0% found this document useful (0 votes)
40 views

Control Charts: 2WS02 Industrial Statistics A. Di Bucchianico

This document discusses control charts, including: 1) The goals of further discussing variable and attribute control charts as well as special charts. 2) The differences between a process being statistically in control versus technically in control. 3) Key aspects of Shewhart control charts including their graphical display, rational subgrouping, and how control limits are determined based on the process mean and standard deviation.
Copyright
© © All Rights Reserved
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
40 views

Control Charts: 2WS02 Industrial Statistics A. Di Bucchianico

This document discusses control charts, including: 1) The goals of further discussing variable and attribute control charts as well as special charts. 2) The differences between a process being statistically in control versus technically in control. 3) Key aspects of Shewhart control charts including their graphical display, rational subgrouping, and how control limits are determined based on the process mean and standard deviation.
Copyright
© © All Rights Reserved
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 72

Control charts

2WS02 Industrial Statistics


A. Di Bucchianico


Goals of this lecture
Further discussion of control charts:

– variable charts

• Shewhart charts

– rational subgrouping

– runs rules

– performance

• CUSUM charts

• EWMA charts

– attribute charts (c, p and np charts)

– special charts (tool wear charts, short-run charts)


Statistically versus technically in control
“Statistically in
control”
• stable over time /

• predictable

“Technically in control”

• within specifications


Statistically in control vs technically in control
statistically controlled process:
– inhibits only natural random fluctuations (common causes)
– is stable
– is predictable
– may yield products out of specification
technically controlled process:
– presently yields products within specification
– need not be stable nor predictable


Shewhart control chart
graphical display of product characteristic which is important for
product quality

Upper
X-bar Chart for yield Control
14,4
Limit
14,2

Centre Line
X-bar

14

13,8

13,6
Lower
0 4 8 12 16 20 Control
Subgroup
Limit


Control charts


Basic principles
take samples and compute statistic

if statistic falls above UCL or below LCL, then out-of-control signal:

X-bar Chart for yield


14,4

14,2
X-bar

14 how to choose control limits?


13,8

13,6
0 4 8 12 16 20
Subgroup


Meaning of control limits
limits at 3 x standard deviation of plotted statistic

basic example:

P ( LCL  X  UCL)  UCL

P (   3 X  X    3 X ) 
P ( 3 X  X    3 X ) 
LCL
P ( 3  Z  3)  0.9973


Example
diameters of piston rings

process mean: 74 mm

process standard deviation: 0.01 mm

measurements via repeated samples of 5 rings yields:

0.01
x    0.0045 mm
n 5
UCL  74  3( 0.0045)  74.0135 mm
LCL  74  3( 0.0045)  73.9865 mm

Individual versus mean
group means

individual
74,03 observations

Centre line

73,97
1 10


Range chart
• need to monitor both mean and variance
• traditionally use range to monitor variance
• chart may also be based on S or S2
• for normal distribution:
– E R = d2 E S (Hartley’s constant)
– tables exist
• preferred practice:
– first check range chart for violations of control limits
– then check mean chart


Design control chart
• sample size

– larger sample size leads to faster detection

• setting control limits

• time between samples

– sample frequently few items or

– sample infrequently many items?

• choice of measurement


Rational subgroups
how must samples be chosen?

choose sample size frequency such that if a special cause


occurs

– between-subgroup variation is maximal

– within-subgroup variation is minimal.

within subgroup
variation

between subgroup

 variation
Strategy 1

• leads to accurate estimate of 

• maximises between-subgroup variation

• minimises within-subgroup variation

process mean


Strategy 2
•detects contrary to strategy 1 also temporary changes of process
mean

process mean


Phase I (Initial study): in control (1)


Phase I (Initial study): in control (2)


Phase I (Initial Study): not in-control


Trial versus control
•if process needs to be started and no relevant historic data is
available, then estimate µ and  or R from data (trial or initial study)

•if points fall outside the control limits, then possibly revise control
limits after inspection. Look for patterns!

•if relevant historical data on µ and  or R are available, then use


these data (control to standard)


Control chart patterns (1)

Control chart of height


10
UCL = 10.00
H eig h t

6 CTR = 0.00
2 LCL = -10.00

-2

-6

-10
0 3 6 9 12 15 18

Observation
Cyclic pattern,

three arrows with different weight


Control chart patterns (2)

Control chart of height


10
UCL = 10.00
Height

6 CTR = 0.00
LCL = -10.00
2

-2

-6

-10
0 4 8 12 16 20

Observation
Trend,

course of pin


Control chart patterns (3)

Control chart of height


10
UCL = 10.00
H eight

6 CTR = 0.00
LCL = -10.00
2

-2

-6

-10
0 4 8 12 16 20

Observation
Shifted mean,
Adjusted height Dartec


Control chart patterns (4)

A pattern can give explanation of the cause

Cyclic  different arrows, different weight

Trend  course of pin

Shifted mean  adjusted height Dartec

Assumption: a cause can be verified by a pattern

The feather of one arrow is damaged  outliers below


Phase II: Control to standard (1)


Phase II: Control to standard (2)


Runs and zone rules
•if observations fall within control limits, then process may still be
statistically out-of-control:
– patterns (runs, cyclic behaviour) may indicate special causes
– observations do not fill up space between control limits
•extra rules to speed up detection of special causes
•Western Electric Handbook rules:
– 1 point outside 3-limits
– 2 out of 3 consecutive points outside 2  -limits
– 4 out of 5 consecutive points outside 1  -limits
– 8 consecutive points on one side of centre line
•too many rules leads to too high false alarm rate



Warning limits

•crossing 3  -limits yields alarm

•sometimes warning limits by adding 2  -limits; no alarm but


collecting extra information by:

– adjustment time between taking samples and/or

– adjustment sample size

•warning limits increase detection performance of control chart


Detection: meter stick production
• mean 1000 mm, standard deviation 0.2 mm

• mean shifts from 1000 mm to 0.3 mm?

• how long does it take before control chart signals?


Performance of control charts
expressed in terms of time to alarm (run length)
two types:
– in-control run length
– out-of-control run length

X-bar Chart for yield


14,4

14,2
X-bar

14

13,8

13,6
0 4 8 12 16 20


Subgroup
Statistical control and control charts
•statistical control: observations
– are normally distributed with mean  and variance 2
– are independent
•out of (statistical) control:
– change in probability distribution
•observation within control limits:
– process is considered to be in control
•observation beyond control limits:
– process is considered to be out-of-control


In-control run length
•process is in statistical control

•small probability that process will go beyond 3  limits (in spite of


being in control) -> false alarm!

•run length is first time that process goes beyond 3  limits

•compare with type I error


Out-of-control run length

•process is not in statistical control

•increased probability that process will go beyond 3  limits (in spite


of being in control) -> true alarm!
•run length is first time that process goes beyond 3 sigma limits

•until control charts signals, we make type II errors


Metrics for run lengths
•run lengths are random variables

– ARL = Average Run Length

– SRL = Standard Deviation of Run Length


Run lengths for Shewhart Xbar- chart
in-control: p = 0.0027

• time to alarm follows geometric distribution:


– mean 1/p = 370.4
– standard deviation: ((1-p))/p = 369.9

UCL

LCL
0.9973 0.99730.99730.0027


Geometric distribution

Geometric Distribution
(X 0.001)
3 Event prob.
2.5 0.0027
2
1.5
probability

1
0.5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
(X 1000)


Numerical values
Shewhart chart for mean (n=1)

single shift of mean:

 P(|X|>
3) ARL SRL

0 0.0027 370.4 369.9

1 0.022 43.9 43.4

2 0.15 6.3 5.3

3 0.5 2 1.4

Scale in Statgraphics
Are our calculations wrong???

ARL Curve for X-bar


400
350
300
250
veragerunlength

200
150
100
50
A

0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Process mean


Sample size and run lengths
increase of sample size + corresponding control limits:

– same in-control run length

– decrease of out-of-control run length


Numerical values
Shewhart chart for mean (n=5)

single change of standard deviation ( -> c)

c P(|Xbar|>3 ARL SRL

1 0.0027 370.4 369.9

1.1 0.0064 156.6 156.1

1.2 0.012 80.5 80.0

1.3 0.021 47.6 47.0

1.4 0.032 31.1 30.6

1.5 0.046 22.0 21.4


Runs rules and run lengths
• in-control run length: decreases (why?)

• out-of-control run length: decreases (why?)


Performance Shewhart chart
•in-control run length OK

•out-of-control run length

– OK for shifts > 2 standard deviation group average

– Bad for shifts < 2 standard deviation group average

•extra run tests

– decrease in-control length

– decrease out-of-control length


CUSUM Chart
plot cumulative sums of observation

change point


CUSUM tabular form

assume

– data normally distributed with known 

– individual observations


i 
C  max 0, X i    0  K   C 
i 1 
C 
i  max 0,   0  K   Xi  C 
i 1 
 
C  C  0; alarm if max C , C
0 0  
i

i H

Choice K and H

•K is reference value (allowance, slack value)

•C+ measures cumulative upward deviations of µ0+K

•C- measures cumulative downward deviations of µ0-K

•for fast detection of change process mean µ1 :

– K=½ |µ0- µ1|

•H=5 is good choice


CUSUM V-mask form

UCL

CL

change point

LCL

Drawbacks V-mask
• only for two-sided schemes

•headstart cannot be implemented

•range of arms V-mask unclear

• interpretation parameters (angle, ...) not well determined



Rational subgroups and CUSUM
• extension to samples:

– replace  by /n
• contrary to Shewhart chart , CUSUM works best with
individuals


Combination
•CUSUM charts appropriate for small shifts (<1.5)

•CUSUM charts are inferior to Shewhart charts for large


shifts(>1.5)

•use both charts simultaneously with ±3.5 control limits


for Shewhart chart


Headstart (Fast Initial Response)

•increase detection power by restart process

•esp. useful when process mean at restart is not equal at target


value

•set C+0 and C-0 to non-zero value (often H/2 )

•if process equals target value µ 0 is, then CUSUMs quickly return
to 0

•if process mean does not equal target value µ 0, then faster
alarm


CUSUM for variability

•define Yi = (Xi-µ0)/  (standardise)


•define Vi = (|Yi|-0.822)/0.349

•CUSUMs for variability are:


i 
S  max 0,Vi  K /   S 
i 1 
S 
i  max 0, K /   V i S 
i 1 
 
S  S  0; alarm if max S , S
0 0  
i

i H

Exponentially Weighted Moving Average chart

• good alternative for Shewhart charts in case of small


shifts of mean
• performs almost as good as CUSUM
• mostly used for individual observations (like CUSUM)
• is rather insensitive to non-normality



EWMA Chart for Col_1
11.4
UCL = 11.00
11 CTR = 10.00
10.6 LCL = 9.00
EWMA

10.2
9.8
9.4
9
0 3 6 9 12 15
Observation


Why control charts for attribute data
•to view process/product across several characteristics

•for characteristics that are logically defined on a classification


scale of measure

N.B. Use variable charts whenever possible!


Control charts for attributes

Attributes are characteristics which have a


countable number of possible outcomes.
Three widely used control charts for attributes:
• p-chart: fraction non-conforming items
• c-chart: number of non-conforming items
• u-chart: number of non-conforming items per unit

For attributes one chart only suffices (why?).


p-chart

Number of nonconforming products is binomially


distributed
n x
P{D  x}    p 1  p 
n x
x  0,1,..., n
 x

D
sample fraction of nonconforming:pˆ 
n
p (1  p )
mean:   p ˆ 
2
variance pˆ

 n
p-chart
m mn
i 1
 i 1
p
i  pˆ D
i
average of sample fractions: m m

Fraction Nonconforming Control Chart:

p 1  p 
UCL  p  3
n
CL  p
p 1  p 
LCL  p  3
n


Assumptions for p chart

• item is defect or not defect (conforming or non-conforming)

• each experiment consists of n repeated trials/units

• probability p of non-conformance is constant

• trials are independent of each other


c-chart
•Counts the number of non-conformities in sample.

•Each non-conforming item contains at least one non-


conformity (cf. p chart).

•Each sample must have comparable opportunities for non-


conformities

•Based on Poisson distribution:

c k
Prob(# nonconf. = k) = e c
k!


c-chart
Poisson distribution: mean=c and variance=c

Control Limits for Nonconformities:

UCL  c  3 c
CL  c
LCL  c  3 c

is caverage number of nonconformities in sample



u-chart
monitors number of non-conformities per unit.

c •n is number of inspected units per


u sample
n
• c is total number of non-
conformities u
UCL  u  3
n
Control Chart for
Average Number of CL  u
Non-conformities Per u
Unit: LCL  u  3
n


Moving Range Chart
use when sample size is 1
indication of spread: moving range

Situations:
automated inspection of all units
low production rate
expensive measurements
repeated measurements differ only because of laboratory error


Moving Range Chart
calculation of moving range:
MRi  xi  xi 1
individual moving
measuremen range
ts
MR
UCL  x  3 UCL  D4 MR
d2
CL  MR
CL  x
LCL  D3 MR
MR
LCL  x  3
d2
d2, D3 and D4 are constants depending number of observations


Example:
Viscosity of Aircraft Primer Paint

Batch Viscosity MR Batch Viscosity MR

1 33.75 9 33.49 0.22

2 33.05 0.70 10 33.20 0.29

3 34.00 0.95 11 33.62 0.42

4 33.81 0.19 12 33.00 0.62

5 33.46 0.35 13 33.54 0.54

6 34.02 0.56 14 33.12 0.42

7 33.68 0.34 15 33.82 0.72

8 33.27 0.41
x  33.52 MR  0.48


Viscosity of Aircraft Primer Paint
since a moving range is calculated of n=2 observations, d2=1.128,
D3=0 and D4=3.267

CC for individuals CC for moving range


0.48
UCL  33.52  3   34.80 UCL  3.267  0.48  1.57
1.128
CL  0.48
CL  33.52
LCL  0  0.48  0
0.48
LCL  33.52  3   32.24
1.128


Viscosity of Aircraft Primer Paint
35
34.5
34
X 33.5
33
32.5
32
0 3 6 9 12 15
1.6
U
C
1.2
L

MR 0.8

0.4

0
0
 3 6 9 12 15

Tool wear chart
known trend is removed (regression)

trend is allowed until maximum

slanted control limits

USL
UCL
reset

LCL
LSL



Pitfalls
bad measurement system

bad subgrouping

autocorrelation

wrong quality characteristic

pattern analysis on individuals/moving range

too many run tests

too low detection power (ARL)

control chart is not appropriate tool (small ppms, incidents, ...)

confuse standard deviation of mean with individual



You might also like