Optimization Editor

The Optimization dialog box appears when you select Optimization.

This dialog box contains controls for:

This page discusses:

Problem Tab

Optimization Type

The list provides you with the following items that define the optimizations types.

Optimized Parameter
It is the Knowledge parameter that is "optimized". This parameter cannot be an integer. Click Select... to display the list of parameters that can be potentially optimized. Once selected, an update is triggered to make sure that the value of the optimized parameter is up-to-date.
Note: Optimizations cannot handle Integer parameters even though the Simulated Annealing algorithm could do it. Any other algorithm relies on the assumption that the objective function, i.e. the function linking the Free Parameters and the Optimized Parameter, is differentiable so that it cannot be a function of integer variables). However, it is possible to force the optimization to handle integer functions of integer variables. For instance if you want to find the minimum of Real.1 = (Integer.1-3)*(Integer.1-3) + 4, you have to define a real number Real.2 and valuate Integer.1 with ceil(Real.2) or floor(Real.2). Therefore, the optimization sees Real.2 as a Free Parameter and Real.1 as an Optimized Parameter.
Important: Free parameters can not be integers.
Free Parameters
They concern the Knowledge parameters changed by the optimization to find the right value i.e. min, max, or target of the optimized parameter.
  • To specify free parameters, click Edit list.
  • To specify bounds for a free parameter, select the parameter you want to add bounds to in the Free Parameters list, then click Edit ranges and step to specify a step.

It is possible to simultaneously valuate the ranges and steps of several free parameters.

  1. Select one free parameter and apply it ranges and/or steps (unchanged).

  2. Select several free parameters and click Edit Ranges and step: The tab is filled with the values of the first selected free parameter. All Apply... options are enabled.

Notes:
  • To apply the same ranges and/or steps to free parameters, their types must be identical. If not, an error tab appears.
  • If one of the Apply this... to the selected parameters check boxes is cleared, the modifications performed (value changed or removed) on the corresponding range and/or the step not be taken into account for the selected parameters. The old configuration is kept for each selected parameter. Therefore, if you modify the value (or removes it by unchecking the first check of the row) of a range or a step, and then clear the corresponding 'Apply this ... selected parameters' check box, the modification are not taken into account. The selected parameters keep their own values.
  • If the Inf. Range check box is cleared, the selected parameters have no inferior range. It is also true for the Sup. Range and the Step.
  • When selecting the free parameters, the associated parameters are highlighted in the tree.
  • It is possible to edit the steps directly in the cells.
Available Algorithms


See Specifying the Algorithm to be Run.
Termination Criteria
If you don't know how the objective function behaves, run the algorithm with the default values. If need be, you can always re-run the algorithm if the process seems to require more iterations.
Optimization Data
To benefit from the optimization curves, save your optimization data. When the Save optimization data box is checked, clicking Run Optimization displays a file selection dialog box.
Run Behavior
  • With visualization update/Without visualization update: You can choose if the optimization is launched with or without visualization update by clicking the appropriate option.
    Notes:
    • In non visualization mode (Without update visualization), performance have been improved. Gains between both modes (With update visualization and Without update visualization) range from 15% to 40% depending on the models and the optimizations.
    • When the With update visualization option is selected, the visualization is updated during the optimization process. If this button is cleared, the visualization is not updated during the optimization process.
  • Disconnect Undo Log: When this option is checked, the optimization is launched without being recorded in the undo log. This new option enables you to optimize medium and large size FEM models. This highlight allows a reduction of computation time and a drastic decrease in memory consumption during the optimization process. Memory consumption must increase linearly and by small amounts during the scenario (due to the optimization historic fill). Computation time of each update during the optimization process remains the same.
  • Update Mode: You can choose which type of update you want to perform. You can perform a lean update in Representation and Product context, which means that you can partially update the product and the representation according to the input/output specifications of the Design Optimization features.
    • Local Update Mode: Choose this update mode when free parameters have a local influence inside the representation or the Product. It has a fast and local influence on the design. Update errors may occur when totally rebuilding the representation.
    • Mixed Variational Update Use: Choose this update mode when the optimization result is influenced by the solving of geometry (sketches with constraints), multi body measures... It is slower than the Local Update mode but allows the recomputation of the variational aspects of the representations as well as the sketches constraints web. This update mode is slower than the Local Update mode and does not update the complete representation. It is available only in a Representation context.
    • Global Update Use: When the optimization result depend on the full representation update. It shows all update errors that can happen during the optimization. It also guarantees that the optimal solution is compatible with the total updated representation, and allows you to trigger additional knowledge capabilities (such as writing in a design table, triggering reactions, triggering VB) during the optimization run.

Constraints Tab

The Constraints tab contains the following options:

Constraints Body Field
  • To create a constraint, click New... (The editor displayed is familiar to those of you who use the Engineering Rules Capture product.)
  • To create a constraint with derivatives, click New Cst With Derivatives.
  • To delete a constraint, click Delete.
Constraints cannot be regrouped. You must enter each constraint one by one. The only operators that you can use when specifying constraints are:
  • == (Simulated Annealing and Algorithm for Constraints and Derivatives Providers)
  • <, > (All algorithms).
When a constraint has been specified, its gap with respect to the value specified in the constraint appears. An icon indicates whether the initial value fulfills the constraint.
Important: An additional column now lists the weight assigned to each constraint. You can modify the weight in this column.
A constraint can be made active (True) or inactive (False) by using the Activity box. A deactivated constraint is ignored by the optimization algorithm.
Constraint Status
Edit: Click Edit... to access the Optimization Constraints Editor to modify the constraint body. Note that this Editor is similar to the Formula Editor.
Important: It is now possible to modify constraints weights by using the Weight box. To do so, select the constraint from the list and enter a new weight in the Weight box.

Note: All constraints support precisions.
  • An equality constraint is always interpreted by the algorithm as a double inequality (using the precision). 
    x+y=10 >


    x+y ≤ 10 + precision
    x+y ≥ 10 - precision
  • Other constraints (inequalities). Although inequalities are treated as is by the algorithms, their status (satisfied or not satisfied) depends on the precision defined. This allows some constraints to be slightly not satisfied from a numerical point of view but satisfied from an engineering point of view.

Computations Results Tab

Constraints Priorities Field
This dialog box lists the constraints and the priorities they are assigned. Note that those priorities are different from the weights used by the algorithms to search for the solutions. You can change these priorities to explore their influence on the already computed results when you look at them with the lexicographic sort. Changing those priorities has no influence on the generated results i.e. they present the existing ones under a new aspect by changing the order of the list.
Important:
  • It is possible to edit constraints priorities in this box. To do so, click the desired cell, and enter the new weight.
  • If you modify a constraint priority, the Sorted results list is automatically updated.
  • Modifying a constraint priority in this tab only impacts the Lexicographic sort and not the algorithm. This may be useful to check the impact of a different weight without running the algorithm.
  • If a priority is equal to 0, the default algorithm is used. 1 is the lowest priority that can be applied. There is no upper limit.
  • Priorities are integers and used in the post-processing of the results. Weights are real numbers (1.2 for example) used during computation.
Settings of the results sort

  • Historic Sort: The results are displayed in the computation order.
  • Lexicographic sort: The results are displayed going from the best to the worst. The priorities assigned to the constraints are taken into account. For more information, see Running a Constrained Optimization With Weights.
  • All: All results are displayed in the sorted list.
  • All constraints satisfied only: Only the results concerning the satisfied constraints are displayed in the sorted list.
  • User-defined: Only the items you selected are displayed in the sorted list.

Sorted Results
This list displays the result of the optimization according to the filters applied and the sorting type (Historic sort or Lexicographic sort). Apply values to parameters enables you to select a row from the list and to apply the values indicated in this row to the parameters.
Curves Field
Select parameters enables you to select the parameters of the optimization file whose evolution are displayed in the curves. For more information, see Interpreting Results.

Show curves... enables you to display the curves showing the parameters during the optimization process.