PROCESS CAPABILITY IMPROVEMENT BY PUTTING

‘STATISTICAL PROCESS CONTROL’ INTO PRACTICE

PIYUSH KUMAR SONI1, IMTIYAZ KHAN2 & ABHISHEK ROHILLA3
1,2,3
Mechanical Engineering, Mandsaur Institute of Technology, Mandsaur, India
E-mail : [email protected], [email protected], [email protected]

Abstract- Quality control helps industries in improvement of its product quality and productivity. Statistical Process Control
(SPC) is one of the tools to control the quality of products that practice in bringing a manufacturing process under control. In
this paper, the process control of a CNC Grinder manufactured at PMT Machines Ltd. Halol, (Gujarat) India is discussed.
The varying measurements have been recorded for a number of samples of a Cam Roller Shoe obtained from a number of
trials with the CNC Grinder. SPC technique has been adopted, by which the process is finally brought under control and
process capability is improved.

Keywords- Statistical Process Control, Range and Standard Deviation, Control Charts, Process Capability.

I. INTRODUCTION optimum cost savings and greater rigor, and produces

products of optimum quality [1].
Statistical Process Control (SPC) refers to controlling
a process (e.g., grinding) based on responding to SPC tools can be used by operators to monitor their
process data with statistical techniques and tools. part of production or service process for the purpose
Statistical process control (SPC) describes a widely- of making improvements [2].
used set of approaches used to detect shifts in
processes in, for example, manufacturing. Among A. Goals of SPC:
these are control charts" [1].  understand the process
 eliminate special cause variation
PMT Machines Ltd., Halol is one of the leading CNC  reduce common cause variation and maintain
Machine Tools manufacturers in India. The SPC a process that is in "statistical control" and
methodology have been adopted for grinding of Cam has high "process capability".
Roller Shoes, a component of one of its customer
companies, Delphi TVS, Chennai (India). B. Errors and variation can arise from two kinds
of causes:
As the thickness of shoes in different samples after  Special Causes: (assignable, bias, local
grinding was found to be out of tolerance limits asked variation), error/variation results in one
by Delphi TVS, the process capability found to be direction (either + or -) and can be traced to
less than the standard value. This required the idea of an assignable, special cause, e.g.,
SPC implementation and the techniques been miscalibrated instrument. It can be detected
practiced. by running known standards and
recalibrating [3].
II. LITERATURE REVIEW  Common Causes: (random, system variation)
error/variation results randomly (without
The use of statistical concepts in the field of quality bias) in both directions (+ and -) and in
emerged in the United States in the beginning of the varying magnitude – due to unknown causes.
nineteenth century. But its democratic use began only Random variation is chronic (continual), e.g.,
in the 1930s. W. Edwards Deming, who applied SPC normal fluctuations in instruments, natural
methods in the US during the Second World War, variation in raw materials.
was the one responsible for introducing this concept
in Japan after the war ended. These methods were not Statistics is more applicable to measuring and
used in France until the 1970s. The 1980s saw the controlling variation from common cause (random)
SPC methods being used frequently, due to the than from special causes [3].
pressure from large clients like automobile
manufacturers and aircraft manufacturers. C. Variations can be reduced by:
 Fundamental Point: Special causes and
Companies who have been operating in the market random causes of variation are treated
for a while already have a quality control process in differently.
place. This process enables a company to meet four
 Juran’s 85% Rule: 85% of variation is
main objectives: higher quality, more effectiveness,
random error in the system and can only be

International Journal of Power System Operation and Energy Management ISSN (PRINT): 2231 – 4407, Volume-2, Issue-1,2
109
 Process Capability Improvement by Putting ‘Statistical Process Control’ into Practice
remedied by management make changes to x is the standard deviation of the sample.
the system. 15% of variation is special cause (pronounced sigma).
and is fixable by the worker.  The arithmetic average (mean) of ranges,

R   i 
D. Usefulness of Control Charting:  R (4)
 n 
Control charts are also known as Shewhart charts or 
process-behaviour charts. Variable control charts are  Process (or population) Standard Deviation,
used to study a process when characteristics is a σ = R / D2 (5)
measurement, for example, cycle time, processing Where, D2 is the factor obtained from tables of
time, waiting time, highest, area, temperature, cost or constants used in constructing control charts.[5]
revenue [4].  Standard Deviation of the sample mean,
σx = σ /√N (6)
Control charts detects special causes of variation,  Average xi (Process mean), x = ∑ xi / n (7)
measures and monitors common causes of variation,
 Upper Control Limit, UCLx = x + 3σx (8)
helps to know when to look for problems and adjust or
when to keep hands off and when to make a  Lower Control Limit, LCLx = x - 3σx (9)
fundamental change [3].  Range charts are constructed immediately below
the xi or x chart. When more than 1 data point
E. Steps in an SPC Program: per day is analyzed and x values are plotted on a
 Identify the cause of variation in order to remedy x chart, the range is the difference between the
it. This is not always obvious; often it is elusive highest and lowest xi in that period (subgroup).
because manufacturing operations are complex -  Range, R = [Highest value – Lowest value] (10)
many interrelated variables. Statistical Control
Charts distinguish between Common causes and  Upper Control Limit, UCLR = D4 R (11)
Special causes of variation.  Lower Control Limit, LCLR = D3 R (12)
 Remove special causes, e.g., recalibrate the
Where, D3 and D4 are the factors obtained from
instrument, store standards to minimize
tables used in constructing control charts.[5]
deterioration, etc. Once a process is free of
special causes, it is said to be STABLE even  Process Capability (Capability Index): Process
though it still has variation due to random causes. capability (Cp) is simply the ability of a process
to meet a customer's product specification. A
 Estimate the Process Capability. process must be in control before Cp is
calculated. Capability Ratio (CR) is simply the
 Establish and carry out a plan to monitor, inverse of Process Capability. The lower is the
improve and assure the quality of the process, CR the more capable is the process.
e.g., charting, maintenance, training and record
keeping, in order to constantly and forever reduce  Process Capability Index (Cpk) is equal to the
variation [3]. lower of CPU (upper process capability) and CPL
(lower process capability). Cpk is a better
 Normally the values cluster about the ‘average measure of process capability than Cp or CR
value’. since Cpk takes into account the actual process


( x  x 2    x n )
Average = x = 1 
x i
(1)
center compared to the target [3].

n n III. METHODOLOGY
Where, n refers to number of data points (usually
called the population), xi refers to the measured 1) Record: Recorded component details. At first,
component details were taken as furnished
dimension of a component of a sample, and x refers
below:
to the average (usually called the population or Component Name : Cam Roller Shoe
process mean). Customer Name : Delphi TVS, Chennai, India.
When all parts are measured, the standard deviation Specifications : 8.07 + 0.02 (as per the customer
calculation becomes, drg). Upper Specification Limit (USL) : 8.09 mm
Lower Specification Limit (LSL) : 8.05 mm

x 
(x i  x) 2
(3)
Several measurements of thickness ground by
the CNC Grinder in different trials have been
n recorded for every component in the observation
tables. One of the observation tables (trial no. 8)
Where, (xi - x ) is the difference between an individual is given as under:
datum and the sample average.

International Journal of Power System Operation and Energy Management ISSN (PRINT): 2231 – 4407, Volume-2, Issue-1,2
110
 Process Capability Improvement by Putting ‘Statistical Process Control’ into Practice

The arithmetic average (mean) of ranges,  R  Ri 

Table 1: Component Trial Report
 n 
COMPONENT DETAILS 
Part 15/09/201 = 0.28 / 16 = 0.0175
TC Trial Date
Name 1
09:00 am Each Process (or population) Standard Deviation,
Trial
– 05:00 Sample 05 σ = R / D2 = 0.0175 / 2.326
Time = 0.0075
pm Size (N)
Per where, D2 = 2.326 (from table of constants, for N = 5)
7 min 40 Dressing
Cycle 0.5mm
sec Amount Therefore, Standard Deviation of the sample mean,
Time
THICKNESS (IN MM) σx = σ /√N = 0.0075 / √5 = 0.0034
S. Ra Average xi (Process mean), x = ∑ xi / n =
N Par Par Par Par Par Mea
nge 8.063
o. t1 t2 t3 t4 t5 n
(Ri
(x1) (x2) (x3) (x4) (x5) (xi) The control limits are,
)
8.07 8.06 8.07 8.08 8.08 8.07 0.0 UCLx = x + 3σx = 8.073
1
8 9 9 0 4 8 15 LCLx = x - 3σx = 8.053
8.07 8.06 8.08 8.07 8.07 8.07 0.0
2 Range = [Highest value – Lowest value]
0 7 0 6 7 4 13
8.07 8.07 8.07 8.06 8.07 8.07 0.0 = 0.040 – 0.015 = 0.009
3
9 2 0 4 0 1 15
8.07 8.06 8.07 8.07 8.08 8.07 0.0 The control limits for Range chart are,
4
8 6 8 3 0 5 14
8.06 8.07 8.08 8.08 8.08 8.08 0.0 UCLR = D4 R = 2.114 X 0.0175 = 0.037
5
8 9 6 4 3 0 18
8.06 8.06 8.07 8.06 8.06 8.06 0.0 LCLR = D3 R = 0 X 0.0175 =
6
7 7 8 5 3 8 15 where, D4 = 2.114 and, D3 = 0 (from table of
8.05 8.05 8.06 8.07 8.06 8.06 0.0 constants, for N = 5) [5]
7
3 8 5 2 2 2 19
8.07 8.07 8.07 8.06 8.05 8.06 0.0 2) Control Charts: Plotted control charts with the
8 help of MS-Excel application software as shown
8 2 3 3 9 9 19
8.06 8.05 8.06 8.07 8.06 8.06 0.0 below:
9
4 8 1 6 6 5 18
8.05 8.07 8.05 8.05 8.06 8.05 0.0
10
4 0 9 0 2 9 20
8.05 8.06 8.04 8.04 8.04 8.05 0.0
11
7 4 8 5 6 2 19
8.04 8.04 8.04 8.04 8.06 8.04 0.0
12
0 6 3 9 2 8 22
8.03 8.03 8.04 8.05 8.04 8.04 0.0
13
9 4 5 6 6 4 22
8.05 8.06 8.04 8.05 8.04 8.05 0.0
14
5 4 9 2 5 3 19 Fig. 1 : X – Chart for table 1.
8.05 8.05 8.05 8.07 8.06 8.05 0.0
15
3 3 8 0 1 9 17
8.06 8.05 8.04 8.05 8.05 8.05 0.0
16
4 7 9 0 5 5 15

Analysis: Here in the above observation record, we

have a number of variable measurement outcomes for
the number of components machined on a CNC
Grinder. To analyze the process capability, the
statistical quality control chart techniques can be
implemented in the following way:
Fig. 2 : R - Chart for table 1

International Journal of Power System Operation and Energy Management ISSN (PRINT): 2231 – 4407, Volume-2, Issue-1,2
111
 Process Capability Improvement by Putting ‘Statistical Process Control’ into Practice
3) Result: As we can observe from the X - chart, the 0.00
10 8.078 8.075 8.078 8.075 8.074 8.076
thickness of component numbers 1, 2, 4, 5, 11, 12, 4
and 13 are out of the control limits, this means that 0.00
11 8.077 8.078 8.078 8.077 8.075 8.077
process is not capable of producing the thicknesses 3
within specification. It is concluded that the process 0.00
12 8.079 8.078 8.079 8.079 8.075 8.078
4
is out of control and not capable to meet specific
0.00
demand of tolerances. Now the root cause of this 13 8.077 8.079 8.077 8.077 8.075 8.077
4
problem should be identified and solved. The CNC 0.00
Grinder was checked by a team of Engineers from 14 8.078 8.078 8.075 8.074 8.075 8.076
4
Design and Production Department and after a 0.00
complete check up, it was found that the tool feed 15 8.077 8.079 8.076 8.079 8.079 8.078
3
arrangement of CNC machine in Z direction was 0.00
16 8.078 8.076 8.079 8.080 8.077 8.078
getting a shock of high tendency whenever feed 4
command for varying depth of cut was given in the ∑Ri
n ∑ xi =
machine program. It happened all mainly because of =
= 129.23
an extreme backlash in drives of its special purpose 0.05
16 3
9
jig. After a close discussion, the team decided to
provide a calibrated scale with the drives in such a 4) Record: A process control is not merely
obtaining the dimensions within specification limits
way that whenever machine starts to follow the
but is said to be accomplished when the dimensions
command to move fast in a direction, the drives
of all samples are obtained with the values nearest to
should take a lead exactly with the calibrated values
the mean value in much lower ranges. The same
of the scale. After the calibrated scale provided to the
procedure of SPC was repeated again to determine
tool drives, all the first three sample trials gave the
the process capability after the remedies were
dimensional values within specification limits (i.e.
applied. One of the observation tables (trial no. 13) is
8.07 + 0.2). It meant that the CNC Grinder was
given as under:
become capable of giving the dimensions of the
components as per the given tolerances.
5) Analysis: To analyze the process capability, the
Table 2. Component Trial Report. statistical quality control chart techniques can be
implemented in the following way:
Component Details
Part 18/09/
TC Trial Date
Name 2011 The arithmetic average (mean) of ranges,

 R   i .
Each
Trial 09:00 am – 03:00  R
Sample Size 5 = 0.059 / 16 = 0.0037
Time pm
(N)  n 
 
Per
Dressing
Cycle 6 min 2 sec 0.5mm Process (or population) Standard Deviation, σ = R /
Amount
Time
D2 = 0.0037 / 2.326 = 0.0016
Thickness (in mm)
S. where, D2 = 2.326 (from table of constants,for N =
N Part 1 Part 2 Part 3 Part 4 Part 5 Mean Ra 5)[5]
o. (x1) nge
(x2) (x3) (x4) (x5) (xi)
(Ri) Therefore, Standard Deviation of the sample mean,
1 8.080 8.078 8.078 8.078 8.076 8.078
0.00 σx = σ / √N = 0.0016 / √5 = 0.00072
4
2 8.079 8.077 8.077 8.076 8.076 8.077
0.00 Average x (Process mean), x = ∑ xi / n = 8.077
3
0.00 The control limits are,
3 8.078 8.078 8.075 8.075 8.074 8.076
4 UCLx = x + 3 σx = 8.079
0.00
4 8.075 8.076 8.075 8.079 8.075 8.076
4 LCLx = x - 3 σx = 8.075
0.00
5 8.080 8.076 8.077 8.076 8.076 8.077 Range = [Highest value – Lowest value]
4
0.00 = 0.004 – 0.003 = 0.001
6 8.079 8.077 8.076 8.077 8.076 8.077
3
0.00 The control limits for Range chart are,
7 8.079 8.075 8.075 8.076 8.075 8.076
4
0.00 UCLR = D4 R = 2.114 X 0.0037 = 0.0078
8 8.078 8.081 8.077 8.077 8.077 8.078
4
0.00 LCLR = D3 R = 0 X 0.0037 =0
9 8.079 8.079 8.079 8.077 8.076 8.078
3 where, D4 = 2.114 and D3 = 0 (from table of
constants, for N = 5)[5]

International Journal of Power System Operation and Energy Management ISSN (PRINT): 2231 – 4407, Volume-2, Issue-1,2
112
 Process Capability Improvement by Putting ‘Statistical Process Control’ into Practice
As we can observe from the X - chart, the thickness Process Capability, Cp = 4.167 ;
of all the components are out of the control limits, Capability Ratio, CR = 0.24 ;
this means that process is capable of producing the
Upper Process Capability, CPU = 2.71 ;
thicknesses within specification limits.
Lower Process Capability, CPL = 5.625 ;
It is concluded that the process is now under control
Cpk = Minimum (CPU, CPL) = 2.71.
and capable of meeting the specific demand of
7) Discussion: Since, the value of Process
tolerances.
Capability Index, as required by the customer, Delphi
Process Capability, TVS, Chennai was greater than 2, and the process
capability index we obtained after the implementation
= (8.09 – 8.05) =
of SPC techniques is 2.71 which is greater enough
4.167 than 2 therefore, we can say that the process is under
6 X 0.0016 control now and capable of producing all the
components under the given specification limits with
Capability Ratio, CR = 1 / Cp = 0.24
the very low normal distribution.
Let, CPU and CPL are Upper and Lower Process
Capabilities, respectively. V. CONCLUSION

CPU = USL - x = 2.71and, CPL = x - LSL = The control limits obtained after the remedial actions
5.625 taken for the grinder are within specification limits
3σ 3σ and the thickness produced in all the components of
every sample lie under the control limits. The
Cpk = Minimum (CPU, CPL)
thicknesses of all the components are located very
= Minimum (2.71, 5.625) = 2.71 close to the process mean. All these results are
positive by which we conclude that the process is
6) Control Charts: Plotted control charts with the
under control. The process capability (Cp) increased
help of MS-Excel application software as shown
from 0.581 to 4.167 which show that implementation
below:
of SPC technique is proved to be successful in
improving the performance of grinding process
thereby making it more capable of producing the
products with right dimensions. Capability Ratio
(CR) is reduced from 1.721 to 0.24 which means that
the process spread now occupies 24 % of the
tolerance. The lower is the CR the more is capable
the process. The provision of a calibrated scale with
the mechanical drives of a CNC Grinder solved the
backlash problems and increased its preciseness. Cpk
is a better measure of process capability than Cp or
CR since Cpk takes into account the actual process
center compared to the target [3]. Here, we got Cpk as
Fig. 3: X – Chart for table 2.
2.125 which is greater than 2. Cpk of value greater
than 2 was required by the customer company, Delphi
TVS, Chennai and their demand has been met
satisfactorily. These all become possible with the
implementation of an SPC technique. Likewise, the
SPC tools can be implemented to solve so many real
life problems of different machines and processes in
future that may come up with meeting the demand of
higher quality and productivity of different
manufacturing processes. The methods developed in
Fig. 4: R – Chart for table 2. the first half of this century by Shewhart and others
are still very useful in many current applications.
IV. RESULTS:
REFERENCES
Upper Control Limit, UCL = 8.079 ;
Lower Control Limit, LCL = 8.075 ; [1] Term papers, Statistical Process Control spc – management
-1, business and market, 34 pages, 2011
Range, R = 0.001 ;
[2] Roberta Russell & Bernard W. Taylor, III John Wiley &
Average Range, R = 0.0037 ; Sons, “Statistical Process Control”, Operations Management
- 5th Edition, Inc.Beni Asllani, University of Tennessee at
Chattanooga, 2006.

International Journal of Power System Operation and Energy Management ISSN (PRINT): 2231 – 4407, Volume-2, Issue-1,2
113
 Process Capability Improvement by Putting ‘Statistical Process Control’ into Practice
[3] SPC92b, file format: microsoft word – “statistical process [5] Institute of Quality and Reliability, “Control Chart Factors”
control (spc) & statistical quality control (sqc)” Tables of Constants for Control charts, Page 1 - 3.
chemrat.com/chemhog2/spc_files/spc92b.doc
[6] The American Heritage Medical Dictionary, 2007,
[4] Rami Hikmat Fouad, Adnan Mukattash, “Statistical Process Published by Houghton Mifflin Company.
Control Tools: A Practical guide for Jordanian Industrial
Organizations” JJMIE, Volume 4, Number 6, ISSN 1995- [7] http://www.businessdictionary.com/definition/
6665 Pages 693 – 700, December 2010. specification-limits.html#ixzz1wLW5GTs8.



International Journal of Power System Operation and Energy Management ISSN (PRINT): 2231 – 4407, Volume-2, Issue-1,2
114