an approach to the study and
understanding of economics. Pareto also used sociology as a foundation for this
theories. He was a professor of Political Economy at the University of Lausanne,
Switzerland, from 1893 to 1907.

Vilfredo Pareto gave us the 20/80 rule, which says that 20% of the “problems are
responsible for 80% of our “trouble”. Pareto’s “rule” holds true on your example
graph. This rule applies in many other setting, too.

The graph has set priorities for you by separating the “trivial many” from the “vital
few”. It tells you to begin to look for causes of your most important (top priority)
problem or category, and to begin to analyze how the process behaves. This can be
done by using several methods, possible the Cause and Effect Diagram, which will
be discussed later.

35
 PARETO CHART
DATA COLLECTION WORKSHEET

Team:___________________ Start Date:______________

Department:______________________________________

Operator(s):______________________________________

______________________________________

______________________________________

Categories Of Data: Frequency Tally Box:

36
Review:

In this lesson I have learned how to construct a Pareto Chart.

The purpose of this is ______________________________.
The person we developed this chart is ________________.
Besides the Pareto Chart, he also popularized a rule of thumb
called the ____ rule. This says that ____ of your problems will
give you _____ percent of your trouble.

When a Pareto Chart is created, choosing a problem to work on

is simple. Just select ____________.

To make a Pareto Chart, determine first your

________________ and decide upon _____________. Then
collect data using a tally sheet which is usually designed in the
team meeting.

Upon completion of data collection, add up each category. Next

draw a _______________. Make a _____________ on the
vertical axis which should go just a little higher than the biggest
category. Show each bar starting with the _________ first and
in ___________ order. Show totals for each category.

The ____________ is important because it gives reference

information such as ________________________.

37
 PARETO DIAGRAMS

A pareto diagram (bar graph) indicates which area is responsible

for the largest percentage of your problems. However, an area
which shows only to be a small percentage of your problems
may be costing you the most money. If possible, research the
facts and classify your problems in dollar amounts.

PARETO PRINCIPLE:

“Vital Few” vs. “Trivial Many”

After acting upon the area you’re determined to be the most

severe problems, construct another diagram to see if
improvements have been accomplished. If so, the percentage or
length of the bar for the problem for which you have been
concentrating will have been reduced or replaced as the most
severe. As changes are made to improve percentages or
proportions of bars will vary depending upon how great the
effect of the changes.

38
 CONSTRUCTING A PARETO DIAGRAM

A. Decide upon classification for your diagram and be sure that

these classes have itemized frequencies of occurrence or itemized
costs.

B. Decide upon a time period for your diagram. All related

diagrams should use the same period of time for ease of
comparison.

C. Total each item, or use 100% as the total and list each item as a
percentage of a 100.

D. Draw horizontal and vertical axis.

E. Segment the vertical axis in the terms you are using (ex. Dollars
or percentages)

F. On the horizontal axis list in order of severity the classification

you have researched. Severity can be thought of as the most
dollars spent or costs per problem, or as percentages.

G. Extend the bars for each class to the appropriate frequency or

percentage.

H. Log any pertinent information on your graph for future reference.

39
 0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000

Sp
Sc li ts
ort
ch
Or
Se
t-u
p
A ir
Tra
p s
No
n-F
ills
No
Fu
tur
eU
se
Me
c

40
Sto ha
ck nic
W al
ill N
ot
Pr
oc
es
s
Wr
TOP 10 SCRAP AREAS

ink
led
Un
de
rga
ge

Ma
te ria
l
Dollars
30

25

20

15
PERCENT

10

0
CUT DIRTY IMP/INC TAGGING MISC. H/F FITTING CUT CRIMPM
RUBBER ASSY FINISH LENGTH DIA

41
 FLOW CHART

Lesson Objectives….

To help you make the most effective use of this manual and your
time in learning the concepts described, we have listed below
your study objectives for this lesson. These objectives tell you
where you are going in this lesson and what you can expect to
learn.

Please read these objectives carefully and keep them in mind as

you read and study this lesson. You should be able to
accomplish each of these objectives upon completion of this
lesson.

• Describe the Flow Chart

• Identify/list the steps in constructing a flow chart

• Describe why the Flow Chart is an important tool.

• Identify ways you will use Flow Charts in your job.

42
Definition:

A Flow Chart is a graphic representation of the sequence of all

events (actions) occurring during the performance or procedure.
It is a “map of the system”.

How To Contruct:

1. Identify the system/process to be charted.

2. Write a simple step-by-step description of the process.

3. Describe on a level of detail

4. Identify each function on the process.

5. Identify the appropriate symbol to use with each function.

6. Draw symbols/functions in sequence.

7. Review the Flow Chart

43
Reminders:

• Keep the wording simple

• Maintain a consistent level of detail.

Recheck periodically

• Use a few easily understood symbols.

• Analyze to insure inclusion of each step.

• Check your flow chart for necessity of each

step.

• Look for complexities and redundancy.

• Are there control points built in to

prevent errors and rejects?

• Is this process actually what is should

be?

• How can this process be more effective?

44
Advantages:

• It is a graphic illustration.

• The visual is easier to grasp than the verbal.

• The information is more focused.

• It reveals interdependence of steps in the

process.

• It eliminates unnecessary steps in the

process.

45
 FLOW CHART
A FLOW CHART IS A PICTORAL
REPRESENTAION OF THE SEQUENCE OF
EVENTS IN LOGICAL ORDER.

FLOW CHARTS MAY INDICATE

PROBLEM SOLVING LOGIC OR PROCESS
FLOW.

46
 PROBLEM SOLVING LOGIC FLOW CHART

PROBLEM: LAMP WILL NOT LIGHT

START

CHECK
PLUG

CHECK
BULB

CHECK
WALL
SOCKET

CHECK
LAMP
CORD

CHECK
LAMP
SWITCH

CHECK
LAMP
SOCKET

47
 PROCESS FLOW CHART

LOAD LOAD
CONVEYOR CONVEYOR
(MANUAL) (AUOTMATIC)

UNLOAD DRY
CONVEYOR
(AUTOMATIC)

UNLOAD
DIP CONVEYOR
(AUOTMATIC) (MANUAL)

TRANSPORT
LOAD TO
COMVEYOR FOUNDRY
(AUTOMATIC)

DRY
2

UNLOAD
CONVEYOR
(AUTOMATIC)

DIP
(AUTOMATIC)

48
 FLOW CHARTS ALSO UTILIZE LOOPS AND BRANCHES

LOOP: A SEQUENCE OF EVENTS WHICH REPEAT UNTIL A

CERTAIN CONDITION IS
MET.

START

STEP #1

STEP #2

NO
DECISION
?

END

49
PROBLEM: DOES CHECKBOOK AGREE WITH BANK STATEMENT?

START

RECORD ALL
CHECKS
WRITTEN THIS
MONTH

SUBTRACT
CHECKS
WRIITEN FROM
BALANCE

NO
DOES
CALCULATED
BALANCA AGREE
WITH BANK
STATEMENT

END

50
 BRANCH: TWO PARALLEL SEQUENCE OF EVENTS; WITH THE
CHOICE OF EXECUTION DEPENDING UPON A CERTAIN
CONDITION.

START

STEP 1

DECISION
?

STEP 2A STEP 2B

STEP 3B
STEP 3A

STEP 4

END

51
 WASHING FLOW CHART

LOAD
CLOTHES

COLORED WHITE
CLOTHES CLOTHES

ADD ADD
MILD DET.
DET. AND
BLEACH

SET SET
TEMP TEMP
TO TO
WARM HOT

CLOSE
DOOR

INSERT
25¢

PUSH
PLUNGER

52
 CHOCOLATE FUDGE FLOW
CHART

START
ADD
VANILLA

COMBINE IN SAUCEPAN:
SUGAR, MILK, CHOC. OR
COCOA, CORNSYRUP& BEAT
SALT.

NO

HEAT & STIR HAS IT LOST

ITS GLOSS?

YES
NO

IS CHOC. & STIR IT

SUGAR
MELTED? QUICKLY

YES

POUR IN PAN
COOK

COOL TO SET
NO

IS BALL
SOFT? SUT IN
SQUARES
YES

REMOVE FROM HEAT &

ADD BUTTER
END

COOL TO LUKEWARM

53
 PROBLEM SOLVING FLOW
CHART
START

UNDERSTAND THE
PROBLEM

INDENTIFY THE
PROBLEM

ANALYZE THE
PROBLEM

PLAN THE
SOLUTION

TEST THE
SOLUTION

NO
SOLUTION
OK?

IMPLEMENT THE
SOLUTION

NO
DID
IMPLEMENTATION
SOLVE THE
PROBLEM?

YES

END

54
 CAUSE AND EFFECT DIAGRAM
(Fishbone Chart)

Lesson Objectives….

To help you make the most effective use of this manual and your
time in learning the concepts described, we have listed below
your study objectives for this lesson. These objectives tell you
where you are going in this lesson and what you can expect to
learn.

Please read these objectives carefully and keep them in mind as

you read and study this lesson. You should be able to
accomplish each of these objectives upon completion of this
lesson.

• Identify/list the steps in constructing a Fishbone Chart

• Describe how/why Fishbone Charts are an effective tool.

• Identify possible uses of fishbone Charts in your area.

• Participate effectively in constructing a Fishbone Chart in

your Team.

55
Fishbone Chart / Ishikawa Diagram

Definition:

This is a method for identifying and charting all of the possible

causes of a problem. This method is important because all of
the possible causes of a problem must be known before a
genuine solution can be identified and implemented. It was
developed in 1950 by Kaoru Ishikawa at the University of
Tokyo.

Purpose:

• To identify all of the possible causes of a problem, and

provide a method for showing (charting) the causes do
that appropriate, effective action can be taken.

• To take the “guess work” out of the problem solving.

• To show the relationship between the causes and the

effect they have on the process.

• To provide a tool to systematically break down the

sometimes complex causes of a problem. The Cause
and Effect Diagram also fulfils the need for sorting
through and prioritizing causes.

• Ro eliminate time-consuming and frustrating efforts at

problem solving.

• To find the REAL cause or causes of a problem.

56
Reminders:

Most problems are the result of a number of different

causes. When a problem is identified, it must be analyzed
to identify/define all of the possible causes before
solutions are identified and put into action.

Two “pitfalls” to effective problem solving:

• Assuming the cause(s) is/are known without

analyzing the problem.

• Identifying a problem and immediately

providing a solution.

A Total Quality Attitude in a Total Quality Culture

demands a Total Quality Approach to Problem Solving.
To get real solutions that are permanent and work requires
finding and addressing all the real causes of the problem.
No short cuts!

57
How to construct:

1. Identify and clearly define the problem, the “effect”, or the quality
characteristic that requires improvement.

2. Write the specific problem / effect or the quality characteristic that

requires improvement in a rectangle or square on the right side of the
page.

Problem –
Specifically and Clearly Stated

3. Draw the main line across the page to the problem identified.

4. Ask the question, “What are the Major Causes and of the Problem?”

In most cases, especially in a manufacturing setting, the four M’s of

production are Materials, Methods, Manpower, and Machines. These
will serve effectively as the MAJOR CAUSES.

5. Place two of the MAJOR CAUSES above the main line, and two
below the main line, with arrows drawn diagonally to the line and
slanted toward the problem.

Manpower Material

Problem
Effect

Machines Methods

58
6. Identify the MINOR CAUSES related to or contributing to the
MAJOR CAUSE. Focus on each MAKOR CAUSE and identify the
factors which contribute to that cause. Draw and arrow pointing to the
MAJOR CAUSE for each MINOR CAUSE. These become the “bones”
in the Fishbone Chart. MINOR CAUSES are identified by asking the
questions “How?” or “Why?” The answer becomes the MINOR
CAUSES.

Manpower Material

Problem
Effect

Minor Causes

Machines Methods

Be sure to show MINOR CAUSES as one word or very brief phrases.

Always avoid phrases such as, “bad part”. Instead, write the name of the
part, and use SUB-CAUSES to describe what characteristic caused it to
be “bad”.

7. List the SUB-CAUSES related to or contributing to each MINOR

CAUSE. Your purpose is to factor down each cause from MAJOR to
MINOR to SUB-CAUSES. Draw an arrow for each SUB-CAUSE to
the MONOR CAUSE to which it relates.

59
 Manpower Material

Problem
Effect

Minor Causes

Machines Methods

Keep asking the question “How?” and “Why?” Make each answer
a SUB-CAUSE and ask the question again. Factor these down to one
word answers, if possible.

8. Review the Fishbone Chart.

• Be sure all items which contribute to (or cause) the problem have
been identified and listed.

9. Establish the Action Plan.

• Identify the top priority “causes”.

• Determine which problem solving method(s) will be used next.

• Use the next problem solving method to continue the process.

60
 CUASE-AND-EFFECT DIAGRAM

A cause-and-effect (or Fishbone) diagram is a simple yet effective tool that

facilitates the investigation into the causes of a problem. Fishbone diagrams are
most effective when used by a group of people working on a problem. While they
are most frequently applied to product quality problems, they are just as
appropriate for other industrial and non-industrial problems.

When cause-and-effect diagrams are used for product quality problems, the goal is
usually to find the sources of variation. The diagram is started as indicated below
with the problem on the right and the basic process components represented by the
main branches (sources of variation).

MATERIAL PEOPLE MACHINE

PROBLEM

MEASUREMENT ENVIRONMENT METHOD

SYSTEM

61
CONTSTRUCTING CAUSE-AND-EFFECT DIAGRAM

A. Decide on a problem (effect) and write this on the right side of

your diagram. Draw an arrow from the left side to your problem.

B. Write the prominent factors which could be causing your problem

above or below the main arrow, directing a branch arrow for each
to your main arrow.

C. Onto each branch arrow, write any factors which could effect
these items. Call these twigs. Add twigs to the twigs if necessary.
Continue in this manner until you have found the causing factor.

• If your diagram has only branched and no twigs, then

probably your knowledge of the process is too limited and
it’s time to enlist the help of others to uncover other possible
causes.

Make the most use of your diagram. Work hard at finding the cause.
After you find what you believe is the causing factor, repeat the steps
you took through your cause-and-effect diagram which led you to this
factor. If you get lost in your diagram or you can’t tie the real factor
down, then the diagram causes are not the real causes of the problem.
Reconstruct the steps you took being sure to include any factors not
previously written in.

Log any specific data (ex. Time, material) relative to the cause right on
the diagram to lead you to prompt action in the correct area.

62
 PROBLEM SOLVING METHOD
BRAINSTORMING

PURPOSE: STEPS/RULES:

Generate new Introduce

Ideas Brainstorming and get
Agreement to use the
Identify: method.
• Problems
• Causes Identify Topic
• Solutions
Select and instruct
recorder(s).

Set a time period.

Establish ground rules:

• All ideas accepted

• No criticism

• No discussion

• Everyone participates

• No evaluation

63
 PROBLEM SOLVING METHOD
BRAINSTORMING

PURPOSE: STEPS/RULES:

Begin idea generation:

• as many as possible

• quantity not quality

• popcorn effect

Allow for occasional

“collection” times

Stay with the process

Until all ideas are
Exhausted, or allotted
Time is used.

64
 PROBLEM SOLVING METHOD
PARETO CHART

PURPOSE: STEPS/RULES:

Set priorities Classify data in categories

20 / 80 Rule Set time period

Determine most Collect data

Important category
Show total from each
Item

Draw graph with scale on

Side

Draw bars

Develop legend

65
 PROBLEM SOLVING METHOD
PARETO CHART

PURPOSE: STEPS/RULES:

Determines all the Draw diagram

Possible causes of
One problem Problem in box on
Right side.

Major Causes-Four M’s

Of production

Minor Causes

Subcauses

Ask “how” and “why”

66
 PROBLEM SOLVING METHOD
FLOW CHART

PURPOSE: STEPS/RULES:

Graphic presentation Identify system/ process

Of process, job, or system to be charted

Map of system Write a sample description

Of each step of the process
Training Tool
No dead ends
Add / delete step
In our process Same degree of detail

Analyze process Use correct symbols:

Determine causes Action/Activity

Decision

Reference letter

Start / Stop

Review flowchart

67
 STATISTICAL METHOD
X, R CHART

PURPOSE: STEPS/RULES:

To understand were Use measurable

The process is running (variable) data

To control the process Decide on time period

For data collection
To know when to adjust
The process and when Decide how many
To leave it alone measurements in subgroup

Analyze how the process Collect data-note changes

behaves in the process

Find X and R for each

Subgroup

Calculation worksheet

Plot data-Range chart first

Add control limits

Use with data collected in

Subgroups

Interpret chart

68
 STATISTICAL METHOD
MOVING RANGE CHART

PURPOSE: STEPS/RULES:

Same as X, R Chart: Use measurable data

To understand were Decide on time period for

The process is running Data collection

To control the process Range is difference

Between 1st & 2nd
To know when to adjust measurement,
The process and when 2nd & 3rd, etc.
To leave it alone
Calculation worksheet
Analyze how the process behaves
Plot data range first

Add control limits

Use individual
Measurements
No subgroup

Interpret chart

69
 STATISTICAL METHOD
p CHART

PURPOSE: STEPS/RULES:

Shows the % defective Use with measurable

(attribute) data-go/no,
Analyze how the good/bad, visual judgement
Process behaves
Determine subgroup size -
Must be at least 25

Decide on time periods

Collect data-
Note changes in process

Determine fraction
Defective

Determine % defective

70
 STATISTICAL METHOD
p CHART

PURPOSE: STEPS/RULES:

Calculation worksheet

Draw scale on Chart

Plot data – % defective

Add control limits

Interpret Chart

71
ANALYZING THE PROCESS

73
 TYPES OF DATA

DATA

YES/NO
GOOD/BAD MEASURABLE
PASS/FAIL

ATTRIBUTE VARIABLE

74
 POPULATION & SAMPLING

Population:

• All items produced of a particular products

Random Sampling:

• A technique used to gain information about population

characteristics.

Sampling Guidelines:

• Select frequency of sampling

• Choose subgroup periods to enhance feedback

• Subgroup should be large enough to have a good chance of

having some non-conformities.

• Select samples in a random manner (use random numbers

table), or collect samples in such a way as to ensure that
every piece of the population has an equal chance of being
selected.

75
Characteristics:

• When producing a product, you must act on the basis of

data.

• Random sampling will result in obtaining representative

results from the population (Central Limit Theorem).

• The Central Limit Theorem explains why random samples

can be used to estimate the characteristics of an entire
population.

• If the population is normally distributed, samples from the

population is normally distributed, regardless of the sample
size.

• If the population is not normally distributed the average (x)

of the sample groups selected from the population will be
normally distributed for sample sizes of thirty or more.

• Critical information which is required to understand the

process can be obtained when a sample size of thirty or
more is drawn from a population. Sample average and
sample standard deviations under the above conditions
approximate population averages and standard deviations.

76
 COLLECTING DATA

Purpose:

Collecting data from the actual process is the first step in

identifying problems in that process. Sample data is used
because of the large number of items being produced.

It would be quite a task, and not very cost effective, to check all
items in the total population being produced. Thus, the use of
sample data.

The key is for each sample selected from the total population to
have an equal chance of being selected. This is what is referred
to as random sampling.

Through the collection of samples you will get the raw data you
need to identify the problems in the process.

You must as four questions before collecting sample data:

• Why do you want to collect the data?

• One purpose for collecting sample data would be

identify what problems you have in the process.

• If you know there are defects occurring in a

process, you want to know what the defects are
and how often they occur.

77
 • In order to find out what is really happening in
the process, you will collect the samples on a
random basis. This will provide a good
representation of what is actually happening in
the process.

• Through this method, you will get an accurate

understanding of the characteristics of the
products you are producing in your process over
a given period of time.

• What type of data will you collect?

In our process, there are a number of variables which

impact the product quality. Considering the variables
which affect your product quality, you must decide what
type of data needs to be collected so that you can identify
where the problem(s) occur(s). These variables include
the quality of the parts or material used in your process,
the machinery involved, and the persons operating the
machines and/or running the process.

Another possibility is the collection of data on the types

of defects which are the results of the total process.

• What are the characteristics of the data you will collect?

When you have established the purpose of your data, and

identified the type of data you will collect, you must
determine the characteristics of the data to be collected.

78
 There are two basic types of data:

• Variable – This is measurable data, such as

weight, size, height, diameter, length, pressure,
etc. You will collect this data from the actual
items produced.

• Attribute – countable data is sometimes called

“go/no go“ data. It is the data we need to collect.
You will be counting the number of defects (or
the number of defectives) in a given sample.

In other words, you have to ask if you can

measure it or if you will just count it.

• Who will collect the data and when will it be collected?

You will answer these questions as you determine the

type of data to be collected and how the process
operates. You can collect samples periodically, and they
may be collected by the person involved in the process.
Records are important. You must be careful to record
the person who collected the data, the date, methods or
instruments used, and the time.

79
 HISTOGRAM

There is a saying that “one picture is worth a thousand words”. A histogram is a graphntation
that can be helpful both in understanding the sample data and in preparing presentations to
communicate to others what the data represents.

CONSTRUCTING A HISTOGRAM

A) Count the data points. Let “N” equal the total.

B) Find the range (R) of the data, (largest sample value minus the smallest sample value)

C) Determine number of classes or cells (K) you will use to construct your histogram using the
following table.

Number of Data (N) Number of Classes (K)

Under 50 5–7
50 – 100 6 – 10
100 – 250 7 – 12
over 250 10 – 20

D) Determine class or cell interval (h) by dividing the range ® by the number of classes or cells
(K). h = R/K

Round “h” off if necessary to make lass division easier.

E) On graph paper draw horizontal and vertical axis.

F) The horizontal axis boundaries will start at the smallest sample value and mark off the class
interval until the largest sample value is surpassed.

G) Place an “X” for each sample value up the vertical axis that corresponds with the class
interval.

H) continue “G” until all sample values have been placed in the appropriate area.

80
 RANDOM SAMPLE OF THE
WEIGHTS OF THE ADULT MALE
228 138 191 182 175 162 146 118 195 241
179 219 182 138 245 212 195 174 193 173
106 178 136 181 180 163 184 251 98 238
235 202 152 201 221 210 182 193 155 191
131 162 146 285 174 196 156 224 172 214
204 222 177 151 191 115 246 164 209 165
186 181 141 194 130 235 168 207 157 200
154 192 153 215 175 120 208 171 232 158

258 161 176 131 202 150 266 167 166 181
160 126 241 121 217 170 149 181 182 228

HISTOGRAM
WEIGHTS OF THE ADULT MALES

X X
X X X X
X X X X X
X X X X X X
X X X X X X
X X X X X X X X
X X X X X X X X X X
X X X X X X X X X X X X
X X X X X X X X X X X X X
X X X X X X X X X X X X X X X
X X X X X X X X X X X X X X X X X X X
300-309

290-299

280-289

270-279

260-269

250-259

240-249

230-239

220-229

210-219

200-209

190-199

180-189

170-179

160-169

150-159

140-149

130-139

120-129

110-119

100-109

90-99

81
 COPPER CUT -LENGTH
SPEC. LIMITS 3,400 ± .010

1 3.402 11 3.400 21 3.403 31 3.398

2 3.401 12 3.401 22 3.400 32 3.403
3 3.399 13 3.402 23 3.399 33 3.397
4 3.400 14 3.400 24 3.400 34 3.396
5 3.402 15 3.401 25 3.398 35 3.396
6 3.398 16 3.399 26 3.399 36 3.395
7 3.397 17 3.399 27 3.403 37 3.404
8 3.399 18 3.401 28 3.397 38 3.405
9 3.402 19 3.400 29 3.400 39 3.402
10 3.398 20 3.401 30 3.401 40 3.404

X
X X X
X X X X
X X X X X
X X X X X X X
X X X X X X X X X
X X X X X X X X X X X
3.394

3.395

3.396

3.397

3.398

3.399

3.400

3.401

3.402

3.403

3.404

3.405

3.406

82
83
1 5

- NORMAL CURVE - PROCESS

CENTERED BUT
NOT CAPABLE

2 6

- PROCESS - OVALITY
CAPABLE BUT - 2 SET – UPS OR LOTS
NOT CENTERED - SHIFT CHANGE

3 7

- PROCESS - OVALITY
CENTERED BUT - 2 SET – UPS PR
USING 100% OF LOTS
TOLERANCE - 2 MACHINES
- OVER ADJUST –
MENT MADE

4 8

- PROCESS NOT - SET – UP PEICES

CENTERED USING INCLUDED
100% OF
TOLERANCE

84
 LSL USL LSL USL
9 12

- GAGING NOT - 100% SORTING

PRECISE ENOUGH OF LOT
- CELL SIZE TOO
LARGE
- EXCELLENT
PROCESS

10 13

- FUDGED DATA, - SAMPLE SIZE

ALL READINGS NOT LARGE
PUT WITHIN SPEC ENOUGH

11 14

- PROCESS IS - 2 CUSTOMERS,
CENTERED, BUT PARTS CLOSE TO
NOT CAPABLE NOMINAL SOLD TO
DIFFERENT CUSTOMER

85
 HISTOGRAM
COPPER CUT -LENGTH
SPEC. LIMITS 3,400 ± .010
1 3.402 11 3.400 21 3.403 31 3.398
2 3.401 12 3.401 22 3.400 32 3.403
3 3.399 13 3.402 23 3.399 33 3.397
4 3.400 14 3.400 24 3.400 34 3.396
5 3.402 15 3.401 25 3.398 35 3.396
6 3.398 16 3.399 26 3.399 36 3.395
7 3.397 17 3.399 27 3.403 37 3.404
8 3.399 18 3.401 28 3.397 38 3.405
9 3.402 19 3.400 29 3.400 39 3.402
10 3.398 20 3.401 30 3.401 40 3.404

AVERAGE (X) = 3.400

STANDARD DEVIATION = .002

3.400
3.399 3.400 3.401
3.399 3.400 3.401 3.402
3.398 3.399 3.400 3.401 3.402
3.397 3.398 3.399 3.400 3.401 3.402 3.403
3.396 3.397 3.398 3.399 3.400 3.401 3.402 3.403 3.404
3.395 3.396 3.397 3.398 3.399 3.400 3.401 3.402 3.403 3.404 3.405

86
 THE NORMAL CURVE

σ= Sigma = Standard Deviation Normal Curve is divided

__ into 6 equal
X = Average segments.

68.26%

95.46%

99.73%

EXAMPLE:
__
X = .500
σ= .011

.497 .498 .499 .500 .501 .502 .503

87
 PROCESS
CAPABILITY

88
INTRODUCTION

When a process is in statistical control and all special causes of variation have been
eliminated, the process capability can be studied. A capable process is one which:

1. Is in statistical control
2. The individual measurements are normally distributed.
3. The process variation consumes no more than 75% of the specification
tolerance. This is a comfortable relationship between process variation and
specification limits.

The Standard Deviation

The standard deviation (σ ) is a measure of variablity. Just as weight may be

measured in pounds, and length in inches, the variation of process is measured in
terms of standard deviation units. The standard deviation takes on the same units
(pounds, inches, etc.) as the measurements. The Greek letter σ (sigma) is used to
indicate one standard deviation. Approximately 99.7% of the area under the
normal curve lies between X plus and minus three standard deviations (Figure 66)

An example of a capable process is shown in Figure 67.

The standard deviation from control charts can easily be estimated by dividing the
average range ( r) by the factor d 2. Refer to Appendix F for factors. The d 2 factor
depends on the sample size. When using a sigma chart, the estimated standard
deviation can be found by dividing the average standard deviation (s) by the factor
c 4.
Example: R = .003 n = 5 σ= R = .0023 = .001
d2 d2
Example: s = .001 n = 5 σ= s = .001 = .0011
C4 .940

89
 99.73%

95 46%

68 26%

-3σ -2σ -1σ +1σ +2σ +3σ

Figure 66. Areas Under the Normal Curve at ±1, ±2, and ±3 Standard Deviations from the Mean.

CONTROL CHART METHOD – CAPABILITY RATIOSS (CR)

AND CAPABILITY INDEXES (Cp/Cpk/CpU)

Steps in Studying Process Capability

1. Make sure the process is in statistical control.

2. Individual measurements, not averages, should be normally distributed
(Figure 68). This is often tested using a frequency distribution or a
histogram for the individual measurements. Another test for the normal
distribution or a histogram is kurtosis (see glossary, Appendix H).

Note: In practice, do not expect the distribution to be perfectly normal. the distribution
should closely approximate a normal shape.

90
 σ = .001

ABOUT
1σ ABOUT
CLEARANCE 1σ
CLEARANCE
LTL -3σ
.496 .4968 +3σ UTL
X .5028 .504
.4998
B/P NOMINAL
.500

Figure 67. An example of a Capable Process (C/P = Blueprint)

X INDIVIDAUL
XXX MEASUREMENTS
XXXXX (NOT AVERAGES)
XXXXXXX
XXXXXXXXX

Figure 68. Frequency distributed of Individuals – Not Averages

91
 3. Calculate the estimates standard deviation (R divided by d2 ).

X
X X X
X X X X X
X X X X X X X
X X X X X X X X X

LTL -3σ = +3σ UTL

.496 .4968 x .5028 .504
.4998

NOMINAL
B/P
.500

Figure 69. Specification and Process information Placed on the Distribution

4. Draw lines on the frequency distribution which reflects the grand average
(X) and each three sigma limit (3σ). Also draw lines on the distribution
which reflect the nominal and both specification limits (Figure 69).

5. If the process is centered as shown in Figure 69, a capability ratio (CR) or

capability index (Cp) can be used. The CR is calculated by dividing 6σ by
the total specification tolerance. The Cp is the reciprocal of the CR,
calculated by dividing the total specification tolerance by 6σ.

92
Capability Ratio

CR = 6σ = .006 = 0.75
Total specification tolerance .008

Capability Index

Cp = Total specification tolerance = .008 = 1.33

6σ .006

Note: The Cp should not be less than 1.33, which means that the process
spread only consumes 75% of the tolerance.

Note: Capability ratios (or capability indexes) should only be used when
processes are centered on specification limits.

Cpk Index

When the process is not centered on specification limits, a Cpk index can be used.
The Cpk index provides a worst case capability index which compensates for
processes which are not centered on specification limits. The Cpk index is the
difference between the grand average (X) and the nearest specification limit
divided by 3σ.

Cpk = USL - X or* X - LSL

3σ 3σ

*Use whichever specification limit is closer to X.

USL – upper specification limit

LSL – lower specification limit
X – the grand average

93
Example: In Figure 70, the upper specification limit (USL) is closer to X so the
Cpk index is:

Cpk = USL - X
3σ

= .504 - .502
3(.001)

= .002
.003

Cpk = 0.67

Note: The Cpk index assumes that the process average is not centered, but it
does fall on or between specification limits.

Z2
Z1

X X Defective Tail
X X X
X X X X X
X X X X X X

X
L = σ= .001’”
.496 X Xu
.502 .504

Figure 70. An Off-Centered Yet Capable Process

For a process to be called capable, the Cpk index should be 1.33 minimum. This
index relates to the goal of a process spread which only consumes 75% of the
specification tolerance.

94
Predicting the Percent Yield of a Process

When the process is not centered on specification limits, the yields (or percent of
good parts) can be predicted using the Table of Areas under the Normal Curve
(Appendix E). This table involves the use of a simple equation to calculate a Z
score, then the Z score is used to find the area.

Note: The areas found in the normal curve (Appendix E) are decimal
fractions. You must convert these areas to a percentage. Once the Z
score is found, it is used to find the area under the curve.

Example: A Z score is calculated to be 2.6. The area related to a Z score of

2.6 is .4953 (see table in Appendix E). The decimal .4953
represents 49.53% of the product is expected to fall between the
specification limit and the grand average.

Referring to Figure 70, there are two Z scores to calculate. The Z score is the
difference between a specification limit (X) and the grand average (X) divided by
one standard deviation (σ).

Z upper = USL - X Z lower = X - LSL

σ σ

= .504 - .502 = .502 - .496

.001 .001

= .002 = .006
.001 .001

Z Upper = 2 Z lower = 6

A value of Z = 2 in the A value of Z = 6 cannot be found in the

Table represents .4772 or table. For practical purposes, when the
47.72% calculates value (e.g., Z = 6.0) is not
found, use the highest Z value in the table.
Therefore, use Z = 4.0 or 49.99%.

95
These two areas (47.72% and 49.99%) represent the expected percentage of good
parts (within specification limits); therefore they can be added.

47.73%
+ 49.99%
97.71%

The percent yield (97.71%) for this process represents the probability of good parts
that can be expected if nothing about the process or specification changes. The
probability of nonconforming parts is equal to 100% minus 97.71% or 2.29% in
this case.

Also because the process distribution is over the upper specification limit, the
2.29% nonconforming parts are expected to be over the upper specification limit.
Prediction and tracking the yield using “Z” scores can often provide a more
realistic approach to capability assessment than the capability indexes (CR, Cp,
Cpk, CpU, and others not covered in this book).

The reason Processes Must Have Clearance

All normal distribution have a six-sigma spread which represents 99.73% of the
distribution. From the center of the distribution (X) there are three equal standard
deviations to the right and three equal standard deviations to the left (Figure 71).

If you have normal distribution, and you know the value of one standard deviation,
then you can accurately predict what the process will do from a given center (X),
which is usually the setup value.

Example: σ = .001 X = .500

The process is in control and the distribution of individuals is normal (figure 72).
If the process is set up on .500, 99.7% of the product will be produced between
.497 and .503 when the standard deviation is .001.

96
 99.73%

95.46%

68.26%

-3 σ -2 σ -1 σ =1 σ +2 σ +3 σ
__
X

SIX SIGMA SPREAD

Figure 71. The Normal Distribution and Six-Sigma Spread

__
-3 σ X = .500 +3 σ
.497 .503

Figure 72. A Process In Control, Normally Distributed, and Centered on

Specification Nominal (.500)

97
 EXTRA 1σ EXTRA 1 σ

-4 σ -3 σ = +3 σ +4 σ
LTL X UTL

NOMINAL
B/P

Figure 74. Extra Clearance at the Tails of the Distribution

SINGLE SPECIFICATION LIMIT

There are standards which only reflect a single limit of tolerance such as:

Geometric Tolerances: Runout, concentricity, parallelism, flatness,

straightness, ect.

Maximum/Minimum Limits: .500 maximum .375 minimum

Surface Finish: AA32

In these cases, process capability concerns itself only with the single limit.
Therefore, a Z score computed between the specification limit and the average of
the process in order to find the percent of product which is expected to be between
the average and the single limit, The other area will represent 50% good product
in most cases (Figure 75).

98
Example: Specification: Runout within .10 total indicator reading TIR).

Grand Average (X) = .008 TIR

σ = .0006

The upper Z score is 3.33 and the area under the normal curve for a Z o f3.33 is
4.996 or 49.96%. Therefore, 49.96% of the product will be between .008 and .010
TIR and the other 50% of the product will b less than .008. Thus, the yield is 50%
+ 49.96% = 99.96% good parts.

99
CONTROL CHARTS

100
 SPC AND THE CONTROL CHART

DATA

FREQUENCY MEASUREABLE
COUNTS

ATTRIBUTE VARIABLE

NUMBER OF
DEFECTS PER
SAMPLE OR
UNIT
NUMBER OF
DEFECTIVE
UNITS
(PASS/FAIL)

C CHART U CHART P CHART NP CHART X/R CHART

NUMBER OF NUMBER OF FIND MEAN
DEFECTS PER DEFECTS PER FRACTION NUMBER AND
SAMPLE UNIT DEFECTIVE DEFECTIVE RANGE

FIXED FIXED FIXED FIXED FIXED

SUBGROUP SUBGROUP SUBGROUP SUBGROUP SUBGROUP
SIZE SIZE SIZE SIZE SIZE

101
 Table for Selection of Subgroup Sizes and Intervals

Appendix

A
No absolute rule can be used for how often we should sample. The realities of the factory layout,
production run size, and the cost of sampling must be balanced with the value of the data
obtained. In general, it is best to sample at close intervals at the beginning and increase the
interval when the process results permit. The following tables (Ref: Mil-Std-414 Variables
Sampling Plans) can be used for estimating the amount and frequency of sampling.

Total No. of Parts

Production Rate per Shift to Be Sampled per Shift
1 to 65 5
66 to 110 10
111 to 180 15
181 to 300 25
301 to 500 30
501 to 800 35
801 to 1300 40
1301 to 3200 50
3201 to 8000 60
8001 to 22000 85

Example: A process produces 3,000 parts per shift. Using the table, we should consider
Sampling 50 parts during the shift. If the control chart uses subgroups of five each,
then 50/5 = 10, and therefore 10 subgroups should be taken during the shift.
Therefore, on an eight-hour shift, samples would be taken every 78 minutes or so.

102
 BENEFITS OF CONTROL CHARTS

1. Control charts are simple and effective – The charts are maintained easily at the job
station by the operator.

2. Control charts give the people closest to the operation reliable information on when
action should be taken and when action should not be taken.

3. A process “in statistical control” is predictable. Producer and customer can rely on
consistent quality and stable costs.

4. Control charts allow for improvements to reduce common cause variation. The effects of
even small changes on the system can be identified quickly and effectively thereby:

A. Increasing the percentage of output that meets customer expectation.

B. Decreasing scrap and rework (Improve cost/unit)

C. Increasing the total yield of acceptable output (Increase productivity)

5. Control charts provide a common language at all levels for communication about the
performance of a process:

A. Shift to Shift

B. Production departments (operators, foreman) to support departments

(maintenance, engineering)

C. Supplier and customer

103
 Preliminary Consideration

1. Establish an environment suitable for action. Communication channels must be open,

fear must be eliminated, and people must be trained.

2. Define the process. The process must be understood relative to other operations/
processes both upstream and downstream. (Secondary and third operations, assembly
equipment).

3. Determine the characteristics to be monitored. Efforts should focus on the changes that
promise the greatest chances for improvement.

A. Repeat problems (characteristics) as shown on inspection reports, scrap, rework.

B. Characteristics critical to fit and function of the parts.

C. Correlation between characteristics. (One tool controls several dimensions,

monitor only one characteristic)

D. Customer needs

E. Review with production employees. They are most familiar with their process.

4. Define the measurement system. Specify what information is to be gathered where,

under what conditions? Accuracy of data requires accuracy in measurement.

5. Minimize unnecessary variation. Avoid obvious problems that can be eliminated without
control charts.

104
 Updating Control Chart Limits

While the revision of control chart limits applies to the initial computation of limits for a given
set of data, the updating of control chart limits pertains to the modification of existing limits on
an ongoing chart.

An obvious time for updating the control limits is following a change in the process. It is a poor
use of a control chart to make it prove what is already known. Instead, one should use the chart
to discover that which is not known. So when a deliberate change is made, it is logical to collect
data following the change and these new data may be used to compute new control limits. While
the new data are being collected following the change and prior to the computation of the new
control limits, the new data may be plotted against the old limits as a means of verifying the
change in process. If a change is indicated, then the new limits are needed. If the new data “fit
in with” the old limits, there is little need to update them.

Another time to consider updating the control limits is when the limits are “trial control limits”.
Trial control limits computed from a limited amount of data. Of course, when there is only a
finite amount of data available this issue is moot. One may always compute control limits using
the data available, and any signals found using trial control limits are likely to be real signals.
Thus, one does not have to wait until a large amount of data has been obtained before computing
control limits. But when the limits for an ongoing control chart are computed from the first few
subgroups, and then additional data is collected, it is generally recommended that the limits be
updated when 20 to 30 subgroups are available. This recommendation of 20 to 30 subgroups is
essentially a piece of insurance. By using 20 to 30 Subgroup Ranges in the computation of the
Average Range the effect of any extreme value will be diluted and minimized. Of course, a
single Subgroup Range that is, say, 10 times bigger than the Average Range will still dominate
the computation and inflate the Average Range, but Subgroup Ranges that are just outside the
control limits will have a fairly small impact when averaged in with 20 to 30 other values.

When updating trial control limits one may also elect to use either the delete or revise approach
or the Median Range approach outlines in the preceding section. The objective is to use the data
available to obtain reasonable and useful limits in as straightforward a manner as possible.

Finally, once the control limits are computed using about 100 observations there will generally
be little change in the limits with further updates. Unless the process is changed, or changes in
some fundamental way, the limits should not need further updates. The practice of automatically
recomputing control limits every time a control chart form is filled up, and then using these latest
limits on the next sheet, may result in the failure to detect low trends in the process. Likewise,
the automatic updating of control chart limits in some computer programs will, by default, use all
of the data available, even though some of those data may no longer be appropriate for the
current process.

The control chart is a versatile tool for use in real-time situations. Therefore, one should always
actively control the manner in which the control chart limits are computed in order to be sure that
the limits are appropriate for the current process. The computation of the limits cannot be
divorced from the context for chart.

105
 VARIABLES CONTROL CHART

The control chart is a line graph that is kept at the work area. It is used to record
the results of each sample. It also shows the control limits, that if exceeded, should
trigger investigation.

CONTROL CHARTS HAVE TWO BASIC USES:

1. As a judgment to give evidence whether a process has been operating in a

state of statistical control and to signal the presence of special cause of
variation so that corrective action can be taken.

2. As an operation, to maintain and monitor the state of statistical control,

by extending the control limits for real-time decisions.

106
 VARABLES CONTROL CHART
(X & R)

DEFINITIONS:
_
X (called X-bar) – sum of the variables divided by the number of variables
(average)
=
X (called X-double bar) – average of the X-bar’s

Range (R) – the difference between the largest and smallest variable.
_
R (called R-bar) – the average of the ranges

Control Limits – these are the limits within which we expect the process
sample averages and ranges to vary within.

UCL_ - Upper Control Limit of the averages.

X

LCL_ - Lower Control Limit of the averages

X

UCL - Upper Control Limit of the ranges

R

USL – Upper specification limit

LSL – Lower specification limit

107
 VARIABLES CONTROL CHART

1. GATHER DATA

A. Subgroup Size – usually fiver pieces but can vary depending on the process.

B. Subgroup Frequency – the subgrouping frequency should make sense in terms of

production to aid in analysis and correction of problems found. (ex. Time
sequence, every hour or production sequence, every 100 pieces)

C. Data collection period should be long enough to capture all likely sources of
variation affecting the process. Control limits should be calculated after 15 to 20
subgroups have been gathered.

2. CALCULATIONS

A. From each subgroup

• SUM
• AVERAGE
• RANGE

B. Control Limits
= _
• X and R
_ = __
• UCLx = X + A2 R
_ = _
• LCLx = X - A2 R
_
• UCLr = D4R

NOTE: Refer to appendix to determine value of constant factors A2 and D4

depending on sample size.

108
 AVERAGE OR MEAN

Definition:

The mean represents the average value for all values contained in a sample group. It is one of
the statistical methods used to describe the measure of location for sample data.
_
X = ΣX
n

Where:
_
X = Average Value

Σ = Sum of (or “add together”)

X = Individual Value

n = The number of individual samples in one subgroup

__________________________________________________________________

RANGE

Definition:

Range is determined by subtracting the smallest value from the largest value in a group of
sample data.

It is used to describe the amount of variation in a group of sample data.

LARGEST VALUE

Minus SMALLEST VALUE

= Range

The Range of your subgroup will provide more information than the Average alone.

109
 DETERMINE
_
X AND R

115 87 100
111 86 92
113 90 98
118 82 112
120 83 97

SUM
_
X

= _
Determine X and R for the above X-Bar and R Values.

= _
X = ΣX =
k

_
R = ΣR =
k

Where k is the total number of subgroups.

110
 _
X AND R CHART

Lesson Objectives…

To help you make the most effective use of this manual and your time in
learning the concepts described, we have listed below your study
objectives for this lesson. These objectives tell where you are going in
this lesson and what you can expect to learn.

Please read the objectives carefully and keep them in mind as you read
and study the lesson. You should be able to accomplish each of these
objectives upon completion of this lesson.
__
• Construct an X, R Chart

• Determine the conditions for selecting a Variable Control

Chart.

• Identify an Out-Of-Control condition using Variable Control

Charts.

• Identify natural variation.

__
• Define how X, R Control Charts can be used to analyze and
control a process.

111
Guidelines:

• Initial Decisions

• Decide what to monitor.

• How many readings in each sample group (3,4,5,etc.)
• Time intervals (15 min., 30min., 60 min., per skid, per bin,
per case)?

• Collect data and make notes.

__
• Find average (X) and range (R) for each sample group.

• Complete a calculation worksheet.

• Plot data – range Chart first (See “Making a Scale”).

• Range Chart must be in control before completing Average

Chart.

• Is the chart process “in-Control”?

• Assignable Cause Variation

• Hugging
• Trends
• Jumps
• Runs
• Cycles

If the process is not stable, identify assignable causes of variation, and

eliminate them to improve the process.

112
Symbols And Definitions:

Variables - Measures values of a characteristic.

Common Cause - Faults of a system, common to a process.

Assignable (Special)
Cause - Specific to a machine, person, material or
method. The operator, supervisor or
technical support person can usually identify
and correct.

Control Chart - A statistical quality control tool used to study

and control processes.

In-Control - A process in which only common causes of

variation are operating; a system which is
predictable.

Out-Of-Control - Both common and special causes of variation

exist; an unstable system.

Natural Tolerance - Natural variability of the process.

Engineering
Tolerance - Specification limits. (Lines [upper and lower
limits] which define how much variation is
acceptable to the customer, internal and
external.)

Control Limits - Lines (upper and lower limits) which define

the amount of variation to be expected if the
process is in control.

Subgroup - Several consecutive samples.

113
Subgroup Size - The number of consecutive samples inspected
at one time.

UCL - Upper Control Limit

CL - Centerline

LCL - Lower Control Limit

X - Denotes an individual sample value.

__
X - The average of two or more measurements.
= __
X - The average of a set of X values, also called
the grand average(The average of the
averages).

R - The difference between the largest and the

smallest values in any set of data.

n - The number of sample measurements in any

given subgroup.
__
R - The average of the ranges.

k - The number of subgroups.

Σ - The sum of a series of numbers.

σ (Sigma) - A measure of the variation within a set of

data.

114
Hugging - A pattern which occurs when the points stay
close to the centerline.

Or

- A pattern which occurs when the points stay

close to the upper control limit, or the lower
control limit.

Jump - Occurs when a tremendous shift occurs

between consecutive points plotted on a
Control Chart.

Trend - Formed when a series of seven or more

consecutive points continue to rise and fall in
one direction.

Cycle - A series of points that display a similar or

repeated pattern of time.

Run - A series of seven or more points above or

below the center line.

115
Making A Scale:

• Always make the Range Chart first (bottom Chart).

• Range Chart starts from zero. Go ahead and put the zero on the
bottom line.

• Scan the range values and pick out the largest number. Compare
this number (largest) to the UCLR. Of these two values, the large
will indicate how high the scale must go. It may be possible to
skip line(s) between each number. Spread out the data. The Chart
is much easier to read this way.

• After the Range Chart is plotted, interpret the results. If the Range
Chart is “In-Control”, you may proceed to the Average Chart.

• The Average Chart does not start from zero. Scan the averages
(not the individual readings) and select the highest and lowest X.
Compare to the UCLX and LCLX as previously done with the
Range Chart.

116
CHOOSING A SCALE FOR THE CONTROL CHART GRAPHS

The choice of scales for the average chart and the range chart is partially a matter of
individual preference. The following guidelines are offered to assist those who feel the
need of some help. They are based on approximations that will yield the right scale,
provided the range of the first subgroup is not extremely large or extremely small.

GUIDELINE 1: Pick a scale that is easy to use. If the forms have every fifth line
emphasized, then the distance between these heavy lines should be made to represent
either one unit, or five units, or ten units. With these choices the small lines will
represent, respectively, 0.2 units, 1 unit or 2 units. (Another possible choice would be to
let each heavy line represent 2 units, so that each small division would correspond to 0.4
units.)

GUIDELINE 2: The spread of values needed for the average chart will generally be
between 3 A2R and 4 A2R, where R is the range of the first subgroup. In the example
above, the first subgroup range was 5 thousandths and A2 = 0.729. Thus 3 A2R = 10.95
thousandths, and 4 A2R = 14.6 thousandths. By letting the distance between the heavy
lines represent one thousandth of an inch, a scale covering nine thousandths could be
placed on the chart shown. Since nine thousandths is close to a smaller value of 10.95,
this was the scale chosen. (An alternative scale would be to let the heavy lines represent
two units each, whereupon the scale would cover eighteen thousandths.)

GUIDELINE 3: The values needed for the average chart should be centered near
the average of the first subgroup.

GUIDELINE 4: The values needed for the range chart will generally begin with
zero and continue to a value that is between 3 and 4 times as large as the range of the first
subgroup.

GUIDELINE 5: Since the ranges will always be expressed as integer multiples of

the smallest unit of measurement, it is convenient to let each small division represent the
smallest unit of measurement. If the measurements are to the nearest tenth of a pound, let
each division represent one-tenth of a pound. If the measurements are to the nearest five
degrees, then let each division represent five degrees. In the example given above, the
smallest unit of measurement was one thousandth of an inch, so each division was set
equal to one-one thousandth of an inch.

These guidelines are based on the assumption that the average and range for the first
subgroup will be fairly typical of what will come. Instead of making this assumption,
you could wait until two or three subgroups had been obtained before setting up a scale
and plotting any points. This was you could use the average range and the average of the
averages to set up the scale according to the guidelines above.

117
 __
How To Construct An X and R Chart:

1. Data must be variable or measurable.

2. Data must be in subgroups.

3. Subgroup size usually consist of three, four or five items.

4. A minimum of 20 subgroups should be used before Control Limits

are established.
__ = __
5. Determine X, R, X, and R values

6. Calculate upper and lower Control Limits.

7. Construct Range Chart and determine if plotted data is with upper

and lower control Limits. Make certain there are no patterns in the
plotted points.
__
8. Construct X Chart and determine if plotted data is with upper and
lower Control Limits.

9. If chart is in control, variation is random.

10. Determine process capability using CpK and Cr formula.

11. If process is stable, construct a Histogram to determine the shape or

distribution of data.

118
Review:

__
In this lesson I have learned to construct an X, R Chart. This
Control Chart uses measurable or ____________________data.
__
X, R Charts analyze where a process is operating. After the
points are plotted (represented measurements)______________
are added to show whether or not the amount of variation in
your process is excessive.

One characteristic of the X, R Chart is that the typical sample

size is either________, _________, or __________. Also, the
minimum number of subgroups is
_________________________________.

___________ Cause variation is within the natural variation of

the process. To change this involves either solving on the spot
or in your team with the guidance of ____________________.

119
 CALCULATION WORKSHEET

CONTROL LIMITS
n=
k=
_ __
R = ΣX
k
= __
X = ΣX
k
__
UCLR = D4R =
__
A2R =
= __
UCLx = X + A2R =
= __
LCLx = X - A2R =
TABLE OF CONSTANTS
n A2 D4
2 1.880 3.268
3 1.023 2.574
4 0.729 2.282
5 0.577 2.114
n = Subgroup Size Part Number _______________
k = Number of Subgroups Job Number________________
Operator___________________
Date of Collection____________

120
 __
GENERAL RULES FOR X & R CHARTS

1. All paces at the top of the Form 1778 Process Control Chart will be completed by the operator
with the correct information each time a chart is made. The start and stop dates will be recorded
in the date space when the chart is started and completed.

2. Zero set procedures are encouraged to simplify calculations and reduce errors.

3. All measurements must be made with the same type calibrated measuring instrument and to
the maximum sensitivity of that instrument. Recorded actual readings – don’t round off!

4. The person doing the sample must initial the entry at the bottom of the charts.

5. A trial scale will be put on the chart after subgroups and points will be plotted.

6. Trial control limits will be calculated and penciled in on the chart after five subgroups. Final
control limits can be calculated and placed on the chart after twenty subgroups if there have been
no major process changes.

7. Samples must be taken in the time frame required or a reason must be written on the chart
explaining why it could not be done.

8. If a subgroup falls out of the control limits or shows a drastic shift in either range or average,
the math and simple measurements should be checked and if necessary the samples retained for
immediate verification by an inspector, facilitator, or another team member. If the original result
proves true, a second subgroup should be taken and recorded immediately. If it confirms a
process change, an assignable cause should be investigated and corrective action taken before
production resumes.

9. If an out of tolerance measurement is found, it should be verified immediately by an

inspector, facilitator, or team member and the manufacturing process stopped or adjusted. All
suspect parts must be segregated to eliminate the out of tolerance parts. Production may begin
when this is done and the process has been corrected.

10. If a change is made in the process, it must be noted on the back of the chart and signed by
the person making the change. Also another subgroup should be taken and recorded to prove the
change corrected the problem or improved the process.

William R. DeSilvey
10/28/93
Rev. 01 06/21/93

121
 MOVING RANGE CHARTS

Lesson Objectives:

To help you make the most effective use of this manual and your time in
learning the concepts described, we have listed below your study
objectives for this lesson. These objectives tell where you are going in
this lesson and what you can expect to learn.

Please read the objectives carefully and keep them in mind as you read
and study the lesson. You should be able to accomplish each of these
objectives upon completion of this lesson.

• Construct a Moving Range Chart.

• Determine the conditions for selecting a Moving Range Chart

• Identify an Out-Of-Control condition using Moving range

Charts

• Define how Moving Range Charts can be used to analyze and

control a process.

• Calculate Moving Range values as the difference between

individual measurements. Record the difference between the
first and second individual values, then between the second
and third, and continue this method until finished (two item
moving Range).
_
• R is the average of all the moving ranges. (Range Charts).
__
• X is the average of the individual measurements. (Individual
Reading Chart)

122
• Calculate Control Limits on the Moving Range Chart and
then on the Individual Readings Chart.

• Select Chart scales for the Moving Range and individual

measurement values.

• After the Range and Individual Reading Charts are in control,

the process will be predictable. Successive Range Chart
values are related since they have a point in common.
Therefore, you must take care in interpreting trends or Out-
Of-Control conditions on a Moving Range Chart.

• If the process is stable, identify the Common Causes present

to reduce variability and improve the process.

123
Definitions:

RUN - A point, or series of points, above or below a

line.

UCL - Upper Control Limit.

LCL - Lower Control Limit.

CL - Centerline

CONTROL - Lines (Upper and Lower Limits) which define

LIMITS the amount of variation to be expected if the
process is In-Control.
__
X - The average of the individual measurements.

R - The difference between the largest and smallest

individual values in any set of data.

n - The number of individual measurements used

when calculating range values.

σ (SIGMA) - A measure of dispersion or variabilty in a set of

data.

124
 CALCULATION WORKSHEET

CONTROL LIMITS
n = The number of individual values used =
when determining a range value
__
R = The sum of the Range values =
The number of Range Values
__
X = The sum of the individual values =
The number of individual values
_
UCLR = D4R =

LCLR = D3R =
__
Ô = R =
d2
__
UCLx = X + 2Ô =
__
LCLx = X - 2Ô =

TABLE OF CONSANTS
n A2 D4 d2 d3
2 1.880 3.268 1.128 0
3 1.023 2.574 1.693 0
4 .0729 2.282 2.059 0
5 0.577 2.114 2.326 0

125
MOVING RANGE CHART

A chart for individual measurements. Principal kinds of data for which this chart should be used
are:
• Accounting figures (shipments, efficiencies, absences, accidents, maintenance
cost, etc.)

• Production data such as temperatures, pressures, gas consumption, etc..

Definitions and directions for making a moving range charts:

1) Start with a series of individual numbers. ≥20 preferred, but not <10.

2) take the difference (range) between the first and second numbers and record it;
then the difference between the second and third numbers, etc. continue to the last
number. (Note: Calculate the differences without regard to sign).

3) Take the averages of the original numbers in the series (X).

4) Take the average of the “Ranges” from step two (2). This is called the moving
range – MR.
Note: MR - (R1 . . . +Rn-1) / (n-1)
___ __ ___
5) Multiply MR by ± 2.66 to obtain control limits – control limits - X ± 2.66 MR.
__
6) Plot the original numbers, X and control limits.

7) Interpret chart – look for trends, whether fluctuations are narrow or wide, if
pattern stay away from one the control limits, and obvious peculiarities in the
pattern (cycles, bunching, etc.).

126
Example/ Problem:

Construct a moving range chart for the following data, and evaluate (discuss) your
results.

TONS OF PRODUCT PRODUCED BY A GROUP OF WORKERS

Tons Produced
JAN (LAST YEAR) 25.0 Change From Preceding Month

FEB 25.3 .3

MAR 33.8 8.5

APR 36.4 2.6

MAY 32.2 4.2

JUN 30.8 1.4

JUL 30.0 .8

AUG 32.3 6.4

SEPT 23.6 8.7

OCT 28.1 4.2

NOV 27.0 1.1

DEC 26.1 .9

JAN (THIS YEAR) 29.1 3.0

FEB 40.1 11.0

MAR 40.6 .5

460.4 53.6

127
 INDIVIDUAL AND MOVING RANGE CHARTS

Individual Charts apply when:

1. The cost of a sample is extremely high. This is often true with

destructive tests.

2. The items being made take a long period of time to produce or to test
or measure.

In these environments, it would be impossible or impractical to take a normal (5)

sample size. The individual chart can be utilized for such situations. However,
there are certain limitations to an individual chart.
__
1. They are not as sensitive as X and R charts to process changes.

2. Large number of samples should be taken (preferably 100 or more)

before any conclusions about process averages and standard deviation
can be made.

Data Collection

Individual values (X) represents sample size of one measurement.

Range – Variation is defined as difference between successive individual

values. (Moving Range – MR)

Control Limits
___
UCLMR = MR (D4)
__ ___
UCLX = X + E2MR
__ ___
LCLX = X - E2MR

128
129
 CHART INTERPRETATIONS (X AND MR)

INDIVIDUAL CHART –
__
The interpretation of individual charts is similar to X and R charts, however
some caution should be taken. As pointed out earlier, the samples should be
taken before any definite statements about the process can be made.

MOVING RANGE (MR) CHART –

Review the MR chart for points beyond the control limits as signs of the
existence of special causes. Note that successive moving ranges are
correlated, since they have a point in common, care must be taken when
interpreting trends because of this.

130
BASIC CHART INTERPRETATION

The Eight Basic Statements of Control

When control charts are used, there is an easy method to determine if the process is in statistical
control or not.

The following eight statements can be used to see if any control chart is in control. The
statements must be answered true or false while looking at the control chart in Figure 41. If all of
the statements are true the process is in control. If any statement is false, the process is not in
control. Several false answers on one process usually means that the process is in a worsened
condition.
______________________________________________________________________________
Statements Answer
______________________________________________ ________________________

1. There are no plotted points outside the control limits

(points plotted on a limit do not count). T F

2. The total number of points above the centerline is close

to the total number of points below the centerline. T F

3. The plotted points seem to be randomly falling over and

under the centerline. T F

4. There are no consecutive runs of 7 or more points on one

side of the centerline. T F

5. Only a few of all of the points are near the control limits. T F

6. There are no upward or downward trends of 6 or more points

heading directly toward either control limit. T F

7. The plotted points do not appear to be hugging the centerline

with little distance between them. T F

8. There are no straight line patterns. T F

131
132
 Notes on Tests for Special Causes
__
1. These tests are applicable to X charts and to individuals (X) charts. A normal distribution is
assumed. Tests 1, 2, 5, and 6 are to applied to the upper and lower halves of the chart separately.
Tests 3, 4, 7, and 8 are to be applied to the whole chart.

2. The upper control limit and the lower control limit are set at three sigma above the centerline
and three sigma below the centerline. For the purpose of applying the tests, the control chart is
equally divided into six zones, each zone being one sigma wide. The upper half of the chart is
referred to as A (outer third), B (middle third) and C (inner third). The lower half is taken as the
mirror image.

3. When a process is in a state of statistical control, the chance of (incorrectly) getting a signal
for the presence of a special cause is less than five in a thousand for each of these tests.

4. It is suggested that Tests 1, 2, 3, and 4 be applied routinely by the person plotting the chart.
The overall probability of getting a false signal from one or more of these is about one in a
hundred.

5. It is suggested that the first four tests by augmented by Teats 5 and 6 when it becomes
economically desirable to have earlier warning. This will raise the probability of a false signal to
about two in a hundred.

6. Tests 7 and 8 are diagnostic tests for stratification. They are very useful in setting up a
control chart. These tests show when the observations in a subgroup have been taken from two
(or more) sources with different means. Test 7 reacts when the observations in the subgroup
always come from both sources. Test 8 reacts when the subgroups are taken from one source at
a time.

7. Whenever the existence of a special cause is signaled by a test, this should be indicated by
placing a cross just above the last point if that point lies above the centerline, or just below if it
lies below the centerline.

8. Points can contribute to more than one test. However, no point is ever marked with more than
one cross.

9. The presence of a cross indicates that the process is not in statistical control. It means that the
point is the last one of a sequence of points (a single point in Test1) that is very unlikely to occur
if the process is in statistical control.

10. Although this can be taken as a basic set of tests, analysts should be alert to any patterns of
points that might indicate the influences of special causes in their process.

Figure 43. Notes on Tests for Special Causes (Courtesy of ASOQ Journal of Quality Technology, October 1984)

133
INTERPRETING INDIVIDUALS AND AVERAGES CHARTS

The previous eight basic statements of control are effective for a “quick look” to determine
whether statistical control exists, but the following is a more definite set of rules which can be
used for individuals and averages charts. Figure 42 represents eight tests for special causes (out-
of-control conditions). The figure divides the plus and minus three standard deviations from the
mean into zones A, B, and C for ease of interpretation. Notes on these tests for special causes
are shown in Figure 43.
__
X Chart Patterns and Possible Causes

134
Range Chart Patterns and Possible Causes

135
Interpreting Average and Range Charts Together

Average and range chars must be interpreted together as well as separately. A stable process will
have points distributed between control limits on the charts randomly. If the process is stable,
the points on the averages chart and the ranges chart should not tend to follow each other. Lack
of stability will sometimes cause the two charts to move together.

Example: If a process is positively skewed, then the points tend to correlate positively with
each other (points are high on both charts). If a process is negatively skewed,
then the points tend not to correlate (points on the averages chart will follow
points on the ranges chart but in opposite directions).

Skewness is a measure of the symmetry of the distribution (see glossary, Appendix H). Positive
skewness is when the distribution slopes downward to the right and negative skewness is when
the distribution slopes downward to the left.

TYPE I AND TYPE II ON CONTROL CHARTS

There are two errors which can occur when interpreting control charts which are similar to
sampling inspection. In sampling inspection the two types of errors are:

Alpha Risk – The risk of rejecting a good lot, which is a risk to the producer.

Beta Risk – The risk of accepting a bad lot, which is a risk for the consumer.

136
In control charting the Alpha Risk is called a Type I error and the Beta Risk is called a Type II
error. The Type I error is a false alarm where the chart appears to be out of statistical control
when it is in control. The Type II error is a false alarm where the chart appears to be in statistical
control when it is not in control.

INTERPRETATION OF RANGE CHARTS

A. Range charts shows uniformity or consistency

B. If the range chart is narrow, product is uniform. If the range chart is wide, the product is out-
of-control, something is operating on the process in s non-conforming manner.

C. Analyze the data points on the range chart for the following:

• Points beyond the control limit.

• Patterns or trends within the control limits.

• Any other non –random pattern.

137
 INTERPRETATION OF RANGE CHARTS

1. Points beyond the control limit.

A. Calculation or plotting error

B. The spread of the distribution (piece-to-piece variability) has increased (worsened) at

that point.

B. Measurement system has changed.

138
 INTERPRETATION FOR RANGE CHARTS

2. Patterns or trends within control limits.

A. Patterns and trends are evidence of non-control or a change in the process spread and
are an early warning of unfavorable conditions.

B. Patterns and trends can also indicate favorable conditions that should be captured for
product improvement.

C. Indicators are:
__
* 7 points in a row on one side of the average range (R). (Run above or below
the R).

* 7 points in a row that are consistently increasing (run up) or consistently

decreasing ( run down).

D. Run above the average range (R) or a run up indicates:

* Greater spread in the output values which could be from special causes.

* Change in measurement system.

__
E. Run below the average range (R) or a run down indicates:

* Smaller spread in output values, that is an improvement.

* Change in measurement system.

139
INTERPRETATION OF RANGE CHARTS

140
 __
INTERPRETATION FOR AVERAGE (X) CHARTS

A. Average chart shows where process is centered.

__ __
B. If the X chart is normal, the center of the process is not shifting. If the X chart shows a trend,
the center of the process is moving gradually up or down. If the X chart is erratic and out-of-
control, something is changing the center rapidly and inconsistently.

C. Analyze the data points on the range chart for the following:

* Points beyond the control limits

* Patterns or trends within the control limits.

* Any other non-random pattern

141
 __
INTERPRETATION OF AVERAGE (X) CHARTS

1. Points beyond the control limits

A. Calculation or plotting error

B. The process has shifted

C. Measurement system has changed

142
 __
INTERPRETATION OF AVERAGE (X) CHARTS

2. Patterns or trends within control limits

A. Patterns and trends are evidence of non-control or a change in capability.

comparison of patterns between the ranges and averages charts may be useful.

B. Patterns and trends can also indicate favorable conditions that should be
captured for product improvement.

C. Indicators are:
__
* 7 points in a row on one side of the process average (X) (Run above or
below the X)

* 7 points in a row that are consistently increasing (run up) or consistently

decreasing (run down)
=
D. Run above or below the X indicates:

* The process average has changed (and may still be changing)

* Measurement system has changed

143
 __
INTERPRETATION OF AVERAGES (X) CHARTS

3. Any other non-random pattern

A. Spread of data points within control limits

B. Cycles

144
 ATTRIBUTE CONTROL CHARTS

Lesson Objectives…

To help you make the most effective use of this manual and your time in
learning the concepts described, we have listed below your study
objectives for this lesson. These objectives tell where you are going in
this lesson and what you can expect to learn.

Please read the objectives carefully and keep them in mind as you read
and study the lesson. You should be able to accomplish each of these
objectives upon completion of this lesson.

• Construct Attribute Control Charts (p, np, c, u)

• Determine the conditions for selecting an Attribute Control

Chart.

• Identify an Out-Of-Control condition using Attribute Control

Charts.

• Define how Attribute Control Charts can be used to analyze

and control a process.

145
Attribute Chart

A defect is a characteristic of a part which is unacceptable. One item,

part, unit, or document may have many defects. Below are two
examples of defects. Complete the list with examples which relate to
your process.

• Number of scratches on one appliance.

• Missing zip code on purchase order.

• ______________________________.

• ______________________________.

• ______________________________.

• ______________________________.

A defect is an entire item, part, unit, or document which is unacceptable.

One item (etc.) may not have multiple defectives. Below are tow
examples of defectives. Complete this list with examples from your
process.

• Warped part

• Cracked glass

• ______________________________.

• ______________________________.

• ______________________________.

• ______________________________.

146
Attribute Chart

Symbols and Terminology:

ATTRIBUTE: - Type of data which is not measured.

A decision of good/bad, go/no is
made. This is sometimes referred to
as countable data.

CONFORMING - Good item (Meets requirements)

NON-CONFORMING - Items which are not acceptable or do

not perform to standard.

IN-CONTROL - Term used to describe a process

where all causes of variant are
inherent in the process (Common
Cause Variation). This usually
refers to a control charts when
points are between control limits
with no abnormal patterns. (A
process which is predictable)

CL - Centerline

LCL - Lower Control Limit

N - The total number of items in a

subgroup.

n - The number of individual items in a

subgroup.

147
Attribute Chart

COMMON CAUSE - Inherent or natural variation in the

process. These represent the “trivial
many.” (Pareto’s 20/80 rule)

ASSIGNABLE CAUSE - Also called “special cause”, refers to

variation in the process which is
local in nature and may be resolved
by the team and the supervisor. The
causes may be specific to a machine,
person, material, or method. This
represents the “vital few.” (Pareto’s
20/80 rule)

148
Attribute Chart

p CHARTS

Definition:

This is an Attribute Charts which is used to measure the percentage of

defective items.

Characteristics:

• Sample sizes may be constant or may vary.

• If the sample size varies more than 25% of the average

sample size, the control limits must be re-calculated.

• This is the only Attribute Charts which expresses the plotted

data as a percent. (%)

How To Construct:

1. Initiate data collection with Team.

• Determine who will collect data.

• Determine how data will be collected.

• Determine sample number in each subgroup (must be at least

25).

• Determine number of subgroups to be used (minimum of 20).

2. Collect Data. The plotted points will be the number of defective for
each sample group.

149
Attribute Chart

3. Compute the average number of defectives for all sample groups

(np).

np = total of defectives all groups = 3+4+11+7+0 = 4.56

total number of sample groups (k) 23

4. Compute UCLnp and LCLnp

Do not convert to a percent. This will be done after all of the

worksheet is complete.

150
Attribute Chart

5. Calculate control limits.

6. Convert p, UCLp, and LCLp to percent (if LCLp is a negative

number (-), use zero (0).
Attribute Chart

151
Attribute Chart

Example Data:

Subgroup Size (n) 150 150 150 150 150 150 150 150

Defectives (np) 4 2 1 1 0 8 12 3

Fraction Defective .027 .013 .006 .006 0 .05 .08 .02

Percent Defective 2.71 .3 .6 .6 0 5 8 2

150 150 150 160 160 150 150 150 150 150

9 6 5 9 8 7 9 12 6 4

.06 .04 .03 .05 .05 .04 .06 .08 .04 .02

6 4 3 5.6 5 4.7 6 8 4 2.6

150 150 150 150 150 150 125

7 8 8 18 11 10 4

.04 .053 .053 .12 .073 .066 .032

4.7 5.3 5.3 12 7.3 6.7 3.2

To find fraction defective and percent defective, see step number three.

152
Attribute Chart

p Chart Worksheet

Process________________________ Team______________________

Sampling
Method________________________ Date_______________________

153
Attribute Chart

np Chart Worksheet

Process________________________ Team______________________

Sampling
Method________________________ Date_______________________

154
Attribute Chart

2. Collect Data
_
3. Compute the average number of defects c.

total number of defects found

number of time periods (k)

or

7+9+6+12+9+16+18+22+15 = 114
22 22
_
c = 5.18

4. Compute UCLc and LCLc =

If LCL is a negative number, use zero.

155
Attribute Chart

c Chart Worksheet

Process________________________ Team______________________

Sampling
Method________________________ Date_______________________

156
Attribute Chart

3. Compute u (average number of defects per unit).

u = defects in one unit (c) =

number of samples per unit (n)

7 = 1.4, 9 = 1.8, 6 = 1.2, 1 = .2, …

5 5 5 5
_
4. Compute u (grand average).

Σ of all Defects
Σ of all units inspected

5. Compute UCLu and LCLu

157
Attribute Chart

u Chart Worksheet

Process________________________ Team______________________

Sampling
Method________________________ Date_______________________

158
Attribute Chart

• Points below the LCL are often desirable however, as

they move closer to the desired zero defects (defective).

• A point on or near the Lower Control Limit indicates that

a desirable condition may exist and should be
investigated.

159
Attribute Chart

Review:

Attribute Charts are used with ____________________ type data,

when making _________________ decisions. The four basic types of
Attribute Charts are ________, __________, _________, _________.
The __________________ and _____________________ are used to
monitor defects per units.

p Charts are different from the other Attribute Charts as the plotted
points are ________________________________________________.
Explain the difference between a defective and a defect:
_________________________________________________________.
_________________________________________________________.

Give an example of how an Attribute Chart will help in your

process:___________________________________________________
__________________________________________________________.

160
161
162
 CHART INTERPRETATION

Interpret for Process Control – p chart

* The purpose of the control chart is to identify when a the

process is operation out-of-control (Special cause variation
present) so that appropriate action me be taken.

* Analyze the control chart for the following:

1. Points beyond the control limits.

2. Patterns or trends within the control limits.

3. Any other non-random pattern.

163
 CHART INTERPRETATION

1. Points beyond the control limits.

* Points beyond the control limits are evidence of non-control at that point and the
presence of special cause variation.

* Points above the upper control limit:

A. Calculation or plotting error

B. Process performance has worsened

C. The measurement system has changed

* Points below the lower control limit:

A. Calculation or plotting error.

B. Process performance has improved.

C. The measurement system has changed.

164
 CHART INTERPRETATION

2. Patterns or trends within control limits

* Patterns and trends are evidence of non-control or a change in the

level of performance.

* Patterns and trends can indicate also favorable conditions that should
be captured for product improvement.

* Indicators are:
_
A. 7 points in a row on one side of the process average (p) (Run above
or below
the average)

B. 7 points in a row that are increasing inconsistently (run up) or

decreasing
inconsistently (run down)

* A run above the average or run up indicates:

A. Process performance has worsened.

B. A change in the measurement system.

* A run below the average or run down indicates:

A. Process performance has improved.

B. A change in the measurement system.

CAUTION: When the number non-conforming (np) is small in each subgroup

(five or less) the above guidelines may not be applicable.

165
CHART INTERPRETATION

166
 CHART INTERPRETATION

3. Any other non-random patterns:

* Non-random patterns such as cycles, the overall spread of data within

the control limits and relationships among values within subgroups
may provide an indication of special cause variation.

* Bout 2/3 of the data points should lie within the middle third of the
region within the control limits.

A. If more than 2/3 of the data points are in the middle third of the
region between the control limits, possible causes include:

- Calculation and/or plotting error

- Each subgroup contains data from two or more process systems

- The data has been edited.

B. If less than 2/3 of the data points lie within the middle third of the
region between the control limits, possible causes include:

- Calculations and/or plotting error

- Each subgroup contains data from two or more process streams

with dramatically different variability. (ex. Mixed material lots)

167
CHART INTERPRETATION

168
 PROCESS CAPABILITY – p CHARTS

1. Eliminate special cause variation

* Process capability is the variation due to common causes, that is, the
variation experienced after all causes of special variation have been
eliminated.

* Special causes of variation must be identified, analyzed, corrected and

prevented from reoccurring to determine process capability.

2. Calculate process capability

* Process capability is reflected in the process average non-conforming

(p) when all special cause variation has been eliminated (ex. All
points are in control).
_
* Historical data may be used to determine process capability (p)
provided that values associated with special cause variation has been
removed from the system. (25 or more subgroups)

3. Evaluate Process Capability

_
* The process capability (p) reflects the future level of performance that
can be expected by management.
_
* Capability (p) will remain constant during future production as long as
system changes are not introduced.

* Management must take steps to improve the process if the average

non-conforming (p) is not acceptable.

4. Improve process capability

* Management action must be directed to the system to reduce inherent

common cause variation.

169
* Tools used by management to identify and eliminate special cause
variation may not be appropriate to identify and eliminate common
cause variation.

* The p Chart is an effective device to monitor and analyze

management action to reduce common cause variation.

* Special causes of variation must be eliminated. The p chart will then

reflect the new process capability (p).

* Management should continue to improve process capability. Process

capability improvement is necessary!

170
171
 MEASUREMENT SYSTEM ANALYSIS

Measuring equipment is subject to variation. Therefore, an analysis of a

process cannot be meaningful unless the measuring instruments used to
collect data are both accurate and repeatable.

Definitions used in measurement system analysis:

1. GAGE ACCURACY – is the difference between the observed average of

measurements and the true average. Establishing the true average is best
determined by measuring with the most accurate measuring equipment
available.

2. GAGE REPEATABILITY – the variation in measuring obtained when

one operator uses the same gage for measuring the identical
characteristics of the same part.

3. GAGE REPRODUCABILITY – the variation in the average of

measurements made by different operators using the same gage when
measuring identical characteristics of the same part.

4. GAGE STABILITY – refers to the difference in the average of at least

two sets of measurements obtained with the same gage on the same parts
taken at different times.

5. GAGE LINEARITY – the difference in the accuracy values through the

expected operating range.

173
 MEASUREMENT SYSTEM ANALYSIS

Preliminary Considerations:

1. How many operators are to be involved?

2. How many sample parts are to be measured?

3. What number of repeat reading will be needed?

Conducting the study: Even though the number of operators, the number of
trials and the number of parts may be varied, one should conduct the study
according to the following steps:

1. Refer to operators A, B, and C, and number the parts 1 through 10

so that the numbers are not visible to the operators.

2. Calibrate the gage.

3. Let operator A measure 10 parts on a random order an enter results

on data sheet.

4. Repeat step 3 with operators B and C.

5. Repeat the cycle, with 10 parts measured in another random order,

for the number of trial required.

6. Steps 3 through 5 may be modified for large size parts,

unavailability of parts or when operators are on different shifts.

7. Enter all observations to derive Gage R and R.

174
175
176
 INTERPRETATION OF R AND R STUDY

GENERAL GUIDELINES

1. Gage R and R (combined)

- Under 10% error - Acceptable

- 10% to 30% error – May be acceptable based upon importance

of the application, gage cost, and cost of repairs.

- Over 30% error – Generally not acceptable. Make every effort to

identify the problem and get it corrected.

2. Reproducibility % is large:

- The operator is not properly trained in how to use and read the
gage instrument.

- Calibration on the gage dial are not clear.

3. Repeatability % is large:

- The gage instrument needs maintenance.

- The gage should be redesigned to be more rigid.

- The clamping or location for gaging needs to be improved.

177
 MEASUREMENT SYSTEM ANALYSIS

Preliminary Considerations:

1. How many operators are to be involved?

2. How many sample parts are to be measured?

3. What number of repeat reading will be needed?

Conducting the study: Even though the number of operators, the number of
trial and the number of parts may be varied, one should conduct the study
according to the following steps:

1. Refer to operators A, B, and C, and number the parts 1 through 10

so that the numbers are not visible to the operators.

2. Calibrate the gage.

3. Let operator A measure 10 parts on a random order and enter

results on data sheet.

4. Repeat step 3 with operators B and C.

5. repeat the cycle, with 10 parts measures in another random order,

for the number of trials required.

6. Steps 3 through 5 may be modified for large size parts,

unavailability of parts or when operators are on different shifts.

7. Enter all observations on data sheet and perform calculations to

derive Gage R and R.

178
APPENDIX

179
 MATHEMATICAL SYMBOLS

SYMBOLS DEFINITION OPERATION EXAMPLE

+ PLUS (ADD) 2+2=4

- MINUS (SUBTRACT) 2-2=0

X TIMES (MULITPLY) 2X3=6

• TIMES (MULTIPLY) 2•2=6

2(X) TIMES (MULTIPLY) 2(3)=6

2A TIMES (MULTIPLY) 2X the value of A

÷ DIVIDED BY (DIVIDE) 6÷2=3

/ DIVIDED BY (DIVIDE) 6/2=3

a a DIVIDED BY b (DIVIDE) 6 =6÷3

b 3

() THE QUANTITY (COMPUTE OPERATION IN ( ) FIRST) 2(1+2)=2(3)

SQUARE ROOT (SQAURE ROOT OF) 9=3

Σ THE SUM OF (ADD ALL VALUES) X, when X =

7,9,10
= 7+9+10
=26

180
 DEFINITION OF SPC SYMBOLS

X the value of an individual measurement or a single observation.

__
X the Mean or Average values of all values for a group of samples.
=
X the Grand Average of all the Averages for a group of samples.

R the Range, which measures the amount of dispersion or scatter contained in a

group of samples.
_
R the Average range of all Ranges for a group of samples.

C the number of Defects per sample.

_
C the average number of Defects per sample or unit inspected.

np the number of Defectives for all sample groups.

__
np the average number of Defectives for all sample groups.

p the fraction or percentage of Defectives for all sample groups.

_
p the average fraction or percentage of Defectives for all sample groups.

u the average number of Defects per sample. A sample in this case may be a unit or
a group of units.
_
u the average number of Defects for all sample groups.

UCL the Upper Control Limit (equivalent to +3σ)

CL the Central Line (equivalent to the Mean or Average Value)

LCL the Lower Control Limit (equivalent to – 3σ)

USL the Upper Specification Limit

LSL the Lower Specification Limit

Σ the symbol used to indicate “the sum of”

k the total number of sample groups or sub-groups.

n the number of individual sample measurements or observation contained in a

sample group.

181
N the total number of sample measurements or observations in a population.

σ the Standard Deviation which measures the amount of dispersion or scatter

contained in a group of samples. This is also referred to as sigma.
_
σ the average Standard Deviation for a group of samples that can be used to
estimate the value of the population’s Standard Deviation.

+3σ three Standard deviations to the right of the Mean (positive)

-3σ three Standard Deviations to the left of the Mean (negative)

MR Moving Range
_
MR the average of a series of Moving Ranges

d2 factor for estimating Standard Deviation

__
A2 factor for estimating X chart limits

D4 factor for estimating Upper Control Limit on R chart

182
 TABLE OF CONSTANTS
CHART FOR CHARTS FOR
AVERAGES INDIVIDUALS
NUMBER OF FACTORS FACTORS
OBSERVATIONS FOR CHART FOR RANGES FOR
IN SUBGROUP CONTROL CONTROL
n LIMITS d2 D3 D4 LIMITS
A2 E2

2 1.88 1.128 0 3.268 2.66

3 1.023 1.693 0 2.575 1.77

4 0.729 2.059 0 2.282 1.46

5 0.577 2.326 0 2.114 1.29

6 0.483 2.534 0 2.004 1.18

7 0.419 2.704 0.076 1.924 1.11

8 0.373 2.847 0.136 1.864 1.05

9 0.337 2.970 0.184 1.816 1.01

10 0.308 3.078 .0223 1.777 .98

11 0.285 3.173 0.256 1.744

12 0.266 3.258 0.284 1.716

13 0.249 3.336 0.308 1.692

14 0.235 3.407 0.329 1.671

15 0.223 3.472 0.348 1.652

16 0.212 3.532 0.364 1.636

17 0.203 3.588 0.379 1.621

18 0.194 3.640 0.392 1.608

19 0.187 3.689 0.404 1.596

20 0.180 3.735 0.414 1.586

183
 FORMULAS

I. Variable Data
__
A. X CHARTS
__
1. X = Average of the values in each sample
= __
2. X = Average of the X value
= _
3. UCL__ = X – A R
X 2
= _
4. LCL__ = X – A R
X 2

B. R CHARTS
_
1. R = Sum of R values divided by the number of samples
_
2. UCL = D R
R 4
_
3. LCL = D R
R 4

C. INDIVIDUAL AND MOVING RANGE

1. X = Individual Value

2. MR = Moving Range
___
3. MR = Sum of the moving range values divided by number of samples.
__ ___
4. UCL = X + E MR
X 2
__ ___
5. LCL = X – E MR
X 2
___
6. UCL = D MR
MR 4

184
II. Attribute Data

185
III. Standard Deviation (Sigma σ)

IV. Capability Ratio (C. R.)

Capability Spread or 6 σ
Specification Spread Tolerance

V. Capability Index (CpK)

Lesser of:

USL – X or X – LSL
3σ 3σ

186
 GLOSSARY OF TERMS

ASSIGNABLE CAUSES

The causes of variation that are not considered to be normal or common causes.
assignable causes of variation can be identified, corrected and eliminated within a
process.

ATTRIBUTE DATA

Data that is countable. For example, data that is classified as Good/Bad,

Accepted/Rejected, Defective/Non-Defective, etc.

AVERAGE

The “sum” of all values in a group divided by the number of values. The Average is also
called the Mean (X).

CAPABILITY

The maximum amount of inherent variation of a process. This is also referred to as the
Process’s Normal Distribution (the values of 6σ).

CAPABILITY RATIO

The ration of the process capability to the design tolerance or specification spread. The
ideal Capability Ratio should be .75 or less in order to maintain an acceptable level of
Statistical Process Control.

CAUSE-AND-EFFECT DIAGRAM

A method of analyzing the potential sources or causes of variation. This is often used to
identify and determine the possible causes for problems.

CHANCE-CAUSE SYSTEM

Various factors contributing to the amount of variation contained in any given process.

CHARATERISTIC

A property or trait that distinguishes one item from other similar items.

COMMON CAUSE

Natural causes of variation that can be expected to occur in any process that cannot be
changed or eliminated unless the process itself is changed.

187
CONSECUTIVE SELECTION

A selection method that is often used for machine-dominant operations whereby several
items in a row are selected from a population.

CONTROL (as applied to “process”)

A condition which monitors, analyzes and maintains the amount of variation contained in
any given process to that which is strictly sue to common or natural causes of variation.

CONTROL CHART

A graphic chart with control limits and plotted values of some statistical measure for a
series of samples or subgroups, over a period of time.

CONTROL LIMITS

Limits on a control chart that serve as a guide and a basis for action to the inherent
variation (Normal Distribution) of the plotted statistic.

CONTROLLED PROCESS (IN-CONTROL)

A process in which all, or nearly all, of the plotted statistics fall within the established
control limits. A process which repeats itself in a predictable manner.

CpK

Capability of a process with respect to the design specifications.

DATA

Facts (numerically expressed) used as a basis for decisions and conclusions.

DEFECT

Each instance that any item or unit fails to meet or conform to a single quality
characteristic or standard imposed on it.

DEFECTIVE

A unit that fails to conform to the stated quality standards thereby making the entire unit
unacceptable. A Defective may contain one or more defects.

188
DISTRIBUTION

The frequency of all possible results.

FISHBONE DIAGRAM

See Cause-and-Effect Diagrams

FRACTION REJECTED

The ratio of Defective pieces to the total number of all pieces inspected, usually
expressed as a decimal.

HISTOGRAM

A pictorial display which illustrates the characteristics of a mass of data.

INSPECTION LOT

A specific quantity or similar units offered for inspection and acceptance at one time.

ITEM

A single member of an inspection lot, usually, though not necessarily, a single article or
unit.

LOT

See INSPECTION LOT

MEAN

The arithmetical “average” that is computed by dividing the Sum of all samples by the
number of samples. Also called the Average values and symbolically expressed as X.

MEDIAN

The “middlemost” value contained in a group of data such that half the values are above
and half are below it.

MOVING RANGE

The difference between successive pairs of numbers in a series of numbers.

189
NON-CONFORMANCE

Any item or group of items that do not meet the established quality standards or
specifications.

NORMAL DISTRIBUTION

The natural distribution of all values that can occur when a process is operating In-
Control. Also referred to as the Normal or Bell-shaped Curve.

ONE HUNDRED PERCENT INSPECTION

The inspection of all items in the inspection lot.

OUT-of-CONTROL (as applied to a process)

The amount of variation that exceeds the limits of it’s Normal Distribution that can be
identified and attributed to a special or assignable cause.

PARETO CHART

A bar chart that is used to define problems and set priorities by listing the various
problems or causes of a particular problem according to their frequency of occurrence.

PERCENT DEFECTIVE

The ratio of Defective pieces to the total number of pieces inspected, multiplied by 100.

POPULATION

The entire group of items from which data or samples are drawn.

PROBABILITY

The chance or likelihood that some future or unknown event will (or will not) occur,
usually expressed as a number between 0 and 1.

PROCESS

An operation or series of operations that are combined to produce a change. i.e., a part,
product, unit, etc.

PROCESS CAPABILITY

A measure of the inherent or natural variation of a process. The Normal Distribution (6

σ) of any given process.

190
RANDOM SAMPLE or SELECTION

Selection of an item from a population such that each has equal chance of being selected.

RANGE

The difference between the largest and smallest values in a sample which indicates the
spread or dispersion in the data.

SAMPLE

A group of items taken from a population (the entire lot) which is used to make decisions
and draw conclusions about the population.

SAMPLING INSPECTION

Evaluation of the quality of material by inspecting only a portion of the material.

SIGMA (σ)

The small Greek letter that is used to denote the Standard Deviation.

SPECIFICATION LIMITS

Limits of the tolerance within which a process should run if its product is to be made
according to design.

STANDARD DEVIATION

A measure of dispersion (scatter or spread) of a set of data or values around its MEAN.

STATISTIC

Numerical data that summarizes a group or series of events

STATISTICAL PROCESS CONTROL

A method of mathematically analyzing process results to determine if the process is

behaving predictably.

191
SATISFACTORY CONTROL
__
X & R Control Charts

A process is operation in satisfactory control with respect to a measurable quality

characteristic if all, or nearly all, of the points are plotted within the control limits and are
randomly distributed around the Mean or central Line (CL); and all, or nearly all, of the
points are within the stated specifications.

X & MR Control Charts

A process is operating in satisfactory control with respect to a measurable quality

characteristic if all, or nearly all, of the points are plotted within the control limits and are
randomly distributed around the Mean or Central Line (CL); and all, or nearly all, off the
points are within the stated specifications.

SUBGROUP

A collection of individual pieces from a common source, possessing a similar set of

quality characteristics from which a random sample is selected.

THREE SIGMA UNITS

Calculated limits based upon observed data. The validity of these limits is based upon
the assumption that the data was selected from a “normal” or nearly normal distribution.

TOLERANCE

The allowable design deviation from a nominal values. The specification limits define
the total allowable tolerance.

UNIT

One of a number of similar items, objects, pieces, etc.

UPPER AND LOWER CONTROL LIMITS (UCL and LCL)

These values are calculated limits, each three “sigma” distance from the Mean or Central
Line, and between which 99.73% of the individual values or items from a Normal
Distribution will fall.

VARIABILITY

The normal scatter or dispersion with in any group of data.

192
VARIABLE DATA

Data that is measurable. That is, one can measure the precise characteristic of a given
item.

193
Table of Areas Under the Normal Curve

Appendix

194