Abstract
We propose method of computer aided design using evolutionary multiple regression analysis in order to reduce weight of car body structure. It is difficult to save weight of car body structure due to trade-off problems between weight and vehicle performances. It is important to clarify a contribution of design variables to weight and multiple vehicle performances in order to solve these problems. As a method of calculation of the contribution, sensitivity analysis and multiple regression analysis are available. However, in these methods, when the number of design variables increases, it is pointed out that the accuracy deterioration due to over fitting. Therefore, we propose the contribution analysis method using a function based on p-value for model evaluation criterion and Genetic Algorithm (GA) for efficient exploration of optimal solution. By minimizing the function that is divided average of p-value by number of design variable, it is possible to select only the contributed variables. And, contribution obtained by this method is converted to mass ratio and standardized to intuitively conceive a structural specification idea from contribution obtained by this method. As a result of applying this method to next car body structure, new design knowledge for weight reduction is obtained.
