Assignment 4B

Process Capability Six Pack Analysis

Group 2

AKHIL RATHEESHAN, ANAND VIJAYAN, SRADDHA MUKUNDAN, DURGA

PRASANNA, JOHN PAUL RAJU
3-12-2023
Executive Summary

Date: 03-12-2023

Introduction:

The study report will provide us with whether the measurement system meets the quality control
standards. The evaluation aims to determine the measurement system's capability to meet the
specifications. The data were collected by providing a single part to five appraisers, and the
measurements were collected once a week for 25 days with a ruler.

Analysis Findings:

Numerical Analysis: The measurement system is incapable because the data is abnormal. The “P-value “
is less than 0.005, showing that the data is abnormal and failed the Box-Cox transformation.

Graphical Analysis: The graphic and numerical analysis also shows abnormal data. The R bar chart shows
that the process is unstable; the histogram analysis clearly states that data distribution is not within the
specification limits. From the probability distribution chart, we can see that the “P-value” is less than
0.005, indicating the abnormality of the data.

Recommendations:

Ensure the instrument is calibrated correctly and has the specified resolution and ensure that the device
is defect-free from wear and tear due to continuous use; make sure that all the operators are properly
trained and follow the standard process for collecting data.

Conclusion: According to numerical and graphical analysis from the 6 Pack variable Analysis study, the
measurement system is incapable.
Objective:

Here, we took a Lego part to check if its dimensions met the specific tolerances. The objective here is to
study the process capability of the part to check whether the process is meeting the required tolerance.

The Specific tolerance of the part has been found using the following equation.
USL−LSL
1.33 =
6σ
Where USL-LSL = Specification Tolerance
σ=Rbar/d2 , where d2 = 2.326 for n=5
substituting these values in the above equation
Specification Tolerance = 6*0.057*1.33
Specific Tolerance = 0.45486

Database

Samples Appraiser A Appraiser B Appraiser C Appraiser D Appraiser E

1 2.9 3 2.9 2.8 3
2 3 2.9 3 3 2.9
3 3 3 3 3 3
4 3 2.9 2.9 2.9 2.9
5 2.9 2.9 2.9 3 2.9
6 2.8 3 3 3 3
7 3 3 3 2.9 3
8 2.8 3 2.8 2.9 3
9 3 3 3 3 3
10 2.9 3 2.9 3 3
11 3 2.9 3 3 2.9
12 2.9 3 2.9 2.8 3
13 2.7 3 3 2.9 3
14 3 2.8 3 3 3
15 2.9 3 2.9 3 2.9
16 2.8 2.9 3 2.9 3
17 3 2.8 2.9 3 3
18 3 3 3 3 3
19 2.8 3 3 3 3
20 3 2.9 3 2.9 2.9
21 2.9 3 2.8 2.9 2.8
22 3 2.9 3 3 3
23 3 3 2.9 3 3
24 2.9 2.9 3 2.9 2.9
25 3 2.9 2.9 2.9 3
Data Collection and Processing

 Take a part that falls in the range of production measurements and keep it as the master part.
 Assign five appraisers to measure the part 5 times using a measurement scale once a shift for
four weeks for 25 days.
 Number of parts used: 1 Part
 Who took the measurements: 5 Appraisers took all the measurements
 Guage : Ruler Scale
 Sample size: 5
 Sample frequency: Once a shift for four weeks

Data Analysis Study

Six Pack Process Capability Analysis Study :

From the above 6-pack analysis, we can see that the process is unstable, and the data is not normal
because the R chart shows an outlier in the 13th observation. It indicates that the process is unstable.

And the P value for the average probability is less than 0.005, proving that the process is abnormal.
The overall Process Spread is larger than the Specification Spread, which means poor capability. The data
that are outside the specification limits represent nonconforming items. Overall and within bell curves
overlap, indicating that the process may be stable, and therefore, Cpk ≈ Ppk.

The center of the data occurs at the peak of the distribution curve. “P-value” is >0.05, so the process
target is not significantly different from the process mean. As a result, Cp and Cpk are not significantly
different.

From the above evidence, we can conclude that the data is not normal, and the process is not stable.

Box-Cox Transformation:

To transform the data to normal, we can do the Box-Cox transformation. It is commonly used in statistics
and data analysis to address issues such as non-normality.

The Box-Cox transformation is beneficial for handling data exhibiting heteroscedasticity or non-constant
variance, contributing to the stabilization of data spread. Its utility is evident when attaining a normal
distribution, and it is essential for statistical techniques like analysis of variance (ANOVA) or linear
regression. Notably, the Box-Cox transformation is designed explicitly for datasets with positive values
and cannot be used for those with zero or negative values. In cases involving such data, alternative
transformations like the Yeo-Johnson transformation can be considered, as they build upon the Box-Cox
approach to address a broader spectrum of data distributions.
After doing the Box-Cox transformation, the outliers in the R chart have been eliminated. However, in the
average prob plot, the “P-value” is still less than 0.005, indicating that even though the process is
statistically stable, the data is still abnormal.

As we can see from the above plot, the P value is still less than 0.005, so

So, let’s find a distribution that matches the shape of our data distribution (a
good fit!), with an acceptable “p-value”
But even after plotting so many distribution plots, there isn’t any distribution fit where the “P-value “ is
less than 0.005. Showing that the data is It's still not normal.
Analysis of Process Capability (Within):
Comparing the within subgroup StDev and Overall StDev in capability plot there is a difference. This
indicates that the process is Not stable.

From the capability plot we get the value of Cpk as 0.37. But the setup bench mark value is 1.33, So according
to the analysis, if Cpk value is less than 1.33 the process has a lot of variations and we need to make ways to
improve to make it stable.

When the CPU value is more than 1, it means that the process variability, or spread, is within permissible bounds in
relation to the specification limitations.In our situation, the CPU is 3.022, higher than the benchmark of 1.33,
indicating a fair degree of variable capability in the process.

In our instance, a CPL of 2.8711 is greater than 1, meaning that the process spread exceeds the specified width.

The high CPL can suggest that there is more variability in the process than anticipated, even though the CPU is
above the benchmark number. It's critical to take your process's particular requirements and product standards
into account. Depending on how well the process complies with the requirements, a high CPU doesn't always
indicate that it's flawless.

Analysis of Process Capability (Within):

Ppk value from the analysis chart is 0.37, which is lower than the benchmark value 1.33 indicating that
process is not consistent or stable enough and which produces the value outside the specified limits.
This can be an indication of unstable controlled process.

To reduce the issue, we should identify what the variation is to make it stable enough.

From the analysis both the Pp = Ppk, By observing that values, we can say the process is centered
between specifications limit.

Comparing Cpk and Ppk:

Control charts are commonly used to evaluate statistical control since they track a process's stability over
time. When a process exhibits statistical stability, it indicates that there are no assignable causes of
variation and that the observed fluctuations are part of the process itself.
Ppk and Cpk values of 0.37, when interpreted together, suggest that the process is having difficulty meeting
requirements.

Sigma Level:

Sigma level is a metric that shows how well a process can operate within parameters. Higher sigma levels are linked
to greater process performance and fewer errors in the Six Sigma approach. The process variation and the
separation between the process mean and the specification limits are used to compute the sigma levels.

Predicted Yield:

An estimate of the percentage of flawless goods or outputs that can be anticipated from a process is called
predicted yield. It is computed in accordance with the standards and the process capabilities.

Process Fall Out:

Probability that the process's output will deviate from its specifications.
Of course! Cpk is a metric used to evaluate how well a process can meet predetermined requirements or tolerances
when assessing process capacity. Since it shows that the process is well-centered within the tolerance limits and
has comparatively little fluctuation when compared to the specifications, a Cpk value of 1.33 is frequently used as a
benchmark for capable processes.
Based on a probability map, a Cpk value of 0.37 indicates a notable difference between the process performance
and the desired capacity. When a process's Cpk is 0.37, it means that its variance is making it difficult for it to
achieve the tolerance requirements.

These values represent the following:

1. The targeted capability level, Cpk = 1.33, indicates that the process should operate well within the given
bounds. It implies that, in comparison to the tolerance width, the process is quite capable and creates few
defects.
2. Cpk = 0.37 (Observed): This value shows that, in comparison to the specification limits, the process
variation is significantly wider. When the process's Cpk is 0.37, it is not operating at the required level of
capability. It implies that there is a greater likelihood that the process will produce products that are not
within the designated tolerance.

The process is not operating at the required capacity level, given the difference between the observed value of 0.37
and the targeted Cpk of 1.33. This discrepancy could point to several problems with the process, like too much
variety, erratic performance, or a lack of control.
It is crucial to locate and deal with the causes of variation, optimize process parameters, strengthen controls, and
reevaluate the process to ensure it satisfies the necessary standards to improve process capability and reach the
intended Cpk level.

Conclusion and Recommendations:

The above study clearly shows that the part cannot meet the required specifications. First, the data was not
normal, and the process was unstable. Both graphical and numerical analyses reveal the presence of
anomalous data. The R bar chart indicates process instability, while the histogram analysis explicitly
indicates that the data distribution exceeds the specification limits. Furthermore, the probability
distribution chart demonstrates a "P-value" below 0.005, signaling the abnormal nature of the data.
All the appraisers need further training and have to check Guage if the calibration of the Guage is accurate or not