Setting Up the SQP Optimization Study in Design Manager

To perform a gradient-based optimization, you set up a SQP optimization study in Design Manager.

You use the parameters that define the dimensions in the CAD model as input parameters for the optimization study. The optimization objective is to minimize the pressure drop.
  1. To create a Design Manager Project:
    1. From the reference simulation gradient_ubend_run.sim, right-click the gradient_ubend_run node and select Create Design Manager Project.
    2. Save the Design Manager project as gradient_ubend_project.dmprj.
  2. To specify the design study type:
    1. Right-click the Design Studies node and select New.
    2. Select the Design Study node and set the Study Type to Optimization.
    3. Select the Optimization Settings node and set the Optimization Type to Sequential Quadratic Programming (SQP).
      SQP is an iterative algorithm, which starts the search of the optimal solution using the last run. Unlike the SHERPA-based search algorithm, SQP does not reuse all the shared designs from previous study runs. SQP can handle a large number of parameters.
    4. Select the Optimization Settings > SQP Method note and keep the default property settings as follows:
      PropertySetting
      Maximum Outer Iterations10

      This value specifies the number of the SQP outer iterations. In each outer iteration, Design Manager recalculates the gradient sensitivities of the parameters.

      10 indicates that after calculating the sensitivity 10 times, the optimization stops.

      Maximum Line Search Iterations5

      This value specifies the largest possible number of design runs without re-calculating the sensitivities.

      5 indicates that Design Manager runs a maximum 5 designs before re-calculating the sensitivities.

  3. To minimize the pressure drop between inlet and outlet, you use all the provided geometric parameters as input parameters for the SPQ optimization study:
    1. To set dimension parameters as input parameters, select the Design Study > Input Parameters node and add all the dimension parameters to the Parameters as follows:


    2. To collectively edit the parameter values , right-click the Input Parameters node and select Edit.
    3. In the Design Study - Input Parameters table, set the Parameter Type and Range for all the parameters as follows:


      The ranges of the parameters are constrained by manufacture feasibility. Simcenter STAR-CCM+ starts with the initial values and search for the local minimum within this range.

    4. To check whether a parameter fulfills the prerequisites for gradient-based optimization, select the relevant Input Parameters > [parameter] node and check whether the Sensitivity property is activated.


      A parameter with a confirmed sensitivity check indicates that the reference simulation can provide both geometric sensitivity and adjoint surface sensitivity for this parameter. Only such parameters can be selected as input parameters for a gradient-based optimization.

    5. To set the pressure drop as a response, select the Design Study > Responses node and add dP to the Responses.
    6. To specify this response as a design study objective, select the Responses > dP node and activate the property Is Objective.


      A response with a confirmed adjoint check indicates that this response has a corresponding Adjoint Cost Function in the reference simulation. Only such responses can be selected as objectives or constraints for a gradient-based optimization.

  4. To output scenes for each completed design run, select the Design Study > Scenes and add Geometry Comparison defined in the reference simulation.
  5. To observe the convergence of each design run, select the Design Study > Plots and add Residuals plot defined in the reference simulation.