Quality control charts

Quality control charts: Background

Quality control charts: Description
3-sigma chart
Quality Control Charts: Performance
Type I and Type II error for the 3-sigma chart
Average Run Length
The OC curve
X, S control chart: unknown parameters
X, R control chart
X, S control charts when the sample size is variable
X, MR control charts

Quality control charts: Background

Statistical Quality Control
Use of statistical and engineering methods that are
used in measuring, monitoring, controlling and
improving quality characteristics

Quality control charts: Background

Methods for quality control and improvement
Design
of
experiments:
An
approach
to
systematically vary the controllable input factors in
the process and determine the effect these factors
have on the output product parameters.

Acceptance sampling: Inspection and classification

of a sample of units selected at random from a
larger batch or lot and the ultimate decision about
disposition of the lot.

Quality control charts: Background

Methods for quality control and improvement

Statistical
process
control:
Techniques
for
monitoring
and
surveillance
of
quality
characteristics. These techniques help to detect
when changes in the inputs are required to bring
the process back to an in-control state.
Regression: Model the relationship between the
influential input variables and the output quality
characteristics (e.g., linear, multivariate, logistic,
time series and differential equation)
Engineering control: Routine adjustment of the
process so that future values of the product
characteristics will be approximately on target

Quality control charts: Background

Methods for quality control and improvement

Design of experiment
Acceptance sampling
Statistical process control (SPC)
Regression
Engineering control

Main statistical process control tools

Histogram
Control
charts
Pareto chart
Cause and effect
diagram

Defect
concentration
diagram
Scatter diagram

Quality control charts: Background

Chance causes (common causes): Small causes of
variability that are inherent to a system or process
and operate randomly or by chance
Assignable causes (special causes): Have relatively
large effects on the process and are not inherent to
it. The circumstances or factors that cause this kind
of variability can be identified (e.g., improperly
adjusted machines, operator errors, or defective
raw materials)
A stable process is in a state of statistical control
and has only chance or common causes of
variability operating on it
A process is said to be out of statistical control if

Quality control charts: Description

The control chart is an online process-monitoring

technique widely used to quickly detect the occurrence of
assignable causes

Once the assignable causes are detected, investigation

and corrective actions should be undertaken before many
nonconforming units are manufactured

To create a control chart, samples have been selected

from the process of interest at a sequence of time points

Depending on the aspects of the process under

investigation, some statistic, such as the sample mean or
sample proportion of defective items is chosen

The values of this statistic is then calculated for each

sample in turn and plotted over time

Quality control charts: Description

Simple control charts have

A Center Line (CL), which can be a process target value or

estimated from the data
Upper Control Limit (UCL) and Lower Control Limit (LCL)
Points within the two limits: In-control. Only reflects random
natural variation
Point outside the limit(s): Out of control. Variation due to an
assignable cause
A TYPICAL CONTROL
CHART
UCL

CL

LCL

Quality control charts: Description

Types of control charts (univariate)
Attribute control charts: Used whenever samples or
subgroups of process outputs are evaluated and the
number of nonconforming items in a sample are counted
(or the number of nonconformities per item are counted)
Variable control charts: Used when the quality
characteristics measured are on the continuous scale
(height, weight, temperature, time)

3-sigma chart
TheX
chart based on known parameter
values
Let X be a random variable for quality characteristic
X
and
let X~N(,)
If
denotes the sample mean for a random sample of
/ n time X
E ( Xn) selected

size
at a Xparticular
point, then following
is true:
P( 3 X X 3 X ) P (3 Z 3) .9974

1.
2.
normally distributed
3 / n

3 / n

3.

Then:
And

LCL =

UCL =

is

3-sigma chart
Example
Piston rings for an automotive engine are produced by a
forging process. Twenty five samples, each of size five,
have been taken when the process was perceived as in
control (see Piston_ring_data). Extensive experience
with this process suggests that when the process is in
control, the inside diameter of the rings is normally
distributed with mean 74 and standard deviation 0.05.
Create a control chart for the mean inside diameter of
the rings.

3-sigma chart
Example
X_bar chart Piston ring diameter - known parameters
74.08
UCL=74.0671

74.06

Sample Mean

74.04
74.02
__
X=74

74.00
73.98
73.96
73.94

LCL=73.9329

73.92
1

11

13

Sample

15

17

19

21

23

25

Quality control charts: Performance

Hypothesis testing and control charts
Ho: Process is in control (sample gives a point inside control
limits)
Ha: Process is out of control (sample gives a point outside
control limits)
Type I error = = P(Fail to reject Ho/Reject Ho)
Type II error = = P(Reject
Ho/Fail to reject Ho)
A TYPICAL CONTROL
CHART
UCL

CL

LCL

Quality control charts: Performance

A chart will be effective if:
It gives a very few out-of-control signals when the
process is out of control
Ho: Process is in control (sample gives a point inside
control limits)
Ha: Process is out of control (sample gives a point outside
control limits)
Low Type I error = = P(Process is out-of-control/
X
Process is in-control)
Example: Assume that you have a process that is currently
in control, with a mean o and known What is the
error for an
chart with 3-sigma limits?

Quality control charts: Performance

AVERAGE RUN LENGTH (ARL)
Expected number of samples taken before a shift (i.e.
from in-control to out-control) is detected
For the in-control process

ARL = 1/

Example: Assume that you have a process that is

currently in control, with a mean o and known X
What is the ARL for an
chart with 3-sigma limits?
How do you judge the performance of the chart
based on the ARL?

Quality control charts: Performance

A chart will also be effective if:
It shows a point outside the control limits as soon as
the process is in control
Ho: Process is in control (sample gives a point inside
control limits)
Ha: Process is out of control (sample gives a point outside
control limits)
Type II error = = P(Process is in-control/Process is out-ofcontrol)
1- = P(Process is out-of-control/Process is out-of-control)
X
Example: Assume that you have a process that is currently
out-of-control,
with known anda mean 1=o + kWhat is the error

Quality control charts: Performance

The OC curve
It is a graph
showing the
relationship
between , k, and n.
It helps us to select
the size of the
sample to be used
in a control chart
What is the risk of
detecting a shift of
1 for a sample of

Quality control charts: Performance

AVERAGE RUN LENGTH (ARL)
Expected number of samples taken before a shift (i.e.
from in-control to out-control) is detected
For the out-of-control process

ARL = 1/1-

Example: Assume that you have a process that is

currently out-of-control, with known anda mean
1=
chart with 3X o + 1What is the ARL for an
sigma limits? How do you judge the performance of
the chart based on the ARL? Assume your sample
size is equal to 5.

X, S control charts: unknown

parameters
Let X be a random variable for quality characteristic
and let X~N(,)
Suppose that k samples are available, each containing
n observations (In practice, select at least 20 or 25
when the process is believed to be in control)

x1 , x2 ,..., xk
Let

an estimator for

be the average of each sample. Then

, the process average is
k

x
i 1

X, S control charts: unknown

parameters
TheX
chart based on estimated parameter
values
Remember the control chart values when and are
3 / n

3 / n
known:
And

LCL =

CL =

UCL =

x 3 chart
x A3 s
x
3
x to
A3 sbe estimated
Whenx and
need
the control
c4 n
c4 n
values are:
LCL =

s
Where

x
n

s
i 1

j 1

si

CL =

UCL =

n 1
,

, and c 4 and

X, S control charts: unknown

parameters
TheS
chart based on estimated parameter
values
Let X be a random variable for quality characteristic
and let X~N(,)
Suppose that k samples are available, each containing
n observations (In practice, select at least 20 or 25
s1 , sthe
sk
2 ,...,process
when
is believed to be in control)
s
Let
be the standard deviation of each
sample (defined in previous slide), and
is the
standard deviation averaged over all samples

X, S control charts: unknown

parameters
TheS
chart based on estimated parameter
values
2
1 c4
s 3
s B4 s
c4

UCL =

CL =
1 c42
s 3
s B3 s
c4
LCL =

Where B3 and B4 are constants that depend on the sample

size (look at the handout Appendix)

X, S control charts: unknown

parameters
Example
Piston rings for an automotive engine are produced by a
forging process. Twenty five samples, each of size five,
have been taken when the process was perceived as in
control (see Piston_ring_data). Create a control chart
for the mean and standard deviation of the inside ring
diameter.

X, S control charts: varying sample

size
If ni
is the number of observations
ithin the
then use
k
k

2
n
x
i i
ni 1 Si

x i1k
S i 1 k

and
ni
ni k
i 1

i 1

sample,
1/ 2

The control chart values ae calculated with the X, S

equations but the constants A3, B3, and B4 will depend on
the sample size used in each individual subgroup

X and R control charts

TheX
chart based on estimated parameter
values
Let X be a random variable for quality characteristic and
let X~N(,)
Suppose that k samples are available, each containing n
observations (In practice, select at least 20 or 25 when
x1 ,process
x2 ,..., xn is believed to be in control)
the
If

is R
a sample
of size n, then the range
x max x min

of the sample is the difference between the largest

, R2 ,...,
Rk
andR1the
smallest
observations; that is,
R R2 Rk
R 1
k
Let
be the ranges
of the k samples. The
average range is

X and R control charts

TheX
chart using range to estimate standard
deviation

3
x
R x A2 R
UCL = d 2 n

CL =

LCL =

3
d2 n

R x A2 R

Values for A2 can be found in the Handout Appendix table

X and R control charts

TheR

chart to monitor variability

UCL =R 3d 3

R
D4 R
d2

CL = R
LCL =R 3d 3 R D3 R
d2
Values for D3 and D4 can be found in the Handout
Appendix table

X, MR control charts
Control charts with individual values
These are charts where the sample size used for
process monitoring is n-1
For these charts, it is possible to use the moving of two
successive observations as the basis of estimating the
process variability. The moving range is defined as

MRi xi xi 1
Where

xi
xi 1

is the value for observation i

is the values for observation i-1

X, MR control charts
Let k be the number of samples (observations) used to
build the control chart. We can estimate process
variability by the average of the k-1 moving range
k
values
MR
MR

i2

k 1

X control chart parameters

UCL x

d2 n

MR x

3
MR x 2.66 MR
d2

CL x
LCL x

3
d2 n

MR x

3
MR x 2.66 MR
d2

X, MR control charts
MR control chart parameters

MR
UCL MR 3d 3
D4 MR 3.267 MR
d2
CL MR

LCL MR 3d 3

MR
D3 MR 0.000 MR
d2