2053867777/2053868265/Rcb Plus Q1 Training Rcb Process Total Quality & Productivity Improvement Principles Tools Techniques Rcb Plus R1060 Act 01mfp770 Singletary, Laura 6772

Key Points about SPC
"A state of statistical control is not a natural state for a process. It is
indeed an achievement, arrived at by elimination, one by one, by
determined effort, of special causes of excessive variation."
- W. Edwards Deming
State of Statistical Control-the condition describing a process from
which all special causes of variation have been eliminated and only
common causes remain.
Economical State of Control-a perfect state of control is rarely attain-
able in a production process. A controlled process is considered to be
one where only a small percentage (less than 5%) of the points go out
of control n~d where out of control points are followed by proper action.
Control Limits-derived from process performance, they define proc-
ess variation by showing natural level of variation to be expected.
Specifications those requirements or needs of the process that are
imposed upon it.
To be in a state of statistical control and to meet specification are two
separate (but related) issues.
For long-term survival and for economic reasons, being in a state of
statistical control is far more important than merely meeting specifica-
tion.
Control charts can be utilized in many different applications as long as
the output you are interested in is quantifiable.
Use of a control chart system to monitor and take action on a process
at the time of production, is far more beneficial than measuring the
product after the fact to determine its acceptability.
Control charts are not a remedy in themselves, it is the action and
decisions that result from them that make the difference.
Ry

Page 6: lss16e00

Simulation of a Process: Quincunx Demonstration
To best illustrate the use of a control chart system to control a process
and the benefits over traditional methods, we will simulate a process
through the use of "Quincunx." Although we are dealing with beads and
the Quincunx, as we go through this exercise try to visualize the applica-
tion to any BL process-whether we are trying to control blend ratios, OV
after Burley Dryer, Gas Consumption, QB-IA differences, PG level at
weighbelt, Daymix Solids, Forming Line OV, or Sheet Weights, Downtime
or any other processes or product quality parameters. We will use these
results to look at the application of control chart concepts and then look at
some specific examples.
Problem
We are a major buyer of red beads looking for a good supplier. We have
found one which we wish to investigate - ABC Bead Corporation. We
have weight requirements for the type of bead we desire: target of 13 oz.
You would like all beads to be 13 oz. But might be willing (reluctantly) to
accept beads with weights as low as~9-oz_ and as high as f7 oz. ABC has
quoted a price of $22 per bead. We will stress to ABC Beadthat we really
would like all beads at 13 oz. We will now watch their process, record
some of the results we see and draw some conclusions.
Questions to Think About
1. Does the process seem to be producing product that meets our
expectations?
2. Does the process exhibit a state of control? Are there special causes
present?
3. Is ABC Bead a good supplier? Why or why not?
4. What can be done to improve the bead weight performance of
ABC Bead?
Procedure
As the process is producing beads, we will take a sample, perform a
weight measurement and record the data on the following sheet. Record
any other significant "events" that occur next to that hour.

Page 7: lss16e00

Simulation Results
TIME
BEAD WEIGHT
NOTES
2330 f~
0030 J
0130 ~~
0230 l~ t l 1~ U
0330
0430 ,
0530 f
0630 .
0730 ir e f t ~ ~~°V
0830
0930
1030 F
P6F J 5`
~~~5~
~i71
1130 `
17,
1230
60
9 .`(~~
1330 f
lo` lv
/0
1430
1530
1630
1730
1830
1930
2030
2130
2230
N
a
G't
W
Go
~ ft'l
CD
Z4-7 ~~
O

Procedure to Derive Control Limits: X Control Charts
0 1. Gather data over representative time period. Plot on run chart.
2. Remove suspicious data.
3. Check to see if data is normal.
4. Calculate average (X ) and standard deviation(s) of the data.
5. Derive UCL (Upper control limit), LCL (Lower control limit).
UCL = X +3 s
LCL = X-3s
6. Check data for control. Delete any data outside the control limits.
If any data is eliminated, repeat steps 4-6 until the data settles down
(all data within the control limits).
7. Determine if X compatible with target.
8. Set up control chart for use on the process.
Worksheet For Simulation
Our objective is to:
1. Derive process control limits for average and range.
2. Determine process capability.
I. Deriving Control Limits (we will follow the 8-step procedure)
N
O
G't
C.1
~
El 4-9 C~D~
N

OVERCONTROL VS UNDERCONTROL
! As you have seen from the simulation, without the use of an accurately
established control chart scheme, it is very easy to commit two types of
errors when interpreting data to act on a process...
Undercontrol-not reacting or taking action when you should be-there
is something special taking place in the process that you should be react-
ing to, but you dismiss it as normal variation.
Overcontrol-reacting or taking action when there is no need to-there
are no special causes present in the system and the value you have just
obtained, although it appears high or low, is to be expected because of the
system causes of variation. Should not overreact.
Both types of errors cause more variation to occur than need be in the
process. The table below summarizes what could happen when trying to
control a process.
True Process Condition
Stable
(Process centered
at target)
Special Cause
Present
(Process not
centered at target)
Make a Overcontrol Correct
Move (Type I Error) Acti o n
A
ti
c
on
Leave Co rrect Undercontrol
Alone Action (Type II Error)
Establishing sound control limits and following control rules through stan-
dardized control moves will absolutely minimize the Type I and Type ff
errors that are made on the RCB processes.
Note:
Whenever tracking data, a data point must either go up or go down
compared to the last one. The purpose of SPC is to interpret the data N
to see if the process truly shifted or are we being tricked by the data O
which reflects random fluctuation. ~
P4-11 -15
~
~

Page 12: lss16e00

CONTROL RULES: TESTS FOR STATE OF
STATISTICAL CONTROL
In a continuous process like the RCB process it is often not economical to
wait for a signal that a single point has gone beyond the control limits. At
the same time we do not want to second guess the logic and probabilities
of the control chart. The following tests for control will allow us to detect
process trends and shifts earlier without altering our chances of taking the
wrong action. These are the control rules adopted for all X control charts.
For simplification, the letters A, B and C have been used to separate the
chart into zones corresponding to 1, 2 and 3 standard deviations.
First Test
One single point falls beyond Zone C (outside of 3s)
F
E
D
X
~ I _ 1i-[ ~ _1iu I_ L u_ i_i
A
a
Q
E
F
Second Test
Two out of three points fall on the same side of A and beyond
Zone B (outside of 2s)
F
E
D
C r - ;
B
I
A
A
D
E
F
4-12 IPF

QB-IA Difference: Production Dust OV
.93
X
3
F a . . I
p . . .~ . . . . . . . . . . ~ . .
C ~ f .
~ . . . . . . .
A ; . 'I
a r ,
C !
E . . . . . .
F
QL ® Ol~ l0 0 Q Gm sm O tL~ Tm ta ~ SO m D Ri n lO S bi ~7
ATA
PG Level in Slurry
5al
X 4-`+9
31-7
D
ATA
F
av
r
E
D
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
. .
.
.
.
.
.
C . . . . . .
' ~ f
~
A
A . . . . . . . . . . . . . . . . . .
~
C . . . .
E . . . . . . . . . . . . . . . . . .
F . . . . . . . . .
1
. . . . .
m1 ®
9tD
®
~
10)
~
oQl
~
O
1~
tO~
tSA
~
fO
~
T7~ ~
.
4-16 11H
E
D
C"`L
®
A
A
!
E
E
D
c v`c-"
a
A
D
E

SEPARATING PROCESS VARIATION:
S YSTEM & SPECIAL CA USES
Key Points About Interpreting Process Variation
1. Variation always exists.
2. Any process has both chance (or system) causes and assignable (or
special) causes which create variation.
3. The major purpose of SPC is to separate results-system variation
from special cause variation-and serve as a basis for action.
4. Eventually, the question for management becomes, how much system
variation are we willing to tolerate?
B
D~
B
~
.
~~ 0 ---------
Is A a special cause? Letting the process "talk
B? to us" via control {imits
C? allows clear separation
D? of causes.
You will remember that the 85/15 rule states:
85% of variation is due to System Causes
15% of variation is due to Special Causes
Since total measured variation can be viewed as coming from two sources
---this results in two types of problems to be solved-using distinctly
different approaches.
4-18 EM
~

Page 19: lss16e00

Type of Problem Definition
is Special Cause
System (Chronic)
Causes
Statistical evidence shows that the result is not
typical for current process. A unique cause is
present and acting on the process.
Level of variation is typical, based on the expected
variation of all process variables. There is no
unique cause currently acting on the process.
85/15: Managing the Improvement Process
The SPC System, if actively utilized, can do two things:
1. Give feedback to attendants (and others) so appropriate actions can
be taken in a timely manner to maintain the process in a state of
control.
2. Give guidance and feedback to system improvement efforts or other
changes made to the process, methods or raw materials. It will objec-
tively tell you what is working and what is not.
Ol The flowchart for improvement is to Stabilize (standardization), Control,
Improve. As we attempt to control and improve the process, this se-
quence can be accomplished through the steps shown below and on the
next page.
"Improvement of Quality only comes about by improving the proc-
ess. Improvement measures can only be taken after a process is in
control." - W. Edwards Deming
E4-19 ~
G.~
N

S
85/15 Principle
Variation in any measured result comes from two sources
- system variation and special cause variation due to
special or unique causes. In most processes 85% of the
variation comes from the system, 15% from special
causes.
System variation - that routine variation caused by ran-
dom fluctuation of system components. If this variation is
to be reduced the process/system needs to be changed.
[t is management's responsibility to lead this work.
Special Cause Variation -that variation of results
caused by atypical performance of the process, material,
man or method. Those closest to the process are best
equipped to react to and address this variation. Typically
15% of the variation in most processes.
S
Example: Line 1 OV
16.9
15.0
18.8
The amount of variation in Line 1 OV found during data collection was
16.9 +/- .63. This variation (a spread of 3.8) can be traced to the two
sources of variation - system and special causes.
What factors contribute to special cause contribution?
What factors contribute to routine system variation?
~
O
C1i
~
E4-23 IN'T
CD
C.7
C'~

Parameter: OV After the Sweco
Control
~
~ ~
AV
- ---- -- - ~
~
~
;
~
~
7
F ?
t. r ...........
0
1. 0
20
30
40
SAMPLE NUMBER
Capability
LSL Nomina2 USL
so
60
2.5 3.5 4.5 5.5 6.5 7.5 6.5 9.5
% ov
70
Statistics: x = 6.03
s = .72 if process is in control
:~
s = .89 currently
Expectations: 5% = +/- 2.0
Conclusions:
4-30 WE
K
S

Case Study #1: (X-chart "regular" method):
~ Line 2 Oil usage
Data was gathered daily by the A shift oiler on number of inches of oil
used daily. (This could have been converted to gallons and tracked in
gallons instead of inches). There were 82 data points collected.
First, a run chart was plotted and suspicious data identified. Notes
in off-standard logs, area log books or on the data log sheets themselves
were used to do this. In this case no suspicious data was identified.
Run Chart for Line 2 Oil Usage
1 /6/91 - 3/28/91
18
16
}
J
¢ 14
O
P
~ -_-,
r
I
0 20 40 60 80 100
SAMPLE hlUMBER
N
0
V1
Cj
~ 4-35 ~
~

Elementary statistics. Since the data appeared to be normal in shape,
statistics were then calculated from the sample data:
mean: 13.2876
standard deviation: 2.14863
n = 82
minimum: 6.73
maximum: 18
median : 13.5
Preliminary control limits. Using the statistics (R, sd,) from the
previous step, "first pass" or preliminary control limits were derived.
UCL=X+(3*sd)
LCL=X-(3*sd)
UCL = 13.2876 + (3 * 2.14863) = 19.7335
LCL = 13.2876 - (3 * 2.14863) = 6.8417
Checking data for outliers and trimming. Plotting the data against
these control limits showed 1 data point beyond the control limits (9th day)
which indicates a special cause occurred on that day.
Line 2 Oii Usage - First Pass
21
12
N
W
S
0
z
H
'I
4
4
Ni
1~'J. ~ I.!~k ~,
9
.r.
...
i
i saiia
r -~
0 20 48 60 88 100
` SAMPLE NUMBER
H

Page 38: lss16e00

Trim and recalculate elementary statistics. The 9th data point was
removed and the mean and standard deviation were recalculated.
2nd pass
X = 13.3683
s=2.0331
Recalculate control limits. Using the 2nd pass statistics, new control
limits were then calculated:
UCL 13.3683 + (3 *2.0331 ) = 19.4676
LCL 13.3683 - (3 >k 2.0331) = 7.269
Plotting the remaining data against these limits showed a data point below
the lower control limit.
Line 2 Oil Usage - Second Pass
21
i
.--~- ;--. -T-- .
18 ~}zi
J}
1
a
Q
W
:> I~f l. I l~~ Ifi( \t i fl it;Yii ~
~ 12
l1.1 ! ~{ 1 .` t 1 -~.
S p L~ ~
Z
0
~
,T.269
20 48 68 88 180
SAMPLE NUMBER
S
4-38 Eff

Page 39: lss16e00

Trim and recalculate elementary statistics. This outlying data point
was removed and for the "3rd pass", the sample statistics were:
S
Line 2 Oil Usage - Third and Final Pass
,
I,
:
_ ,
8 20 40 68 80
SAMPLE NUMBER
3rd pass
X= 13.4447
s = 1.9251
Recalculate control limits. Using the 3rd pass statistics, new control
limits were then calculated:
UCL 13.4447 + (3* 1.9251) = 19.22
LCL 13.4447 - (3 * 1.9251) = 7.67
Plotting the remaining data against these limits showed there were no
outliers - all data was within the control limits.
.3.
.i.
-i.
-a,
W4-39 Ct~
~

Page 40: lss16e00

The proposed control chart can now be set up with the center line at
13.44, UCL at 19.22, LCL at 7.67 and the zones evenly spaced between
the limits.
19.22 UCL
C
17.28 - - - - - - - - - -
B
15.36 - - - - - - - - - - - - - - -
A
13.44 x
A
11.52 B - - - - - - - - - - - - - - -
9.60 - - - - - - - - - - - - - - -
C
7.67 LCL
Process capability: The final trimmed data histogram appeared to be
normal in shape, therefore we can use the final R(13.44) and standard
deviation (1.92) for our Cpk calculations.
Line 2 Oil Usage Trimmed
24 r
8 F-
4 I-
r
~
J
er ~
8.5 10.5 12.5 14.5 16.5 18.5
INCHES REMOVED DAILY
4-40 WE

When we see this pattern, the moving range method to derive control
limits needs to be used. The model is very similar to the flow chart de-
scribed earlier. The biggest difference is in using the moving range (differ-
ence in value between two successive data points) to get a more accurate
estimate of standard deviation. If we did not use this technique, we would
end up with estimated control limits that would be too wide to fit our
process.
We will use the procedure to calculate control limits:
Procedure for Moving Range Method to Calculate
Control Limits for X Chart
Step
1. Run chart. Check for suspicious data
2. Histogram. Check for normality
3. Calculate moving ranges for all data points
4. Calculate average moving range (9R)
5. Calculate upper control limit for moving ranges
UCL (M ranges) = D4* MR
6. Check list of moving ranges to see if any exceed the
UCL (m ranges). If any do, remove the moving range
and the corresponding raw data point.
7. Continue steps 3-7 until all moving ranges fall within
limits.
8. Calculate Xof remaining data points
9. Calculate control limits for raw data using standard
deviation estimated by moving range.
UCL = X+ (E2 * MR)
LCL = X- (E2 * MR)
E2 = 2.66
10. Check raw data to insure it falls within the control limits.
If it does not, eliminate the data point and go back
through steps 3-9 until all data "settles down" --
falls within control limits.
4-42 EM

Step 2:
The histogram was produced and showed that these differences do form a
normal pattern of variation centered roughly around 0.
Frequency Histogram
Line 3 QB - IA Difference Original Data
1ee ~-
s0(-
20 F
-3.9 -1.9 0.1 2.1 4.1
QB - IA
Step 3
Calculate moving ranges for the 261 data points...
Data Movi ng Ra nge
-.12
-.29 .17
1.82 2.11 _
.53 1.29
1.11 .58
.27 .84
2.03 1.76
-.70 2.73
51 1
21
. . ~
.78 .27
0 CfT
0
4-44 W1

Page 45: lss16e00

Step 4
S
MR=.997
Step 5
UCLR,,ar,9e=D4*MR
= (3.269) (.997)
= 3.259
Step 6
Moving Range Chart for Line 3 QB - IA Difference
3
i
!
"
~
~
~
"
~..
.
~...:~.. .~....
~. ,.~i
F
~
0 5e 1@8 156 288 258 300
SAMPLE NUMBER
1 point exceeds control limit for moving ranges. Moving range and raw
data point (QB-IA difference value) removed.
N
C?
Cd[
C.3
09
~
2.
.s.
m
4-45 LrI
on

Step 10
X Chart for Line 3 QB - IA Difference
25@
-,-r,
3@0
e
~~,-.-r.----,-,
50 100
1.50
2@@
SAMPLE NUMBER
Looking through the 260 remaining QB-IA differences shows 5 data points
exceeding these limits. Remove these values and recalculate steps 3-10
r P
X=-.165
. MF-t =.943
UCL nnR = (3.269) (.943) = 3.08
1 moving range exceeds this value. Remove it and recalculate.
Moving Range Chart for Line 3 QB - tA Difference
~
0
z
a
~
a
z
M
~
0
E
1@@ 15@ Z@@ 25@ 388
0 56
0
SAMPLE NUMBER
W
~
4-47
~
C7~
0

All raw data falls within these control limits. That is, the variation that we
should expect to see whenever we get Line 3 QB reading and a matching
~ IA OV value and subtract them is a difference from 2.31 to -2.62. We
can now set up a chart to monitor this parameter and initiate the appropri-
ate control moves (including calibration) when the chart signals an atypical
process condition.
Line 3(QB - IA) Difference in OV Readings
2.31
1.49
.666
-.156
-.978
-1.80
-2.62
C
A - - - - - - - - - - - - - - -
A
- - - - - - - - - - - - - - -
B
- - - - - - - - - - - - - - -
C
UCL
X
LCL
The final histogram showing the variation in difference values that we
should expect to see if the Line 3 QB meter is functioning typically.
Line 3 QB - tA Difference Trimmed Data
8e
60
20
r
r-r, -. . , ~
r
r i I
-2.6 -1.6 -0.6 0.4 1.4 2.4
DIFFERENCE
~ 4-49 ~
cn
2V

Page 50: lss16e00

c Charts: When to Use, Model
Introduction
Sometimes when you look at a histogram of data it does not form the
typical normal pattern but takes on the skewed shape that we looked at in
an earlier section.
Normal Skewed (right)
In applications like these, it is inappropriate to use an x control chart to
monitor a parameter. Instead, the c chart must be used. (Doesn't matter
if it is skewed left or skewed right). You are likely to see this where there
is an artificial barrier that prevents the data from "spreading out" in both
directions or where you are dealing with discrete or counting type data or
in cases where there is a large opportunity for something to occur but you
don't see it occur that much.
4-50 Kff
Zti?
0
Ul
~
c7~
CD
M
CJ

Page 51: lss16e00

J
When To Use c Charts
Histogram shows skewed pattern- to the left or right
Dealing with counting - type data (discrete) and...
Opportunity for "occurrence" is large, but actual
occurrences" are low
Examples: Process stage downtime, number of accidents,
pounds of waste,number of hogsheads refed/day,
number of damaged hogsheads/boxes
Procedure For c Charts
1. Gather data based on a constant "window of time".
Plot run chart.
Remove suspicious data.
2. Histogram. Check to insure it is skewed in shape.
3. Calculate average (Z-)
4. Calculate control limits
UCL - c+ 3.=c
LCL= c-3.F-
5. Check data for "outliers" - any points beyond the control
limits
6. Remove outliers and repeat steps 3-5 until all data falls
within control
7. Set up the control chart with the center line at the median
and control limits at the final ones from step 6
~ 4-51 ~
~
~

Page 52: lss16e00

Case Study #3 (c Chart): RCB Floor Wastes
Scrap materials that are swept from the floor in Bay 4 and Packing were
weighed up once per shift by the cutter helper. The data was collected
over several weeks. The parameter that we are interested in tracking is
the total pounds per shift.
Step 1
Data (total pounds per shift) was gathered over a 5 week period of time
and plotted on a run chart.
Run Chart for RCB Floor Wastes
,
es r
. 2 c® 6® 80 s90
SAriPLE NUf18ER
Step 2
The histogram for the shift total waste values definitely showed the
skewed pattern. In this case, skewed to the right.
RCB Floor Waste
e®
40
s. ®
6
20
40
60
80
100
LHS OF WQS7E
4-52 ~

Page 53: lss16e00

Step 3
From the data, the calculated average pounds per shift was
~=23.1
Step 4
UCL=c+ 3.j~
= 23.1 +3~23.
= 37.5
LCL = c- 3 fe
= 23.1 --~(23.1)
=8.7
Step 5
From the graph, you can see 7 points that fall outside the upper control
limit. We assume that in each of these cases a special cause was present
in the process to produce this result. Since we are attempting to describe
system variation only with the control limits, these 7 points need to be
removed.
C Chart for RCB Floor Waste
1.ee f-
se[-
26 ~-
~]~ J 1~ . ~ - Mg
4 ~' ~' A I W\IIWV V Q Y
0
20
4@
Be
SAMPLE NUMBER
I
se
n
186
O
Ut
w
~
0 11
~
4-53 ~
~

Page 54: lss16e00

Step 6
Remove the 7 data points and recalculate c and the control limits
Second pass
c = 20.4
UCL=20.4+3 20.4=33.9
LCL=20.4-3 20.4=6.8
C Chart for RCB FLoor Waste
20 F
eL
~ JI~
~ i "i4
I ~ ~
e 20 48 6e ee
GpMP! F N/ IMRFP
There are now 3 points beyond the current limits. These need to be
eliminated and a new c calculated.
3rd ass
c = 19.8
UCL=19.8+3 19.8=33.1
LCL=19.8-3. 19.8=6.5
C Chart for RCB Floor Waste
soe
6e
~
¢ 60
~
a
~
~t 40
m
~ ~..-.
------------------
~ 1
z® f~V
eL
e 20 40 60 80
All data points now fall within the control limits. Now a control chart can
be set up to track, control, and improve the performance of this parameter.
4-54 K~E

Page 55: lss16e00

Step 7
The median of the final data set from step 6 is 19.
S
33.1
19
6.5
UCL
c
LCL
Note:
With this type of control chart only two of the control rules can be used:
single point beyond a control limit, 7 points on the same side of the center
[ine.I.
~~---~ ~
~
~ 4-55 ~
Gt~

X-RChart
0 When to use:
Used where you want to track average level and range of variation at
a point in time.
Need to gather data in "subgroup" form.
Example:
Suppose we wanted to track sheet weights on the forming line by taking
3 samples across the sheet every hour.
Sheet Weights
Hour Weig hts _
_(SubgroM Front Edge Middle Back Edae X g
7:00 11.8 10.9 11.4 11.7 .9
8:00 12.1 11.7 11.9 11.90 ,4
9:00 11.1 10.4 11.5 11.00 1,1
10:00 12.0 11.0 12.3 11.77 1.3
! 11:00 11.9 10.0 11.7 11.20 1,9
You would only use this chart if you were interested in and could
take control moves on both the X as well as the R.
You would have 2 charts to monitor an X portion and R portion.
UCL
X
LCL
X
UCL
R
X
~
0
~
~ LCL Go
~
~ 4-57 ~g
~
C

Page 58: lss16e00

Control Rules:
X: Use all 4 rules to interpret each data point
R: Use only the two rules that are used on the c chart
Procedure For Developing X-R Chart Control Limits
1. Gather data, organized by subgroup. Plot on run chart.
2. Calculate average and range for each subgroup.
3.Checkto see if subgroup averages are normal. (Histogram)
4. Calculate average range (R)
5. Derive UCL, LCL for range. Check_data for control. Delete any data
outside control limits. RUCL = D4 R, RLCL = D3 R. Repeat steps
3 and 4 if any points were eliminated.
Subgroup
Size
D3
D4
2 - 3.267
3 - 2.574
4 - 2.282
5 - 2.114
6 - 2.004
7 .076 1.924
8 .136 1.864
6. Calculate Grand Mean (X) of data remaining after range trimming.
Calculate standard deviation (sX) of subgroup averages.
7. Derive UCL, LCL for X.
UCL= X+3sX
LCL = 7-3sX
8. Check data for control. Delete any data outside control limits.
Repeat steps 5,6, and 7 if necessary. Continue this until data
"settles down." (All data within control limits.)
9. Determine if X is compatible with desired target.
If not-need to reconcile.
10. Set up control chart for process control.
R chart: Solid line at X, solid lines at UCLx, LCLX
R chart: Solid line at R, solid lines at UCLR, LCLR
4-58 WE

Page 59: lss16e00

Control Charts for Diagnostic Work
40 The most frequent use of control charts will be for on-fine SPC efforts as a
part of the RCB SPC System or with the SPC System created in mainte-
nance or QA through their Departmental Implementation efforts. How-
ever, control charts can also be a very effective tool to be used by indi-
viduals or teams in their off-line SPC efforts. In cases where problem-
solving work has been done to change/modify equipment, methods or
materials and you are trying to find out if the implemented change has
truly had a significant effect. It allows us to go beyond opinion or
feelings but rather use facts and data to objectively assess process
improvement efforts.
7~1
O
C1t
C.~
~
4-59 ~
~
r~

Page 60: lss16e00

Procedure
Procedure to use control chart technique to check for
significant change
1.Collect data (using appropriate technique - x, c, p, X-R)
before change is made or on one of the two "conditions"
you are comparing.
2.Derive control limits using the standard procedure.
3.Collect data for after the change is made or for the other
"condition" you are comparing.
4.PIot the data from step 3 on the control chart already
derived in step 2. Interpret for control.
(using appropriate rules)
5.lnterpret Results. If the second condition plots out of
control, then there is a significant difference between the
two alternatives. If it falls in a state of statistical control,
then there is no difference between the alternatives.
yw-. ~- ~J a:uG-f ~--
4-60 Eff

Page 61: lss16e00

CONTROLLING TO TARGET:
40 One of the benefits of a TQI SPC System such as the RCB Process
System is that it puts the tools and methods in the hands of attendants,
supervisors and others to consistently control a parameter around a
target. The target and the way in which it is controlled is the same
regardless of shift, regardless of what attendant is operating the process
that day or who is supervising that process stage. As we saw with the
concept of the Loss Function, the big advantage once everyone is operat-
ing within the system is that you can reduce variation and then set the
target to optimize cost and quality of the process and product.
Example: Forming Line 2 OV
Shown below is a histogram of OV readings from Line 2 gathered every
15 minutes over a period of several days. From the histogram, it appears
the data contains several populations. (Doesn't appear to be single
peaked but several peaks).
Frequency Histogram
Forming Line 2 OV
88
68
20
I
LJ
14 15 16 17 18
% OV
Looking at a run chart of this time period explains why there are several
populations. Although during this time period the fine was "running to one ~
target" (there were no instructions from the cutter man to adjust fine O
target), without the tools of SPG and everyone using the same system to ~
react to data - the result is many populations of product created. GO
~ 4-61 ~
~
~

Page 62: lss16e00

Forming Line 2 OV
iT F
r
,
t
14r
0 50 1.. 15® 2.0 250 300 35.
SAMPLE ltUKBER
Using the same scale and amount of variation - look what happens to
overall OV variation if the tools of SPC were used to react to OV data.
Frequency Histogram
Forming Line 2 OV Controlled to Target
~ 1 , ,~\\ !
I
14 15 16 17 1a
i ov
With this improved level of variation:
The target can be o ptimized to a g reater extent
True process capa bilities can be d etermined
The cost of low OV results (dryin g costs) can be reduced
O
The risk of high OV product failu re s can be reduced
The risk of z-test fa ilures can be re duced C3
4-62 EM