MOGA & MOGT Optimisation Strategies and SOM Results Representation

Multi-disciplinary & multi-objective design optimization tools are used more and more in order to help CAE designers and managers in their quest for product higher quality and returns. As far as an optimisation problem is concerned, several strategies can be used to select candidates for evaluation, i.e. designs on the Pareto frontier, as for instance, Preliminary Exploration Methods, DOE, Local Refinement Methods, Special-Purpose Plug-ins or Multi-Objective Optimization Methods that include genetic algorithms and evolution strategies. The goal of the Multi-Objective Optimization Methods consists in locating the designs on the Pareto frontier automatically, letting the user to pick the desired trade-off among them by means of Decision Making tools. The present paper illustrates the comparison between the results achieved by means of the MOGA (Multi Objective Genetic Algorithm) and the MOGT (Multi Objective Game Theory) optimisation strategies. The aim of the application consists in a construction machinery cab vibro-acoustic performance optimization. A 3D cavity representing the real cab has been modelled by means of a (Ansys) FE structural mesh. Starting from the cab vibration load experimental acquisition, a (Sysnoise) BEM coupled analysis has been carried out in order to evaluate the cab inner vibro-acoustic field as a function of the physical properties of each structural element. The multi-objective design optimisation code (modeFRONTIER) drives the analysis process flow taking into account the cab parameter structural modifications and carrying out the vibro-acoustic field optimisation. The less tested and more innovating MOGT strategy shows itself to be a robust and fast multi objective optimisation tool too when combined with Evolutionary Algorithms. The recently developed results representation by means of SOM (Self-Organizing Maps) shows itself to be a powerful analysis tool. It allows a clear and fast qualitative comprehension of the relations between optimisation process design variables and objectives.