Army Meta-Model Analysis
Alternate Designs
Introduction
Operating in a world of uncertainties, our military must make vital decisions based on the data and factors presented to them on a daily basis. For example, there are changes to personnel, supplies, training, and equipment constraints; additional to the common factors there are environmental effects that are not probable to predict consistently. Therefore, leaders need to not just develop meta-models, but they should use alternative designs to generate new meta-models to gain better insight on various situations.
Scenario
The objective for this analysis will be to use all of the tools provided in our meta-model analysis which are the Ruby script, AoSummary.rb, provide by Professor Sanchez. Also, we have been …show more content…
in 2012, which offers 512 Nearly Orthogonal Nearly Balanced (NOB) design points for the given intervals of the five factors. Also, leadership is required to use the Frequency Domain based design spreadsheet, rotate_5_factor.xlsx, provided by Professor Sanchez; which gives complete orthogonality for all main effects, quadratic effects, and two-way interactions. Additionally, the Frequency Domain design has been pre-rotated and stacked five times. Since the Frequency Domain design has been structured based off of the cosine function, leadership is required to properly scale and develop suitable input ranges for the AoSummary.rb model. Finally, each experiment has a run-length of 1,100 days and the truncation point of the Ruby script, AoSummary.rb model will reduce that to 1,000 days of operation. All the while, keeping in mind that there is no budget limit, hence, the constraint for number of experiments has been …show more content…
Consists of executing the given AoSummary.rb model with those 405 design points outlined in Figure 2.
2) A Nearly Orthogonal Nearly Balanced design (NOB).
Stated above in question 1, this design has 512 design points. The NOB_Mixed_512DP_v1.xlsx, spreadsheet provides a 300 mixed factor design with 10 blocks of 20 k-level discrete factors with 100 continuous factors. The pairwise correlation of this design is just over 3 percent. While experimenting with this design, there were two data sets created and both correlations plots can be seen in Figures 4 and 5.
As seen in Figure 3, it was designed by using all five factors in the continuous factor range of the NOB spreadsheet. Therefore, it fills all of the space and the correlations are below the 10 percent threshold. In Figure 4, the NOB was designed by placing initial stock, repair cycle and mean repair time in the continuous factors cells, and maintenance personnel was placed in the six level 2 factor as well as the breakdown rate was placed in the two level 1 factor of the NOB spreadsheet. This design in Figure 4 was purely generated to see if a suitable meta-model could predict “lower decile” knowing that less white space would be filled using the factors in the discrete
Introduction
Operating in a world of uncertainties, our military must make vital decisions based on the data and factors presented to them on a daily basis. For example, there are changes to personnel, supplies, training, and equipment constraints; additional to the common factors there are environmental effects that are not probable to predict consistently. Therefore, leaders need to not just develop meta-models, but they should use alternative designs to generate new meta-models to gain better insight on various situations.
Scenario
The objective for this analysis will be to use all of the tools provided in our meta-model analysis which are the Ruby script, AoSummary.rb, provide by Professor Sanchez. Also, we have been …show more content…
in 2012, which offers 512 Nearly Orthogonal Nearly Balanced (NOB) design points for the given intervals of the five factors. Also, leadership is required to use the Frequency Domain based design spreadsheet, rotate_5_factor.xlsx, provided by Professor Sanchez; which gives complete orthogonality for all main effects, quadratic effects, and two-way interactions. Additionally, the Frequency Domain design has been pre-rotated and stacked five times. Since the Frequency Domain design has been structured based off of the cosine function, leadership is required to properly scale and develop suitable input ranges for the AoSummary.rb model. Finally, each experiment has a run-length of 1,100 days and the truncation point of the Ruby script, AoSummary.rb model will reduce that to 1,000 days of operation. All the while, keeping in mind that there is no budget limit, hence, the constraint for number of experiments has been …show more content…
Consists of executing the given AoSummary.rb model with those 405 design points outlined in Figure 2.
2) A Nearly Orthogonal Nearly Balanced design (NOB).
Stated above in question 1, this design has 512 design points. The NOB_Mixed_512DP_v1.xlsx, spreadsheet provides a 300 mixed factor design with 10 blocks of 20 k-level discrete factors with 100 continuous factors. The pairwise correlation of this design is just over 3 percent. While experimenting with this design, there were two data sets created and both correlations plots can be seen in Figures 4 and 5.
As seen in Figure 3, it was designed by using all five factors in the continuous factor range of the NOB spreadsheet. Therefore, it fills all of the space and the correlations are below the 10 percent threshold. In Figure 4, the NOB was designed by placing initial stock, repair cycle and mean repair time in the continuous factors cells, and maintenance personnel was placed in the six level 2 factor as well as the breakdown rate was placed in the two level 1 factor of the NOB spreadsheet. This design in Figure 4 was purely generated to see if a suitable meta-model could predict “lower decile” knowing that less white space would be filled using the factors in the discrete