Abstract
In this paper, the method, which estimates the design parameters (thickness, shape of cross section, length of partition, position of flange etc.) using neural network, is described. We deal with simple joined part models composed of 2 beams as an antomobile body structure. At first, some learning data are prepared by changing the design parameters. Next, the relation between joint stiffness and design parameter is constructed by using neural network. Once the trained neural network is given, the design parameter can be estimated by only inputting the some values of joint stiffness into the trained neural network. It was shown that the design parameters were obtained from the values of joint stiffness by using the trained neural network.
