SOM - Self organizing maps

Abstract

A industry today is challenged by a large number of complexes and often conflicting constraints and requirements such as reduce the cost and weight of vehicles, compress vehicle development cycle time, and improve product performances, e.g., Safety, NVH, Durability, etc. More recently, multidisciplinary design optimization (MDO) is a systematic tool to integrate all the attributes in vehicle design and find a compromise solution to satisfy those stringent performances and requirements. In addition, as most computer aided engineering (CAE) simulations are computation intensive, special optimization methods and processes are often required. This presentation will focus on historical developments and applications of optimization and robustness methods for vehicle designs. It will address significant technologies, such as advanced model bias updating method, data mining based design space identification involving large-scale engineering problems, and score function based reliability design method considering data uncertainty. Lastly, a vehicle example of minimizing the weight and satisfying safety and NVH requirements is presented to demonstrate the proposed methodology.