Professional Documents
Culture Documents
Statistical Process Control SPC: Quality Management in The Bosch Group - Technical Statistics
Statistical Process Control SPC: Quality Management in The Bosch Group - Technical Statistics
•••••••••
•••••••••
2021-02-10 - SOCOS
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Booklet No. 7
2021-02-10 - SOCOS
Edition 11.2020
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
All minimum requirements specified in this booklet for capability and performance criteria correspond
to the status at the time of printing (issue date). [CDQ 0301] is relevant for the current definition.
2021-02-10 - SOCOS
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Table of Contents
1 Introduction ..................................................................................................................................... 5
2 Quality Control ― Application of the Control Chart ........................................................................ 7
2.1 Basic Rules ............................................................................................................................. 7
2.2 Necessary Documents............................................................................................................ 8
2.3 Short Control Loop ― Directly at the Machine ..................................................................... 8
2.4 Reaction Plan ......................................................................................................................... 9
2.5 Further Measures ................................................................................................................ 11
2.5.1 Sorting Inspection ................................................................................................... 11
2.5.2 Decision about the sorted Parts ............................................................................. 11
2021-02-10 - SOCOS
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
1 Introduction
Statistical Process Control (SPC) is a procedure for open or closed loop control of manufacturing
processes, based on statistical methods.
Random samples of parts are taken from the manufacturing process according to process-specific
sampling rules. Their characteristics are measured and entered in control charts.
According to [CDQ 0301], the use of the Solara® / qs-STAT® / procella® / O-QIS® software package is
prescribed within Bosch. It calculates capability and performance indices and operates control charts
in accordance with the used evaluation strategy.
Statistical indicators are calculated from the measurements and used to assess the current status of
the process. If necessary, the process is corrected with suitable actions. Statistical principles must be
observed when taking random samples.
measure characteristic
- General improvement
- Ongoing care Maintain and evaluate quality
control chart
Shortest possible
Production process control loop
influenced by:
The control chart method was developed by Walter Andrew Shewhart (1891-1967) in the 1920´s and
described in detail in his book “Economic Control of Quality of Manufactured Product” [Shew 1931].
SPC is used at RB in a common manner in all divisions. The procedure is defined in [CD 00301] in
agreement with all business divisions.
Current questions on use of SPC and related topics are discussed in a work group (Functional Council).
Results that are helpful for daily work and of general interest can be summarized and published as QA
Information.
SPC is an application of inductive statistics. Not all parts have been measured, as would be the case for
100% inspection. A small set of data, the random sample measurements, is used to estimate
parameters of the entire population.
In order to correctly interpret results, we have to know which mathematical model to use, where its
limits are and to what extent it can be used for practical reasons, even if it differs from the real situation.
Statistical process control is based on the concept that many inputs can influence a process. The “5 M´s”
– man, machine, material, milieu, method – are the primary groups of inputs. Each “M” can be
subdivided, e.g. milieu in temperature, humidity, vibration, contamination, lighting, ....
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Despite careful process control, uncontrolled, random effects of several inputs cause deviation of
actual characteristic values from their targets (usually the middle of the tolerance range).
The random effects of several inputs ideally result in a normal distribution for the characteristic. Many
situations can be well described with a normal distribution for SPC.
A normal distribution is characterized with two parameters, the mean 𝜇 and the standard deviation 𝜎.
Fig. 1.2
µ - 3 µ - 2 µ - 1 µ µ + 1 µ + 2 µ + 3
In SPC, the parameters 𝜇 and 𝜎 of the population are estimated based on random sample measure-
ments and these estimates are used to assess the current status of the process.
An essential advantage of quality control charts is the fact that the original data and usually also the
conditions under which they were determined are documented and remain available. In [Shew 1939]
Shewhart explicitely lists the advantages of the original data.
The measurement results are represented by numerical values (and units of measurement).
The original data are documented in the order in which they were determined.
The conditions under which the data were determined and the measurement process are
reproducible.
Especially the chronological order of the data is an essential information which is lost during any kind
of further data compression.
NOTE: [Booklet 3] shows three different representations of data with increasing values, a random sequence of
values and a sawtooth-like progression, which differ only in their chronological order. The corresponding
distribution model is always the same.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
It cannot be completely avoided that in this chapter also topics are touched, which go clearly beyond
the topic of this Booklet on SPC. Meant are for example
2021-02-10 - SOCOS
2. When adjusting the machine or the tool, it is unavoidable to check continuously until the correct
centered location (according to step 1) is achieved. The inspection interval begins then with the entry
of the first measurement results (results of the first sample) in the control chart as first “OK sample”.
Since the parts, produced from the beginning of the intervention or adjustment until achievement of
the center position, can show greater deviations from the target value, they have to be kept separately
and be sorted out, if necessary.
3. After each inspection interval a sample is taken, the considered characteristic is measured and the
measurement results are entered in the control chart.
4. In case of a tool breakage or other reasons for a process intervention one must assume, that the
parts produced since the beginning of this inspection interval are potentially defective/nonconforming.
They are treated according to Section 2.3.
6. All correctives measures on the process (e.g. according to Chapter 10) and on the parts (Section 2.5)
are documented.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Hint: Rework and repair processes as well as sorting inspections involve further quality risks, e.g. due
to
• mix-ups of products or components incl. mixing of variants,
• damage of parts by disassembly o rhandling,
• causation of further failure modes, e.g. flashing with wrong software,
• logistics mistakes, quantity deviation, wrong packaging,
• exceedance of maximum storage periods due to violation of the “First In, First Out” (FIFO)
principle.
The creation and maintenance of reaction plans can be supported by providing plant, area and process-
specific catalogs. Within the scope of the obligatory use of Q-DAS software, the maintenance of action
plans/catalogs and the documentation of measures in the software must be ensured.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Reaction in SPC
2021-02-10 - SOCOS
yes
b
yes SPC,
Single value out of Sort 100 % back control chart,
tolerance zone? to last sample reaction plan
no
c no Correction in SPC,
Statistic beyond case of changed control chart,
alarm limits? process reaction plan
behavior
yes
d Documentation
Further actions
Fig. 2.4: Illustration of the reaction when the control chart responds
a) Intervention necessary?
After each entry in the SPC control chart, it has to be checked whether intervention is necessary. An
"intervention" is an appropriate reaction to a statistically unusual event, i.e. one of the control criteria
listed below is met (e.g. a control limit has been exceeded).
Control criteria:
A single value is outside the tolerance zone.
A sample’s statistic (e.g. 𝑥̅ or 𝑠) is outside the control limits.
Optional: An unusual sequence of points is observed ― rule of seven (i.e. seven points in a row
on one side of the average line, seven measurement values with continuously increasing or
decreasing tendency). For processes with systematic changes of the location ( 𝑥̅ ), e.g. for
processes with a trend, this rule cannot be applied.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
In this case, it has to be checked whether the process behavior has actually changed, and, if this is
confirmed, after determination of the cause (according to the reaction plan) the original process state
has to be restored.
NOTE: Due to the operation characteristics of the control charts there is an error probability of about 1 % for
the exceedance of control limits, although the process behavior is unchanged (s. Chapter 13).
d) Sorting inspection
All parts produced since the last sampling are 100 % inspected and sorted (s. Section 2.5).
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
The measures are documented. Details are regulated in [CDQ 0503] and [CDQ 0509].
NOTE: In the broadest sense, a sorting inspection is rework of products with unclear status.
To avoid such risks, appropriate safeguarding measures and approvals are necessary. In particular,
effective measures must prevent
further processing,
unintended use, and
the delivery
of nonconforming or incorrectly labeled products or products whose status is not identifiable
(potentially nonconforming products). See [CDQ 0503]. For this purpose, blocking, identification,
sorting and sorting-out must be clearly regulated.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
NOTE: Terms like analysis, tracking and evaluation are often used synonymously in everyday language. In the
context of this booklet they have slightly different meanings.
Analysis: Decomposition, systematic investigation of an issue and its causes
Monitoring: tracking, observation, surveillance, ongoing investigation, activities to find out something
Evaluation: Summary, condensation, presentation of the essentials, interpretation, evaluation of data,
derivation of measures
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Example:
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
3 Planning
The planning takes place within the framework of “Management of Characteristics“ according to
CD 00301. Responsibilities are defined there.
The definition of the inspection strategy requires, among others, the analysis of the cause-effect
relationships between process characteristics and product characteristics, taking into account distur-
2021-02-10 - SOCOS
bance / noise and control variables. Knowledge of these relationships is a prerequisite for the process
to be influenced (controlled) with regard to the inspected characteristic (product characteristic).
Usually, production equipment can directly be influenced so that the inspected characteristic (product
characteristic) can be changed in the desired way (short control loop).
If a characteristic cannot be measured directly, it may be possible to determine a substitute inspection
characteristic that is known to be related to it.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
NOTE: IMT equipment according to the definition in “Terms and Definitions” is subject to control. IMT equipment
serves for the proof of conformity in the framework of “Management of Characteristics”, for instance.
Software, used for the creation of measurement results is also subject to this mandatory control. Such software
must be validated in appropriate form.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
In short-term studies (studies of machine capability), characteristics of product parts are evaluated
which have been manufactured in a continuous production run in an uninterrupted sequence, so that
possibly only the influence of the machine is active.
In contrast, the parts to be measured originate in studies of long-term capability (process capability)
from a larger, more representative period for the series production, so that possibly all influences on
the process, which are to be expected, take effect.
Particularly during series rump-up, there are frequently not enough product parts available nor can
enough parts be taken out of the manufacturing process over a sufficient period of time. Nevertheless,
as an alternative, or in addition to the machine capability, at least a preliminary conclusion about the
2021-02-10 - SOCOS
expected manufacturing process capability can be demanded (see “Initial Process Capability” [AIAG
PPAP] and “Preliminary process capability” [VDA-4]). In this case, a short-term study is conducted,
which can differ from the long-term study in the following points.
Type of sampling: The parts to be studied can be taken out of the manufacturing process in
shorter intervals, if necessary in extreme cases, one immediately after the other.
Number of parts: It is permissible if there are not sufficient parts available, to take less than the
required 125 parts for the long-term study.
Limits for capability and performance indices: The increased limit of 1.67 applies if there are
more than 125 parts. If there are less than 125 parts, the limit is raised depending on the number
of parts at the same value as in the long-term study with reduced quantities.
Designation of statistical indices: Capability indexes are designated with 𝐶𝑝−𝑆𝑇 und 𝐶𝑝𝑘−𝑆𝑇 and
performance indexes with 𝑃𝑝−𝑆𝑇 und 𝑃𝑝𝑘−𝑆𝑇 (short-term).
NOTE: The general definition of SPC does not presume process capability. However, if capability is not given,
then additional actions are necessary to ensure that the quality requirements for manufactured products are
fulfilled.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
If a characteristic can be measured, then a control chart for continuous characteristics must be used.
Normally the 𝑥̅ − 𝑠 chart with sample size 𝑛 = 5 is used.
NOTE: In the past, also control charts for discrete characteristics have been used, where the number or
proportion of nonconforming units (np chart, p chart) or the number or proportion of defects per unit (c chart,
u chart) is used as the characteristic to be controlled.
Due to the following disadvantages, control charts for discrete characteristics are no longer up-to-date:
The existence of defects is a prerequisite for the functioning of these types of charts. Thus, they stabilize
rather than eliminate the undesirable condition.
For small error proportions, unrealistic sample sizes of several hundred parts are required.
Due to the large random variation ranges of the characteristics, the response sensitivity is limited, i.e.
until significant changes are detected, more entries are required than with a control chart for continuous
characteristics.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
yes
Fig. 3.5.1: Flow chart for selecting the quality control chart
When using the Shewhart chart, tolerance-related control limits can be used to assist in decision
making in the framework of a sorting procedure (refer to Section 2.4). These limits are then referred
to as alarm limits.
It is not permissible to draw tolerance-related control limits (alarm limits) as lines in Shewhart control
charts.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Properties
Average chart For the same sample size, better sensitivity or selectivity than the 𝑥̃ − 𝑅
and the individual data charts with respect to a shift of the average
location as well as to an increase of the standard deviation 𝜎.
For 𝑛 ≥ 4 insensitive to deviation of the population from the normal
distribution
Preferred chart for use with computers
Chart for moving Can also be used for destructive or complicated testing
indices
This chart has a delayed reaction to suddenly occurring process
2021-02-10 - SOCOS
changes.
The control limits are the same as for “normal” average charts
Average chart with Preferred chart for processes with systematic changes of the average
extended limits
Acceptance control Due to the tolerance-related control limits, the process is controlled
chart worse than with process-related control limits.
Individual data chart Presentation of all individual data from the random sample, so the
chart is especially easy to understand and to fill out, lending it suitable
for simple documentation of the manufacturing process.
No calculation is necessary to use the chart.
The individual data chart reacts to both a shift of the average location
as well as to an increase of the standard deviation of a characteristic.
Compared to the 𝑥̅ − 𝑠 chart and the 𝑥̃ − 𝑅 chart less sensitivity or
selectivity with respect to a shift of the average location as well as to an
increase of the standard deviation 𝜎.
A decrease of 𝜎 is not indicated by going below a limit value.
More sensitivity to deviations of the population from a normal
distribution than with the average chart.
Potential confusion of users from the fact that, for example, the 𝑈𝐶𝐿
increases with larger sample sizes n.
Evaluation, such as calculation of 𝐶𝑝𝑘 , is more complicated, because the
original data is usually not recorded numerically (e.g. only points on the
graph).
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
2. Sampling intervals for ongoing process control can be defined in the following manner, selecting
the shortest interval appropriate for the process.
3. Take a final sample at the end of a series, before switching to a different product type, in order
to confirm process capability until the end of the series.
Note: The inspection interval is defined based on quantities (or time periods) in a manner that detects
process changes before defects are produced. More frequent testing is necessary for unstable
processes.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
NOTE: The first parts produced after the set-up process can hardly be called a random sample. But here it is
rather a matter of proving that the process is mastered and that all products meet the requirements.
2021-02-10 - SOCOS
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
NOTE: If, in addition to the control limits LCL and UCL, the characteristic limits LSL and USL ("tolerance limits")
are also displayed in the control chart, there is a risk that the statistics are erroneously compared with LSL and
USL (s. also Section 4.3.2).
Note: In order to be able to sort out bad parts after a control limit was exceeded, it is essential to be
able to trace the production quantity manufactured since the last random sample was taken.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
coded form).
Fig. 4.3.1:
64,0
63,5
Illustration of the “rule of seven”
UCL
63,0
Note: The rule of seven, in the simplest
Mean x_bar
62,5
62,0
case (seven points on one side of the
average line), is based on the assumed
61,5
random process behavior and the at least
61,0
LCL approximately normal distribution of the
60,5
random sample statistics. Of course, it
60,0 cannot be used for processes with
0 5 10 15 20 25
systematic changes of the average.
Sample No.
Since for average charts and stable process behavior, each point has a 50 % probability of being above
or below the average line, independent of the previous point, the probability of seven points in a row
being on one side of the average line is (0,5)7 = 0,008 < 1 %.
This is why it is called an “unusual sequence of points”.
Use of the rule of 7 in this case (the same as comparing a random sample statistic with the control
limits) is a statistical test with a proba-bility of error (type I error) of 1 %.
For the given assumptions, the probability that at least 7 points in a row form a continuously increasing
or decreasing trend is also very low.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
USL
Fig. 4.3.3
UCL
This figure illustrates the proportions
for an average chart with 𝑛 = 5.
When 𝐶𝑝 = 1.33 the distance of the
control limits 𝑈𝐶𝐿 − 𝐿𝐶𝐿 corres-
LCL ponds to about 30 % of the
tolerance.
LSL
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
Comparison of process-related control limits 𝐿𝐶𝐿 and 𝑈𝐶𝐿 with tolerance-related control limits (alarm
limits) per Chapter 8 allows us to assess the amount of variation and the quality capability of a process.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
4.4 Documentation
Define when the evaluation results and/or overall summaries are to be submitted. Define the type and
location for archiving. Archiving durations are defined in [CD 02981], for instance.
It is important to assess the graphs of the measured values themselves, the means and the standard
deviations. Their curves can often deliver information on process performance characteristics (e.g.
trends, cyclical variations).
In general, the parameters of the resulting process distribution are already known from the capability
study (Section 3.4.).
5.2 Disturbances
If non-random influences (disturbances) occur frequently during the trial run, then the process is not
stable (not in control). The causes of the disturbances must be determined and eliminated before
process control is implemented (repeat the trial run).
Hint: Due to the software support available today, the procedures in connection with the preparation,
management and evaluation of control charts allow a better adaptation to process-specific conditions
(e.g. process models) than is possible with manual calculation procedures. However, they also
inevitably bring along higher requirements regarding statistical knowledge and the handling of
statistical software. The requirements for personnel and training must take this into account.
In every division and every plant a fully trained SPC specialist should be available as contact person.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
or
b) 𝑠̅ 1
𝜎̂ = 𝑎 with 𝑠̅ = 𝑚 ∙ ∑𝑚
𝑗=1 𝑠𝑗 0.552 +0.452 +⋯+2.552
𝑛 𝜎̂ = √ = 1.41
25
𝑛 𝑎𝑛
3 0.89
5 0.94
See Chapter 9 for additional values 7 0.96
or
c) 𝑅̅ 1 1+1+⋯+6
𝜎̂ = with 𝑅̅ = ∙ ∑𝑚
𝑗=1 𝑅𝑗 𝑅̅ = 25
=2.96
𝑑𝑛 𝑚
𝑛 𝑑𝑛
3 1.69
5 2.33
See Chapter 9 for additional values 7 2.70
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Relation between the variance (standard deviation 𝜎𝑥 ) of the single values (original values, individuals)
𝜎
and the variance (standard deviation 𝜎𝑥̅ ) of the mean values: 𝜎𝑥̅ = 𝑛𝑥
√
2021-02-10 - SOCOS
80
UCL
70
Characteristic value x
60
LCL
50
40
0 5 10 15 20 25
Fig. 7: Schematic about the functionality of a mean chart with 𝑛 = 5. In order to illustrate the relation-
ships, the mean values (points) as well as the individual values (crosses) from each sample are
represented. The means like the individual values show a variation around the process average 𝜇;
1
however, the variation of the means is smaller by the factor than that of the individual values.
√5
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
2.58 2.58
𝑈𝐶𝐿 = 𝐶 + ∙ 𝜎̂ 𝑈𝐶𝐿 = 62 + ∙ 1.35 = 63.6
√𝑛 √5
Average Chart
2.58 2.58
𝐿𝐶𝐿 = 𝐶 − ∙ 𝜎̂ 𝐿𝐶𝐿 = 62 − ∙ 1.35 = 60.4
√𝑛 √5
Also possible for normal distributions:*
𝑈𝐶𝐿 = 𝐶 + 𝐴∗ ∙ 𝑠̅ 𝑈𝐶𝐿 = 62 + 1.23 ∙ 1.27 = 63.6
𝐿𝐶𝐿 = 𝐶 − 𝐴∗ ∙ 𝑠̅ 𝐿𝐶𝐿 = 62 − 1.23 ∙ 1.27 = 60.4
2021-02-10 - SOCOS
2.58 2.58
𝑈𝐶𝐿 = 𝐶 + ∙ 𝑐𝑛 ∙ 𝜎̂ 𝑈𝐶𝐿 = 62 + ∙ 1.2 ∙ 1.35 = 63.9
√𝑛 √5
Median Chart
2.58 2.58
𝐿𝐶𝐿 = 𝐶 − ∙ 𝑐𝑛 ∙ 𝜎̂ 𝐿𝐶𝐿 = 62 − ∙ 1.2 ∙ 1.35 = 60.1
√𝑛 √5
Also possible for normal distributions:*
𝑈𝐶𝐿 = 𝐶 + 𝐶𝐸 ∙ 𝑅̅ 𝑈𝐶𝐿 = 62 + 0.59 ∙ 2.96 = 63.7
𝐿𝐶𝐿 = 𝐶 − 𝐶𝐸 ∙ 𝑅̅ 𝐿𝐶𝐿 = 62 − 0.59 ∙ 2.96 = 60.3
Note: Use of the median-R chart is only appropriate when charts are manually filled out,
without computer support.
𝑈𝐶𝐿 = 𝐶 + 𝐸𝐸′ ∙ 𝜎̂ 𝑈𝐶𝐿 = 62 + 3.09 ∙ 1.35 = 66.2
Individual Data
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
The sample means of such skewed distributions and small 𝑛 are no longer necessariy normally
distributed. This also applies to the mixture distribution. It can be sensible then to use a Pearson chart.
2021-02-10 - SOCOS
Compared to the corresponding Shewhart chart, this has the advantage of slightly wider control limits.
They consider the skewness of the distribution of the original values and are therefore asymmetrical
to the mean value. However, the disadvantage is that the control limits are more complicated to
calculate and can practically only be done with the help of a computer.
In this case, a distribution from the family of Pearson distributions is best fitted to the sample means.
The control limits of the Pearson chart then correspond to the 0.5% and 99.5% quantiles of this
distribution, for instance.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
3 7 4 𝑥̅1 = 4.7
3 7 4 9 𝑥̅2 = 6.7
3 7 4 9 2 𝑥̅3 = 5.0
3 7 4 9 2 8 𝑥̅4 = 6.3
Of course, moving averages calculated in this manner are not mutually independent. That is why this
chart has a delayed reaction to sudden process changes. The control limits correspond to those for
“normal” average charts:
2.58 2.58
𝐿𝐶𝐿 = 𝐶 − ∙ 𝜎̂ 𝑈𝐶𝐿 = 𝐶 + ∙ 𝜎̂
√𝑛 √𝑛
This approach for moving sample measurements can also be applied to the variation, so that an 𝑥̅ − 𝑠
chart with a moving average and moving standard deviation can be used.
After intervention in the process or process changes, previously obtained measurements may no
longer be used to calculate moving indices.
NOTE: In case of the der 𝑥̅ chart with moving average there is a greater probability for the occurence of “unusual
sequences of points”. This is why the rule of seven must not be applied to this chart type.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Due to this recursive calculation all previous sample results 𝑥𝑖 are contained in the test statistic 𝑦𝑖 . The
2021-02-10 - SOCOS
larger λ is, the less the effect of previous values 𝑥𝑖 on the current test statistic 𝑦𝑖 .
Example: 𝑦4 = 𝜆 ∙ 𝑥4 + (1 − 𝜆) ∙ 𝜆 ∙ 𝑥3 + (1 − 𝜆)2 ∙ 𝜆 ∙ 𝑥2 + (1 − 𝜆)3 ∙ 𝜆 ∙ 𝑥1 + (1 − 𝜆)4 ∙ 𝐶
For 𝜆 = 0.2 is (1 − 𝜆)3 ∙ 𝜆 ≈ 0.1. The third-last sample value is therefore only considered with a
weight of approx. 10 %.
The control limits of this chart are recalculated for each sample value.
𝜆 𝜆
𝐿𝐶𝐿 = 𝐶 − 2,58 ∙ √2−𝜆 ∙ [1 − (1 − 𝜆)2∙𝑖 ] ∙ 𝜎̂ 𝑈𝐶𝐿 = 𝐶 + 2.58 ∙ √2−𝜆 ∙ [1 − (1 − 𝜆)2∙𝑖 ] ∙ 𝜎̂
They are therefore variable in time. Their distance from each other initially increases, but asymp-
totically approaches a constant value dependent on 𝜆:
𝜆 𝜆
𝐿𝐶𝐿 = 𝐶 − 2.58 ∙ √2−𝜆 ∙ 𝜎̂ 𝑈𝐶𝐿 = 𝐶 + 2.58 ∙ √2−𝜆 ∙ 𝜎̂
After every intervention the control limits are reset to their initial values and the numbering of the
samples starts again at 𝑖 = 1.
The calculation of the control limits presupposes a known and constant process standard deviation 𝜎̂.
The number 2.58 is the 99.5% quantile of the standard normal distribution.
A typical value of λ is λ = 0.2. Due to the expression (1 − λ)2∙i under the square root the control
λ
limits reach their respective constant values EG = C ± 2.58 ∙ √2−λ ∙ σ
̂ after a few samples, and the
faster, the greater λ is. This is the reason why often these constant control limits are used. On the
other hand, the smaller λ is, the greater the sensitivity of the EWMA card to small mean value
deviations.
Advantage Disadvantages
The EWMA chart reacts sensitively to The test statistic is not immediately
small changes of the average location. understandable.
When variable control limits are used,
this can be irritating for the user.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
The control limits to monitor the variation Note: Formula a) must be used in
(depending on 𝑛 ) relate to 𝜎̂ and 𝑠̅ and the case of moving s calculation.
likewise 𝑅̅ (= “Central line”). Calculation of 𝜎̂ per Section 6.2 a).
∗
𝑈𝐶𝐿 = 𝐵𝐸𝑜𝑏 ∙ 𝑠̅ 𝑈𝐶𝐿 = 2.05 ∙ 1.27 = 2.6
2021-02-10 - SOCOS
′ ′ ∗ ∗
𝑛 𝐵𝐸𝑢𝑛 𝐵𝐸𝑜𝑏 𝐵𝐸𝑢𝑛 𝐵𝐸𝑜𝑏 𝐷𝐸𝑢𝑛 𝐷𝐸𝑜𝑏
3 0.07 2.30 0.08 2.60 0.08 2.61
5 0.23 1.93 0.24 2.05 0.24 2.10
7 0.34 1.76 0.35 1.88 0.34 1.91
See Chapter 9 for additional values
The standard deviations s cannot become less than zero. They are subject to a distribution whose
density function is strongly asymmetrical (right-skewed) for small sample sizes. The mean value 𝑠̅
which is entered as a dashed line in the chart diagram is therefore below the middle between 𝐿𝐶𝐿 and
𝑈𝐶𝐿. As a tendency, slightly more values (points) can be expected below this line than above. This also
applies when using the ranges R and the average range 𝑅̅.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
7.2 Control Limits for Processes with Systematic Changes in the Average
If changes of the mean need to be considered as a process-specific feature (trend, lot steps) and it is
not economical to prevent such changes of the mean, then it is necessary to extend the “natural
control limits”.
The procedure for calculating an average chart with extended control limits is shown below.
The overall variation consists of both the “inner” variation of the random samples and of the “outer”
variation between the random samples.
𝑚−1
with 𝑁 = 𝑛 ∙ 𝑚
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
𝑘𝐴 is called the delimitation factor (for calculating 𝜎̂ refer to the formula below).
Control limits for corresponding 𝑠 or 𝑅 charts are calculated according to Section 7.1.3.
NOTE: Tolerance-related control limits can be used to support the decision according to Section 2.4. These limits
are then called alarm limits (LAL, UAL).
Average chart
1 𝑝 = 1 % and 𝑃𝐴 = 99% yield:
𝑘𝐴 = 𝑢1−𝑝 + ∙ 𝑢𝑃𝐴
√𝑛
𝑛 𝑘𝐴 𝑛 𝑘𝐴
3 3.67 7 3.21
4 3.49 8 3.15
5 3.37 9 3.10
6 3.28 10 3.06
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
9 Tables
Table 1 Table 2a
Constants to estimate the Constants to estimate the control limits of
standard deviation (Section 6.2) the individual data chart (Section 7.1.1)
n 𝑎𝑛 𝑑𝑛 n 𝐸𝐸′
2 0.798 1.128 2 2.807
3 0.886 1.693 3 2.934
4 0.921 2.059 4 3.023
5 0.940 2.326 5 3.089
6 0.952 2.534 6 3.143
7 0.059 2.704 7 3.188
2021-02-10 - SOCOS
Table 2b
Constants to determine the control limits (Section 7.1)
Location Variation
𝑥̃ − 𝑅 𝑥̅ − 𝑠 𝑥−𝑅 s Chart R Chart
∗ ′ ′ ∗ ∗
n 𝐶𝐸 𝐴 𝐸𝐸 𝐵𝐸𝑢𝑛 𝐵𝐸𝑜𝑏 𝐵𝐸𝑢𝑛 𝐵𝐸𝑜𝑏 𝐷𝐸𝑢𝑛 𝐷𝐸𝑜𝑏
2 1.614 2.283 2.487 0.006 2.807 0.008 3.518 0.008 3.518
3 1.019 1.678 1.734 0.071 2.302 0.080 2.597 0.080 2.614
4 0.683 1.398 1.468 0.155 2.069 0.168 2.245 0.166 2.280
5 0.593 1.225 1.328 0.227 1.927 0.242 2.050 0.239 2.100
6 0.471 1.105 1.240 0.287 1.830 0.302 1.924 0.296 1.986
7 0.437 1.015 1.179 0.336 1.758 0.350 1.883 0.341 1.906
8 0.371 0.944 1.133 0.376 1.702 0.390 1.764 0.378 1.846
9 0.354 0.886 1.098 0.410 1.657 0.423 1.709 0.408 1.798
10 0.311 0.837 1.069 0.439 1.619 0.451 1.664 0.434 1.760
′
𝑈𝐶𝐿 = 𝐶 + 𝐶𝐸 ∙ 𝑅̅ 𝑈𝐶𝐿 = 𝐵𝐸𝑜𝑏 ∙ 𝜎̂ 𝑈𝐶𝐿 = 𝐷𝐸𝑜𝑏 ∙ 𝑅̅
𝑥̃ ′
𝐿𝐶𝐿 = 𝐶 − 𝐶𝐸 ∙ 𝑅̅ 𝐿𝐶𝐿 = 𝐵𝐸𝑜𝑏 ∙ 𝜎̂ 𝐿𝐶𝐿 = 𝐷𝐸𝑢𝑛 ∙ 𝑅̅
∗
𝑈𝐶𝐿 = 𝐶 + 𝐴∗ ∙ 𝑠̅ 𝑈𝐶𝐿 = 𝐵𝐸𝑜𝑏 ∙ 𝑠̅
𝑥̅
𝐿𝐶𝐿 = 𝐶 − 𝐴∗ ∙ 𝑠̅ ∗
𝐿𝐶𝐿 = 𝐵𝐸𝑢𝑛 ∙ 𝑠̅
𝑈𝐶𝐿 = 𝐶 + 𝐸𝐸 ∙ 𝑅̅
𝑥
𝐿𝐶𝐿 = 𝐶 − 𝐸𝐸 ∙ 𝑅̅
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Causes
C1 Batch change in the materials S2 Work piece clamping device, breakage
C2 Batch change in processing S3 Work piece limit stop defective
S4 Work piece drive unit defective
K1 Calibration not correct
K2 Reference standard defective V1 Incoming goods are dirty
2021-02-10 - SOCOS
Action
A1 Termination of the inspection during M6 Machine cleaned
a sample, since the values up to now M7 Machine serviced/repaired
have all been outside the plausibility
area. P3 Inspection equipment changed
P4 Inspection controller changed
A2 Termination of the inspection after
completing a sample, since it is beyond P5 Inspection equipment adjusted
the control limits and it is expected that P6 Inspection equipment cleaned
it will affect subsequent characteristics. P7 Inspection repeated
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
11 Reaction Catalog
A reaction catalog needs to be created especially for the specific process being controlled (machine,
test location).
The event codes listed in the previous sections are only meant to be examples, which in this case refer
to a mechanical manufacturing process.
Goods: Machine:
11. 100% sorting, OK-proportion delivered 41. Adjust machine
12. Goods delivered with concession 42. Machine defect mech./hydr./pneum.
13. Goods held; rework necessary 43. Machine defect electrical
14. Scrap 44. Change dressing tile
45. Change dressing rollers
Measurement system:
Tool:
51. Change of gage caliper
21. Dress grinding wheel
52. Measurement control/compensation
22. Change grinding wheel
General:
Work piece: 61. Contamination
31. Change chuck 62. Preliminary working/delivery
32. Stop dog 63. Other causes (to be documented)
•••••••••
•••••••••
2021-02-10 - SOCOS
•••••••••
•••••••••
Quality Control Chart Part Flange 1 ... ... 027 Department W410
Operation 1130, precision-turn Contol chart by
for Continuous
Quality Management Characteristic Diameter Date 10.09.2019
Characteristics Issue no. 1
x1 63 63 63 61 62 61 62 63 64 63 62 60 62 60 58 61 60 62 62 60 62 58 62 61 59 Machine ..., Spindle
x2 63 63 63 60 62 62 61 62 64 63 64 61 65 62 63 63 61 61 63 61 63 63 63 64 65 Machine no. ... ....
x3 62 62 61 59 63 63 62 63 63 61 64 63 64 65 63 62 62 62 63 65 64 61 64 64 64
x4 63 63 63 63 63 63 61 62 63 60 65 62 63 65 61 64 59 61 63 65 60 62 64 63 62 USL 48.69
x 62.6 62.8 62.4 61.0 62.4 61.8 61.6 62.2 63.4 62.0 63.4 61.8 63.2 63.0 61.4 62.4 60.4 61.6 62.8 62.8 62.2 61.6 63.6 62.4 62.0 Reference value 48.00
Booklet No. 7 ― Statistical Process Control
s 0.55 0.45 0.89 1.58 0.55 1.30 0.55 0.84 0.55 1.41 1.34 1.30 1.30 2.12 2.07 1.14 1.14 0.55 0.45 2.28 1.48 2.30 1.14 1.82 2.55 Unit 0.01 mm
65 Sample size 5
64 Frequency every 150 parts
Additional test EII
63 Control limits for o o o
~
̅ − 𝒔 Chart
x x x
62 Process related o
x
39
Tolerance related o
61 UAL 64.6
60 UCL 63.6
Target value 62.0
59 LCL 60.4
0 5 10 15 20 25
LAL 59.4
3
Control limits for o s o R
UCL 2.6000
Central line 1.300
s 2 LCL 0.3000
Evaluation
x 62.3
1 s 1.27
̂ 1.35
Cp 1.73
0 Cpk 1.66
0 5 10 15 20 25
Date 11.10. 11.10. 11.10. 11.10. 12.10. 12.10. 12.10. 12.10. 13.10. 13.10. 13.10. 13.10. 14.10. 14.10. 14.10. 14.10. 15.10. 15.10. 15.10. 15.10. 18.10. 18.10. 18.10. 18.10. 19.10. Corrective action (see overleaf)
Time 6:50 11:40 14:40 20:35 6:25 10:50 14:20 21:05 6:20 10:15 15:00 21:35 7:10 11:50 16:05 21:30 6:50 12:10 16:15 21:20 6:45 12:05 16:20 21:15 6:50 Evaluated
Name Checked
Booklet No. 7 ― Statistical Process Control
1 Fig. 13
0.9
0.8
indicates process shift
Probability that chart
0.7
The curve is steeper at 𝑛 = 5 than at
2021-02-10 - SOCOS
0.6
0.5
𝑛 = 3, so the control chart with a larger
0.4 sample size reacts more sensitively to a
0.3 change in the process location. A high
0.2
responsiveness of the average chart
0.1
0
must therefore be "bought" by the
0 0.5 1 1.5 2 2.5 3 3.5 4
sample size. The curve for the average
Process shift in sigma units
chart is also steeper than that of the
xbar-s; n=3 xbar-s; n=5
median chart.
A special feature is not easily recognizable in the presentation. It is the fact that the curve at the lower
end still indicates an intervention probability of approx. 1 % at zero displacement. Thus, even with a
completely stable centered process, there is a probability of 0.5 % in each case that a sample mean 𝑥̅
is below 𝐿𝐶𝐿 or above 𝑈𝐶𝐿. This error probability is also called a type-1 error.
The above table indicates the maximum number of control limit violations that can occur purely by
chance, although the process has not changed. Where 𝑚 is the number of samples.
EDITORIAL NOTE: Statistically speaking, the situation corresponds to an urn model (bowl model) in which 99 %
of the balls bear the inscription "unchanged process", the remaining 1 % of the balls bear the inscription
"changed process" (see [Booklet 2]). If 100 balls are taken out of the urn, with a probability of 99.5 %, no more
than 4 will have the inscription "changed process". The number 4 is the upper threshold value of the binomial
distribution for a probability of 99.5 %. In the case of the s chart, only the exceeding of the upper control limit is
considered, with the associated intervention probability of 0.5%.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
The term maturity often appears in connection with an assessment model (maturity model) that is
used to assess the "maturity" (quality capability) of an organization, system or process, for example,
by means of a catalog of questions or criteria.
NOTE: The mental link with the development of fruits or living organisms certainly plays a role here.
The assessment is carried out, for example, using a multi-level scale (1, 2, 3 or red, yellow, green). An
2021-02-10 - SOCOS
even number of levels (e.g. four levels: not or merely rudimentary, largely not, largely, almost complete
or complete) has the advantage that there is no "medium" rating. The assessor must therefore decide
in the middle range for the worse or better half of the possibilities. A larger number of steps than about
three or four steps makes it more difficult to differentiate between them. A quantitative scale has the
advantage that the results can be presented quite clearly in the form of a network diagram (radar
chart).
Review
To ensure the consistent realization and effectiveness of the control loops as well as the effectiveness
and efficiency of the inspection processes in production, it makes sense to carry out regular reviews.
Maturity Assessment
Carrying out assessements is recommented, in order to
check the process of quality planning process regarding its application,
to assess the maturity of its implementation and
to derive targeted improvement potentials.
The aspects to be evaluated include, for example, the quality characteristics and the respective
associated inspection strategy, the measurement and test processes as well as the implemetation of
the quality control loops (big, small).
“SPC-Assessment” is just another name for such a maturity assessment. A major difference is the "flight
altitude", i.e. the focus and the level of detail. Typical main and sub criteria are, e.g.
Personnel and Qualification (qualification of users)
Planning (measurement/test processes, machine and process characteristics, machine
capability)
Realization (application of the control chart)
Documentation (of actions)
Effectiveness (short-term process capability, long-term evaluation)
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
List of Symbols
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
′ ′ ∗ ∗ ′
Konstanten: 𝑎𝑛 , 𝑐𝑛 , 𝑑𝑛 , 𝐴, 𝐴′ , 𝐴∗ , 𝐵𝐸𝑢𝑛 , 𝐵𝐸𝑜𝑏 , 𝐵𝐸𝑢𝑛 , 𝐵𝐸𝑜𝑏 , 𝐶𝐸 , 𝐷𝐸𝑢𝑛 , 𝐷𝐸𝑜𝑏 , 𝐷𝑜𝑏 , 𝐸𝐸 , 𝐸𝐸′
Other symbols used only in individual chapters or the use of symbols with different meanings are defined
in the appropriate context.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
English German
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
alarm limits
Tolerance-related control limits whose violation generally requires increased monitoring of the
process.
2021-02-10 - SOCOS
determination
Activity to determine one or more characteristics and its characteristic values [ISO 9000]
capability
Suitability of a object (for example product, service, process, person, organization, system, resource)
for realization of a result, which will fullfill the requirements of this result (with reference to [ISO
9000])
capability
Suitability of an organization, a system or an process for realisation of a product, which will fullfill the
requirements of this product [ISO 22514-1]
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
EDITORIAL NOTE: In the case of characteristics limited on one side above (only USL given), such as roughness
(Rz), shape and position (e.g. roundness, rectangularity), it does not make sense to assume 𝐿𝑆𝐿 = 0 and
𝑈𝑆𝐿
consequently set 𝐶 = .
2
characteristic
Feature for the recognition and distinction of units. [DIN 55350-12]
characteristic
Distinguishing property
NOTE 1: A characteristic can be inherent or assigned.
NOTE 2: A characteristic can be of qualitative or quantitative nature.
NOTE 3: There are different classes of characteristics, e. g.:
physical, for example mechanical, electrical, chemical or biological characteristics;
2021-02-10 - SOCOS
conformity
Fulfilling a requirement [ISO 9000]
conformity evaluation
Systematic examination about the degree to which an entity fulfills special requirements
[ISO 3534-2]
continuous characteristic
Characteristic whose characteristic values are the measuring values of a physical quantity (for example,
weight, length, current, temperature)
EDITORIAL NOTE: Often imprecisely called “variable characteristic”; see also “continuous scale” [ISO 3534-2]
continuous characteristic
Quantitative characteristic whose co-domain (range of values) is uncountably infinite. [DIN 55350-12]
EDITORIAL NOTE: The value of such a characteristic is always given as a product of numerical value
and unit.
E XAMPLES:
Length: 12.54561… m,
Diameter: 3.532… mm.
The numerical value of the characteristic "length", for example, can take on every value between 12
and 13; mathematically expressed: every real number on the interval between 12 and 13. I n reality, no
value of such a continuous characteristic can be measured with infinite accuracy. The number of
decimal places is always limited by the number of available digits that can be displayed. In addition,
properties of the measurement process can limit the number of useful decimal places.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
control chart
Chart on which some statistical measure of a series of samples is plotted in a particular order to steer
the process with respect to that measure and to control and reduce variation. [ISO 3534-2]
countable characteristic
Special discrete characteristic whose co-domain is given by the set of natural numbers including zero
(0, 1, 2, …) or a subset of this set.” [DIN 55350-12]
discrete characteristic
Characteristic whose characteristic values are counted measurands in a countable unit (e. g. good/bad,
right/wrong, red/green/blue)
EDITORIAL NOTE: Often imprecisely called “attributive characteristic”; see also “discrete scale”
[ISO 3534-2].
2021-02-10 - SOCOS
discrete characteristic
Quantitative characteristic whose co-domain is finite or countably infinite. [DIN 55350-12]
EDITORIAL NOTE (E XAMPLES)
Mass in kg (without decimal places).
Body weight in cm (without decimal places).
Number of "6" when rolling a die 100 times. The result can take on all of the values from 0 to 100.
The number is finite.
Number of lightning strikes in Germany during a given year (one can also locate and, of course,
record atmospheric lightning using antennae). This number can take on a countably infinite
number of values (0, 1, 2, 3, …), although very large values will appear with decreasing probability.
entity
That which can be individually described and considered [ISO 3534-2]
EDITORIAL NOTE: Not to be confused with “unit” (see [VIM])
enumeration
Determining the measurand “number of ellements in a set”, for instance by counting. [DIN 1319-1]
estimator
Statistic used in estimation of the parameter Θ. [ISO 3534-1]
estimation
Procedure that obtains a statistical representation of a population from a random sample drawn from
this population.
NOTE 1: In particular, the procedure involved in progressing from an estimator to a specific estimate constitutes
estimation.
[ISO 3534-1]
estimate
Observed value of an estimator. [ISO 3534-1]
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
influence quantity
Quantity that, in a direct measurement, does not affect the quantity that is actually measured, but
affects the relation between the indication and the measurement result. [VIM]
inspection
Conformity assessment through observation and assessment, accompanied ― if applicable ― by
measurement, testing or comparison [ISO 3534-2]
measurement result
2021-02-10 - SOCOS
Set of quantity values that are assigned to a measurand along with any available relevant information
[VIM]
measuring instrument
Device, which is used alone or in conjunction with additional facilities for the performance of
measurements [VIM]
measurand
Quantity which is to be measured [VIM]
measurement
Process in which one or more quantity values, that can be reasonably assigned to a quantity, are
experimentally determined
NOTE 2: A measurement means comparing sizes and includes counting
[VIM]
measurement process
Set of activities for determining a quantity value [ISO 9000]
measurement uncertainty
Not negative parameter that characterizes the variation of values that is attached to the measurand
on the basis of information used [VIM]
measurement value
Quantity value which represents a measurement [VIM]
measuring system
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Combination of measuring devices and often other devices and, if necessary reagents and utilities,
which are arranged and adapted to provide information to obtain readings within certain intervals for
quantities of certain types
NOTE: A measuring system can consist of a single measuring device
[VIM]
median of a sample
Among the values ordered in ascending numerical order and numbered "1" to "𝑛", for odd 𝑛 the actual
𝑛+1
value with the number 2 , for even 𝑛 usually the mean value of the two actual values with the
𝑛 𝑛
numbers 2
and 2
+ 1.
Example: For a sample of size 5 ordered by ascending numerical value, the median is the middle of the
5 values.
nominal characteristic
Qualitative characteristic whose values have no quantitative relation to each other. [DIN 55350-12]
EDITORIAL NOTE, Example: The characteristic "color" with the values "red", "yellow", "green".
The values of these characteristics can only be classified and differentiated (is equal, is not equal).
“The value of a nominal characteristic is often also called attribute. A nominal characteristic with only two,
mutually exclusive values is called a dichotomous characteristic, a binary characteristic or an alternative
characteristic. It can only take on one of two mutually exclusive values.”
Examples:
good / bad,
inside / outside the tolerance range (OK / not OK),
defective / not defective,
present / not present.
Such data is, for instance, gained using limit gauges or by visual assessment with boundary samples. It is,
however, also possible to define a “discrete” continuous characteristic by first measuring it and then classifying
the measured value using a category “inside/outside the tolerance range”. However, this is rarely sensible due
to the associated loss of information. If measurement values exist for a continuous characteristic, they should
be used in their original form.
ordinal characteristic
Qualitative characteristic whose values have a quantitative relation to each other. [DIN 55350-12]
EDITORIAL NOTE, Example:
The characteristic "clothing size" with the values XXS, XS, S, L, XL, XXL, …, 7XL,
The characteristic "(European) shoe size" with the values 14, 15, 16, …, 48, 49.
The values of these characteristics can be classified and differentiated, e.g. using the relationships “smaller
than”, “is equal” or “greater than”.
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
population
Population of the considered entities [ISO 3534-2]
EDITORIAL NOTE: The term “population” refers to a limited or unlimited number of observable units that are to
be considered concurrent within the framework of an existing statistical problem. Such observable units can, for
example, come from "observations” or results from “tests” conducted under the same conditions.
Examples of finite populations are the number of
students in a school,
eligible voters within a state,
television viewers who watched the final match of the last Football World Cup,
parts in a delivery of goods,
products manufactured within one shift at factory XY.
Examples of (theoretically) infinite populations are the number of
points observed when rolling dice,
2021-02-10 - SOCOS
process
Set of associated or mutually influencing activities, which uses entries to achieve an intended result
[ISO 9000]
EDITORIAL NOTE: In the context of this Booklet, a process is generally understood to be a sequence of activities
and/or procedures in which materials or preprocessed parts/components are further processed and a product
is produced. These can be manufacturing or assembly processes, for example. But also a measuring process is
a sequence of activities and/or sequences in which something is produced, namely measuring results.
parameter
Index of a family of distributions
NOTE 1 The parameter may be one-dimensional or multi-dimensional.
NOTE 2 Parameters are sometimes referred to as location parameters, particularly if the parameter corresponds
directly to the mean of the family of distributions. Some parameters are described as scale parameters,
particularly if they are exactly or proportional to the standard deviation of the distribution. Parameters that are
neither location nor scale parameters are generally referred to as shape parameters.
[ISO 3534-1]
EDITORIAL NOTE 1: Parameters of location (expected value, median), parameters of variability (variance,
standard deviation, coefficient of variation) and shape parameters (skewness, kurtosis, excess) are also called
functional parameters.
EDITORIAL NOTE 2: Colloquially the term process parameter is occasionally used. The standard [ISO 3534-1]
refers this term exclusively to parameters of statistical distributions. In connection with processes [ISO 22514-1]
uses the term process characteristic.
performance index
Parameter that indicates the performance measure with respect to stipulated specifications
[ISO 22514-1]
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
population parameter
Summary measure of the values of some characteristic of a population
EXAMPLE Population mean = µ; population standard deviation = σ
NOTE Population parameters are usually symbolized by lower case Greek letters in italics.
[ISO 3534-2]
process characteristic
Inherent characteristic of a process [ISO 22514-1]
EDITORIAL NOTE: Process characteristics are necessary characteristics to ensure the conformity of the product
2021-02-10 - SOCOS
characteristics. These are characteristics that are transmitted to facilities and equipment, and not on the
product.
For the term “inherent”, see also the editorial note with regard to the term “product characteristic”.
process quality
Required quantified output from process, which is equal to the quality objective (usually fractions
nonconforming e.g. defined in ppm). [CD 00301]
product
Result of a process [ISO 22514-1], [ISO 3534-2]
product characteristic
Inherent characteristic of a product [ISO 22514-1]
EDITORIAL NOTE: “Inherent” means “an inherent unit” (for example, physical properties such as weight, size,
power consumption of a product); therefore an inherent characteristic may be a quality characteristic, but not
a “mapped” characteristic (such as, e. g., price, the owner).
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
[ISO 22514-1]
qualitative characteristic
Characteristic whose values are allocated to a scale without defined distances.
[DIN 55350-12]
quality capability
Suitability of an organization or parts of an organization (e. g. people, procedures, processes,
equipment) for realizing a result that will meet the quality requirements of this result (with reference
to [ISO 22514-1, 3.3.2] and [ISO 9000, 3.6.12]).
A process is quality capable if it is able to fully comply with the specified requirements.
quality characteristic
Inherent characteristic of a product, a process or system related to a requirement [ISO 22514-1]
EDITORIAL NOTE: For the term “inherent”, see also the editorial note for the term “product characteristic”.
quantitative characteristic
Characteristic whose values are allocated to a scale with defined distances. [DIN 55350-12]
EDITORIAL NOTE: All physical quantities are quantitative characteristics, for example length, volume,
angle, weight, mass, temperature, tension, current, time, speed.
quantity
Property of a phenomenon, a body or a substance wherein the property has a value which can be
expressed by a number and a reference [VIM]
quantity type
Aspect which is common with comparable quantities [VIM]
quantity value
Numerical value and reference, which together specify a quantity quantitatively [VIM]
random cause
Cause of the variation which is constantly inherent in a process [ISO 22514-1]
random sample
Sample, which has been selected randomly [ISO 3534-1]
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
range
Greatest minus smallest single value: 𝑅 = 𝑥𝑚𝑎𝑥 − 𝑥𝑚𝑖𝑛
reaction plan
Reaction plan Action or series of steps prescribed in a control plan in the event abnormal or
nonconforming events are detected. [IATF 16949]
requirements
Requirement or expectation that or which is stipulated, commonly provided or mandatory
[ISO 9000]
sample
Subset of a population which consists of one or more selection units.
[ISO 3534-1]
EDITORIAL NOTE: In contrast to the population a sample is a real and therefore finite number of “things” or
events. Examples of this are the set of
vehicles that passed through the highway tunnel near Leonberg on a specific day,
results observed when rolling one die 10 times,
results obtained when conducting 25 measurements against a standard of length,
50 parts made while testing a machine’s capacity.
By the way, the German term for “sample” (“Stichprobe”) originates from the practice of “piercing” grain sacks
and cotton bales during quality inspection. A sample consists of one or several units that were “drawn” from a
real or fictitious population according to the random principle. The number of these elements is called the
sample size. The properties of the sample are supposed to represent the population. Random sampling
presupposes that each element of the population is given the same chance (same probability) to be picked for
the sample. In general, it is rarely possible to apply the random principle in a nearly ideal manner (flipping a
coin, roulette, drawing the lottery numbers). The idea is especially problematic with regard to fictitious
populations; “drawing” the sample is only possible in a figurative sense.
sample mean
Average, arithmetic mean. Sum of random variables in a random sample divided by the number of
terms in the sum. [ISO 3534-1]
sample size
Number of sampling units in a sample [DIN 55350-14]
sampling unit
One of the individual parts into which a population is divided. [ISO 3534-1]
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Control chart with Shewhart control limits intended primarily to distinguish between the variation in
the plotted measure due to random causes and that due to special causes. [ISO 3534-2]
sorting inspection
This is a 100% inspection and means that objects/units/items are inspected (visually or using technical
means) with respect to defined characteristics and sepa-rated according to the individual results (e.g.
conforming / nonconforming).
specification
Document that specifies requirements
NOTE: A specification may refer to activities (for example process document, process specification and test
specification), or products (for example, product specification, performance specification and drawing).
[ISO 9000]
2021-02-10 - SOCOS
specification interval
Area between the limits maximum value and minimum value [ISO 22514-1]
EDITORIAL NOTE: The limits are also referred to as specification limits.
specification limit
Limiting value stated for a characteristic [ISO 3534-2]
stable process
Process which is only subject to random variation causes
NOTE 2: A stable process behaves in general as if samples drawn from the process are at any time simple
random samples from the same population.
[ISO 22514-1]
EDITORIAL NOTE: The definition of the term from the original English version ISO 3534-2 (2006) was adopted in
modified form in the English version ISO 22514-1 (2014); The German version DIN ISO 22514-1 (2016) uses only
the term “stable process”.
statistic
Completely specified function of random variables
NATIONAL FOOTNOTE: Statistics characterize properties of a frequency distribution
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
[ISO 3534-1]
target value
Preferred value or reference value of characteristic which is specified in a specification
[ISO 3534-2]
(specified) tolerance
Difference between maximum value and minimum value [ISO 3534-2]
EDITORIAL NOTE 1: The specification of a tolerance consists of a number and a unit of measurement, e.g. 0.01
mm. The term tolerance interval or specification interval refers to the range between the minimum value and
the maximum value. Apart from the limited resolution of the measuring process, the tolerance interval contains
an infinite number of values.
EDITORIAL NOTE 2: Colloquially, phrases such as "exceeding tolerance" or "adhering to tolerance" are often
used. This usually means that a measured value of a characteristic lies within or outside the tolerance interval.
EDITORIAL NOTE 3: Since according to the standard a specification is a document, the term specification interval
is strictly speaking wrong. What is meant is the (specified) tolerance interval defined in the specification for a
characteristic.
tolerance zone
Area of permitted values between lower and upper limiting value. [DIN 55350-12]
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
variance of a sample
The deviations of the individual values from the arithmetic mean are summarized and the sum is
1
devided by the number of values reduced by one: 𝑠 2 = 𝑛−1 ∙ ∑𝑛𝑖=1(𝑥𝑖 − 𝑥̅ )2
variation
Difference between values of a characteristic [ISO 22514-1]
2021-02-10 - SOCOS
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Literature
Corporate directives, only internally available
[CD 02981] CD 02981 Information Governance
[CD 00301] CDQ 0301 Management of Characteristics
[CD 00501] CDQ 0501 Identification and Traceability
[CD 00503] CDQ 0503 Rework and Repair
[CDQ 0509] CDQ 0509 Concessions
[Booklet 3] Booklet No. 3, Basics Concepts of Techn. Statistics, Evaluation of Measurement Series
[Booklet 8] Booklet No. 8, Measurement Uncertainty
[Booklet 9] Booklet No. 9, Machine and Capability Capability
[Booklet 10] Booklet No. 10, Capability of Measurement and Test Processes
[AIAG PPAP] AIAG Core Tools, Production Part Approval Process (PPAP)
[AIAG SPC] AIAG Core Tools, Statistical Process Control (SPC)
[AQDEF] Advanced Quality Data Exchange Format of the Automotive Industry, Q-DAS GmbH,
Weinheim, Germany
[IATF 16949] IATF 16949, Automotive Quality Management System Standard ― Quality manage-
ment system requirements for automotive production and relevant service parts
organizations
[DIN 1319-1] Basic concepts in metrology, Part 1: General concepts (under revision)
[DIN 55350-11] DIN 55350-11, Concepts for quality management ― Part 11: Supplement to DIN EN
ISO 9000
[DIN 55350-12] DIN 55350-12, Concepts in the field of quality and statistics ― Concepts relating to
characteristics
[DIN 55350-14] DIN 55350-14, Concepts in the field of quality and statistics ― Concepts of sampling
[ISO 3534-1] DIN ISO 3534-1, Statistics — Vocabulary and symbols — Part 1: General statistical
terms and terms used in probability
[ISO 3534-2] DIN ISO 3534-2, Statistics — Vocabulary and symbols — Part 2: Applied Statistics
[ISO 7870-2] Control charts ― Part 2: Shewhart control charts
[ISO 7870-5] Control charts ― Part 5: Specialized control charts
[ISO 7870-6] Control charts ― Part 6: EWMA control charts
[ISO 7870-8] Control charts ― Part 8: Charting techniques for short runs and small mixed batches
[ISO 9001] DIN EN ISO 9001, Quality management systems — Requirements
[ISO 11462-1] Guidelines for implementation of statistical process control (SPC) — Part 1: Elements
of SPC
[ISO 11462-2] Guidelines for implementation of statistical process control (SPC) — Part 2: Catalogue
of tools and techniques
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
[ISO 22514] DIN ISO 22514-2, Statistical methods in process management — Capability and
performance — Part 2: Process performance capability and process capability
statistics of time-dependent process models
Vienna Zurich
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
Index
long ........................................................... 12
5 short ........................................................... 8
5 M .................................................................. 5 countable characteristic ............................... 47
counts ........................................................... 18
A
acceptance control chart......................... 35, 45 D
alarm limits ........................................ 10, 25, 45 delimitation factor ........................................ 35
AQDEF...................................................... 14, 45 determination ............................................... 45
assessment ....................................................41 discrete characteristic................................... 47
attribute ........................................................49 distribution parameters ................................ 26
2021-02-10 - SOCOS
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
measurement ................................................48 R
process ......................................................48 random
result .........................................................48 cause......................................................... 52
uncertainty .......................................... 16, 48 sample ...................................................... 52
value ..........................................................48 range ............................................................. 53
measuring reaction plan ............................................. 8, 53
instrument ................................................48 requirements ................................................ 53
measuring and test equipment .................48 responsibilities .............................................. 14
system .......................................................48 responsiveness ............................................. 40
N resulting process distribution ....................... 53
review ........................................................... 41
normal distribution ......................................... 6
risks ............................................................... 11
2021-02-10 - SOCOS
O robustness .................................................... 15
rule of seven ................................................. 24
operation characteristic ................................40
original data..................................................... 6 S
P sample........................................................... 53
mean......................................................... 53
parameter................................................ 50, 51
size ...................................................... 19, 53
Pearson, distribution family ..........................30
sampling
performance index ........................................50
interval ..................................................... 21
population .....................................................50
unit ........................................................... 53
process ..........................................................50
Shewhart ......................................... 5, 6, 19, 53
average......................................................27
smooting factor ............................................ 32
capability index .........................................51
software .................................................. 14, 26
characteristic .............................................51
sorting inspection ................................... 11, 54
characteristic in control ............................51
SPC ............................................................ 5, 55
in control ............................................. 51, 54
characteristics .......................................... 15
quality .......................................................51
specification .................................................. 54
stable .........................................................54
interval ..................................................... 54
product ..........................................................51
limit........................................................... 54
product characteristic ...................................51
standard deviation of a sample .................... 55
Q statistic.......................................................... 54
quality statistical quality control .............................. 55
capability ...................................................52
T
characteristic .............................................52
target value ................................................... 55
quality capable process .................................52
tolerance ....................................................... 55
quality control chart ............................ 5, 23, 52
interval ..................................................... 55
quantity .........................................................52
zone .......................................................... 55
type ...........................................................52
trial run ......................................................... 26
value ..........................................................52
type-1 error................................................... 40
•••••••••
•••••••••
Booklet No. 7 ― Statistical Process Control
U V
upper specification limit ................................55 variance of a sample ..................................... 56
variation .................................................. 27, 56
variation control chart .................................. 33
2021-02-10 - SOCOS
•••••••••
•••••••••
2021-02-10 - SOCOS
•••••••••
•••••••••
2021-02-10 - SOCOS
•••••••••
•••••••••
2021-02-10 - SOCOS
•••••••••
•••••••••