Module III - Statistical Process Control
Module III - Statistical Process Control
Module III - Statistical Process Control
PPT/L-TCIO-1
Chapter III –
Statistical Process
Control
Operational quality and Lean
management - S7 EENG 4
Title of Lesson
Title of Lesson 1
ID document: e.g. PPT/L-TCIO-1
Chapter organization
1 – Introduction / reminders about dispersions
3 – Capabilities
4 – Principles of SPC
- Control charts
- Control charts with measures
- Attribute control charts => next lecture
- EWMA and CUSUM boards
Title of Lesson 2
ID document: e.g. PPT/L-TCIO-1
● Make the difference between a gap between average and tolerance interval a gap
between dispersion and tolerance interval
1 – Introduction / reminders
Title of Lesson 4
ID document: e.g. PPT/L-TCIO-1
Thickness
Title of Lesson 6
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 7
ID document: e.g. PPT/L-TCIO-1
t i n g ,
The objective is to get the expected thickness of the piece of wood and a smooth surface
e s e t d
s a m l w o o
Cutting tool
h t h e , f i n a m e
n w i t n u t s h
High speed
e s a Front roller
Eve r t e
Rear roller
h a y s t
rotation tool
a s f o e a l w t i o n
w e l l o t b o d u c
as Final wood
w i l l n s p r
e s s
thickness
m a s
i c k n a k e Piece of wood
th y o u m
i f
Title of Lesson 8
ID document: e.g. PPT/L-TCIO-1
-hardness of the wood piece -dust between table and piece of wood
Etc. Etc.
Title of Lesson 9
ID document: e.g. PPT/L-TCIO-1
- Materials
- Manpower/people
- Machine/equipments
- Methods
- Mother nature (environment)
Title of Lesson 10
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 11
ID document: e.g. PPT/L-TCIO-1
Special causes
● Special (or accidental) causes are identifiable causes, often irregular and
unstable, and thus difficult to predict. The appearance of a special cause
requires intervention on the process.
Title of Lesson 12
ID document: e.g. PPT/L-TCIO-1
Special causes
● Special causes can be classified into two categories:
- Causes that affect the setting, the position of the monitored value (ex : tool
misalignment)
- Causes that affect the dispersion and therefore the capability of the
machine (ex : lubrication of the machine)
Title of Lesson 13
ID document: e.g. PPT/L-TCIO-1
All common causes together make a process following the Gauss curve
Title of Lesson 14
ID document: e.g. PPT/L-TCIO-1
(tolerance interval)
Title of Lesson 16
ID document: e.g. PPT/L-TCIO-1
3 – Capability
Title of Lesson 17
ID document: e.g. PPT/L-TCIO-1
3 – Capability
A – Introduction
Title of Lesson 18
ID document: e.g. PPT/L-TCIO-1
1 finished product
= 10 components
Title of Lesson 19
ID document: e.g. PPT/L-TCIO-1
Let’s bet :
simulation :
Result :
Conclusion :
Title of Lesson 20
ID document: e.g. PPT/L-TCIO-1
Capability – Introduction
Now you are aware that :
Title of Lesson 21
ID document: e.g. PPT/L-TCIO-1
Capability - Definition
Title of Lesson 22
ID document: e.g. PPT/L-TCIO-1
Capability – Introduction
● The indicator Cp (Process Capability) compares the expected performance
of the process (tolerance interval) and the actual performance on it
(dispersion). We can therefore define the calculation of the capability by :
● TI : tolerance interval
● S: estimate of the standard deviation on a sufficient number of measurements
Title of Lesson 23
ID document: e.g. PPT/L-TCIO-1
Capability – Introduction
Tolerance interval Tolerance interval
Dispersion Dispersion
● On the right, a capable process: the dispersion is less important than the
confidence interval.
Title of Lesson 24
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 25
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 26
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 27
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 28
ID document: e.g. PPT/L-TCIO-1
3 – Capability
- Short-term indicators
- Performance rate
Title of Lesson 29
ID document: e.g. PPT/L-TCIO-1
Capability – Introduction
● The imposed capability assessment method is derived from the 2016
automotive industry standard IATF 16949 (formerly ISO/TS 16949). It defines
two types of indicators:
Title of Lesson 30
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 31
ID document: e.g. PPT/L-TCIO-1
𝑇𝑜𝑙𝑒𝑟𝑎𝑛𝑐𝑒 𝑖𝑛𝑡𝑒𝑟𝑣𝑎𝑙 𝑇𝐼
𝐶𝑚 = 𝐶𝑝 = =
𝑆ℎ𝑜𝑟𝑡 𝑡𝑖𝑚𝑒 𝑑𝑖𝑠𝑝𝑒𝑟𝑠𝑖𝑜𝑛 6𝜎𝑆𝑇
Title of Lesson 32
ID document: e.g. PPT/L-TCIO-1
n approximately = 50
Title of Lesson 33
ID document: e.g. PPT/L-TCIO-1
Cpk =
Cpk = 3 10,2+3X0,05
0,3/0,15=2
= 10,35
9,5 10 10,5
10,2
Cp = 3 Cp = 3
Cpk = 2
Cpk = 1
Title of Lesson 35
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 36
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 37
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 38
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 39
ID document: e.g. PPT/L-TCIO-1
Time
Short term capability
(can be during just one day)
Title of Lesson 40
ID document: e.g. PPT/L-TCIO-1
Long-term performance Pp
● The dispersion of the measurements over the long term is evaluated by :
Title of Lesson 41
ID document: e.g. PPT/L-TCIO-1
Long-term performance Pp
● The Pp is calculated :
- The sample must take into account the variations of instructions as for
example the adjustments
Title of Lesson 42
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 43
ID document: e.g. PPT/L-TCIO-1
Dispersion Dispersion
Title of Lesson 44
ID document: e.g. PPT/L-TCIO-1
● The greater the deviation, the greater the difference between Pp and Ppk.
Title of Lesson 45
ID document: e.g. PPT/L-TCIO-1
● We can have acceptable Cm, Cmk, Cp, Cpk, Pp and Ppk (>1,33) but the
process can still be improved.
Title of Lesson 46
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 47
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 48
ID document: e.g. PPT/L-TCIO-1
A – Introduction
Title of Lesson 49
Hystory of Quality : increasing efficiency to achieve quality
Shewhart Crosby
Taylor Juran
Deming Ishikawa
ID document: e.g. PPT/L-TCIO-1
SPC – Introduction
In a traditional manufacturing process, a final control of the product is
performed. This type of control has many disadvantages:
Informations on
performance Action on the
production
Title of Lesson 52
ID document: e.g. PPT/L-TCIO-1
SPC – Introduction
Controlling the manufacturing process with statistical techniques has the
following advantages: Deep knowledge of the whole
population characteristics with the
measuring of few samples only
No additional costs
Title of Lesson 55
ID document: e.g. PPT/L-TCIO-1
SPC – Introduction
● Even when a machine is perfectly adjusted to the target coast, there is a
dispersion of values around it. This dispersion is due to common causes.
● The set of common causes represents the natural dispersion of the process.
Most of the time, we have no choice but to live with them and it is not
necessary to intervene on the process.
● Examples:
- Gap in a slide(=coulisse)
- Variation of the hardness in a material
- Variation in temperature
- Dimensional variation of the pieces
Title of Lesson 56
ID document: e.g. PPT/L-TCIO-1
SPC – Introduction
● On the other hand, a special cause can have a great influence on the process
and requires an intervention on it. A special cause is an assignable cause, so
one can intervene on it.
● Examples:
- Tool fatigue
- Chips(=copeaux) under a stop
- Material change
- Poor lubrication
Title of Lesson 57
ID document: e.g. PPT/L-TCIO-1
(tolerance interval)
Title of Lesson 58
ID document: e.g. PPT/L-TCIO-1
SPC – Introduction
● Depending on the variations that the process will present, we have to ask
ourselves if we should intervene or not :
What criteria will give the statistic
- "Am I in the presence of a signal?" (special cause)
proof
In this that
case, we must the process has drift
intervene.
- "Am I inand
the presence of noise?it’s
that (common necessary
cause) to
In this case, the intervention is not necessary.
take actions ?
Title of Lesson 59
ID document: e.g. PPT/L-TCIO-1
SPC – Introduction
● The control chart is the tool that will allow us to make the right decisions.
Title of Lesson 61
ID document: e.g. PPT/L-TCIO-1
- Introduction
- Structure
- Driving method
Title of Lesson 62
ID document: e.g. PPT/L-TCIO-1
Upper limit
Title of Lesson 63
Principle :
ID document: e.g. PPT/L-TCIO-1
(tolerance interval)
With the help of statistics, looking at the mean and the range of a sampling of few pieces
will confirm or deny that the production is still on the target.
Title of Lesson 64
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 65
Mathematical property
ID document: e.g. PPT/L-TCIO-1
Mean of N samples
Population
𝝈
σ’=
√𝑵
σ
If we take many times N pieces among the whole population, the mean value of the N pieces follows a normal law with :
- X chart
- R chart
C – Measurement control
charts
- S chart
Title of Lesson 67
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 68
ID document: e.g. PPT/L-TCIO-1
● The X chart is intended to control the position of the mean of a small sample
(n usually between 3 and 10) taken in a predefined interval.
● The R chart is intended to control the position of the range of a small sample
(n usually between 3 and 10) taken in a predefined interval.
- Step 1: Data collection : A sample is regularly taken and its value or its
difference from the target value is noted.
The mean and range of the measurements are calculated.
This step allows to set the limits within which it is natural to find common
causes.
Title of Lesson 69
ID document: e.g. PPT/L-TCIO-1
…..
x̅ R̅
Title of Lesson 70
ID document: e.g. PPT/L-TCIO-1
Table of coefficients
Title of Lesson 71
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 72
ID document: e.g. PPT/L-TCIO-1
ഥ±𝑨 ×𝑹
𝑼𝑳𝑪, 𝑳𝑳𝑪 = 𝑿 ഥ
𝟐
Title of Lesson 73
ID document: e.g. PPT/L-TCIO-1
● When the standard deviation of the population is known, the control limits are
defined as follows:
ഥ 𝟑 ഥ±𝑨×𝝈
𝑼𝑳𝑪, 𝑳𝑳𝑪 = 𝑿 ± ×𝝈=𝑿
𝒏
The ratio can be taken over by the coefficient A (see coefficient table).
The target for X charts is the mean of all values of the sample.
Title of Lesson 74
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 75
ID document: e.g. PPT/L-TCIO-1
LLC = 0 X 0,73 = 0
Title of Lesson 76
ID document: e.g. PPT/L-TCIO-1
Resume manufacturing
stop and change the setting
Title of Lesson 77
ID document: e.g. PPT/L-TCIO-1
Graph Description Decision
Process under control :
-The R curves oscillate on
either side of the mean
PRODUCTION
-- 2/3 of the points are in
the central third of the
chart
Upper limit case Lower limit case
Point out of bounds - Short-term capability - Short-term capability
Table of
(Out of control) deteriorates. The origin must ameliorates
The last point crossed be found and actions taken - Measuring system is blocked
one of the control limits - Measurement error
Title of Lesson 79
ID document: e.g. PPT/L-TCIO-1
- Tool wear/fatigue
Title of Lesson 80
ID document: e.g. PPT/L-TCIO-1
- Change in material
- Change of load(=charge)
Title of Lesson 81
ID document: e.g. PPT/L-TCIO-1
Frequency of sampling
● Determining the frequency of sampling is mostly a matter of common sense.
Other factors, such as sample size, control time or control costs can also be
decisive. The goal is to obtain a correct picture of the process.
● When starting the process analysis with a control chart, it is wise to have a
relatively high sampling frequency and to decrease it when the process
stabilizes.
Title of Lesson 82
ID document: e.g. PPT/L-TCIO-1
Frequency of sampling
● The following points should be noted:
Title of Lesson 83
ID document: e.g. PPT/L-TCIO-1
Rules of verification
Title of Lesson 84
ID document: e.g. PPT/L-TCIO-1
Target
Correct setting et the end
Correct setting
LLC
Title of Lesson 85
ID document: e.g. PPT/L-TCIO-1
Synthesis :
Driving/Monitoring methods
Title of Lesson 87
ID document: e.g. PPT/L-TCIO-1
Driving methods
● 5 steps allow to summarize the control method with the control charts:
- 2. Measure the parts and transfer the points on the control chart
- 4. If necessary, intervene
Title of Lesson 88
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 89
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 90
ID document: e.g. PPT/L-TCIO-1
● When n > 5, S chart is more relevant than R chart because it uses information
of each measured value while R chart uses only information of Min and Max
value in the sample.
Title of Lesson 91
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 92
ID document: e.g. PPT/L-TCIO-1
LLC = 0 X 0,29 = 0
Title of Lesson 93
ID document: e.g. PPT/L-TCIO-1
CUSUM chart
Title of Lesson 94
ID document: e.g. PPT/L-TCIO-1
CUSUM chart
● The CUSUM chart is more efficient for detecting small or medium amplitude
drifts (1 to 2σ drift) on the average. To detect slow and progressive drifts in a
continuous manufacturing process, the CUSUM chart is very efficient because
it detects a drift on an average number of successive samples.
● The Cusum chart is not relevant for range drifts if the average remains the
same.
Title of Lesson 95
ID document: e.g. PPT/L-TCIO-1
CUSUM chart
● As with the EWMA map, we have a link between past and present averages:
With as parameters :
k : filter coefficient
Title of Lesson 96
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 97
ID document: e.g. PPT/L-TCIO-1
CUSUM chart
● Analyzis :
As long as Si+ and Si- are between h and -h: the process is under control. If they
are not between, the process is out of control.
● We will therefore use the following table to define these different values:
Remenber :
k is link to the deviation (expressed in standard deviation unit) you want to be warned about : the lower it is the more sensitive it will be
After a new setting on the machine, reset the Si calculation : S0+= S0- =0
Title of Lesson 98
ID document: e.g. PPT/L-TCIO-1
Title of Lesson 99
ID document: e.g. PPT/L-TCIO-1
EWMA chart
The EWMA (Exponentially Weighted Moving Average) control chart allows to
detect small deviations from the target. Indeed, each point established
previously is related to the following point; thus, for each value, the average
weighted by a coefficient λ is obtained :
With as parameters:
λ : constant less than or equal to 1. Generally : λ = 0.33
: initial value = target
: the average of two successive observations
EWMA chart
We finally have:
The lower the λ, the greater the detection of small drifts but the less sensitive
we will be to sudden and large drifts. λ is therefore chosen according to the
knowledge of the process: most often λ = 0.33.
EWMA chart
The bounds of the control chart are set to [target ± 3 σ], so if the means are
randomly distributed with a standard deviation equal to then the standard
deviation of the Xi distribution is :
With as parameters :
EWMA chart
One thus takes for lower and upper limits of the limit of when one makes
tend i towards +/- infinity (this limit being reached very quickly when i
increases).
EWMA chart
● The limits therefore depend on the number of the sample, but they
converge very quickly to a straight line when i increases, which is the reason
for the mathematical simplification. A complementary reasoning in deviations
from the target can be done in the case of small series when the process
performs a repetitive step.
Documents Management
Qualité 1 – Chapitre 3 – Coûts d’obtention de la qualité
Title: ID No.:
(COQ)
Root:
Description:
Sébastien
1.0 30/08/2021 Initial document
Lasserre