Sei sulla pagina 1di 7

Quality Engineering Lab 4 Assignment Seneida Biendarra

1.
Residual Plots for Burning Rate
Normal Probability Plot Versus Fits
99
20
90
10

Residual
Percent

50
0
10
-10
1
-20 -10 0 10 20 15 20 25 30 35
Residual Fitted Value

Histogram Versus Order


10.0 20

7.5
Frequency

Residual 10

5.0
0
2.5
-10
0.0
-10 -5 0 5 10 15 20 1 5 10 15 20 25 30 35
Residual Observation Order

From the residuals plots, we can see that the data is random, normally distributed, and has
equal variance.

ANOVA: Burning Rate versus Batch, Process


Factor Information
Level
Factor Type s Values

Batch Fixe 4 1, 2, 3,
d 4
Proces Fixe 3 1, 2, 3
s d
Analysis of Variance for Burning Rate
Source DF SS MS F P Batch: the p-value is within a=0.05 and is within
the statistical likelihood, will probably not effect
  Batch 3 230.3 76.77 1.0 0.40 the final outcome.
0 6
  2 676.1 338.0 4.4 0.02
Process: the p-value is less than 0.05, is
Process 3 1 1 statistically significant and must be accounted
for in the outcome
Error 30 2301. 76.71    
Quality Engineering Lab 4 Assignment Seneida Biendarra

3
Total 35 3207.      
6
Model Summary
R-
S R-sq sq(adj)

8.7583 28.26 16.30%


8 %
Residual Plots for Burning Rate

Boxplot of Burning Rate


60

50
Burning Rate

40

30

20

10
Process 1 2 3 1 2 3 1 2 3 1 2 3
Batch 1 2 3 4

There are a series of deductions can be made from the visual interpretation of this
data set:
Batch 1: Process 1 has a statistically different outcome that could indicate that the
system is “out of control”
Batch 2: All 3 proceedings have similar ranges and means
Batch 3: Process 3 is significantly different from all other data sets.
Batch 4: Process 1 in all four batches is the lowest mean and could indicate that there
is a source of interference between the Batch and Process.
Quality Engineering Lab 4 Assignment Seneida Biendarra

Boxplot of Burning Rate


60

50
Burning Rate

40

30

20

10
Batch 1 2 3 4 1 2 3 4 1 2 3 4
Process 1 2 3

ANOVA: Burning Rate versus Batch, Process


Factor Information
Level
Factor Type s Values

Batch Fixe 4 1, 2, 3,
d 4
Proces Fixe 3 1, 2, 3
s d
Analysis of Variance for Burning Rate
Batch: the p-value is less than 0.05, is
Source DF SS MS F P statistically significant and must be accounted
for in the outcome
  Batch 3 230.3 76.77 4.06 0.01
8 Process: the p-value is less than 0.05, is
  Process 2 676.1 338.0 17.8 0.00 statistically significant and must be accounted
3 7 0 for in the outcome
  6 1847. 307.8 16.2 0.00 Batch*Process: the p-value is less than 0.05, is
Batch*Process 3 8 8 0 statistically significant and the interaction
Error 24 454.0 18.92     between the two factors must be accounted for
in the outcome
Total 35 3207.      
Quality Engineering Lab 4 Assignment Seneida Biendarra

6
Model Summary
R-
S R-sq sq(adj)

4.3493 85.85 79.36%


3 %
Residual Plots for Burning Rate

2. In this problem, there is human factor that must be accounted for. Because the results of the
study could be skewed by the skills and ability of the operator, we must use an additive model
to remove the interaction of the operator on the outcome.

Residual Plots for SurfaceFinish


Normal Probability Plot Versus Fits
99
20
90
10
Residual
Percent

50 0

10 -10

-20
1
-20 -10 0 10 20 40 60 80
Residual Fitted Value

Histogram Versus Order


6.0
20
4.5
Frequency

10
Residual

3.0 0

1.5 -10

-20
0.0
-20 -10 0 10 20 2 4 6 8 10 12 14 16 18 20 22 24
Residual Observation Order
.

Before we begin, we can confirm that the data meets the assumptions needed to use an ANOVA model.
The Residual charts show us that the data is normally distributed, random, and has equal variance.

Now, we will conduct an ANOVA, including the operator in the model. This is because not only will the
operators have different experience levels, they might also have a certain skillset that gives them more
experience on one machine over another.
Quality Engineering Lab 4 Assignment Seneida Biendarra

ANOVA: SurfaceFinish versus Machine, Operator


Factor Information
Level
Factor Type s Values

Machine Fixe 4 1, 2, 3,
d 4
Operato Fixe 3 1, 2, 3
r d
Analysis of Variance for SurfaceFinish
Source DF SS MS F P

  3 3617. 1205. 7.6 0.00


Machine 7 9 6 2

  2 996.6 498.3 3.1 0.06


Operator 6 6
Error 18 2835. 157.5    
1
Total 23 7449.      
3
Model Summary
R-
S R-sq sq(adj)

12.550 61.94 51.37%


1 %
Residual Plots for SurfaceFinish

The objective going forward is to determine:

- Which machine produces the best outcome


- Does which operator is running the machine have an impact on the final outcome?
Quality Engineering Lab 4 Assignment Seneida Biendarra

Boxplot of SurfaceFinish
100

90

80
SurfaceFinish

70

60

50

40

30
Operator 1 2 3 1 2 3 1 2 3 1 2 3
Machine 1 2 3 4

Using the boxplot, we can break down the performance of each variable in context.

Machine 1: The range of means are statistically different, and the ranges are all vary. This machine lacks
both accuracy and precision, which could be further exasperated depending on the operator

Machine 2: While the means for each operator are statistically different, the range of each is similar.
This shows that while the machine lacks accuracy (ie., the mean is not accurate), it has precision (there is
a low standard deviation and variance). The operator could account for the difference in means.
Machine 2 has the highest likelihood of producing the highest quality product.

Machine 3: The range of means are statistically different, and the ranges are all vary. This machine lacks
both accuracy and precision, which could be further exasperated depending on the operator

Machine 4: Judging by the range of the boxplot values, Machine 4 has high consistency across all the
operators. Not only are the means statistically similar to one another, they all have a similar range.
However, assuming that a higher number indicates higher performance, this machine is not performing
well. This could simply be because of an assignable cause, because otherwise this machine looks like it
would run well if the cause is identified and removed.
Quality Engineering Lab 4 Assignment Seneida Biendarra

Boxplot of SurfaceFinish
100

90

80
SurfaceFinish

70

60

50

40

30
Machine 1 2 3 4 1 2 3 4 1 2 3 4
Operator 1 2 3

From the boxplot, we can make several assumptions about each operator:

Operator 1: This operator has the greatest amount of knowledge in running machines 1,2 and 3.
However, performance on machine 4 is lacking

Operator 2: This operator has similar Surface Finish performance on machines 1 & 2 but a much lower
mean for machines 3 & 4

Operator 3: While this operator does not have particular expertise overall, they have the most
consistent performance across all machines.

An overall observation on the effect of machine/operator combinations is that there is a significant


difference between high and low performance. This could be an indication that operator perform well
once they are properly trained, but that they have similarly low performance if they are not.

Potrebbero piacerti anche