Using Statistical Approaches for Correcting Manufacturing Faults Soft

Drinking Water Process as Example

Salih Yassin Mohammed Ragab1

Mechanical Engineering Department, Higher Institute of Scientific and Technology, Tobruk, Libya

altorosco@yahoo.com

Abstract.

Statistical approaches can be used to identify and correct manufacturing faults.

Statistical process control (SPC) is a method of monitoring and controlling manufacturing
processes to ensure that they are operating within acceptable limits. SPC involves collecting
data from the process, analyzing it using statistical methods, and then taking corrective action
if necessary. This data can be used to identify potential problems in the process and take
corrective action before they become serious issues. Additionally, SPC can be used to identify
trends in the process that could lead to future problems. By understanding these trends,
manufacturers can take proactive steps to prevent them from occurring in the first place. In this
study we used General factorial design by Minitab software for correcting level faults accrued
during filling soft drink under three factors first one is percent of carbonation second is the
pressure of drink and third levels is line speed. All results discussion based on P-Value
indicator.
.

Keywords: SPC, General factorial, Minitab, Factors, P-Value.

1 General information

1.1 Introduction

A general factorial design GFD is an experimental design in which two or more independent variables are
manipulated simultaneously to observe their combined effect on a dependent variable. This type of design is
used to determine the main effects and interactions between the independent variables. It is also used to
identify the most important factors that influence the dependent variable.

Design of experiments (DOE) is a systematic approach to the design and analysis of experiments. It is used
to determine the effects of different factors on a given process or system. DOE involves the use of statistical
methods to identify and quantify relationships between input variables and output variables. It can be used
to optimize processes, improve product quality, reduce costs, and increase efficiency. DOE can also be used
to identify potential problems in a system before they become serious. In this study, we focus on using the
approach of GFD to evaluate the problem of filling deviation or the differences in the level of filling in the
bottles during process of filling the soft drinking water as shown in fig 1. For more understanding of
manufacturing of soft drinking water here, we explains the main production processes:
1. Pre-Production: Before the actual manufacturing process begins, the ingredients and packaging materials
must be sourced and prepared. This includes sourcing the raw materials such as sugar, carbon dioxide, water,
and flavorings, as well as packaging materials such as cans or bottles.

2. Production: The production process begins with mixing the ingredients together in a large tank. Heating
the mixture to create a syrup then cooled and filtered. Then adding Carbon dioxide to give the drink its
fizziness and adding flavorings for taste. Pasteurizing the syrup to kill any bacteria that may be present.

3. Packaging: packaging the syrup into cans or bottles depending on the type of drink, applying Labels to
each can or bottle before they are sealed shut and sent off for distribution.

Such as, all liquid filling processes there are some parameters or factors may influences on the filling
process as well as the quality of production and products.

2 Literature review

Factorial designs involve manipulating two or more independent variables simultaneously in order to
determine their combined effect on the dependent variable.[1] This type of design allows researchers to
identify interactions between variables that may not be evident when only one variable is manipulated at a
time. For example, if two independent variables (A and B) are manipulated together, an interaction between
them may be observed if the effect of A on the dependent variable is different when B is present than when
it is absent.

The general factorial design can also be used to study the effects of multiple levels of one or more
independent variables on the dependent variable.[2] This allows researchers to identify any differences in the
effects of different levels of an independent variable on the dependent variable. For example, if one
independent variable has three levels (low, medium, and high), then a factorial design can be used to
determine whether there are any differences in the effect of each level on the dependent variable.

In addition to its ability to identify interactions between variables and differences between levels of an
independent variable [3], another advantage of using a factorial design is that it allows researchers to control
for extraneous variables by randomly assigning participants into different conditions or groups. This helps
ensure that any observed effects are due solely to manipulation of the independent variables rather than other
factors such as participant characteristics or environmental conditions.

Overall, factorial designs provide researchers with a powerful tool for studying complex relationships
between multiple independent variables and a single dependent variable. They allow researchers to identify
interactions between variables as well as differences between levels of an independent variable while
controlling for extraneous factors. As such, they are widely used across many fields and can provide valuable
insights into how different factors influence outcomes in experiments.

2
3 Design of Experiment

3.1 Methodology and ANOVA Results

One of the strongest indicators that indicate poor manufacturing and that the product is poor, which
can be seen with the eye, is that one dozen of soft drinks contain in each bottle a different amount of juice
than the other bottle. Several inputs or variables can be controlling during the manufacturing process to
control quality and prevent a difference in the level of juice inside the bottles; the engineer of process can
control three variables: the percent carbonation (A), the operating pressure in the filling line (B), and number
of filled bottles each minute (C).

We assuming that we can controlling the percent of carbonation (8, 10, 12 and 14 percent) also we can
controlling the pressure of drink to (100, 150, 200 psi) and four levels of line speed (180, 200, 220, 240 bpm)
we will run two replicates of factorial design in these three factors in random order. The response is the
average deviation from the target fill into filled level for each bottle. If the filling level over than the target
level the deviation will be positive, and if the filling level under the target level the deviation will be negative.
Table No 1 shows the summary of the multilevel factorial design.

Table 1. Summary of the multilevel factorial design

Multilevel Factorial Design

Design Summary
Factors: 3 Replicates: 2
Base runs: 48 Total runs: 96

Base blocks: 1 Total blocks: 1

Number of levels 4 3 4

Table 2. Analysis of Variance

Source DF Adj SS Adj MS F-Value P-Value
Model 48 268.833 5.6007 0.96 0.550
Blocks 1 0.000 0.0000 0.00 1.000
Linear 8 40.229 5.0286 0.87 0.552
carbonation 3 4.333 1.4444 0.25 0.862
pressure 2 20.646 10.3229 1.78 0.180
speed 3 15.250 5.0833 0.88 0.461
2-Way Interactions 21 99.625 4.7440 0.82 0.687
carbonation*pressure 6 15.854 2.6424 0.45 0.838

3
 carbonation*speed 9 41.583 4.6204 0.80 0.622
pressure*speed 6 42.187 7.0312 1.21 0.318
3-Way Interactions 18 128.979 7.1655 1.23 0.275
carbonation*pressure*speed 18 128.979 7.1655 1.23 0.275

Error 47 273.000 5.8085

Total 95 541.833

Table 3. Model Summary

S R-sq R-sq(adj) R-sq(pred)
2.41009 49.62% 0.00% 0.00%

Table 4. Coefficients
Term Coef SE Coef T-Value P-Value VIF
Constant 0.792 0.246 3.22 0.002
Blocks
1 0.000 0.246 0.00 1.000 1.00
carbonation
8 0.000 0.426 0.00 1.000 1.50
10 0.333 0.426 0.78 0.438 1.50
12 -0.083 0.426 -0.20 0.846 1.50
pressure
100 0.458 0.348 1.32 0.194 1.33
150 0.177 0.348 0.51 0.613 1.33
speed
180 -0.000 0.426 -0.00 1.000 1.50
200 0.583 0.426 1.37 0.177 1.50
220 -0.042 0.426 -0.10 0.923 1.50
carbonation*pressure
8 100 0.000 0.603 0.00 1.000 2.00

4
 8 150 0.031 0.603 0.05 0.959 2.00
10 100 -0.583 0.603 -0.97 0.338 2.00
10 150 0.073 0.603 0.12 0.904 2.00
12 100 -0.292 0.603 -0.48 0.631 2.00
12 150 0.365 0.603 0.61 0.548 2.00
carbonation*speed
8 180 0.042 0.738 0.06 0.955 2.25
8 200 -0.708 0.738 -0.96 0.342 2.25
8 220 0.917 0.738 1.24 0.220 2.25
10 180 -0.125 0.738 -0.17 0.866 2.25
10 200 1.125 0.738 1.52 0.134 2.25
10 220 -1.083 0.738 -1.47 0.149 2.25
12 180 0.292 0.738 0.40 0.694 2.25
12 200 -0.292 0.738 -0.40 0.694 2.25
12 220 0.833 0.738 1.13 0.265 2.25
pressure*speed
100 180 -0.250 0.603 -0.41 0.680 2.00
100 200 -0.583 0.603 -0.97 0.338 2.00
100 220 1.042 0.603 1.73 0.090 2.00
150 180 1.156 0.603 1.92 0.061 2.00
150 200 -0.052 0.603 -0.09 0.931 2.00
150 220 -1.052 0.603 -1.75 0.087 2.00
carbonation*pressure*speed
8 100 180 1.46 1.04 1.40 0.169 3.00
8 100 200 0.46 1.04 0.44 0.663 3.00
8 100 220 -1.17 1.04 -1.12 0.269 3.00
8 150 180 -0.20 1.04 -0.19 0.850 3.00
8 150 200 -0.82 1.04 -0.79 0.434 3.00
8 150 220 1.68 1.04 1.61 0.115 3.00
10 100 180 -1.12 1.04 -1.08 0.287 3.00
10 100 200 -0.13 1.04 -0.12 0.905 3.00
10 100 220 -0.92 1.04 -0.88 0.384 3.00

5
 10 150 180 -1.41 1.04 -1.35 0.184 3.00
10 150 200 1.97 1.04 1.89 0.065 3.00
10 150 220 0.30 1.04 0.29 0.773 3.00
12 100 180 0.08 1.04 0.08 0.937 3.00
12 100 200 -0.58 1.04 -0.56 0.579 3.00
12 100 220 0.79 1.04 0.76 0.452 3.00
12 150 180 0.80 1.04 0.77 0.446 3.00
12 150 200 0.51 1.04 0.49 0.627 3.00
12 150 220 -2.49 1.04 -2.39 0.021 3.00

Table 5. Fits and Diagnostics for Unusual Observations

Obs deviation Fit Resid Std Resid
27 -3.00 1.00 -4.00 -2.37 R
41 -3.00 1.00 -4.00 -2.37 R
75 5.00 1.00 4.00 2.37 R
96 5.00 1.00 4.00 2.37 R
R Large residual

4 Discussion

The ANOVA is summarized in table 2. Strongly we can see that not all factors can affect the
fill level. A p-value of 0.05 or lower is generally considered statistically significant. From table 3
based on A p-value we can see:

Table 6. Summarized Coefficients

100 220 1.042 0.603 1.73 0.090 2.00
150 180 1.156 0.603 1.92 0.061 2.00
150 200 -0.052 0.603 -0.09 0.931 2.00

6
 150 220 -1.052 0.603 -1.75 0.087 2.00

10 150 200 1.97 1.04 1.89 0.065 3.00

10 150 220 0.30 1.04 0.29 0.773 3.00

12 100 180 0.08 1.04 0.08 0.937 3.00

12 100 200 -0.58 1.04 -0.56 0.579 3.00

12 100 220 0.79 1.04 0.76 0.452 3.00

12 150 180 0.80 1.04 0.77 0.446 3.00

12 150 200 0.51 1.04 0.49 0.627 3.00

12 150 220 -2.49 1.04 -2.39 0.021 3.00

From table 5 we can see the P-Value for some interactions between two values close to 0.05 that mean is
the pure factors are not significant but there interaction is very strong and the most significant point is
when carbonation is 12% with 150 psi pressure and 220 bpm, at P value equal 0.021 < 0.05. Therefore, the
process engineer has to focus around this point for adjustment the process.

5 Conclusion

From all above results and based of P-value, we conclude that each factor alone is not significant and
Modifying it individually would be technically infeasible, Accordingly, it will be practical to adjust
the production line at the appropriate point previously mentioned, the best and fastest procedure from
a practical point of view. Imagine that we want to modify this line and make sure that at every point
we will take hundreds of hours and stop production to adjust the line. Thanks to statistical tests, we
were able to obtain the critical point that the maintenance and modification process should center
around.

References

1. Statistics for Experimenters: An Introduction to Design, Data Analysis, and Model Building. John Wiley & Sons
Inc., New York, NY.Box, G.E.P., Hunter, W.G., and Hunter, J.S. (2005).
2. Design and Analysis of Experiments (8th ed.). John Wiley & Sons Montgomery, D.C. Inc., Hoboken, NJ.(2013).
3. Response Surface Methodology: Process and Product Optimization Using Designed Experiments (3rd ed.)Myers,
R.H., Montgomery, D.C., and Anderson-Cook, C.M.. John Wiley & Sons Inc., Hoboken, NJ. .(2009).
4. Carbonated Soft Drinks: Formulation and Manufacture" by Duncan Manley and Peter Fox
5. The Science of Soft Drinks: From Production to Consumption" by John Mlynczak and John Ritzenthaler, Jr.
6. Design for Manufacturability: How to Use Parallel Engineering to Rapidly Develop Low-Cost, High-Quality
Products for Lean Production, by DFM Solutions Group.
7. The Lean Six Sigma Pocket Toolbook: A Quick Reference Guide to 100 Tools for Improving Quality and Speed,
by Michael L. George.
8. Introduction to Manufacturing Processes, by John A. Schey, McGraw-Hill, 2000