Revised September 7, 2021
RM13006 Process Control Methods
AESQRM006202109
SAE Industry Technologies Consortia provides that: “This AESQ
Reference Manual is published by the AESQ Strategy Group/SAE ITC to
advance the state of technical and engineering sciences. The use of
this reference manual is entirely voluntary and its suitability for
any particular use is the sole responsibility of the user.”
Copyright © 2021 AESQ Strategy Group, a Program of SAE ITC. All
rights reserved.
No part of this publication may be reproduced, stored in a
retrieval system, distributed, or transmitted, in any form or by
any means, electronic, mechanical, photocopying, recording, or
otherwise, without the prior written permission of AESQ Strategy
Group/SAE ITC. For questions regarding licensing or to provide
feedback, please contact
[email protected].
Aerospace Engine Supplier Quality (AESQ) Strategy Group The origins
of the AESQ can be traced back to 2012. The Aerospace Industry was,
and still is, facing many challenges, including:
• Increasing demand for Aero Engines
• Customers expecting Zero Defects
• Increasing global footprint
The Aero Engine manufacturers Rolls-Royce, Pratt & Whitney, GE
Aviation and Snecma (now Safran Aircraft Engines) began a
collaboration project with the aim of driving rapid change
throughout the aerospace engine supply chain, improving supply
chain performance to meet the challenges faced by the industry and
the need to improve the Quality Performance of the supply
chain.
Suppliers to these Engine Manufacturers wanted to see greater
harmonisation of requirements between the companies. Each Engine
Manufacturer had Supplier Requirements that were similar in intent
but quite different in terms of language and detail.
This collaboration was formalized as the SAE G-22 Aerospace Engine
Supplier Quality (AESQ) Standards Committee formed under SAE
International in 2013 to develop, specify, maintain and promote
quality standards specific to the aerospace engine supply chain.
The Engine Manufacturers were joined by six major Aero Engine
suppliers including GKN, Honeywell, Howmet Aerospace, IHI, MTU and
PCC Structurals. This collaboration would harmonise the aerospace
engine OEM supplier requirements while also raising the bar for
quality performance.
Subsequently, the Aerospace Engine Supplier Quality (AESQ) Strategy
Group, a program of the SAE Industry Technologies Consortia (ITC),
was formed in 2015 to pursue activities beyond standards writing
including training, deployment, supply chain communication and
value-add programs, products and services impacting the aerospace
engine supply chain.
AESQ Vision To establish and maintain a common set of Quality
Requirements
that enable the Global Aero Engine Supply Chain
to be truly competitive through lean, capable processes
and a culture of Continuous Improvement.
ii
The SAE G-22 AESQ Standards Committee published six standards
between 2013 and 2019:
• AS13000 Problem Solving Requirements for Suppliers (8D) • AS13001
Delegated Product Release Verification Training Requirements (DPRV)
• AS13002 Requirements for Developing and Qualifying Alternate
Inspection Frequency Plans • AS13003 Measurement Systems Analysis
Requirements for the Aero Engine Supply Chain • AS13004 Process
Failure Mode & Effects Analysis and Control Plans • AS13006
Process Control
In 2021 the AESQ replaced these standards, except for AS13001, with
a single standard, AS13100.
The AESQ continue to look for further opportunities to improve
quality and create standards that will add value throughout the
supply chain.
Suppliers to the Aero Engine Manufacturers can get involved through
the regional supplier forums held each year or via the AESQ website
http://aesq.saeitc.org/.
AESQ Reference Manuals AESQ Reference Manuals can be found on the
AESQ website at the following link:
https://aesq.sae-itc.com/content/aesq-documents AESQ publishes
several associated documents through the SAE G-22 AESQ Standards
Committee supporting deployment of AS13100. Their relationship with
APQP and PPAP is shown in Figure 1.
Figure 1: AESQ Standards and Guidance Documents and the link to
AS9145 APQP / PPAP
This Reference Manual (RM) has been developed by the AESQ Process
Control Methods Working Group, a group of Senior Industry
Specialists from leading Aerospace companies, to promote the
correct application of process control. Aerospace products are such
that quality issues can be high profile and cause reputational
damage to the producer, customer, and the industry. They also cause
disruption to operations. Therefore, specialists from the leading
Aerospace companies collaborate to improve the industry’s adoption
and application of process control.
This Reference Manual includes both statistical and non-statistical
tools for the application of control activities in the factory, and
a range of statistical methods for process study of stability and
capability leading to process improvement.
It also discusses process control from a principles level to help
practitioners apply the techniques in the diverse array of
manufacturing processes and environments. Common pitfalls and
barriers are also discussed.
Many of the graphics in this guidance are produced using Minitab -
a recognized statistical software application.
1
2
RM13006 - Process Control Methods
12. BENEFITS OF STATISTICAL PROCESS CONTROL (SPC)
................................................ 68 12.1 Background
.........................................................................................................................
68 12.2 Benefits
...............................................................................................................................
68 12.3 Resistance to SPC
...............................................................................................................
69 13. METHODS AND FORMULAE
..............................................................................................
71 APPENDIX A PROCESS CONTROL METHODS ASSESSMENT CHECKLIST
......................................... 76 APPENDIX B PROCESS
CAPABILITY PLAN - EXAMPLE FORM
............................................................. 79
APPENDIX C TRAINING SYLLABUS
........................................................................................................
80 APPENDIX D ACKNOWLEDGEMENTS
....................................................................................................
84 Figure 1 AESQ Standards and Guidance Documents and the Link for
AS9145 APQP / PPAP ............ iii Figure 2 A Traditional View of
Quality (Anything within Tolerance is Equally Good)
..............................6 Figure 3 Taguchi’s Loss Function
(Any Deviation from Target Incurs Some Loss)
................................6 Figure 4 A Simple Control System
.......................................................................................................7
Figure 5 Process Control Overview
......................................................................................................7
Figure 6 3 Step Process for Process
Control........................................................................................9
Figure 7 The Deming (PDCA) Cycle
..................................................................................................
10 Figure 8 A Control Chart
....................................................................................................................
17 Figure 9 Variable Control Chart Selection
..........................................................................................
18 Figure 10 Process Showing No Signs of Special Cause Variation
........................................................ 19 Figure
11 Tests for Special Cause Variation
........................................................................................
20 Figure 12 Run Chart with Non-Statistical Limits
...................................................................................
21 Figure 13 Pre-Control Chart for Bilateral Tolerance
..............................................................................
23 Figure 14 Pre-Control Chart for Unilateral
Tolerance............................................................................
23 Figure 15 Fuel Air Bracket Example
.....................................................................................................
24 Figure 16 Attribute Control Chart Selection
..........................................................................................
26 Figure 17 P Chart of Defectives
...........................................................................................................
27 Figure 18 P Chart with Varying Sample Sizes
......................................................................................
27 Figure 19 C Chart
................................................................................................................................
28 Figure 20 C Chart
................................................................................................................................
29 Figure 21 Individuals Control Chart
......................................................................................................
30 Figure 22 Process Checklist Format Example
......................................................................................
31 Figure 23 Process Capability Index Cp/Pp
...........................................................................................
34 Figure 24 Elements of Process Capability Index (Cpk/Ppk)
..................................................................
35 Figure 25 High Capability - Practically Stable
.......................................................................................
38 Figure 26 Use of Ppk
...........................................................................................................................
39 Figure 27 Points Well Outside Control Limits
.......................................................................................
40 Figure 28 Binomial Capability Study
....................................................................................................
41 Figure 29 Poisson Capability Study
.....................................................................................................
41 Figure 30 A Non-Normal Distribution
...................................................................................................
42 Figures 31 and 32 A Bimodal Process Due to Oscillation
.......................................................................
43 Figures 33 and 34 A Bimodal Process Due to Step Changes
.................................................................
43 Figures 35 and 36 Normality Assessment (Process Approximately
Normal) ........................................... 45 Figures 37
and 38 Normality Assessment (Non-Normal Process)
........................................................... 45
Figures 39 and 40 Normality Assessment (Bimodal Distribution)
............................................................ 46
Figure 41 Effect of Taking Averages on a Flat (Uniform)
Distribution ....................................................
47 Figure 42 A Non-Normal (Skewed) Process Using an I-Mr Control
Chart ............................................. 48 Figure 43 A
Control Chart Using Transformed Data
.............................................................................
48 Figure 44 A Control Chart of Non-Normal Data with Appropriate
Limits ................................................ 49 Figure
45 Distribution Identification Using Minitab Software
.................................................................
50 Figure 46 Process Capability Analysis Using a Weibull
Distribution ......................................................
50 Figure 47 Probability Plot of Original Data (Left) and
Transformed Data (Right) ................................... 51
Figure 48 Capability Analysis of Transformed Data. The Capability
Is Not Ideal. .................................. 51 Figure 49
Common Sources of Variation
.............................................................................................
52 Figure 50 Variation Within and Overall is Similar
..................................................................................
54 Figure 51 Xbar-R Chart Produced with Data from Figure
50.................................................................
54 Figure 52 Pattern of 20 Holes
..............................................................................................................
55 Figure 53 X Bar and R Chart of Pattern of 20 Holes
.............................................................................
56 Figure 54 I-MR Chart with Pattern of 20 Holes
.....................................................................................
56 Figure 55 A 3-Way Control Chart with Pattern of 20 Holes
...................................................................
57 Figure 56 Capability Analysis with Pattern of 20 Holes
.........................................................................
58
3
4
1. THE IMPORTANCE OF PROCESS CONTROL
If a process is in state of statistical control, it is likely to
behave in a stable and predictable manner. This means that the
process will perform at a reasonable performance level, provided
the process’s capability is good, thus providing benefit for the
producer. The process will produce less ‘surprises,’ and many
aspects of operational planning can become more straightforward as
a result.
Additionally, for product features that influence performance, a
state of statistical control will offer the ability to maintain the
process around the optimal design nominal. Thus, providing benefit
for customers and users of the product.
Stability provides the potential for reliable planning. Instability
causes un-predictable performance that is difficult to plan
for.
But a state of statistical control is not necessarily a natural
state. Processes that are not maintained and controlled will
naturally decline over time. For this reason, methods of process
control are needed.
2. KEY PRINCIPLES FOR PROCESS CONTROL
2.1 Key Principles
Process control tools can be used for a number of purposes such as
performance calculations, root cause analysis, stability
assessments, etc. The tools can be very useful. However, it is
their application for the control of processes that maximizes their
benefit, through being able to control quality proactively, thus
avoiding quality issues.
The following principles underpin the use of the tools. All are
important:
Principle 1 - On Target with Minimum Variation
A process with excessive variation will invariably lead to
problems. The sources of variation should be managed proactively
and in a systematic way. For all operations this will be through
management of the process itself, but also foundational activities
such as maintenance of equipment, training and competency,
standardization of methods, correct measurement, etc. High quality
tends to result from a well-managed and stable manufacturing
environment.
Many product features have a design nominal that, if deviated from,
causes a loss in the performance of the end product. For these
features a process maintained ‘on target’ will perform better than
one allowed to run ‘off target’ regardless of the conformance to
specification. This concept is known as Taguchi’s Loss Function
(see Figure 3).
Additionally, even processes without a performance related nominal
will benefit from being ‘centralized’ between specifications due to
the reduced likelihood of non-conformance.
5
RM13006 - Process Control Methods
Figure 2 - A Traditional View of Quality (Anything within Tolerance
is Equally Good)
Figure 3 - Taguchi’s Loss Function (Any Deviation from Target
Incurs Some Loss)
Principle 2 - Move from Inspecting in Quality to Controlling
Quality
Reliance on inspection does not provide the optimal conditions for
quality control. Even when inspection is introduced at the point of
process, a ‘bad’ result is often detected too late to prevent
further non-conformance due to buildup of work in process (WIP) or
inability to spot trends if the data format is not appropriate.
Inspection is rarely 100% effective due to gauging and process
variations, and human factors.
To understand and control the process, it should be viewed using
tools that offer the correct level of granularity to highlight
trends and events and manage variation ‘on target’. Tools such as
SPC charts (variable control charts) offer a far higher level of
granularity than pass/fail inspection results.
Principle 3 - A Short Cycle Closed Loop Control System is
Vital
A closed loop system (shown in Figure 4) involves the capture of
information from the process, analysis of the information, a
decision against some criteria (typically on whether a process
anomaly is present), and a reaction to any such anomaly. The links
between each of these activities need to be in place and be as
short as possible in order to make decisions and actions
timely.
Such control systems can be operated manually or built into the
manufacturing process using automation. An example of an automated
system is an in-cycle probing routine used in an NC machine
tool.
6
Figure 4 - A Simple Control System
If the system is manually operated, ideally the measurer, decision
maker and action taker will be the same person. If not then
necessary communication channels, roles, responsibilities will need
to be defined, agreed, and maintained.
Systems that are overly reliant on end of line inspection are
compromised in all respects. They have severe delays, reliance on
distant communication that is almost always too late to achieve
anything constructive. At best, end of line inspection causes
issues to be seen late, at the point where customer disruption is
inevitable.
Principle 4 - The Operator Can Only Control if They Can See How the
Process is Behaving
Often the process operator is in the possession of some process
information. But if this information is not presented in an
appropriate manner the operator will be unable to see any changes
and trends. Then they will be unable to act on them. An example of
information that is difficult to process is a Coordinate
Measurement Machine (CMM) inspection report. The operator can
recognize non-conformances easily enough, however. the amount of
numbers and the discrete nature of each report means the data is
not stitched together to show the process behavior. This will
result in the operator only being able to detect non-conformance
thus making control of the process reactive. On the other hand, a
process control chart allows the operator to see the behavior of
the process, and if it changes significantly the operator can take
appropriate action to address the issue.
2.1.1 Overview of Process Control
Process Control has three main facets that are: Product Capability,
Process Control Methods, and Foundational Activities (see Figure
5). High performance is not achievable without all three elements
being in good order.
Figure 5 - Process Control Overview
Measure process
7
2.1.1.1 Level 1 - Importance of Product Capability
Process Capability (and thus Product Capability) is designed in
during the selection and development of the manufacturing method.
It is fundamental, because once designed in it can be very
difficult to change. The capability should be high enough in the
short-term that inevitable drifts and shifts over time do not
result in non- conformance or deviations from the design nominal
that result in a meaningful performance loss. Factors that may
result in additional variation and process movements include
multiple machine tools, batch to batch variations, operator to
operator variations, tooling variations, raw material variations,
etc.
The process designer should anticipate the potential effects of
these factors when designing the process.
The better the short term capability the more tolerant the process
will be to the sources of variation that affect the process in the
medium and long term. A high capability such as a Cpk of 2.0 will
allow the process to drift slightly without meaningful risk of
non-conformance.
Process control will not fix an incapable process.
2.1.1.2 Level 2 - Importance of Process Control Methods
Once the manufacturing method is selected and the potential sources
of variation have been determined, the process designer will
develop process control systems that detect anomalies when they
occur. The process and the product (process inputs and outputs)
will be considered. A range of process control tools may be used
(statistical and non-statistical). In many situations, control of
process inputs will be preferable to monitoring of outputs, however
this will be situation specific. These controls will ideally be
closed loop systems.
2.1.1.3 Level 3 - Importance of Foundational Activities
The management of Foundational Activities provides the basis for
stable operating conditions making process control achievable.
These activities include, but are not restricted to: machine tool
capability, condition and maintenance, standard methods,
measurement systems, training and competence, factory environment,
and raw material quality. It is expected that these be
appropriately managed.
Regardless of the process capability and process control system, a
process deployed into an environment which is unstable will cause
significant problems. The result will likely be continual issues
and frustration.
A stable environment will provide the conditions for anomalies to
be the exception rather than the rule.
3. APPLYING PROCESS CONTROL
3.2 Process Control Activities
The Process Control Activities fall into three key steps (see
Figure 6).
1. Process Control Method Selection - The selection of appropriate
process control tools and methods for each item in the Control
Plan.
2. Process Analysis and Improvement - Analytical study of the
process to prove the effectiveness of the process controls
described in the Control Plan. This involves the study of process
stability, capability and any actions needed to address
shortfalls.
The analytical study involves the following:
• The planning of the data that will be used to understand process
control and capability, and any predefined acceptance criteria for
control items. And the generation of a data collection plan.
• The execution of the data collection plan and application of
visual tools to view initial data.
8
RM13006 - Process Control Methods
• The analysis of the process data using statistical techniques to
describe process stability and capability, study the effect of the
sources of variation, and understand the nature of any
shortfalls.
• The actions to address any shortfalls in process stability,
capability, or input variation.
3. Process Monitoring and Control - The application of the controls
during continued production to detect issues and maintain process
stability and capability.
Figure 6 - 3 Step Process for Process Control
3.3 Process Control in Process Design and Quality Planning
Process design and Quality Planning are concurrent activities.
Process Control can be considered as much part of the process
design as it is part of the Quality Planning activity. This point
is often missed when one views tools such as PFMEA and Control
Plans exclusively through a Quality Planning lens.
If the Quality Planning process is viewed without consideration for
the process design activity, or one takes an overly document
centric view, one could conclude that process control is only
decided after the PFMEA activity. However, in reality one would
begin to design the control system proactively as early as
possible. This will often be done through applying pre-existing
methods and considering past experience, often reapplying methods
from similar products/processes.
Some controls will be based on part family standards and process
best practices.
By the time the PFMEA is undertaken, the control system will mostly
be decided upon. The controls provide the basis for scoring the
detection in the PFMEA. The development of improvements is then
based on the risk profile for the process. Additional controls and
improvements may be developed based on this.
During process design and development, the capability of the
process should be assessed to establish whether the process has
sufficient capability to be adequately controlled within the
specification limits or close to a target value. Ideally realistic
tolerances will have been agreed during product development, based
on customer needs and historic capability information.
In this early stage of development, the producer will likely be
running the process on a limited run of product with fewer sources
of variation present than would be expected in full manufacture.
For example, a single machine with limited strip and reset of the
process, and little raw material changes.
For this reason, the producer will need to estimate the likely
effects on the process capability when the process goes into full
manufacture, and judge the required capability for the initial
proving run. Capability at the proving run is of no use if it
cannot be translated into capability in volume manufacture.
Once the capability and stability are proven the process will be
operated in serial manufacture using the adopted control system.
The capability and stability may be assessed at various stages of
product implementation and production.
Choice of Capability Metric
For initial capability the process may typically be run on a single
machine tool and the product will likely be run on a continuous
production run. For this reason, the capability may be reasonably
well estimated by the Cp/Cpk metrics which are based on the
analysis of variation in the short term, within subgroups from part
to part variations of individuals.
9
RM13006 - Process Control Methods
For a full production run the data may have been derived from many
subgroups (with sources of variation between each subgroup) or over
a longer time period in which some natural process drifts and
shifts occur. In this scenario the Cp/Cpk indices will be biased
towards the short term (within subgroup) variation and may give
overly optimistic results. For this reason, while the Cp/Cpk
capability metrics can be informative they should be used in
conjunction with the Pp/Ppk performance metrics which use an
overall estimate of variation. In certain cases the difference
between the short term (within subgroup) variation and the overall
variation are such that a Between/Within capability study may be
required.
For data involving multiple machine tools, the Cp/Cpk metrics may
not make sense if the machines have systematic differences between
them. In this case again the Pp/Ppk metrics may be more
appropriate. However, there may be a case, where differences
between machine tools are significant, to assess the capability of
each machine separately.
3.4 Process Control in Continuous Improvement
The development of the control system in improvement may be done
proactively (for example as an outcome of a PFMEA activity), or as
a reaction to quality problems. When done proactively it follows a
similar approach to that taken in process design and development,
however the process control system will more likely be
developed/refined following a process data study and PFMEA
activity, to a pre-existing manufacturing process.
In problem solving it may be more or less regimented depending on
the nature of the problem, the methodology used and whether the
cause of the problem is obvious or not.
In continuous improvement activity usually some type of methodology
will be used. Most methodologies follow a sequence of Plan, Do,
Check, Act (known as the Deming cycle, see Figure 7). In the early
phases, process data may be examined to understand the nature of
the problem and decide on a course of action. The stability and
capability of the process will be assessed. The work will be
planned with an idea as to what the expected outcome will be.
Sometimes this will involve modification of the control system.
Once the work has been done, the result of the actions will be
checked and compared against the expected outcome. Action will be
taken based on this. Usually some form of Adopt, Adapt or Abandon
decision for the change. This will be a data driven cycle.
Figure 7 - The Deming (PDCA) Cycle
3.4.1 Communication and Workforce Engagement
In certain situations, the closed loop system will involve multiple
personnel. For instance, the person monitoring the process may not
be the person responsible for making adjustments. In these
situations, the responsibilities need to be made clear and
particular attention will need to be given to the engagement of all
personnel in the process.
A RACI analysis may be worthwhile to clarify who is Accountable for
the control systems operation, Responsible for each activity within
it and those Consulted and Informed periodically during its
operation.
10
RM13006 - Process Control Methods
Process control and foundational activities are best sustained when
the workforce is highly engaged in the operation of the controls,
understand the importance of them, and are involved in the
improvement of the systems.
4. PROCESS CONTROL METHODS OUTLINE
4.1 Nine Recognized Process Control Methods
The AESQ recognize the following Process Control Methods. Table 1
gives a simple summary. More comprehensive guidance follows in
Section 5.
These are listed in a sequence that roughly aligns with the
robustness or precision of each method. But their selection will
depend on a number of factors. It is a case of selecting the right
tool or tools for the job.
Table 1 - Overview of Recognised Process Control Methods
Method
Application
Example
See Also
Error/Mistake Proofing To avoid defects caused by inadvertent
errors. The most robust and preferred method. Mistake Proofing
devices build quality into a process in order to prevent and/or
detect errors prior to defects being made.
Typical reaction:
Some error proofing devices prevent the possibility of entering an
error state, so no Reaction Plan is required.
Some error proofing devices such as alarms and buzzers require the
operator to stop and investigate the error cause. This reaction may
involve following a prescribed recovery plan that eliminates the
error condition or escalates the situation to an engineer or
supervisor to determine next steps.
One-way fit of a die insert to prevent incorrect orientation during
loading.
Use of a physical device to prevent installation of an oil- feed
tube into the wrong port.
Use of electrical devices such as proximity switches and cameras to
ensure proper alignment and orientation prior to the operation
proceeding.
Section 5.1
See Also
Control Charts for Variable Data To monitor process inputs or
process outputs that are continuous in nature for the purpose of
establishing and maintaining a state of statistical control (also
referred to as process stability).
Typical reaction:
Variable Control Charts alert the operator to “out of control”
process behavior (special causes). If these occur, action is taken
to identify the causes and bring the process back into statistical
control. Recovery actions may be prescribed, or technical support
may be provided depending on the situation.
Dimensional product features are plotted on Control Charts at the
point of process and monitored by the operator. The operator takes
action to investigate and remedy issues when special causes are
detected.
The pressure drop in a vacuum furnace is monitored on a Control
Chart to warn of developing issues. The operator responds to
special causes by performing equipment diagnostic checks.
Section 5.2
Run Charts with Non-Statistical Limits To monitor process inputs
that require adjustment within acceptable operating limits in
response to natural drift. Likely to be used when statistical
limits offer little practical benefit or lead to false signals of
special cause.
To control conditions that follow a specific “profile” during the
operation of the process.
Typical reaction:
Similar to Control Charts these Run Charts will have rules applied.
Rules will typically be based on limits requiring some action
(e.g., tool change). While these limits may not be statistically
determined in the same way a Control Chart is, the Reaction Plan is
similar to the ones used for Variable & Attribute Control
Charts.
The viscosity of the slurry used in an investment casting process
is monitored. When a limit is reached, the operator adds water to
the mixture to correct for evaporation over time.
A highly capable characteristic of a machined part where tool wear
is expected and can be tolerated to a point to maximize its
effective use. The operator changes the tool at a predetermined
dimension before the dimension becomes nonconforming.
Furnace Run Charts tracking thermocouple temperature levels
throughout a cycle for heat treat and brazing processes. Each point
in the cycle will have a normal operating window beyond which
investigation occurs. Most likely to use software enabled system
linked to the equipment.
Section 5.3
See Also
Pre-Control Charts To keep a capable process on target when the
process has a tendency to move from the nominal value. Where
processes are not sensitive to small changes, the use of a
statistical Control Chart offers little additional value.
When simple operating rules are beneficial.
Typical reaction:
Pre-Control Charts have "warning limits". The action required is
either one of further monitoring or action to investigate the
reason for the process running off target. The reaction will depend
on the ruleset being used.
Correct setup of a fuel control valve grinding process is confirmed
by running the process and making adjustment until process is
centered. Once centered, the process is monitored and only adjusted
when Pre-Control rules are broken.
Monitoring of the outside diameter of an air cycle machine shaft
where the operator controls adjustments using a machine offset in
response to signals on the Pre-Control Chart.
Section 5.4
Life/Usage Control Processes that degrade over time where the
useful life or usage is known. Limits to operation (time or number
of cycles) will be set conservatively to avoid
nonconformances.
Typical reaction:
The operator may be provided with a machine cycle counter. The
reaction is to change the item that has reached its life limit at
that point.
If cutting tool usage is monitored electronically, the machine may
be programmed with control criteria, e.g., programmed not to allow
further use of the tool after a certain number of cycles or hours
use.
A forging die is run for a predetermined number of cycles before
being removed for refurbishment/disposal. The life and die change
are managed to coincide with batch changes.
Cutting tools with known wear characteristics are run for a
specific cutting time. The tool life is electronically monitored by
the Computer Numerical Control program to prevent overuse.
Section 5.5
See Also
Attribute Control Charts For monitoring quality levels of product
or process attributes where the output is based on counts
(typically defects) or classification (typically defectives). Used
for recognizing changes in quality level due to special causes of
variation.
Typical reaction:
Similar to Variable Control Charts. The action may be to stop the
affected process or to investigate and resolve the problem.
Inspectors counting solder defects on a printed circuit board use a
chart that monitors the number of defects per board. When a special
cause is detected, the soldering process owner is informed and
investigates the cause of the issue. The charts are reviewed by the
operations management to identify opportunities for improvement,
and to confirm results of improvement initiatives.
Section 5.6
Visual Process Check and Checklist Checking process attributes and
recording them as meeting the requirements to run the
process.
Typical reaction:
If the checklist cannot be completed, action will be taken to
correct the gap. The process is not started. The execution of the
process check should be audited for compliance.
A forging die is periodically examined by an operator for evidence
of damage, wear, or scoring. The operator uses a checklist to
record the result of the check.
An operator of a process with a lengthy setup operation uses a
checklist to confirm each step of an operation is completed before
running the machine. The checklist may also include safety
items.
Section 5.7
First Piece Check To validate the setup and quality of a process
prior to the production run.
Typical reaction:
If the criteria applied to the first-piece check are not met, the
reason for the failure will be investigated. Once corrective action
has been implemented the first-piece check will be repeated to
validate the setup. Any activity of this kind should be documented
for traceability.
A Coordinate Measuring Machine check of the first part in a batch
of parts off a forming press is performed following change of press
tooling. If the part meets the requirements, the process is allowed
to run, and is then controlled using other Process Control Methods
during the production run.
Section 5.8
See Also
Test Piece evaluation Commonly used along with process parameter
control to provide validation of product quality. Typically, a
destructive examination. It should be noted that a destructive
examination processed with a batch of material is more inspection
than control; so it needs to be used along with effective process
input control.
Typical reaction:
For a test specimen that does not meet specifications upon the test
conducted, the Reaction Plan will typically instruct the test
operator to engage the appropriate engineer (e.g., Materials,
Quality or Manufacturing Engineer) who will investigate the cause
of the failure (process parameter inputs, furnace run schedule,
etc.) as for clues to why the test specimen failed to meet the
test. The product will be quarantined.
A piece of test material processed along with a batch of carburized
gears in a heat treatment cycle is tested in a laboratory.
Tensile strength destructive examination of a test specimen used in
a heat exchanger vacuum braze process.
Section 5.9
4.2 A Note on Automation
Process Control Methods can be incorporated using automation to add
reliability and access to information at the earliest possible
opportunity (e.g., in-cycle machine/part probing, automatic process
compensations).
5. PROCESS CONTROL METHODS FURTHER EXPLANATION
5.1 Error/Mistake Proofing
Error proofing is the use of an automatic device or method that
either makes error impossible or makes its occurrence immediately
apparent. Error proofing should be chosen when the process is at
risk of human error. The process risk analysis (PFMEA) should
identify where human error is a potential cause of failure, where
it has a high impact (severity) or may not be easily detected
(detection). Safety related risks often require mistake proofed
solutions.
Error proofing devices can take four forms. The hierarchy of these
is:
1. Elimination - design the product or process hardware/software in
such a way that an error is not possible.
2. Control - prevent an error being made by detecting it before it
has an effect.
15
RM13006 - Process Control Methods
3. Signal - provide an immediate and obvious warning to prevent or
highlight an error.
4. Facilitation - methods of guidance that make error less likely
or will catch it.
NOTE: Error proofing methods are not industry specific. Some
industrial sectors have a particularly well developed
mistake-proofing culture often extending into product as well as
process design. The automotive industry is very well known for its
use of error proofing both from the manufacturing processes to the
operation of the final product.
Examples:
• Guide Pins used to assure a one-way fit of a tool, fixture, or
part to prevent incorrect orientation.
• An alarm used to alert an operator that a machine cycle has been
attempted with a misaligned tool. The operator can take action to
correct the problem.
• A limit switch used to detect correct placement of a work
piece.
• Counters can be used to help an operator track the correct number
of components needed in an assembly.
• A checklist used to assure all key steps are completed by the
operator to prevent missing something that could cause an escape
and/or defect. This approach is also described further in Section
5.7 - Visual Process Check & Checklist.
• Use of machine probing as either a control during manufacturing
to check a size before final cut or as a signal after final cut to
detect an anomaly or identify that an adjustment may be
needed.
• Use of a Stopper Gate (physical barrier) affixed to a Fan
Compressor assembly fixture to ensure an oil fill tube is installed
in the correct port when there are multiple ports to choose
from.
• Asymmetrical design of a nameplate that assures it is installed
in only one possible orientation preventing backwards or upside
down installation.
• A left/right two button hand operated system with foot switch
operation to ensure hands are free prior to cycling a forging
press.
• Automated weighing of a part or batch to ensure part is
completely processed or batch is complete and present before moving
to the next operation.
To ensure error proofing devices are robust, it is good practice to
check that the failure of the device does not cause a problem (test
to see what happens if the device fails to detect the error).
Depending on the result (and the criticality of failure), revisit
the design and maintenance requirements of the device and improve
it.
If it is not possible to have an automated error proofing device,
some of the other methods included in this standard may offer an
adequate level of protection.
For further reading on the subject of Error/Mistake-Proofing the
following may be referred to:
“Poka-Yoke,” by Productivity Press, ISBN 0-915299-31-3
“Mistake-Proofing for Operators: The ZQC System,” by Productivity
Press, ISBN 1-56327-127-3
5.2 Control Charts for Variable Data
A control chart is a tool used to monitor and visually assess the
behavior of a process over time. The control chart shows process
data and ‘control limits’ which provide an approximation of the
natural range of the process due to ‘common causes’ of variation.
These limits (and other tests) are then used to detect abnormal
events and trends (‘special causes’ of variation). The response to
common cause issues and special cause issues are typically
different, making the correct choice of approach important.
16
RM13006 - Process Control Methods
• Response to special causes of variation should be to immediately
respond and investigate their cause. Provided the process was
previously ‘in control’ the question is usually one of
understanding what has changed.
• If common cause variation is behind the issue, then the
fundamentals of the process should be understood, and the process
changed to improve its capability.
The ‘control limits’ are derived using the process data and not
product or process tolerances, thus minimizing the risk of
responding in the wrong way (such as missing a signal to
investigate or adjusting the process when it was not needed).
Assuming the process is capable the control chart will allow
special causes of variation to be detected even if the data fall
within the specification limits, meaning problems can be recognized
earlier than if traditional inspection methods were used.
Figure 8 - A Control Chart
This section outlines four recognized control charts for variable
data and provides guidance as to when they may be used. The list is
not exhaustive. There are many more types of control charts not
covered here that may be used for specific situations.
Figure 9 and Table 2 outline the basis for variable control chart
selection.
71645750433629221 581 Observation
Table 2 - Variable Control Charts
Chart Its Use
Xbar and R
Xbar and S
Monitoring and control of characteristics on products being
produced at a volume where typically a sample (subgroup) will be
taken periodically to maintain quality.
Example: From a high volume process, five parts per hour are
sampled from the line and measured. The average and range is
plotted to understand if the process has changed (due to moving off
target or through an increase in variation).
Can also be used for multiple similar products where it can used to
plot ‘deviation from target’ thus avoiding the need for multiple
charts.
The X bar chart displays the average of the subgroup. The R or S
chart displays the variation within the subgroup (either the Range
or Standard Deviation).
An X-Bar and R chart is used for subgroups of 3 to 8.
An X-Bar and S chart is used when subgroup size exceeds 8.
NOTE: The variation within the subgroups is assumed to be
representative of the overall variation (no between batch effects
expected). When this assumption is not met the process may appear
out of control when in fact it is not. Consult an experienced
practitioner if this appears to be the case.
Individual and Moving Range
Monitoring and control of characteristics on individual products
being produced from continuous processes at a rate where
subgrouping of data is not feasible.
Monitoring and control of process characteristics.
Can also be used for short run applications where there is product
mix with similar characteristics (may be known as part families).
In this situation the variability for all parts should be similar;
used to monitor part families.
The Individuals chart displays the actual measured value (or
deviation from target).
The Moving Range chart plots the difference between consecutive
points (short-term variation).
NOTE: The variation from item to item is assumed to be
representative of the overall process variation (no batching
effects or systemic drifts/wear expected). When this assumption is
not met the process may appear out of control when in fact it is
stable. Consult a process control specialist if this appears to be
the case.
18
Within control chart or ‘Three Way’ chart
Characteristics where the variation within the subgroup is not
representative of the overall variation between them, usually the
case when monitoring processes with ‘batching’ effects or multiple
characteristics (a group of identical features) within a part are
studied where the assumptions for an Xbar/R or S chart are not
met.
The subgroup average is plotted on the Xbar chart.
The variation between consecutive subgroup averages is plotted on
the Moving Range chart.
The Variation within the subgroup is plotted on the R or S
chart.
NOTE: Higher subgroup sizes may lead to higher sensitivity to
‘special causes’ on R and S charts. Expected patterns within parts
and batches can sometimes show signals that have no practical
significance. Guidance may be sought from an experienced SPC
practitioner if this appears to be the case.
There are eight industry standard tests for statistical control; to
determine if the process data contains evidence of special causes
of variation.
A process can be judged to be in statistical control (i.e., only
common causes of variation present) when there is an absence of the
patterns shown in Figure 11. An example of a stable process is
shown in Figure 10. It should be noted when seeking to improve a
process that the more tests used, the more signals will be
detected. It may be worth using a selected few when starting out
using control charts.
For process control purposes manufacturers often select the most
appropriate tests for the process being operated, taking into
account the actions that would be needed when they occur. Tests
most frequently used by operators are Tests 1 and 5 (Figure 11),
however, software applications make the use of all tests relatively
simple.
Figure 10 - Process Showing No Signs of Special Cause
Variation
2825221 91 61 31 0741 Observation
In di
vi du
al V
al ue
M ov
in g
Ra ng
19
5.3 Run Charts with Non-Statistical Limits
Some processes have characteristics that naturally drift in a
certain direction as the process runs (i.e., the drift is a ‘common
cause’ in the process). These processes when viewed on commonly
used control charts tend to break tests for special cause long
before the drift becomes a meaningful issue. Operation of
traditional statistical control limits may then provide little
benefit when compared to the characteristics’ ‘loss function’ and
the cost and other implications of adjustment or reset. The more
frequently the process is sampled the smaller the differences
between measurements, which tends to exacerbate the issue.
Processes where this behavior may exist naturally are chemical
etching (concentration changes), investment casting slurry control
(through evaporation) and in some cases machining cutting tools (if
they exhibit significant wear/drift with use).
An approach to manage this variation is to set limits on a time
series chart. This limit will be set such that it detects drifts to
avoid problems, but not so soon as it becomes uneconomic to adjust.
This type of control is generally only useful when operated at the
process rather than at an end of line inspection.
With appropriately set limits this method can be used effectively
to control quality even using simpler measurement systems than
downstream measurement equipment such as a CMM.
The following six step approach can be used:
1. Determine the variable to be monitored.
2. If the variable is an input or process variable, study, and
quantify its relationship to the process outputs.
3. Establish the optimal process limits to be applied. In most
cases this should be done using process data, to best ensure the
limits are not too wide to allow a non-conformance.
20
RM13006 - Process Control Methods
4. Establish the adjustment to be made when the limit is reached.
For example, this may be to adjust towards a lower limit, or an
optimal setting, or in the case of a cutting tool, replace it. This
reaction will be documented in the Control Plan and process
instructions.
5. Operate the process and plot the measurements.
6. If the process limit is reached, adjust/set the process (see
step 4). Confirm the adjustment has had the desired effect. If so
continue. If not take action to understand why.
Figure 12 demonstrates how a chart of this type may be used. The
process drifts upwards so a lower limit is not discussed within
this example (for simplicity). It may, however, be wise to have one
to mitigate other risks.
Figure 12 - Run Chart with Non-Statistical Limits
Process improvements can be made using the data from the run chart,
for example in the following ways:
• Use process data and related process output to determine tighter
reaction limits.
• Incorporation of automatic adjustments to the process to tighten
the adjustment interval. This will decrease the spread between the
limits.
• Make changes to the process or tools that decrease the rate of
change of the process variable being controlled.
• Optimize the initial location for the process to increase the
time between adjustments.
Features controlled in the way described should typically have a
relatively flat ‘loss function’ when compared to the cost of reset
or adjustment. The designer should be consulted where implications
of process drift is not understood.
201 81 61 41 21 08642
1 0.075
1 0.050
1 0.025
1 0.000
set-point' is reset to 'process reached process When limit is
Diameter
21
Processes with systematic drift and infrequent ‘large adjustments’
may produce distorted capability analysis. There are two reasons
for this.
1. The within subgroup range is typically small relative to the
overall variation, resulting in Cp metrics being overly optimistic
and not representative of the spread of the process.
2. The distribution of the data may not fit a distribution well
enough to make accurate capability predictions. Both Cp and Pp
derived capability may be inaccurate and alternative methods (e.g.,
non-normal methods such as Johnson Transformation, Box-Cox
Transformation (see Section 7. Guidance for Non-Normal Data) may be
required. If these methods do not help, then the process
performance may need to be characterized by other means.
5.4 Pre-Control Charts
5.4.1 Background
Pre-Control is a method for monitoring and controlling the process
within specification limits. It may be particularly useful when
applied to process outputs or parameters that have a tendency to
drift but for which the process is not overly sensitive to small
changes. For example, a measurement taken on a ground feature where
the grinding wheel wears over time.
Pre-Control may also be useful where it is important to maintain a
capable process centered or ‘on target’, when detection of process
‘special causes’ are less important.
NOTE: The use of Pre-Control dates back to the 1950s. The merits of
its use are often debated, with some favoring and some opposing its
use. There are definitely valid arguments for and against which
should be considered.
Pre-Control uses a chart that monitors items by classifying the
measurements into colored zones (Red, Yellow, or Green). Decisions
are made whether to adjust or stop the process based on where in
these zones the measurements lie.
The advantages of Pre-Control are its simplicity and that it drives
behavior towards on-target thinking.
NOTE: It is commonplace for the bands to be set as follows (see
Figure 13):
• Green - the central 50% of the tolerance band (or 50% tolerance
around a specific target).
• Yellow - outer quartiles (or remainder) of the tolerance
band.
• Red - outside the tolerance.
Where tolerance is unilateral, the chart will have a single green,
yellow, and red zone (see Figure 14).
5.4.2 Method
Following setup, a qualification phase runs according to a
predefined ruleset to ensure the process is ‘on target’. Typically,
qualification is passed after five consecutive units are produced
in the green zone.
Three styles of Pre-Control exist:
1. Classical Pre-Control: Rules based around sampling two
consecutive items periodically from a production run:
• Single item in Yellow - continue to run (but check subsequent
item).
• Both items in Yellow - stop and investigate. Correct the
process.
• Single item in Red - stop and investigate. Correct the
process.
22
RM13006 - Process Control Methods
2. Two Stage Pre-Control: Based on a single item being sampled
periodically.
• A single measurement in the yellow zone triggers measurement of
additional items.
• A single Red will trigger process to be stopped and
corrected.
3. Modified Pre-Control:
• A standard control chart with colored zones applied as described
for Classical Precontrol (but to control limits, not
tolerances).
With the exception of modified Pre-control, the limits and rules
are not statistically derived. Opponents argue there is a risk of
process tampering (over-control), if applying Pre-Control to an
incapable process; or missing special causes that would be detected
by statistical control charts. It is therefore not advisable to use
Pre- Control on processes with poor capability or in situations
where small changes in process need to be recognized.
Figure 13 - Pre-Control Chart for Bilateral Tolerance
Figure 14 - Pre-Control Chart for Unilateral Tolerance
NOTE: If analyzing the capability of a process that uses
Pre-Control methods, a statistical control chart should be
constructed to ensure the process is stable prior to analysis of
capability and communication of capability indices such as
Cp/Cpk.
Despite the concern of an unstable process on capability, a measure
of goodness such as extended period in green zone on a Pre-Control
Chart may serve as satisfactory evidence of capability to meet
customer requirements if the customer permits this. This is more
likely for minor characteristics than for KCs or special
characteristics such as those categorized as Major or
Critical.
For further reading on the subject of Pre-Control refer to
Implementing Six Sigma (2nd Edition) - Breyfogle 2003. ISBN
0-471-26572-1).
5.4.3 Pre-Control Example
An aerospace manufacturer produces a Fuel Air Bracket (see Figure
15) with a key feature having an engineering tolerance of 0.386 ±
0.005 inches. The central 50% of the total tolerance (±0.0025
inches) defines the green zone.
23
Figure 15 - Fuel Air Bracket Example
The engineer defines the zones on the Pre-Control chart. The edges
of the green zone are known as Upper and Lower Pre-Control limits
(UPC and LPC).
UPC limit = 0.386 + 0.0025 = 0.3885 inches.
LPC limit = 0.386 - 0.0025 = 0.3835 inches.
The control method selected is two stage Pre-Control.
Set-Up Procedure
Following successful setup, the process operator runs five parts
and records the dimensions of the features being controlled. If all
five parts fall within the green zone on the Pre-Control chart (UPC
= 0.3885 inches and LPC = 0.3835 inches) the setup is judged to be
targeted properly and sample measurements are taken at a frequency
of 20% (check every 5th part). This measurement frequency is for
the purpose of maintaining process control and does not relate to
product inspection frequency.
Executing the Pre-Control Monitoring Technique
The 10th piece comes up for inspection. It has a measured value of
0.387 inches. This is within the Pre-Control (UPC and LPC) limits,
and the operator continues with production. The next piece to be
inspected is the 15th. Its measurement is 0.3854 inches, well
within the Pre-Control limits so the operator continues. The 20th
part measures 0.3892 inches. This value is outside the UPC limit.
The reaction plan referenced in the Control Plan determines that
the operator now measures the next part produced, in this case the
21st. This part measures 0.3867 inches, again outside the UPC
limit. The operator stops the process and investigates according to
the prescribed reaction plan.
Pre-Control Rule 1: If the measured value is within the green zone
(Pre-Control limits UPC and LPC) the operator may continue to check
every 5th part (apply a 20% monitoring frequency).
Pre-Control Rule 2: When two consecutive measured values fall
outside the same Pre-Control limit (UPC and LPC), the operator
should react making an appropriate process adjustment. The reaction
plan reference in the Control Plan (refer to AS13004) should
describe the actions required.
Pre-Control Rule 3: When one measurement violates one Pre-Control
limit and the following part violates the opposite Pre-Control
limit, the variability may have increased. The operator should
investigate the cause engaging support if needed (e.g.,
Quality/Manufacturing Engineer). The reaction plan referenced in
the Control Plan (refer to RM13004) should describe the actions
required.
5.5 Life/Usage Control
Processes may have factors that are dynamic in nature and change
through use or over time. Such processes may require control
methods that prevent the process (or its factors) reaching a
condition that will adversely affect the product of the process.
Such controls can be placed on, e.g., chemicals, wearable items
such as cutting tools, and other consumables.
KEY
24
RM13006 - Process Control Methods
The control criteria for life/usage controls may be defined in many
ways. Control is often not simply a question of ‘how old’. Examples
of control criteria are: number of parts processed, total running
time, number of cycles, once opened use by date, weight of parts
processed and surface area processed.
Examples of control application include:
• A cutting tool has a maximum operating time. The tool life is
recorded on a machine readable chip. The machine program includes
code that checks the life of the tool prior to use. When cutting
tips are replaced and the tool is set a pre-setting operation
resets the readable chip to zero.
• A peening operation has media that is controlled based on the
total equipment running time. A timer is installed on the equipment
to indicate how close the process is to a media change. In addition
to this method of control, the process also has assessment for
media quality and uses test pieces to qualify the process for
correct operation.
• The concentration of a chemical etch bath is routinely maintained
with an auto-dosing system. However, once a month the entire system
is emptied, cleaned out, and refilled. To keep the planning of this
control simple this is done at a defined time regardless of use -
for example the morning of the first Monday in every month.
A life/usage limit may also incorporate a check and reset. For
example, a wearable item may be tested after a number of cycles and
found to have not reached a point where change is required. The
tool may be returned for use for a defined number of cycles. It
should be noted that this does not imply the tool will be run to
the point of failure.
The life/usage limits should ideally be determined to maximize the
process quality. Statistical studies and experiments will allow the
life to be optimized for other factors such as cost. These studies
may be performed on test pieces and scaled to the production
process. The life/usage limits should be validated however usually
at process qualification.
NOTE: These guidelines and examples do not replace specific process
standards or customer requirements that may exist to govern the
life/usage controls.
5.6 Control Charts for Attribute Data
Attributes are characteristics, or conditions characterized as
present or not-present or counted, typically through some form of
inspection or check. A number of charts may be used depending on
the attribute being studied.
NOTE: Process control via attributes is less effective than
variable methods. Some checking methods may provide attribute data
despite being variable in their nature. An example is a hole size,
that may be checked via variable methods or attribute (e.g., plug
gauge). If an attribute method were selected based on its speed and
simplicity, it should be on the basis that the process is proven
capable, because an attribute go/no-go gauge will not give early
warning of emerging issues, the way a variable gauge does. A robust
control strategy in the case of hole size may be to use a variable
tool measurement device such as a presetter to assure the quality
of the tool, and an attribute style plug gauge as a quick
conformance check but with a periodic sample taken from production
for variable measurement.
Figure 16 and Table 3 outline the basis for attribute control chart
selection.
25
Table 3 - Attribute Control Charts
Scenario
A process that observes discrete values, such as pass/fail,
go/no-go, present/absent, or conforming/non- conforming.
For example, a circuit card could consist of a number of solder
joints that either conform or do not conform to a set
standard.
Appropriate:
When it is important to control the number or % of defects over a
given time period, lot to lot, or unit to unit such as measuring
improvement over time, when go/no-go gauges are employed or when
visual inspections are used.
Not Appropriate:
Cannot be used for establishing process control or process
capability in the same way as variables data due to the scale not
being continuous. Measures of performance and stability can be
undertaken with a view to directing improvement activities, but
true process control needs to be done through process variables,
inputs, and foundational activities.
Not appropriate for rare events.
P Chart
Plot the percent defective - classifying product as good or bad
with changing or constant subgroup size.
Plot the monthly percent defective rate of a critical supplier;
plot the On Time Delivery performance of a critical supplier.
NP Chart
Plot the number defective - classifying parts as good or bad with
constant subgroup size.
A machining cell produces fuel control valves in standard lot sizes
of 50. Final Inspection performs a 100% inspection of the product
and plots the number of valves that are determined to be
nonconforming.
C Chart
Plot the count of defects based where the same area of opportunity
(constant subgroup size) exists.
An aerospace manufacturer produces one type of heat exchanger for a
customer. After vacuum braze a leak check is performed. A C chart
is used to plot the number of leaks requiring weld repair.
U Chart
Plot Defects Per Unit (DPU) based on counts and varying or constant
area of opportunity (changing or constant subgroup size) the
defects come from.
An aerospace manufacturer operating Production Part Approval
Process (PPAP) tracks the DPU on a monthly basis for all the
inspected PPAP packages. An accompanying Pareto Diagram suggests
the categories driving the DPU rate are poor PFMEAs, part marking
errors and poorly written Control Plans. Projects are established
to address these issues in order to reduce the overall DPU rate
shown on the Uchart.
26
Figure 17 - P Chart of Defectives
Example: The non-conformities from a series of batches of 50 parts
are monitored by the manufacturer on a P- Chart (Figure 17). The
manufacturer observes an overall defective rate of 2.2%. The
manufacturer concludes from the control chart that - despite the
variability from batch to batch - the rate of defectives is
statistically stable over time.
Figure 18 - P Chart with Varying Sample Sizes
2825221 91 61 31 0741
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00
1 .00
P Chart of % Yield
RM13006 - Process Control Methods
Example: A manufacturer monitors the yield % in their goods
produced per week on a P Chart (Figure 18). The weekly output
varies. The manufacturer concludes that the process yield is not
stable over time and seeks to understand the cause of the ‘bad’
weeks.
C Chart Example:
Figure 19 - C Chart
Example: A manufacturer produces a similar quantity of product each
day. The number of defects noted from a visual inspection area is
plotted on a C Chart (Figure 19) in order to understand the process
performance and behavior over time. In this case the supplier notes
a run of improved performance between days 12 and 22, and an
increase in defects on day 30. In reaction to the defect rate on
day 30 the manufacturer launches a problem solving activity.
NOTE: The use of NP charts and U charts are not illustrated in this
document. Implementing Six Sigma - Breyfogle 2003. ISBN
0-471-26572-1 may be referred to for explanation and examples of
their use.
The tests for special causes of variation for attribute control
charts are as follows:
• One or more points beyond a control limit.
• A run of eight or more points on the same side of the center
line.
• Six points in a row increasing or decreasing.
• Fourteen points in a row alternating up and down.
It is considered good practice to use a Pareto chart to support
attribute methods to allow further prioritization and insight on
the defects/defectives within the attributes plotted.
Assumptions for Attribute Charts
Attribute control charts have assumptions that need to be met in
order for the chart to function correctly. If these assumptions are
not met, then the control limits for the chart may be
incorrect.
A C chart works best with a minimum average defect rate per
subgroup of approximately 4, and a minimal number of zero values.
Where this is not met the chart’s usefulness will be
compromised.
2825221 91 61 31 0741
20
28
RM13006 - Process Control Methods
The calculations for the control limits for a C chart are based on
the Poisson distribution.
A P Chart and NP Chart assume that the defectives are randomly
dispersed and independent. In situations where defectives are
generated in clusters the limits generated may be too narrow to
reliably represent the common cause variation of the process.
The calculations for the control limits for the P Chart are based
on the binomial distribution.
Use of Variable Methods for Attributes
In some scenarios, attribute data may be monitored quite adequately
using variables control charts. For example, the Right First Time
measure of a manufacturing operation whilst based on an attribute
(good/bad), may be expressed as a ratio and plotted on a simple
individual’s control chart. In many cases an Individuals chart is
simpler to interpret and construct than attributes charts. Also of
consideration is the sample sizes used, that when large may result
in tighter control limits that result in the majority of data
showing as ‘out of control especially when defective items occur
naturally in clusters. The individual’s chart may help put the
process in a better perspective.
A Note on Rare Events
For rare/infrequent events, attribute control charts can give less
definitive results. The absence of events/defects/failures for
example will have an adverse effect on the control limits and
averages. In these cases, a time between failures may be a more
useful measure to track. Mean Time Between Failure (MTBF) is a
commonly used measure of equipment reliability for example.
Figure 20 - C Chart
29
RM13006 - Process Control Methods
Example - A manufacturer plots the failures of a machine tool,
counting how many failures were experienced over a 100 day period
(Figure 20). The chart is not very informative.
Figure 21 - Individuals Control Chart
Example: The manufacturer plots the time between failures for the
data on an Individuals chart (Figure 21). The chart is much more
informative. The average days between failures of 7.7 days and the
control limits can help guide the manufacturer on equipment
reliability and maintenance activity planning.
5.7 Visual Process Check and Checklist
A visual process check provides positive confirmation of goodness
either prior to allowing a process to run, or during its
operation.
The process checks need to become part of routine operation. The
personnel conducting the check will ideally understand the
importance of the check and also understand the reaction if the
check fails against the criteria. In many cases the check will
confirm that a particular step of the sequence has been done
correctly.
The checks may be conducted by a single person, however on
important items or high consequence failure items the method may
use two persons who jointly confirm that the correct condition is
achieved. An example of this approach is the standard pre-flight
checks that are undertaken by pilot and co-pilot when preparing for
a flight. One pilot calls out the check, the other performs the
check and confirms as correct, and then the first records the check
on a checklist before proceeding.
An example is shown in Figure 22.
To increase robustness, a “double scrutiny” and/or “buddy check”
may involve two personnel to positively confirm an action or result
of a check; or the check may be performed by someone independent of
the operation.
A single person check may have some inherent risks of error. A
preferred approach is automation or error proofing devices, (see
Section 5.1 - Error/Mistake Proofing). Prior to finalizing the
check, it is advisable to confirm the PFMEA risk level - as the
method of control relates to the detection score in the PFMEA
(refer to RM13004 for guidance).
1 0987654321
1 8
1 6
1 4
1 2
1 0
30
RM13006 - Process Control Methods
Pre-Operation Process Checklist Note to operator: Use this
checklist prior to execution of the process operation and sign off
each item below. Part No: 123456-78 Process operation number: 110
Run date: 08/12/2016 Process step name: Machine air holes
in Fuel/Air bracket
Check item number
Reaction (if Fail) Sign off (initial and date)
1 Health/Safety check Stop and isolate equipment. Contact cell
leader
2 Work instructions are latest version
Contact Manufacturing Engineer - obtain instructions
3 Machine asset care checks complete and correct
Raise issue with cell leader
4 Gages in calibration Contact Quality engineer
5 Fixture damage check Contact Manufacturing Engineer
6 CNC programme correct (as per instruction)
Contact Manufacturing Engineer
8 Etc.
5.8 First Piece Check
The objective of a first piece check is to validate the set-up and
quality of a process prior to the full production run. Alongside
other controls it serves to verify and confirm the integrity of the
production system (man, machine, fixture, tool, NC program, etc.)
at a point in time, and hence to avoid economic damage of
non-conformance (through timely action to ensure process
conformance).
Prerequisite to a first piece check should be the adherence and
confirmation that all other foundational control requirements are
met (e.g., calibration, machine tool diagnostics, tooling within
prescribed life limits, acceptable parameter settings, consumables
level, etc.) typically approved through positive confirmation (see
Section 5.7).
As a general rule, all manufacturing processes can be subject to
first piece inspection.
It may be called out in a control strategy:
• Whenever a new production lot is started.
• Following maintenance/repairs of measurement systems and
production equipment, as well as after software updates of
production equipment control systems.
• At a defined interval (e.g., at the start of each shift).
• When tools used to produce the component contour are replaced
(e.g., diamond rolls, profiled grinding/cutting wheels,
etc.).
31
First-piece checking/inspection may be independent from the
production method in a number of ways:
• Inspection by an operator other than the person having performed
the operation (two person rule); thus, avoiding risks due to bias
and other human factors.
• Inspection using another inspection tool or inspection method
(where possible); thus avoiding/highlighting measurement
discrepancies.
If independent inspection is to be used the method should be at
least as good as the production method, free from bias and have
adequate resolution to make the decisions valid. Tighter limits may
apply to first piece checks, and this should be considered when
evaluating such measurement equipment.
In order that the process is correctly judged as sufficiently good
to continue additional criteria may be applied. Such criteria
should have a rational and/or scientific basis for its application.
For instance, a process capability study or designed
experiments.
Example 1: A machined dimension with a known adequate level of
capability, achieved at first part check may be deemed sufficient
if within 50% of process tolerance; a measurement close to normal
limits of operation may result in adjustment and further
measurement to bring the process on target.
Example 2: A process with a tendency towards upward drift may have
a zone in the lower region of the specification band that provides
a standard for process acceptance of the first item. Continued
conformity as the process drifts naturally through use is provided
by a tool life/usage control. The zone has been determined through
a previous tool wear study. If the measurement is outside this
zone, the operator refers to a process guidance document
(referenced in the Control Plan) to determine appropriate action
(e.g., tool replacement, or adjustment to the tool life/usage
standard).
A first piece check strategy may extend to multiple parts -
depending on process risk and behavior. For example, a very large
batch of parts, a rapidly cycling process or high cost parts may
require inspection of the first five parts (Pre-Control may be
beneficial (see Section 5.4)).
It is good practice to require formal record keeping for approval
of first piece checks (e.g., a signature, and/or
countersignature/inspection report).
NOTE: The method should be used in conjunction with other methods
to make the control strategy robust to variations that may occur as
production continues.
NOTE: First Piece Check should not be confused with First Article
Inspection (FAI). For further information in FAI, refer to
AS9102.
5.9 Test Piece Evaluation
Some characteristics and properties that are created or changed
through processing may not be directly measurable other than
through destructive or damaging testing. Use of test pieces
processed alongside the product may help to determine the result of
the process and also its stability. These test pieces are tested
following processing to validate the products of the process and/or
confirm the effectiveness of the other process controls.
Such processes should be highly controlled through process
parameter controls and monitoring and may be categorized as ‘fixed
processes’ or ‘special processes’ often with regulatory control
requirements.
A test piece/coupon should be to a defined standard (thus
minimizing the variation in the test material itself).
In some instances a test piece may be operated within a first piece
check to qualify the process setup prior to the full production run
(see Section 5.8).
Examples of processes that use representative test pieces include
the following:
• Heat treatment operations
32
• Mechanical property testing using test bars
• Surface contamination coupons in heat treat or thermal
processes
• Coupons determining material removal rates in etch and
electro-polish processes
• Cast coupons determining chemical analysis of parts from
melts
• A forging that has extra material outside the finished part
envelope that will be removed for testing
Once a result has been obtained from a test piece the result can be
analyzed with a variety of process control tools such as control
charts (variable and attribute) and run charts.
Acceptance of process results by the use of test specimens or
coupons is typically approved and agreed to by the customer.
NOTE: There may be regulatory, customer, product specifications,
and other requirements that address the extent to which test piece
evaluation, or requirements are permissible and established as part
of process qualification. Equivalence between test piece and
physical product should be understood.
6. PROCESS CAPABILITY INDICES
Process Capability is the ability of a process/product to
consistently meet a specification or customer requirement.
Various indices are computed to assess the Process Capability of a
given product characteristic.
The definition and calculation of these is often misunderstood and
thus misinterpreted. The methods described within this section are
based on recognized industry methods. Software tools such as
Minitab calculate capability in line with these methods and
additionally cater for some specific scenarios that exist such as
batch processing where information may be sought about the
capability both within and between batches of production.
Process Capability can be assessed for Variable and Attribute
data.
6.1 Fundamentals for Variable data
At the heart of capability for variable data, is the need to manage
process variation and location to align with customer specification
to ensure that requirements can be continually met.
Variability of the process is calculated through statistical
methods; these methods aim to anticipate the total process
variation rather than just the range seen in the data collected for
the capability study. A process spread of six standard deviations
is used to represent this spread. This six standard deviation range
theoretically covers 99.73% of the area under a normal distribution
curve. Data is assumed to be normally distributed (symmetrical,
bell shaped).
Many processes have a tendency - even naturally - to periodic drift
or shift. Therefore, borderline capability is not desirable for
either supplier or customer. A capability of 1.33 is often seen as
a minimum to assure continued conformance while allowing for minor
process drift. However, depending on the process, a higher level of
capability may be required. Products with large numbers of
characteristics that cannot be controlled independently may require
some additional margin for small drifts that may occur through
production.
For any capability calculation to be reliable, it is important that
the process be in a state of statistical control thus behaving in a
predictable manner - otherwise any perceived goodness may be
short-lived. It is possible for a process with a ‘good’ capability
index to be producing non-conforming product if a state of control
is not reached. Process stability is therefore a prerequisite to
capability calculation.
33
Capability Indices Cp and Pp
Cp and Pp indices are simply a ratio of specification width to
process variation thus calculating the ‘potential of the process if
centered’. The indices increase if variation is reduced. A Cp or Pp
of exactly 1.0 indicates that six standard-deviations of process
variation match the width of the specification. Such a process if
centralized within the specification would be intolerant to even
minor drift over time. Not an ideal situation.
Figure 23 - Process Capability Index Cp/Pp
The process shown in Figure 23 has a Cp or Pp>1. The process is
less variable than allowed by the specification.
Cp and Pp use different methods for estimating process variability.
Cp uses ranges of the data within subgroups (or difference between
individual values) to estimate the process variation. A statistical
constant d2 is used to adjust for the subgroup size. This method
estimates the standard deviation of the process rather than
calculating by the more involved ‘root sum of squares’ method
(which is used to calculate Pp).
The average range over d2 method generates the estimate denoted by
sigma hat (Equation 1).
(Eq. 1)
The root sum of squares method generates the standard deviation
denoted by s (see Equation 2).
= ∑ (−)2 =1 −1
(Eq. 2)
Cp is typically used to assess short term (within subgroup)
capability whereas Pp is used to assess longer term (overall)
capability.
LSL USL
RM13006 - Process Control Methods
These are incorporated into the formulae (Equations 3 and 4) as
follows:
(Eq. 3)
(Eq. 4)
For a stable continuous process behaving in a random manner, Cp,
and Pp calculations can be expected to deliver similar
values.
Capability Indices Cpk and Ppk
In order to estimate the likely performance - against a
specification - of the process Cpk and Ppk indices are used. These
indices are similar ratios to Cp and Pp but additionally take into
account the process location.
Cpl and Cpu, and Ppl and Ppu measure capability against each of the
specification limits. The ‘l’ and ‘u’ indices will be equal only if
the process is centered. The Cpk or Ppk is the smaller of the upper
and lower values.
The ‘l’ and ‘u’ indices can be used to determine how the process is
located relative to specifications, however, a visual assessment of
the capability histogram is usually preferred to understand this
situation.
The formulae for these indices is shown (Equations 5 to 10).
(Eq. 5) (Eq. 8)
(Eq. 6) (Eq. 9)
(Eq. 7) (Eq. 10)
LSL USL
3σ 3σ
RM13006 - Process Control Methods
The process shown in Figure 24 has a Cp of approximately 1.0 but
due to being too close to the upper specification limit (with the
tail of the distribution outside it) the Cpk is <1 If the
process average is outside the specification, the Cpk will be
negative.
NOTE: It will not be possible to calculate Cp or Pp indices for
processes with unilateral (single sided) tolerances as the
tolerance width cannot be defined. However, Cpk and Ppk can be
calculated from the Cpl/Ppl or Cpu/Ppu (whichever can be
calculated).
Table 4 provides guidance on approximate expected performance
levels at various levels of Process Capability. The performance
rates assume a process perfectly centered between two specification
limits. This table assumes a normal distribution.
Table 4 - Expected Performance for Cpk
CPK Sigma Level
(assumes “centered” process)
%YIELD (assumes “non-
REJECT RATE Parts Per Million
(assumes “non-centered” process with 1.5 sigma
shift) 0.50 1.5 86.64 133614 49.87 501350 0.67 2.0 95.45 45500
69.16 308417 0.80 2.4 98.36 16395 81.59 184108 1.00 3.0 99.73 2700
93.32 66811 1.20 3.6 99.97 318 98.21 17865 1.33 4.0 99.994 63
99.377 6227 1.50 4.5 99.9993 6.8 99.865 1350 1.67 5.0 99.99994 0.57
99.977 232 1.80 5.4 99.999993 0.067 99.9952 48 2.00 6.0 99.9999998
0.002 99.99966 3.4
For the Cpk and Ppk calculations in this section, the process is
assumed normally distributed. If the data are non-normal (skewed
for example) alternative methods can be used (see Section 7 -
Guidance for Non-Normal Data).
NOTE: The descriptions in this section are fundamentals. Some
additional methods for specific situations are described in Section
9 - Scenarios requiring specific analysis methods.
Some characteristics may benefit from being ‘targeted’ to a
particular nominal value. These are usually characteristics that
influence performance of the product, that have a loss associated
with deviation from target even within the specification. These
characteristics may have additional requirements communicated by
the customer. For these types of characteristics, it is important
to examine the location of the process relative to this target. It
should be noted that due to the calculation methods, high Cpk/Ppk
indices do not necessarily imply the process is on target as their
calculations use the distance of the process mean to the
specification limits. The nominal location is not considered in the
calculation.
A target-based process capability index (Cpm) may be used in these
situations. Cpm is not covered in this RM but is described in
statistical texts and provided in statistical software
applications.
Data Collection and Sample Size Considerations (Added)
A process capability index can mathematically be produced on any
dataset with two samples or more. However, the confidence one would
have in the capability metric will depend on the amount of data
that has been gathered and how representative it is of the study in
question.
The studies that are more likely are the initial process study and
the ongoing performance study.
36
Consideration should be given to the following
• The items and the period over which the data will be
collected
• The method of collection (either manual or automatic)
• The inspection method
• Interim review to act on obvious signals and trends (prior to
full st