Optimizations with Different Variable Types
From gtwiki
| «««(?) |
Contents |
Optimizer Target Settings
The Product Engineering Optimizer workbench tool "Optimization" can be used to breed several solutions to a complex configuration of variables. The fitness is tested by honing the output value into one variable. For example, in the circle example, the summation of all of the point differentials is the target number. If this summation is zero, the circle has found the most optimal orientation.
Multiple Variable Types
The problem with the PEO is that only one value can be the target value. When users try to target two or more values, some clever mathematics must be used, like in the circle example. This becomes a greater problem when two differing variable types require optimization. The user cannot add X units of inches to Y units of degrees and expect a resolved solution. And in most cases, these values are set to a "Target" value minimization, where arithmetic solutions are impossible to conceive.
Absolute Value Workaround
The "absolute value" workaround is an adequate way to return several desired output types in one formula.
Users just need to encapsulate the difference in desired and resultant values in an absolute value command. The formula for accomplishing this is as follows:
abs((Resultant_Value - Desired_Value) / 1unit)
where 1unit is the current singular unit type for that variable at this session's settings.
For example, if at the end of the optimization the user desires for variables to end up with the following values:
Length variable: L1 --> 23in
Length variable: L2 --> 12.875in
Angle variable: RES_ANG --> 32.843deg
Length variable: TopDim --> 4.5in
The Formula for determination should look like this:
abs((L1 - 23.0in)/1in) + abs((L2 - 12.875in)/1in) + abs((RES_ANG - 32.843deg) / 1deg) + abs((TopDim - 4.5in) / 1in)
The list can have N number of absolute value calculations, so the workaround is quite robust. For questions on its use, feel free to email God.
