Performing an NLPQL Optimization

The NLPQL (Nonlinear Programming by Quadratic Lagrangian) method can be used for both Response Surface Optimization and Direct Optimization systems. It allows you to generate a new sample set to provide a more refined approach than the Screening method. Available for continuous input parameters only, NLPQL can handle only one output parameter goal. Other output parameters can be defined as constraints. For more information, see Convergence Rate % and Initial Finite Difference Delta % in NLPQL and MISQP and Nonlinear Programming by Quadratic Lagrangian (NLPQL) in DesignXplorer Theory.

To generate samples and perform an NLPQL optimization:

  1. In the Outline view for the Optimization cell, select the Optimization node.

  2. In the Properties view, enter the following input properties:

    • Method Name: Set to NLPQL.

    • Verify Candidate Points: Select to verify candidate points automatically at end of the optimization update.

    • Finite Difference Approximation: When analytical gradients are not available, NLPQL approximates them numerically. This property allows you to specify the method of approximating the gradient of the objective function. Choices are:

      • Central: Increases the accuracy of the gradient calculations by sampling from both sides of the sample point but increases the number of design point evaluations by 50%. This method makes use of the initial point, as well as the forward point and rear point. This is the default method for preexisting databases and new Response Surface Optimization systems.

      • Forward: Uses fewer design point evaluations but decreases the accuracy of the gradient calculations. This method makes use of only two design points, the initial point and forward point, to calculate the slope forward. This is the default method for new Direct Optimization systems.

    • Initial Finite Difference Delta (%): Advanced option for specifying the relative variation used to perturb the current point to compute gradients. Used in conjunction with Allowable Convergence (%) to ensure that the delta in NLPQL's calculation of finite differences is large enough to be seen above the noise in the simulation problem. This wider sampling produces results that are more clearly differentiated so that the difference is less affected by solution noise and the gradient direction is clearer. The value should be larger than both the value for Initial Finite Difference Delta (%) and the noise magnitude of the model. However, smaller values produce more accurate results, so set Initial Finite Difference Delta (%) only as high as necessary to be seen above simulation noise.

      For a Direct Optimization system, the default percentage value is 1. For a Response Surface Optimization system, the default percentage value is 0.001 (minimum = 0.0001, maximum = 10).

      For parameters with Allowed Values set to Manufacturable Values or Snap to Grid, the value for Initial Finite Difference Delta (%) is ignored. In such cases, the closest allowed value is used to determine the finite difference delta.

    • Allowable Convergence (%): Stop criterion. Tolerance to which the Karush-Kuhn-Tucker (KKT) optimality criterion is generated during the NLPQL process. A smaller value indicates more convergence iterations and a more accurate (but slower) solution. A larger value indicates fewer convergence iterations and a less accurate (but faster) solution.

      For a Direct Optimization system, the default percentage value is 0.1. For a Response Surface Optimization system, the default percentage value is 0.0001. The maximum percentage value is 100. These values are consistent across all problem types because the inputs, outputs, and gradients are scaled during the NLPQL solution.

    • Maximum Number of Iterations: Stop criterion. Maximum number of iterations that the algorithm is to execute. If convergence happens before this number is reached, the iterations stop. This also provides an idea of the maximum possible number of function evaluations that are needed for the full cycle. For NLPQL, the number of evaluations can be approximated according to the Finite Difference Approximation gradient calculation method, as follows:

      • For Central: number of iterations * (2*number of inputs +1)

      • For Forward: number of iterations * (number of inputs+1)

    • Maximum Number of Candidates: Maximum number of candidates that the algorithm is to generate. The default is 3. For more information, see Viewing and Editing Candidate Points in the Table View.

  3. Specify optimization objectives and constraints.


    Note:  For NLPQL, exactly one output parameter must have an objective defined. Multiple parameter constraints are permitted.


  4. Specify the optimization domain.

    • In the Outline view, select the Domain node or an input parameter or parameter relationship under it.

    • In the Table or Properties view, define the selected domain object. For more information, see Defining the Optimization Domain.

  5. Click the Update button on the toolbar or right-click the Optimization node in the Outline view and select Update.

    The result is a group of points or sample set. The points that are most in alignment with the objective are displayed in the table as the candidate points for the optimization. In the Properties view, the result Size of Generated Sample Set is read-only. This value is always equal to 1 for NLPQL.

  6. In the Outline view, select the Optimization node and view convergence details under Optimization Status in the Properties view. The following output properties are displayed:

    • Converged: Indicates whether the optimization has converged.

    • Number of Iterations: Number of iterations executed. Each iteration corresponds to one formulation and solution of the quadratic programming subproblem, or alternatively, one evaluation of gradients.

    • Number of Evaluations: Number of design point evaluations performed. This value takes all points used in the optimization, including design points pulled from the cache. It can be used to measure the efficiency of the optimization method to find the optimum design point.

    • Number of Failures: Number of failed design points for the optimization. When a design point fails, NLPQL changes the direction of its search. It does not attempt to solve an additional design point in its place and does not include it on the Direct Optimization Samples chart.

    • Size of Generated Sample Set: Number of samples generated in the sample set. This is the number of iteration points obtained by the optimization and should be equal to the number of iterations.

    • Number of Candidates: Number of candidates obtained. This value is limited by the Maximum Number of Candidates input property.

  7. In the Outline view, select the Domain node or any node under it to review domain information in the Properties and Table views.

  8. For a Direct Optimization system, in the Outline view, select the Raw Optimization Data node. The Table view displays the design points that were calculated during the optimization. If the raw optimization data point exists in the design points table for the Parameter Set bar, the corresponding design point name is indicated in parentheses in the Name column.


    Note:  This list is compiled of raw data and does not show feasibility, ratings, and so on for the included design points.



Release 18.2 - © ANSYS, Inc. All rights reserved.