Abstract
This paper deals with an optimum design method for a machine tools bed using neural networks. A machine tools bed often has a rib structure and its rib layout correlates with its natural frequency or rigidity. Some natural frequencies and rigidities are obtained by FEM and learned by the neural networks to recall that of unlearned rib layouts. The natural frequency or rigidity of all prospective lib layouts are estimated in short time and the optimum rib layout that shows high objective function can be chosen from them. Two optimum designs for low natural frequency and high rigidity respectively were carried out using the whole machine tools model assuming a machining center and the simple bed model. 4500 lib layouts were used to search the optimum one and the time to search became less than 1/50 comparing with the case of full search method using FEM. The estimation error is approximately 6% and the proposed method is effective for the bed design. Moreover, it is confirmed that the optimum design using the simple bed is effective to improve the performance of the whole machine tools model.
