203 遺伝的アルゴリズムとニューラルネットワークを用いたエンジンマウントの最適設計

抄録

In the case of the engine mounting design, the decoupled layout is very important for the isolations of engine vibration and shock torque. On the other hand, the optimum layout is difficult to select many parameters such as mounting position, cushion rubber spring rate, etc. So, the Genetic Algorithm so called GA is newly applied for the stiffness matrix calculation and the many parameters optimization. The MPOD, stands for Most Probable Optimal Design based on the Holographic Neural Network, is also tried to compare the accuracy against GA. The optimized results by GA were well matched with the theoretical calculation, which needs the special vibration analysis technique and also takes over two weeks. But, it takes only one hour, and the inexperienced engineers can easily obtain the optimized result. By the confirmation test of FEM, the engine lateral vibration level at 25Hz dropped below 1/5 and its effects were significant.