Correlation and Simulation Process Improvement for Automotive Axle 2007-01-1205
Previous studies showed that a good correlation of axle vertical bending test can be obtained by choosing a proper correlation strategy that adapts both displacement and strain as correlation targets . However, due to experimental error, as well as some environment noises, it was found that the correlation problem was highly nonlinear. This creates a large difficulty in searching for the global optimal solution with a gradient-based optimization technique that largely depends on the selection of initial search points. The confidence for finding a global optimal solution is highly limited without a search of the complete design space. Additionally, certain aspects of axle system modeling, such as bearing stiffness effects on the correlation result, are yet to be found. This study is a continuous investigation of CAE correlation of the axle vertical bending test by focusing on the improvement of the correlation process and confidence. A new optimization algorithm using design exploration techniques was adopted for the optimal solution search. The study found that the previous 2-step search process could be replaced by a one-step global search process. Additionally, different exploring search techniques were compared and application guidelines were provided. The new correlation process is now implemented in the product development process.