Summary

This course provided a comprehensive exploration of parametric variation analysis and optimization using Ansys optiSLang® process integration and design optimization software through an example problem. The optimization was performed on a CFD simulation solved using Ansys Fluent® fluid simulation software. Below are some of the key takeaways:

• Ansys optiSLang software is a Process Integration and Design Optimization (PIDO) tool that automates simulation workflows, enabling engineers to maximize innovation, shorten development cycles, and achieve robust designs.
• The software offers user-friendly wizards that allow us to create new projects or workflows, parametric solver systems, and set up sensitivity, optimization, or robustness analyses.
• Process integration connects Ansys, third-party and in-house tools to automate simulation workflows.
  • In this case, Ansys Discovery™ 3D production simulation software, Ansys Fluent Meshing, and Ansys Fluent Solver were connected into a seamless workflow.
  • This automated workflow enabled the evaluation of multiple design variations without manual intervention.
• Sensitivity Analysis evaluated the importance of input parameters in relation to output parameters and generated a Metamodel of Optimal Prognosis (MOP).
  • Based on the designs evaluated in the Design of Experiments (DoE) generated by the software, different metamodels are created utilizing AI/ML algorithms.
  • MOP is the metamodel with the highest Coefficient of Prognosis (CoP), which is the measure of a metamodel’s predictive quality.
  • Sensitivity analysis was performed using the Adaptive Metamodel of Optimal Prognosis (AMOP) approach which adds new designs automatically in interesting regions of the design space to improve the metamodel quality.
  • CoP of the MOP in the example problem was well above 90%.
• In Ansys optiSLang software, there are two main optimization approaches: MOP optimization and direct optimization.
  • MOP Optimization: This approach uses metamodels to substitute for the solver and allows for quick evaluations as the design evaluation using a metamodel takes only seconds.
  • Direct Optimization: Evaluates designs using the solver runs directly, which eliminates approximation errors but results in longer runtimes.
  • If required, these two methods can also be integrated into a single optimization process, commencing with MOP optimization and following up with direct optimization
• Optimization was performed to obtain the optimal values of the vane angles for the duct to minimize pressure drop in the duct.
  • One-Click Optimization (OCO) method was used, which is a hybrid approach that dynamically selects the best algorithms for single- or multi-objective optimization.
  • Since optimization using MOP did not yield satisfactory results, direct optimization was subsequently performed to obtained improved results.
• The software offers interactive postprocessing tools for sensitivity analysis and optimization, which provide valuable insights into design performance, parameter importance, and convergence behavior.