MEMORIAL UNIVERSITY OF NEWFOUNDLAND

ENGI 7930 – Instrumentation and Exp. Design

Lab 2: LabVIEW Temperature Measurement

Project Team:
Shahir Haque – 201655149
Rayhanul Islam – 201558095
Noah Jacob Fultz – 201509056

Professor: Oscar DeSilva

1
Contents
List of Figures .......................................................................................................................................................................... 3
List of Tables ........................................................................................................................................................................... 4
Concrete Production Simulator .............................................................................................................................................. 5
Concrete Production Simulator ½ Factors ............................................................................................................................ 18
Drilling Advance Rate Study .................................................................................................................................................. 29
Polymer Conversion Experiment .......................................................................................................................................... 40
BlueBall Simulator ................................................................................................................................................................. 47
References ............................................................................................................................................................................ 57

2
List of Figures
Figure 1: Concrete Production Simulator main effects........................................................................................................... 6
Figure 2: Concrete Production Simulator half normal plot..................................................................................................... 7
Figure 3: Concrete Production Simulator pareto chart. ......................................................................................................... 8
Figure 4: Concrete Production Simulator ANOVA table. ........................................................................................................ 8
Figure 5: Concrete Production Simulator adjusted and predicted. ........................................................................................ 9
Figure 6: Concrete Production Simulator box-cox plot......................................................................................................... 10
Figure 7: Concrete Production Simulator normal plot of residuals. ..................................................................................... 11
Figure 8: Concrete Production Simulator residuals vs predicted. ........................................................................................ 12
Figure 9: Concrete Production Simulator residuals vs run. .................................................................................................. 12
Figure 10: Concrete Production Simulator equation in terms of coded factors................................................................... 13
Figure 11:Concrete Production Simulator equation in terms of actual factors.................................................................... 14
Figure 12: Concrete Production Simulator interaction between mix time and yield strength. ........................................... 15
Figure 13: Concrete Production Simulator interaction between yield strength and mix time. ........................................... 16
Figure 14:Concrete Production Simulator interaction between mix and dry time .............................................................. 17
Figure 15: Concrete Production Simulator interaction between mix and temperature. ..................................................... 18
Figure 16:Concrete Production Simulator main effects for ½ factors. ................................................................................. 19
Figure 17: Concrete Production Simulator half normal plot for ½ factors. .......................................................................... 20
Figure 18:Concrete Production Simulator pareto chart for ½ factors. ................................................................................ 20
Figure 19: Concrete Production Simulator ANOVA data for ½ factors. ............................................................................... 21
Figure 20:Concrete Production Simulator adjusted vs predicted for ½ factors. .................................................................. 21
Figure 21: Concrete Production Simulator equation in terms of coded factors................................................................... 22
Figure 22: Concrete Production Simulator equation in terms of actual factors................................................................... 23
Figure 23: Concrete Production Simulator box-cox plot for 1/2 factors. ............................................................................. 24
Figure 24:Concrete Production Simulator normal plot of residuals for 1/2 factors. ............................................................ 25
Figure 25: Concrete Production Simulator residuals vs predicted for 1/2 factors. .............................................................. 25
Figure 26: Concrete Production Simulator residuals vs run for 1/2 factors. ........................................................................ 26
Figure 27: Concrete Production Simulator interaction between mix and time for ½ factors. ............................................. 27
Figure 28: Concrete Production Simulator interaction between mix and dry time for ½ factors. ....................................... 28
Figure 29: Concrete Production Simulator interaction between mix and temperature for ½ factors. ................................ 29
Figure 30: Input data for drilling advance rate study. .......................................................................................................... 30
Figure 31: Matlab interface for drilling advance rate study with both numerical and categorical inputs. .......................... 30
Figure 32: Drilling advance rate study half normal plot. ...................................................................................................... 31
Figure 33: Drilling advance rate study pareto chart. ............................................................................................................ 32
Figure 34: Drilling advance rate study ANOVA data. ............................................................................................................ 32
Figure 35: Drilling advance rate study ANOVA data. ............................................................................................................ 33
Figure 36: Drilling advance rate study equation in terms of coded factors. ........................................................................ 33
Figure 37: Drilling advance rate study normal plot of residuals. .......................................................................................... 34
Figure 38: Drilling advance rate study residuals vs predicted. ............................................................................................. 34
Figure 39: Drilling advance rate residuals vs run. ................................................................................................................. 35
Figure 40: Drilling advance rate study predicted vs actual. .................................................................................................. 35
Figure 41: Drilling advance rate study bx-cox plot................................................................................................................ 36
Figure 42: Drilling advance rate study interaction between md type and drill load. ........................................................... 36
Figure 43: Drilling advance rate study 3D plot of relation between flow rate and drill load. .............................................. 37
Figure 44: Drilling advance rate study 3D plot of relation between flow rate and drill load. .............................................. 37
Figure 45: Drilling advance rate study point prediction data. .............................................................................................. 38
Figure 46: Drilling advance rate study point prediction test in agreement. ......................................................................... 38
3
Figure 47:Drilling advance rate study point prediction data for mud type B. ...................................................................... 39
Figure 48: Figure 46: Drilling advance rate study point prediction test in agreement for mud type B. ............................... 39
Figure 49: Input data for polymer conversion experiment................................................................................................... 40
Figure 51: Matlab interface of polymer conversion experiment. ......................................................................................... 41
Figure 52: Polymer conversion experiment half normal plots, for both responses. ............................................................ 41
Figure 53: Polymer conversion experiment ANOVA data, for both responses. ................................................................... 42
Figure 54: Polymer conversion experiment predicted vs adjusted, for both responses. ..................................................... 42
Figure 55: Polymer conversion experiment equation in terms of actual factors, for both responses. ................................ 42
Figure 56: Polymer conversion experiment normal plot of residuals, for both responses. ................................................. 43
Figure 57: Polymer conversion experiment residuals vs predicted, for both responses. .................................................... 44
Figure 58: Polymer conversion experiment predicted vs actual, for both responses. ......................................................... 44
Figure 59: Polymer conversion experiment box-cox plot, for both responses..................................................................... 45
Figure 60: Polymer conversion interaction terms, for both responses. ............................................................................... 45
Figure 61: Polymer conversion experiment 3D plot of temperature against catalyst, for both responses. ........................ 46
Figure 63: Blueball experiment main effects. ....................................................................................................................... 48
Figure 64: Blueball simulator ANOVA data. .......................................................................................................................... 49
Figure 65: Blueball simulator ANOVA summary. .................................................................................................................. 50
Figure 66: Blueball simulator half-normal plot. .................................................................................................................... 50
Figure 67: Blueball simulator pareto chart. .......................................................................................................................... 51
Figure 68: Blueball simulator box-cox plot. .......................................................................................................................... 52
Figure 69: Blueball simulator normal plot of residuals. ........................................................................................................ 53
Figure 70: Blueball simulator residuals vs predicted. ........................................................................................................... 54
Figure 71:Blueball simulator residuals vs run. ...................................................................................................................... 55
Figure 72: Blueball simulator predicted vs actual. ................................................................................................................ 55
Figure 73: Blueball simulator 3D plot of factor H against A. ................................................................................................. 56

List of Tables
Table 1: Concrete Production Simulator response data. ........................................................................................................ 5
Table 2: Concrete Production Simulator response data for ½ factors.................................................................................. 19
Table 3: Response data for drilling advance rate study. ....................................................................................................... 30
Table 4: Response data for polymer conversion experiment. .............................................................................................. 40
Table 5: Blueball experiment response data ........................................................................................................................ 47

4
Concrete Production Simulator

Table 1: Concrete Production Simulator response data.

5
After naming all the factors and parameters, the abovementioned table was formed that consists of 32 test runs with all
possibilities for 25. The responses were found using Concrete Production on the Matlab DOE application.

Figure 1: Concrete Production Simulator main effects.

The table shows the standardized effects, sum of squares, and % contribution for all the possible intercept terms for the
tests.
6
 Figure 2: Concrete Production Simulator half normal plot.

The Half-Normal Plot is generated where the points close to the line is considered less significant, and the points far
away from the line are the factors which cause the main effect.

7
 Figure 3: Concrete Production Simulator pareto chart.

The above figure is the Pareto Chart which shows factors that are above or below the t-value. The t-value limit is
2.05553 and the factors above the t-value show significant effect. The Pareto Chart agrees with the Half-Normal plot as
the factors A, B, E, D, AB are causing the effects.

Figure 4: Concrete Production Simulator ANOVA table.

8
The Anova table was generated with acceptable p-value % significance of 5%. Only the significant model terms (with p-
value < 0.05) were listed in this Anova table. With p-value less than 5%, these effects are caused due to main effects.

Figure 5: Concrete Production Simulator adjusted and predicted.

The adjusted and predicted R-Squared values of regression are fairly close and it is reasonable to assume a good fit.

9
 Figure 6: Concrete Production Simulator box-cox plot.

The Box-Cox plot for power transformation shows the current lambda and best lambda values are really close, with
values 1 and 1.03 respectively. The recommended transformation is ‘None’.

10
 Figure 7: Concrete Production Simulator normal plot of residuals.

Since the majority of the points lie on the line, the normality of the residuals are acceptable. However, one point between
-2 and -1 of internally studentized residuals is away from the line. The majority errors follow Gaussian distribution.

11
 Figure 8: Concrete Production Simulator residuals vs predicted.

All the predicted values have internally studentized residual values between 2 to -2 and the values does not show any
triangle shape narrowing down the residual values.

Figure 9: Concrete Production Simulator residuals vs run.

12
The Residual vs Run graph agrees with the Residual vs Predicted graph.

Figure 10: Concrete Production Simulator equation in terms of coded factors.

13
 Figure 11:Concrete Production Simulator equation in terms of actual factors.

The figure shows the final equations in terms of Actual factors and final equation in term of Coded factors.

14
 Figure 12: Concrete Production Simulator interaction between mix time and yield strength.

The graph above shows the interaction between Mix and Time (x-axis) vs Yield Strength (y-axis)

15
 Figure 13: Concrete Production Simulator interaction between yield strength and mix time.

The graph above shows the interaction between Time and Mix (x-axis) vs Yield Strength (y-axis)

16
 Figure 14:Concrete Production Simulator interaction between mix and dry time

The graph above shows the interaction between Mix and Dry Time (x-axis) vs Yield Strength (y-axis)

17
 Figure 15: Concrete Production Simulator interaction between mix and temperature.

The graph above shows the interaction between Mix and Temperature (x-axis) vs Yield Strength (y-axis)

Concrete Production Simulator ½ Factors

Question 2 was done similar to Question 1, except to have ½ number of factors, the team chose 25-1 instead of 24. This is
because we have 5 parameters, and thus we need to have 2 to the power of 5, and then we can tweak that to find our
desired amount of test runs. All the graphs are very similar, except now we have 16 results than 32.

18
Table 2: Concrete Production Simulator response data for ½ factors.

Figure 16:Concrete Production Simulator main effects for ½ factors.

19
 Figure 17: Concrete Production Simulator half normal plot for ½ factors.

This Half-Normal Plot shows D and E are on the line, however this was not the case when we took 2^5 factors. Factors
that are deviated from the line, are considered to cause significant main effect.

Figure 18:Concrete Production Simulator pareto chart for ½ factors.

20
A, B, AB are drastically beyond the t-value limit. E and D values are above the line as well. Factors beyond the limit
shows significant effects.

Figure 19: Concrete Production Simulator ANOVA data for ½ factors.

All the p-values are less than 0.05 except D and E, which was not the case when experimented with 32 runs. With 32
runs D and E also had below 0.05 p-value. With 16 runs, these are considered insignificant.

Figure 20:Concrete Production Simulator adjusted vs predicted for ½ factors.

The adjusted and predicted R-Squared values of regression are fairly close and it is reasonable to assume a good fit.
21
Figure 21: Concrete Production Simulator equation in terms of coded factors.

22
 Figure 22: Concrete Production Simulator equation in terms of actual factors.

The figure shows the final equations in terms of Actual factors and final equation in term of Coded factors.

23
 Figure 23: Concrete Production Simulator box-cox plot for 1/2 factors.

The Box-Cox plot for power transformation shows the current lambda and best lambda values are really close, with
values 1 and 0.95 respectively. The recommended transformation is ‘None’.

24
 Figure 24:Concrete Production Simulator normal plot of residuals for 1/2 factors.

Since the majority of the points lie on the line, the normality of the residuals are acceptable with no major drastic deviation

Figure 25: Concrete Production Simulator residuals vs predicted for 1/2 factors.

25
Residual and predicted run vs internally studentized residual. The graph is similar to the previous run, with residual
values between 1.8 to -2.2.

Figure 26: Concrete Production Simulator residuals vs run for 1/2 factors.

Run and Run number vs Residual values. The graph is very similar to the previous experiment of 32 runs.

26
 Figure 27: Concrete Production Simulator interaction between mix and time for ½ factors.

The graph above shows the interaction between Mix and Time (x-axis) vs Yield Strength (y-axis). This graph shows a
steeper slope than the previous experiment, this is because of the difference in number of factors.

27
 Figure 28: Concrete Production Simulator interaction between mix and dry time for ½ factors.

The graph above shows the interaction between Mix and Dry Time (x-axis) vs Yield Strength (y-axis)

28
 Figure 29: Concrete Production Simulator interaction between mix and temperature for ½ factors.

The graph above shows the interaction between Mix and Temperature (x-axis) vs Yield Strength (y-axis).

Drilling Advance Rate Study

In this study, the drilling advance rate is simulated using Design Expert. The study involves four factors relating to the
drilling process. The factors are as follows:

Drill load (N)

Mud flow rate (ml/s)

Rotational speed of the shaft (rpm)

Drilling mud type

Experimental Data

The output response for the experiment is the measured advance rate of the drill. The response equation used in the
model is in the form of a Natural Log Equation in the form y’=ln(y + k) . The experiment conducted had (2 4 =) 16 runs
with all possible combinations considered. The Input data is shown in the table below.

29
 Figure 30: Input data for drilling advance rate study.

Each factor is named, and the high and low values are input into the Design Expert interface. The experiment is then run,
and the results are obtained as shown below.

Table 3: Response data for drilling advance rate study.

The response was calculated for each run of the experiment using the Matlab
Simulator and the recorded in the datasheet.

Figure 31: Matlab interface for

drilling advance rate study with both
numerical and categorical inputs.

30
 Figure 32: Drilling advance rate study half normal plot.

The Half-Normal Plot is then generated. The points along the line are considered less significant and can be
characterized as noise. The points which deviate from the line are the factors which cause the main effect.

31
 Figure 33: Drilling advance rate study pareto chart.

The obtained Pareto Chart shows the factors whose t-values are below and above the t-value limit. (2.2281). Factors
with t-values beyond the limit show significant effects. The above Pareto Chart agrees with the Half-Normal Plot
obtained for the experiment as indicated by the factors causing the main effects. (C, A, D, BD)

Figure 34: Drilling advance rate study ANOVA data.

32
An Anova table is generated with a significance level of 0.05. The significant model terms are A, C, D and BD. Since their
p-values are less than 0.05, there is a high probability that these effects are not due to noise but the main effects.

Figure 35: Drilling advance rate study ANOVA data.

The adjusted and predicted R-Squared values of regression are fairly close. Their difference is small and it is reasonable
to assume a good fit. Additional data points are not required.

The signal to noise ratio is indicated by the “Adeq Precision” field. The obtained ratio of 33.09 which is greater than 4
which is the minimum desirable ratio for an adequate signal. All factors indicate a god model thus far.

Figure 36: Drilling advance rate study equation in terms of coded factors.

The final equation in terms of the coded factors are shown above. The Equation was not available in actual factors.

Results and Discussion

Before finalizing the selection of the model, some further analysis was performed verify the validity of the model. The
following checks are performed:

a) The normality of residuals

33
 Figure 37: Drilling advance rate study normal plot of residuals.

Since most points are close to the line the line, the model exhibits good normality of residuals. This implies that the
majority of errors follow a Gaussian distribution.

b) Equality of variance

Figure 38: Drilling advance rate study residuals vs predicted.

34
 Figure 39: Drilling advance rate residuals vs run.

Since the points in the Residual vs Predicted plot are well dispersed and appear to be scattered randomly, an acceptable
equality of variance can be assumed. This means for different levels of each factor, there is equal standard deviation for
errors.

c) Accuracy of prediction equation

Figure 40: Drilling advance rate study predicted vs actual.

The Predicted vs Actual Plot shown above shows how accurate the prediction equation can predict the output. The
points are within reasonable distance from the line which signifies accurate predictions.

35
d) Transformations

Figure 41: Drilling advance rate study bx-cox plot.

The experiment was run for the first time with no transformation and the results recommended a natural log
transformation. This was changed and all the results shown above are obtained after the recommended transformation
was applied. The equation is transformed into the form y’ = (y + k).

Interaction Terms

Figure 42: Drilling advance rate study interaction between md type and drill load.

The interaction graph shown above illustrates the effects between Factor D and Factor B on the response. The
interaction was studied for a rotational speed of 20000 rpm shown as a red square and a flow rate of 6 ml/s show as a
green triangle. There are significant interactions between the factors as seen in the graph.

36
 Figure 43: Drilling advance rate study 3D plot of relation between flow rate and drill load.

A 3D Surface graph is created which shows how the response (advance rate of drill) changes with the selected variables.
This is done by choosing two factors that are numeric and by varying the remaining factors to see how it affects the
reponse. The figure above shows how the response changes for a selected mud type A at rotational speed of 10000 rpm
for a range of flow rate and drill load values. It is observed that the maximum advance rate is greater than the predicted
value for these factors.

Figure 44: Drilling advance rate study 3D plot of relation between flow rate and drill load.

The response surface shown above is for mud type B and rotational speed of 10000rpm. Comparing the response
surface of both mud types shows mud type B achieves a larger advance rate for the range of flow rates and drill loads.

37
Point Prediction Check

To verify the accuracy of the results the Point Prediction tool was used to compare the values obtained from the Design
Expert software and the DOE Simulator.

The check is to verify that the predicted response from Design Expert agrees with the response calculated using the DOE
simulator.

The first check is done using all low level values for each factor for Mud type A. The Point Prediction tool data is shown
below.

Figure 45: Drilling advance rate study point prediction data.

The predicted output should be between 1.30158 mm/min and 2.25747mm/min.

Figure 46: Drilling advance rate

study point prediction test in
agreement.

The results obtained from the Design Expert agree with the simulator results which gave a response of 1.725 mm/min

Another check was done for factors with their low levels for Mud type B.

38
 Figure 47:Drilling advance rate study point prediction data for mud type B.

The predicted output should be between 1.30158 mm/min and 2.25747 mm/min.

Figure 48: Figure 46: Drilling advance rate study point prediction test in agreement for mud type B.

The results obtained from the Design Expert agree with the simulator results which gave a response of 2.639 mm/min.

39
Polymer Conversion Experiment
In this study, a Polymer Conversion experiment is simulated using Design Expert. Polymers are substances made up of
many smaller units called monomers. The monomer is converted to a polymer by a chemical process. The rate of
Polymer conversion can be studied by conducting an experiment with the following 3 variables used in the conversion
process. The factors are as follows:

• Temperature (deg-C)
• Time (min)
• Catalyst (%)

Experimental Data

The output response for the experiment is the conversion rate and the molecular activity. The response equation used in
the model is in the form of a Square Root Equation in the form y’= √(y + k) . The experiment conducted had (23 =) 8 runs
with all possible combinations considered. The Input data is shown in the table below.

Figure 49: Input data for polymer conversion experiment.

Each factor is named, and the high and low values are input into the Design Expert interface. The experiment is then run,
and the results are obtained as shown below.

Table 4: Response data for polymer conversion experiment.

40
 The response was calculated for each run of the experiment using the
Matlab Simulator and the recorded in the datasheet.

Figure 50: Matlab interface of polymer

conversion experiment.

Figure 51: Polymer conversion experiment half normal plots, for both responses.

41
 The Half-Normal Plot for both responses are shown above. The points along the line are considered less
significant and can be characterized as noise. The points which deviate from the line are the factors which cause the
main effect.

Figure 52: Polymer conversion experiment ANOVA data, for both responses.

An Anova table is generated with a significance level of 0.05 by selecting the Factors whose contribution to is less than
5% from the Effects List. The contribution of those factors are most likely due to noise. The significant model terms are
A, C, AC and BC, and B and C for Conversion and Activity respectively. Since their p-values are less than 0.05, there is a
high probability that these effects are not due to noise but the main effects.

Figure 53: Polymer conversion experiment predicted vs adjusted, for both responses.

For the Conversion response, there appeared to be significant difference in the predicted and adjusted R-squared values
and for Activity response, the R-squared values are in reasonable agreement. The discrepancy is attributed to the limited
number of runs performed in this experiment.

Figure 54: Polymer conversion experiment equation in terms of actual factors, for both responses.

The final equation in terms of the actual factors are shown above.

42
Results and Discussions

a) The normality of residuals

Figure 55: Polymer conversion experiment normal plot of residuals, for both responses.

From the Normal Plot of Residuals for both the Conversion and Activity response, a trend is seen in the residual values.
The are all located along the line with some points away from the line. The normality of residuals are acceptable. This
implies that the majority of errors follow a Gaussian distribution.

b) Equality of variance

43
 Figure 56: Polymer conversion experiment residuals vs predicted, for both responses.

Since the points in the Residual vs Predicted plot are well dispersed and appear to be scattered in an upwards and
downwards manner, an acceptable equality of variance can be assumed. This means for different levels of each factor,
there is equal standard deviation for errors.

c) Accuracy of prediction equation

Figure 57: Polymer conversion experiment predicted vs actual, for both responses.

44
The Predicted vs Actual Plot shown above shows how accurate the prediction equation can predict the output. The
points are within reasonable distance from the line which signifies fairly accurate readings although it should be noted
there are some inaccuracies involved in the prediction of the output.

d) Transformations

Figure 58: Polymer conversion experiment box-cox plot, for both responses.

The Box-Cox plot for Power Transform is a tool used to check whether the results may be optimized by transforming the
response equation. Initially a square-root transform was used for both responses in this problem due to the non-linear
form of the response.

Interaction Terms

Figure 59: Polymer conversion interaction terms, for both responses.

45
 The Interaction terms were only present for the Conversion response and not for the Activity response. The
interaction shown between factors are at the minimum temperature of 40 deg-C and for low level Catalyst (black
square) and high level Catalyst (triangle). The Conversion response shows significant interaction between B and C as the
time goes from 15 to 35 mins for both levels of Catalyst.

For interactions between A and C, low level Catalyst shows significant interactions between factors but the high level
does not have very significant interactions.

Figure 60: Polymer conversion experiment 3D plot of temperature against catalyst, for both responses.

The plot above shows a 3D response surface for both responses.

46
BlueBall Simulator

Table 5: Blueball experiment response data

The table above lists the experimental values; the responses were obtained using the matlab BlueBall simulator.

47
 Figure 61: Blueball experiment main effects.

The table above displays the contribution of each response term by giving a percentage. A term can be classified as an
error if it has a p-value of more than 5%, which is information attained by using the ANOVA table on the following page.

48
 Figure 62: Blueball simulator ANOVA data.

From this ANOVA table, we see that there are some effects listed that can be considered error. These are A-A, H-H, A-H,
and BD. The two-factor effects BE and DE are close to being able to be considered as error, but not quite. All three-and-
higher factor terms are considered negligible due to the principle of sparsity of effects.

Below is a summary of the ANOVA table.

49
 Figure 63: Blueball simulator ANOVA summary.

Figure 64: Blueball simulator half-normal plot.

The half normal plot shows all significant terms, where any term laying on the ling is insigificant/caused by noise. The
farther away from the line, the more significant the effect. Here we see that F, D, B, E, EF, C and BC are all significant
effects.

50
 Figure 65: Blueball simulator pareto chart.

This pareto chart follows a rough exponential decay. It is very visually useful, as It shows exactly the line for which terms
may be deemed significant vs. insignificant. Confirming what we know from the half normal plot, effects F, D, B, E, EF, C,
and BC are the most significant. As suspected, the farthest effect from the line, BC, is also the farthest from the line on
the half normal plot, though this chart allows us to more readily compare effects.

51
 Figure 66: Blueball simulator box-cox plot.

This box-cox plot is also very visually useful, showing the optimal value at the minimal position. It does this by fitting the
data to a quadratic, in order to ensure a minimum value can be found. The data fits the curve surprisingly well.

52
 Figure 67: Blueball simulator normal plot of residuals.

Another way to visualize this data is by analyzing the normal plot of residuals. In this plot, the distance from the line
gives the R2 value, though it may be hard to read if there is a lot of data.

53
 Figure 68: Blueball simulator residuals vs predicted.

The residuals vs. predicted plot can determine the strength of prediction in a model. A uniform distribution is
indicative of an accurately predicting model, while a larger scatter is less accurate in modeling the error. Here we see
that the data is reasonably distributed, but if the experiment was conducted again, a tighter residuals vs normal graph
would be sought.

54
 Figure 69:Blueball simulator residuals vs run.

Very much the same as the residuals vs predicted plot, the residuals vs run plot sees the residuals error points
compared with the runs instead of residual values.

Figure 70: Blueball simulator predicted vs actual.

The predicted vs. actual graph is a great way of determining if your final design is accurate enough. If the scatter is very
large and inconsistent, the model may need to be reworked. While this model isn’t the most accurate, it does follow the
predicted line quite well trend-wise. This allows us to infer that residuals vs predicted plot may, in fact, be permissible
here, depending on the specific requirements of the design.

55
 Figure 71: Blueball simulator 3D plot of factor H against A.

The 3D surface is a representation of the response surface of the H factor against the A factor, showing both the
target error and its distribution over the surface. This comparison can be made between other factors as well; this
depends on the information desired from the experiment.

56
References
[1] O. DeSilva, “ENGI 7930 Course Notes,” Brightspace [Online].
Available: https://online.mun.ca/d2l/le/content/352111/viewContent/3299559/View?ou=352111

57