Control Charts crgraph.

com

Introduction
A Control Chart shows the time course of a process characteristic. For this purpose,
data can be taken continuously or in periodic samples. The prerequisite is that the
process capability has previously been confirmed. The Control Chart is also well
known as Statistical Process Control SPC.
There are control charts for continuing quantitative measurands and for attributive
characteristics (test with counts). The sole representation of a continuing measurand
is defined as an single measurement chart.

Continuing measurements Attributive characteristics

Single measurement chart (only p-chart proportion of
measurand on its own) defective units
𝑥̅ /𝑠 - control chart (additionally with np-chart number of defective units
scatter distribution)

𝑥̅ /𝑅 - control chart (additionally with u-chart Number of errors per unit

distribution of range) for multiple possible
errors

Purpose and benefit:

The quality control chart helps in recognizing systematic influences of a process, as
well as disturbance variables and environmental influences. In the SPC, warning and
control limits calculated statistically from the process data are determined and
entered. An intervention is only required if the control limits are exceeded, leading to
a more stable process (see illustration):

UCL (upper control limit)

UWL (upper warning limit)
Quality feature

Central line (mean value)

Nominal value

LWL (lower warning limit)

LCL (lower control limit)

Time scale

If one or more values of a random sample is/are outside the control limits, this defect
must be identified and rectified as quickly as possible before the process exceeds the
tolerance limits. Note: The tolerance is normally not shown in the control chart.

© Copyright CRGRAPH 2022

 Control Charts crgraph.com

Further interventions must be made if the following conditions are met:

One point 6 points in

outside succession
 3s rising or falling

2 out of 3 9 points in
points in succession on
succession the same side
outside  2s

14 points in
4 out of 5 succession
points in alternating above
succession and below the
outside  1s midline

8 points in
succession
outside  1s

An expansion to 8 rules is also referred to the so called "Western Electric Rules".

The causes of these deviations must be ascertained. Measures carried out in a
corresponding way are noted on the SPC. The production quantity since the last
random sample was taken must be traced.

Basics and examples

x - Quality control chart

In the following illustration, the length of a transverse member was monitored using
the SPC:

UCL
OE
OEGG

OWG
OWG
UWL
385 385
Pro d u k tm e rk m a l

380 380 mean

Mittel
Mittel

375 375
UWG
LWL
UWG

UE
UEG
LCLG

0 5 10 15 20 25 30 35 50 100 150 200

Abs olute Häufigkeit laufende Num m er

© Copyright CRGRAPH 2022

Control Charts crgraph.com

What are referred to as the lower/upper warning limits LWL/UWL and control limits
LCL/UCL are shown on the right.

LWL - UWL 95.45 % Average value ±2·s‘

LCL - UCL 99.73 % Average value ±3·s‘

The areas ±2 s‘ and ±3s‘ correspond to a probability range of 95.45% and 99.73%.
(Note: In sources from the German literature, warning and control limits above 95%
and 99% probability are often defined).
These limits do not reflect the tolerance range, but rather only the observed frequency
distribution (histogram on the left in the previous illustration) of the particular random
sample parameter that is being monitored. The tolerance limits are never specified
on the process control chart. The collected process data does however form the basis
for the process capability investigation in relation to the tolerance (see Cp/Cpk).
The warning and control limits are calculated periodically based on the most recent
process data. Intervention of correction of the process only takes place once the
control limits have been undershot or exceeded.

x/s - Quality control chart

In the 𝐱̅/𝐬 - quality control chart acc. to Shewhart, the data is subdivided into random
samples. Within these random samples, it is possible to create a standard deviation
for each, the profile of which is represented alongside the average value. The warning
and control limits for the average values are brought into relation with a subdivided
sample size n‘ here (standard group size n‘ = 5):

LWL - UWL 95.45 % Average value  2 s' / n'

LCL - UCL 99.73 % Average value  3 s' / n'

Warning and control limits for the standard deviation are calculated here using the
²-distribution:

 02,02275, n '−1  02,97725, n '−1

LWL = s UWL = s
n'−1 n'−1

 02,00135, n '−1  02,99865, n '−1

LCL = s UCL = s
n'−1 n'−1

The standard deviation should always be as small as possible, so the LWL and LCL
are not used as a rule.

Quality control charts for qualitative features

Qualitative features are understood to be events such as faulty/good, yes/no or
missing/not missing, etc. In this case, large sample ranges are required so that
defects can be detected. If there are very low defect probabilities, consequently, the
samples must be taken over a long period and this can be a disadvantage. The
following differentiation is made:
© Copyright CRGRAPH 2022
Control Charts crgraph.com

p-card The proportion of faulty units in the sample is entered here

u-card Here, the number of defects per unit in the sample is entered

p-card
The sample ranges do not have to be constant, although a fluctuation of more than
25% is not recommended. The relative proportion of faulty units is:
n* n' p p : relative defect proportion
p = = n* : number of faulty units
n' n'
n‘ : sample range (faulty and intact units)

Depending on whether the relative defect proportion (p-card) or the absolute defects
are noted on the control chart, there is also what is referred to as the np-card for the
latter case. In the np-card, the random sample range is constant.
The average defect proportion in several random samples is then:

1 m ni*
n
: number of defects per random sample
p= *
i m : number of random samples
ntot i =1 ntot : number of total units

The control limits are calculated by way of approximation to the normal distribution
where:
u u
LCL = u − 3 UCL = u + 3 u : Mean sample size
n n
Here too, LCL is generally not required, particularly since negative values are
mathematically possible.
u-card
In the u-card, the random samples must consist of units with several components or
defect possibilities. The sample size can be different, but should not fluctuate by
more than 25%. The number of defects per unit is:
c
u = c : number of defects in the random sample
n' n‘ : sample range

Analogously to the u-card, there is also the c-card, in which the absolute number of
defects is shown here. The random sample range is constant.
In the u-card, the average number of defects is:

1 m ci : number of defects per random sample

u =
ntot
 ci
i =1
m : number of random samples
ntot : number of total units

The control limits are calculated as follows:

u u
UEG = u − 3 OEG = u + 3 n : Mean sample size
n n

LCL, however, is usually not used and can also be negative.

© Copyright CRGRAPH 2022

Control Charts crgraph.com

Using Visual-XSel

All control charts are available as templates. Click to the icon in the main guide…

or use menu File / Templates…

© Copyright CRGRAPH 2022

Control Charts crgraph.com

Use the Paste – link for your data in

the clipboard, which has to be copied
before (e.g. from Excel).

After data transfere start the macro

Then the charts on the right main window has update with the new data

In case of the Control_Chart_x_s.vxg on the 2nd page the "Western Electric Rules"
will be shown, if relevant conditions where met here.

The example has at point 17 one point is outside ± 3s

© Copyright CRGRAPH 2022