00138 ­b34bfe1c47dc055c8d02cda05623df

139

Optimization and Sensitivity Analysis

conclusions conceming highly significant variables with respect to expected cost per time unit by changing and not changing the highly significant variables and noting the effects on expected cost.

A major obstacle to industrial implementation of the LV model is the large number of terms and difficulties in their estimation. Our results indicate that one could use as few as four input variables and observe relatively smali changes in the cost response relative to the fuli model. This study provides a basis for the investigation of the use of cost ratios rather than actual cost as a further aid to implementation.

Table 8. Results of Mis-specification Test

Inputyariables    Results

X T₀ T. T, E C_q    C,    W a    b    Y    A Cost    % change

($)    {$)    ($)    ($)    ($)    ($)    ($)    over base

expected +    ---++    +    +    +    +    +    -

relalion

with cost

baseline 007    0.6 0.3 0.2 0.1    10 50    10 0.3 0.1 20 0.75    16.07 modify

non-

significant

variables

maximize    0.07    0.54 0.27 0.18 0.11    10    50    11    0.33 0.11    22 0.75    1 16.62

cost

minimize    0.07    0.66 0.33 0.22 0.09    10    50    9    0.27 0.09    18 0.75    15.52    7%

cost modify

significant

variables

maximize 0.07    0.6 0.3 0.2 0.1    11 55    10 0.3 0.1 20 0.68    18.10

cost

minimize 0.06 cost

0.6 0.3 0.2 0.1    9    45    10    0.3 0.1 20 0.83

14.20    24%