Evolutionary neural networks for product design tasks

  • Authors:
  • Angela Bernardini;Javier Asensio;José Luis Olazagoitia;Jorge Biera

  • Affiliations:
  • The Automotive Technological Innovation Centre of Navarre, CITEAN, Pamplona, Spain;The Automotive Technological Innovation Centre of Navarre, CITEAN, Pamplona, Spain;The Automotive Technological Innovation Centre of Navarre, CITEAN, Pamplona, Spain;The Automotive Technological Innovation Centre of Navarre, CITEAN, Pamplona, Spain

  • Venue:
  • HAIS'12 Proceedings of the 7th international conference on Hybrid Artificial Intelligent Systems - Volume Part I
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Standard development process of a product involves several CAE and CAD analyses in order to determine parameter values satisfying technical product specifications. In case of nonlinear behavior of the system, the computational time may quickly increase. In the current study, a new methodology that integrates Neural Networks (NN) and Genetic Algorithms (AG) is introduced to analyse virtual models. The proposed tool is based on different computational , mathematical and experimental methods that are combined together to distil a single tool that permits to evaluate in a few seconds how behaviors of certain product vary when any design parameter is altered. As example, the methodology is applied to adjust design parameters of an exhaust system, showing same accuracy range than FEA, but strongly reducing the simulation time.