Use of multi-variate data analysis, fast algorithms and grid automation in modeFRONTIER for multi-objective optimization

A. Clarich, R. Russo (ESTECO)

Performing a system optimization that involves a large number of design variables and computationally expensive analysis is possible but not always practically feasible. As a solution, the optimization process can be made more efficient if the designers can analyze the problem characteristics a priori and use the information to reduce the problem complexity.

This paper will illustrate the use of Multivariate Data Analysis and CAP (Computer Aided Principle) techniques available in modeFRONTIER code in different phases of the optimization process, and here applied to the CFD design of a yacht sail flying shape. The data analysis techniques can help the designer with several aspects, including the reduction of the number of computations required during optimization by analyzing the local influence of variables on objectives, and supporting the final design selection from Pareto set.

At the same time, to reduce the overall time needed for the simulations, the Grid Tool of modeFRONTIER is used, allowing to automatically distribute the CFD simulations on the available machines in the network, optimizing their load balance, and fully exploiting the available computational resources.

Finally, efficient and fast algorithms like FMOGAII (based on Response Surfaces) or MOGT (Game Theory) can be used to obtain the optimal results with a reduced number of simulations.

The case study chosen to illustrate the use of multivariate data analysis techniques and efficient multi-objective algorithms is the design of a yacht sail flying shape. The complete optimization process flow is built in the commercial process integration and optimization software modeFRONTIER. The design variables involved are primarily the geometric parameters to change the shape of the sail. For design changes and evaluation, modeFRONTIER is coupled to the commercial mesh morphing tool ANSA to modify automatically geometry and mesh, and to the commercial computational fluid dynamics tool ANSYS CFX for aerodynamic analysis, defining in a modular workflow the complete process chain.