A Study on the Improvement of a PVC Pipe Production

Abstract

This paper explains our project about a six sigma application to improve a PVC pipe production process.
We began with introducing the company that we chose which is the National Industry Company (NIC)
and their production process of both PVC pipes and fittings. Then we used the DMAIC methodology to
define the problem that they are facing which is defective pipes and fittings and in the measure phase we
collected data regarding both productions and calculated the average defective pipes per month with their
associated costs and decided to focus on pipe production lines. After that, in the analyze phase, we looked
into the factors and found the root cause of the defective pipes are torque and feeder speed levels. In
addition, we used Minitab to draw X bar-R charts and found that the process was out of control, hence we
used contour plots to find the optimal settings of both torque and feeder speed levels. Finally, in the
improve phase, we removed the out of control points that were out of the optimal settings found earlier
and the process became in control indicating 89% improvements. Finally, we provided an out of control
action plan for when an outlier appears on the X bar charts during the monitoring of the process.

Keywords: Pipes, defect rate, torque, feeder speed.

1. Introduction

A Plastic manufacture company was established in the 1960’s in Kuwait which is one of the well-known companies
that focuses on manufacturing and marketing of construction materials. It has two main factories which are located
in Mina Abdullah and Sulaibiya; they also own 16 smaller factories which produce plastic, paint, ceramics, and
interlocks. We chose to apply six sigma application on the production process of both pipes and fittings in this
company as we found that they have a range of defective pipes and fitting products with an average of 2-2.2 % per
month which is causing loss of customer ratification and loss of profit as well. Our main goal is to minimize the
defective pipes produced by the company by 10% to improve their production line in order to gain customer
satisfaction and to reduce their losses and increase their profits.

2. Literature Review

Nowadays, manufacturing industries are widely spread and since we will apply six sigma on a manufacturing
industry, then most of our articles are under this topic. Gupta et al, (2017) applies six-sigma to decrease the
variations in bead splice of a tire manufacturing company in India. After using DMAIC methodology the standard
deviation decreased significantly from 2.17 to 1.69 and the process capability increased from 1.65 to 2.95 indicating
a capable process with a lower defective rate. Moreover, Girmanova et al, (2017) implemented six-sigma to improve
the metallurgic product quality of a company. They used both DMAIC and failure mode and affect analysis and the
sigma level increased to 13%, lowering the defective rates and processing cost. Raman and Basavaraj, (2019) used

l
six-sigma methodology to decrease the capacitor rejection issues in a manufacturing company in India. DMAIC
strategy was used and great improvements in quality product were seen as well as a high reduction in cost and an
increase in customer satisfaction.

In addition, Abbes et al. (2018) conducted six-sigma study on a small-medium sized enterprise to improve the
quality of their clothes production by using DMAIC. In return, their process capability increased from 0.20 to 1.47
and the DPMO (defects per million) decreased from 780,000 to 308,000. Similarly, Prabu et al. (2013) applied six-
sigma on a manufacturing company in India to overcome ovality in stage casting components. They used DMAIC as
well as failure mode and affect analysis to improve the sigma level from 3.90 to 3.97. Zasadzień, (2015) applied six-
sigma methodology to improve the machine failures and decrease their downtimes in the production process. After
using DMAIC analysis, the machine failures reduced from 23 factors to only 2 factors and the downtime duration
decreased from 18 to 9 hours. This resulted in an increase of machine availability and in a decrease of downtime
duration as well as the removal of failure.

To add to the above article, Küçük and Orbak, (2011) studied the reduction of transportation costs in a pipe
manufacturing company by using six-sigma methodology. The reason that transportation costs are high is because of
the loading and unloading time of the vessel. The daily rate of loading the vessel is 1253 ton of pipes and the
average waiting time was 26%. After applying the best solution they were able to load 2470 tons of pipes daily with
an average waiting time of 13% only. More studies were recognized by Amri, (n.d) Amrianalysed a lean Six Sigma
for improving pipeline project performance at a Project Site ABC base on Balance Score Card Framework. The
company’s objective is to reduce the reject rate of pipeline welding process and reject material for pipe replacement.
Before implementation, welding had a defect rate of 14% and material had a defect rate of 7%. After improvements
the defect rate for rejected welding decreased to 3% and material rate decreased to 0%, savings for the project
increased to $23,000 for welding and $16,000 for rejected material. Likewise, Rehman et al, (2012) conducted a Six
Sigma process on a cell site construction of a telecom company to find the factors that are causing the cell sites to
break down and require maintenance. After the implementation of the six-sigma, cell sites breakdown and
maintenance reduced to 48 defects in 80 cell sites resulting in a 5.02 sigma level and increasing company savings to
$0.45 million.

In addition, Choomlucksana et al, (2015) studied a case study about manufacturing and they applied six-sigma to
improve the efficiency of the production of sheet metal stamping. By using Poka-Yoke, 5s and lean manufacturing
principle the processing time of a polishing stage was reduce by 62.5%. John and Areshankar, (2018) studied the
reduction of the bearing end plate reworks in a machining process, by reducing the rework due to thickness and
diameter variation. By using the six-sigma methodology the process met their requirements, improving thickness
and diameter and reducing rework. Oguz et al., (2012) applied Six Sigma to the concrete- panel production system
in a multi-housing complex project to decrease the number of projects behind the schedule. By using the
combination of DMAIC methodology and by taking the variation of panel production as a critical total quality
(CTQ) they were able to increase the panel production from 18/day to 75/day which decreased the number of
projects delayed.

3. Methodology

We used the approach of six sigma application in specific DMAIC methodology tool in order to improve the
company’s production process and these phases are explaine below.

3.1 Define

After visiting NIC Company, we found that they have two main production lines, one for pipes and the other for
fittings; hence we provided a flowchart for both production processes as shown in figure 1 below. The production
process of both products start off similarly by oppening the resin bags, mixing them with the required materials,
heating them and storing them for later use. Then when they are needed they are reheated and each production has
its own separate line, for instance, the fittings go to the injection modling process and the pipes go to the extrusion
line process. After both products are manufactured they are tested in the lab for their quality and if they are of
acceptable quality then they are sent to the warehous, otherwise they are recycled.

l
 Resin bag arrivals

Machine opens the bags of resin and stores in

silos

A mixer mixes the resin with additives and

pigments

Heating for 3-4 hours

Cooling and storing into silos

Reheating for 3-4 hours

YES Injection molding line? NO

Compound is melted Extrusion process

Compound is injected into mold

Vacuum cooling

Compound cools down Pipe is pulled by haul-off

Fitting is ejected
Printing machine

Cutting machine

Socketing machine

Tested in quality control lab

NO Accepted? YES

Recycled Transferred to warehouse

End

Figure 1: Process Map

We also illustrated the initial stage of both pipe and fitting industry to the final stage and this is shown in table 1
below. The company recieves their resins, pigments, lubricants and stabilizers from SABIC Company in Saudi
Arabia and these products are used for both iinjection molding and extrusion process to produce pipes and fittings
which are distributed to Sewage Companies.

l
Measure:

In this stage, we measured the defect rates in both the uPVC pipes and the fittings. We also drew some bar graphs to
identify whether the production lines are producing a higher defect rate or whether specific products are.

Defects Percentage
1.00%
Average Defect 0.80%
0.60%
0.40%
0.20%
0.00%

July
Mar

June
ch

May

Aug
Apri
l
ry

ry
Janua

Februa

Figure 2: Average defect percentage of each month in 2022

From the graph above, we understand that throughout the year 201, the minimum defect rate is seen in March
(0.33%) and the maximum in January (0.89%). Therefore, we decided to look at the daily rejection summary for the
entire month of January since it has the highest defect rate.

Average defect percentage in each extrusion line

2.50%
2.00%
1.50%
1.00%
0.50%
Average
defect

0.00%
Extrusi

Extrusi

Extrusi

Extrusi
on

on

on

on

Figure 3: Average defect percentage in each extrusion line in

January 2022

Figure 3 above, shows the average defect percentage in each extrusion line regardless of the type of pipe being
produced. We can see that extrusion line 3 has the highest rejection percentage which is 2.01% compared to the
other extrusion lines and hence we calculated it cost as shown in table 3 below.

l
 Table 3: Cost of each pipe in extrusion line 3

Extrusion line 3 products Rejected quantity Cost/unit (KWD) Total cost (KWD)

225x10.0x6000 Blue 1 10.9 10.9

160x7.7x6000 DG 7 8.8 61.6

200x14.0x6000 DG 9 9.5 85.5

200x4.5x6000 LG 31 8.2 254.2

315x7.7x6000 LG 13 12 156

Sum 568.2

Table 3 shows the calculation of the cost of the rejected pipes that were produced in extrusion line 3 and we found it
to be a total of 568.2 KWD. This means NIC Company has lost 568.2 KWD in one month on 61 defective pipes
produced by extrusion line 3 only.

Average defect percentages in each molding lines

2.50%
2.00%
1.50%
1.00%
Average

0.50%
defect

0.00%
Moldi

Moldi

Moldi

Moldi
line

line

line
ng

ng

ng

ng
1

Figure 4: Average defect percentage in each molding line in January 2022

Figure 4 above, shows the average defect percentage in each molding line and we can see that molding line 6 has the
highest rejection percentage which is 2.22% and the cost associated with this line is 17.87 KWD as shown in figure
5 below.

Table 4: Cost of each fitting in molding line 6

Cost/unit
Molding line 6 Rejected Total cost
products quantity (KWD) (KWD)

110 COUPLER LG 35 0.45 15.75

110 COUPLER GB 4 0.53 2.12

Sum 17.87

l
Table 4 shows the cost associated with molding line 6 and we can see that the total cost is 17.87 KWD. Therefore,
comparing the costs of the highest defective extrusion line with highest deective molding line, we can see that
altough molding line 6 has a higher defect rate of 2.22%, it costs only 17.87KWD, where as extrusion line 3 costs
568.2 KWD and this is why we chose to focus on extrusion line 3, in specific the 200mm pipe with 4.5mm wall
thickness.

Process Capability Report for Average

LSL USL
Process Data Overall
LSL4.5 Within
Target*
Overall Capability
USL5.2
Pp 0.94
Sample Mean4.9088
PPL 1.10
Sample N102 PPU 0.79
StDev(Overall)0.123502 Ppk 0.79
StDev(Within)0.107064 Cpm *
Potential (Within) Capability
Cp 1.09
CPL 1.27
CPU 0.91
Cpk 0.91

4.48 4.64 4.80 4.96 5.12 5.28

Perfo rmance
Observed Exp ected Overall Exp ected Within
PPM < LSL 9803.92 466.30 67.18
PPM > USL 9803.92 9191.18 3265.78
PPM Total 19607.84 9657.48 3332.97

Figure 5: Process capability for average wall thickness of 200 mm pipe

We can see from figure 5, that the average wall thickness of some pipes are out of the acceptable range of
specification limits (4.5mm up to 5.2mm) and the actual process capability is 1.09 which is acceptable but since the
process capability index is 0.91, which is not equal to the actual process capability value then this means that the
process is not centred and the average value of the wall thickness is below the lower specification limit. We can also
see that the expected overall defective parts per million is 9,657 which is very large and by multiplying the process
capability index value by three we understand that the company is operating at 2.73 sigma level currently which is
very low and requires improvements.

3.3 Analyze:

In this phase we want to find the root causes of the problems so we had to check the factors affecting the extrusion
process and we found that torque and feeder speed were the two factors that are closely related to our defect which is
less wall thickness. In addition, we understood that every two hours the quality control inspector checks the pipe’s
wall thickness at eight different points (as shown in figure 6 below) using a vernier calliper and if one point is above
or below the specification limit then it is recycled.

Figure 6: Pipe’s dimension

l
We constructed an Ishikawa diagram to show the main reasons of the 200mm defective pipes and these were
categorized into different headings which are; machine, method, manpower, environment and material. All the
reasons below contribute to less wall thickness; however, after talking to the manager of the production process, we
understood that the main root cause was the method where the feeder speed and torque factors exist.

Figure 7: Ishikawa diagram

Response surface Methodology:

In order to check whether these two factors affect our response, we collected data regarding the current factor levels
and wall thickness measurements for the 200mm pipe, and used Response Surface Methodology.

Response surface methodology is used to determine the effect of several factors and their interaction on one or more
response variable. It uses the actual data which is measured throughout the laboratory control checks, which is the
total opposite of the design of experiment that requires setting high and low levels for the factor, implementing them
and measuring the response variables that they produce. RSM uses a regression equation that relates the independent
input variables to the output variables in order to optimize the response.

Hypothesis Testing

H0: Torque, feeder speed and their interaction has no significant effect on the wall thickness of the 200mm pipe at
each specific point.

H1: Torque, feeder speed and their interaction has a significant effect on the wall thickness of the 200mm pipe at
each specific point.

l
 Table 5: Response surface methodology results:

Point P-value R-squared R-squared

(adjusted)

A 0.234 6.76% 1.9%

B 0.207 7.17% 2.29%

C 0.086 9.43% 4.72%

D 0.36 5.47% 0.55%

E 0.58 3.85% 0%

F 0.58 3.81% 0%

G 0 20.37% 16.23%

H 0 23.19% 19.19%

From table 5, we can make a decision to reject the null hypothesis for the specific point’s c, g, and, h since the p-
value is relatively small and the maximum error that we might face in rejecting H 0 is only 8.6%. In addition, the r-
squared values range from 9.43% up to 23.19% so we can say that between 9.43% up to 23.19% of the variations in
the wall thickness of the pipes are explained by the torque and the feeder speed. Hence, we conclude that torque,
feeder speed and their interaction significantly affects the wall thickness of the 200mm pipe at these points (c, g
and, h).

Contour Plots:

In the figures 8,9, and 10 below, we drew contour plots for the significant points (c,g, and, h) that we found in the
200mm pipes to find the optimal settings of the two factors in order to minimize the defective pipes. From customer
specification limit, we know that the pipes lower specification limit is 4.5mm and the upper specification limit is
5.2mm and after observing all the graphs shown in figures 8, 9, and 10. We perceived that on average the optimum
limit for the torque is from 60% to 62.5% and the optimum limit for the feeder speed is from 28rpm to 29.5rpm to
avoid defective pipes due to large/small wall thickness.

l
Figure 8: Contour plot for point c of the 200mm pipe

Figure 9: Contour plot for point g of the 200mm pipe

Figure 10: Contour plot for point h of the 200mm pipe

l
X-bar & X-bar R Charts:

X-bar and X-bar R charts were used to check whether the current process is in control or out of control for the
significant points found earlier and the equations with their specific values for A2,D3,D4 were derived from
(Web.mit.edu, 2019). The Xbar and Xbar-R charts shown in figures 11,12, and 13, are for the points c, g, and h of
the 200mm pipe and they all have a minimum of two outliers per chart with point h having the most outliers.
Therefore, we can understand that the process is out of control and we will check the torque and feeder speed level
for each outlier to see whether they are defective because their settings are out of the optimal ranges we found from
the contour plots.

Xbar-R Chart of C
UCL=5.209
5.1

4.8 _ X=4.713

4.5
1 1

4.2 LCL=4.217
13579 11 1315171921
Sample

1.00
UCL=0.862

0.75
0.50
_
0.25R=0.264
0.0022 LCL=0

13579 11 1315171921
Sample

Tests performed with unequal sample sizes

Figure 11: X-bar R chart for point c on 200mm pipe

Xbar-R Chart of G
1
1 UCL=5.261
5.2

5.0

_ X=4.803

4.8

4.4
LCL=4.344
4.6
1 3 5 7 9 11 1315171921
Sample

1
1.2 1

0.9
UCL=0.797
0.6

_ R=0.244
0.3 LCL=0

0.0 13579 11 1315171921

Sample

Tests performed with unequal sample sizes

Figure 12: X-bar R chart for point g on 200mm pipe

l
 Xbar-R Chart of H
1
5.2 UCL=5.2204
1

5.0
_ X=4.8352

4.8
4.6
1
1
LCL=4.4500
4.4
1 3 5 7 9 11 1315171921
Sample

11
1.00

0.75 UCL=0.669

0.50

_ R=0.205
0.25

LCL=0
0.00 13579 11 1315171921
Sample

Tests performed with unequal sample sizes

Figure 13: X-bar R chart for point h on 200mm pipe

3.4 Improve:

In this phase we iterated the Xbar and Xbar-R charts several times for the 200mm pipe to remove all the defective
points that were caused by the settings of torque and feeder speed levels in order to finalize the control limits of
phase I and these graphs after improvements are shown below.

Xbar-R Chart of C
UCL=5.0322
5.0

4.8 _ X=4.7427

4.6

LCL=4.4533
4.4 1
12345 6 7 8 9 1011
Sample

UCL=1.061
1.00

0.75
_ R=0.502

0.50

0.25

0.00LCL=0
12345 6 7 8 9 1011
Sample

Figure 14: X-bar R chart for point c after improvements

l
 Xbar-R Chart of G
1

5.10 UCL=5.1063

4.95
_ X=4.8189

4.80
4.65
4.50LCL=4.5316

12345 6 7 8 91011
Sample

UCL=1.053
1.00

0.75
_ R=0.498

0.50
0.25

0.00LCL=0
12345 6 7 8 91011
Sample

Figure 15: X-bar R chart for point g after improvements

Xbar-R Chart of H
UCL=5.0728

5.0

_ X=4.8567

4.8

LCL=4.6407
4.6 1

1 2 3 4 5 6 7 8 91011
Sample

0.8UCL=0.7920

0.6
0.4_
R=0.3745

0.2

0.0 LCL=0

12345 6 7 8 91011
Sample

Figure 16: X-bar R chart for point h after improvements

After the iterations were made we finalized the upper and lower control limits as shown in figures 14, 15 and, 16, to
be used for phase two. We found that all the out of control points were caused by the torque and feeder speed
settings and after removing them we can see then nearly all the data points are within the control limits and the
process is now in control. As a matter of fact, the one/two outliers that are seen in points c, g, and h are caused by
chance and not due to the factors settings. Therefore, changing the optimal settings of torque from 59% up to 64% to
60% up to 62.5% is an effective solution.

Process Capability:

After improvements we recalculated the capability of the process and its index and found them to be 1.34 and 1.07
which means that the process is at an excellent state and is centered. In addition, the total defective parts per million
has been reduced to 989 indicating that we succeeded in reducing the defects by 89.7%.

l
 Process Capability Report for Average

LSL USL
Process Data Overall
LSL4.5 Within
Target*
Overall Capability
USL5.2
Pp 1.29
Sample Mean4.92003 PPL 1.55
Sample N55 PPU 1.03
StDev(Overall)0.0904954 Ppk 1.03
StDev(Within)0.0872833 Cpm *
Potential (Within) Capability
Cp 1.34
CPL 1.60
CPU 1.07
Cpk 1.07

4.5 4.6 4.7 4.8 4.9 5.05.15.2

Perfo rmance
Observed Exp ected Overall Exp ected Within
PPM < LSL 0.00 1.73 0.75
PPM > USL 0.00 988.07 669.13
PPM Total 0.00 989.80 669.87

Figure 17: Process Capability report after improvements

Out of Control Action Plan:

After finalizing the control limits for both pipes, they can be used now to monitor the process and check to see if an
out of control point is detected. If a defective point is detected then the Out of Control Action Plan shown in figure
18, below can be used. If an out of control signal is seen then the data must be checked whether it was entered
correctly or not, then if it was entered correctly the failure test should be checked. From their onwards specific
actions must be taken for the average test failure and for the range test failure. If the average test failed, then the
torque and feeder speed levels should be checked and if they were not set to their correct levels then they should
check whether this is the third adjustment or not. If it is, then the process engineer should be contacted, otherwise,
the factor levels should be set to their optimal levels. However, if the range test failed, then the defective points out
of the control limits should be checked for their causes and if they were caused by torque and feeder speed then by
setting these factor levels back to their optimal settings the problem can be fixed.

l
 Out of control signal on X-bar chart

No
Are the data entered correctly? Edit data to enter it correctly.

Yes

Range Check the torque and feeder speed levels for the defective points
Which test failed?

Average

Yes

Were torque and feeder speed ranges correct?

No

No Reset feeder speed and torque ranges to 28rpm-29.5rpm and 60%-60.25%

Is this the 3rd adjustment?

Yes

Contact Processes engneering

Enter comments in the log describing the action

Figure 18: Out of Control Action plan

4. Results and Analysis

After careful observations, we found that pipes were defective due to less wall thickness and to find the cause of
this we used response surface methodology to test specific factors significance on the eight points of the pipe’s wall
thickness. The results we gained from the response surface methodology provided us with the specific points that
were being affected by both feeder speed, torque and their interaction. Then contour plots were constructed for
these specific points to find the optimal settings of feeder speed and torque. Furthermore, control charts were
established to see which points were out of control and to check whether the torque and feeder speed are out of the
optimal settings or not. We found out that the torque and feeder speed ranges were too broad and this was the major
cause of the defective pipes. In addition, we analyzed the capability of the process and understood that it was not
working efficiently and was not centered but was producing on average less than the lower specification limit of
the customer. Furthermore, the defective data points on the X-bar and X-bar R charts that were out of the optimal
settings were removed and the ones within the optimal settings, were considered to be out of control because of

l
natural causes and is kept in the readings. This step was done several times until the out of control points were all
due to natural causes and the graphs were constructed again to show that now the process is in control. Moreover,
the capability analysis was recalculated and showed an increase in the capability index indicating a centered
process able to meet customer’s satisfaction. Also, the number of defective parts per million decreased significantly
from 9657 to 989 showing that the system improved by 89% and is producing less defective pipes. So, we learned
that to not produce defective pipes the torque and feeder speed levels should be maintained within their optimal
settings.

5. Conclusion

To conclude, we used six sigma methodology on a pipe production line in a plastic manufacturing company and we
found that the problem they are facing is having 2.02% defective pipes per month. After conducting the phases of
DMAIC we understood that the torque and feeder speed levels were causing the pipes to have less wall thickness
and in return causing them to be defective. Therefore, we used Minitab to draw contour plots and were able to find
the optimal settings for the torque and feeder speed level which when used provides us with an in control process
plotted on an X bar-R chart. Hence, 89% improvements were achieved by providing the optimal settings of both
factors.

Questionnaire ;

Sure, here’s a questionnaire on the improvement of PVC pipe production:

1.How long have you been involved in PVC pipe production?

a. Less than a year
b. 1-5 years
c. 5-10 years
d. More than 10 years

2.What are the key challenges you face in PVC pipe production?
a. Inconsistent pipe quality
b. Equipment breakdowns
c. Lack of skilled labor
d. High production costs
e. Other (please specify)

3.What is the average production rate of your PVC pipe production line?
a. Less than 100 kg/h
b. 100-500 kg/h
c. 500-1000 kg/h
d. More than 1000 kg/h

4.How do you ensure the consistency of PVC pipe quality during production?
a. Regular testing of raw materials
b. Monitoring of production parameters
c. In-process quality checks
d. Final product inspection
e. All of the above
l
5.Have you implemented any process improvements to increase production efficiency?
a. Yes
b. No

6.If yes, which process improvements have you implemented?

a. Equipment upgrades
b. Optimization of production parameters
c. Staff training on best practices
d. All of the above

7.How do you manage waste generated during the PVC pipe production process?
a. Recycling of waste material
b. Disposal in landfills
c. Incineration
d. Other (please specify)

8.How do you maintain the equipment used in the PVC pipe production process?
a. Regular maintenance by in-house staff
b. Outsourcing of maintenance to third-party service providers
c. Combination of in-house and third-party maintenance
d. No regular maintenance

9.How important is sustainability in your PVC pipe production process?

a. Very important
b. Somewhat important
c. Not important

10.What sustainability measures have you implemented in your PVC pipe production process?
a. Use of recycled materials
b. Energy-efficient equipment
c. Waste reduction programs
d. All of the above