Best practices for design optimization

For part, sheet metal, and assembly models, you should verify that the following properties are available for optimization:

• Verify that the Variable Table contains the design dimensions that you want to change with design optimization. You can open, review, and edit the Variable Table contents using the Tools tab→Variables command.

• You must set upper and lower limits on the domain of possible values for any design variable that you want to allow to change during optimization. Use the Variable Rule Editor dialog box, which is available within the Variable Table. You can define a range of values for a variable. This allows you to restrict design changes to a controlled set of values.

  For more information, see Define limits for a variable.

• For the most accurate results, set the precision in the File Properties dialog box to the highest possible value (the most decimal places) on the Units tab and in the Advanced Units dialog box.

• Verify that the correct material is applied to the model in the Application menu→Info→Material Table.

• For part models, verify that the Physical Properties dialog box contains the updated properties and units that you want to reference in design optimization. This dialog box is available using the Inspect tab→Physical Properties group→Physical Properties command.

• For assembly models, you can use the Physical Properties Manager dialog box to access the properties for individual parts. This dialog box is available using the Inspect tab→Physical Properties group→Physical Properties Manager command.

• The geometry that you want to optimize must be included in the solved study.

• Optimization is supported in synchronous and ordered modes. In synchronous mode, you need to identify the Design Intent relationships you want to preserve or relax so that the part geometry can be modified to satisfy the optimization requirements. In an assembly, you need to identify the Design Intent relationships for all of the parts that can be modified (due to the inclusion of peer variables) when the optimization is solved.

• Often, changing one design variable automatically results in changes in other quantities. If design variable A affects both B and C, and you are primarily concerned with A, then only A should be specified as an optimization variable.

• To find the best design it is often necessary to change several design features simultaneously. If several variables need to be changed, then you should specify those variables as input to a single optimization run. For a part with two design variables A and B, the best value for variable A often depends on the value of variable B. You should not try to find the optimum design by first optimizing variable A and then optimizing variable B, because the value of variable A is liable to be non-optimal when B has changed.

• However, if the optimum value for a variable is obvious, then the variable should be set to its optimum value before optimization, and it should not be entered as an optimization variable

• You can specify an arbitrary number of variables, but in practice it is unwise to use an excessive number of optimization variables. The size of the search space increases rapidly with the number of variables, and this similarly increases the time required for optimization.

  Secondly, when using a large number of variables (more than ten), it is difficult to define the variables in such a way that no invalid designs are generated.

• In situations where a large number of design variables need to be optimized, it may be beneficial to break the problem down into two optimization runs. Sometimes the best values for one set of design variables are independent of the best values for another set of variables. In this situation, the problem is more efficiently solved by running the optimizer twice. In the first run, only the first set of design variables should be chosen as optimization variables, and in the second run, only the second set should be selected.