Optimization Module

New Functionality in Version 5.2a

Constraint Method Setting in Pointwise Inequality Constraints

In version 5.2, the method used for assembling pointwise inequality constraints was silently changed from elemental to nodal. This, in practice, removed a lot of redundant constraints. Including each constraint once instead of several times in an optimization problem does not change its solutions, but the SNOPT solver’s termination criterion is sensitive to the number of constraints. Therefore, in order to allow direct comparisons between solutions in different versions, a Constraint method setting has been added to the Pointwise Inequality Constraint features. The default method is still Nodal. Choose Elemental to recreate the behavior from versions older than 5.2.

Defaults changed in Global Least-Squares Objective

When the experimental Data source in a Global Least-Squares Objective feature is set to Result table, no table will be selected by default. In previous versions, the first table in the list under Results>Tables was arbitrarily used as default.

Similarly, the default for the Time column and Parameter column settings are now None instead of the first column in the selected table.

These changes lead to more predictable behavior when deleting tables and table columns. Some Java® and MATLAB® code recorded in previous versions may be affected, producing an error message when executed. The solution in these cases is to insert lines explicitly setting the desired table and column. For example:

model.physics("opt").feature("glsobj1").set("table", "tbl1");

model.physics("opt").feature("glsobj1").set("parameterColumn", "1");

Time-dependent Least-Squares Optimization

In time-dependent least-squares optimization problems, time lists are modified while solving if they contain time values that are outside the time range defined by the time list of the corresponding Time Dependent study. The modified time lists do not contain time values that are outside of the time range defined by the corresponding Time Dependent study. Such modified time lists are used in the optimization iterations when least-squares objectives are evaluated. Before, the modified least-squares time lists were used only in the last iteration. Hence, running such a model can now return a solution that differs in this way. It is, however, possible to use the unmodified least-squares-defined time list by setting the new property of Time-Dependent Solver, Exclude times outside General parameter value lists, to off.