Design Manager: Gradient-Based Optimization of a U-Bend
Gradient-Based Optimization is an iterative method that optimizes a set of geometric features towards an engineering objective.
In Simcenter STAR-CCM+, you perform this type of optimization using the Sequential Quadratic Programming (SQP) study in Design Manager, which optimizes selected parameters based on two parametric gradients-the Geometric Sensitivity and the Surface Sensitivity.
The geometric sensitivity, computed by the Geometric Sensitivity mesher, defines the dependence of selected parameters on the geometric features.
The surface sensitivity, computed by the Adjoint Surface Sensitivity model, identifies the areas of the geometry which are more critical (with large surface sensitivity values) or less critical (with small surface sensitivity values) with respect to the adjoint cost function.
At each SQP outer iteration, for the newly optimized geometry, the Geometric Sensitivity mesher executes a remesh and generates the new geometric sensitivities. The Adjoint model recomputes the surface sensitivities. Design Manager computes the new parametric gradients from the new values for the next optimization iteration and modifies the parameters.
Within the inner line search iterations, no new values are computed. The optimum search continues with the old values from the last outer iteration.
In this tutorial, you set up a gradient-based optimization study of a u-bend. The goal of the study is to optimize a set of geometric parameters in order to minimize the pressure drop between the inlet and the outlet of the u-bend.
The image below shows the optimized geometry after 10 outer iterations:
