基于神经网络与遗传算法的移动硬盘耐撞性能多目标优化设计

Multi-objective Optimization for Mobile Hard Disk Crashworthiness Based on ANN and GA

针对移动硬盘复杂的内部结构,遵循质量等效原则,建立有限元模型,结合均匀设计试验方法,进行移动硬盘跌落冲击有限元仿真。将试验数据作为训练样本,建立移动硬盘传动臂厚度、悬臂厚度及枢轴轴承刚度与磁盘接触面应力、磁头与磁盘间距之间的BP神经网络,运用遗传算法对移动硬盘结构参数进行多目标优化。研究表明:在考虑耐撞性能方面,运用该方法建立的设计变量与跌落冲击响应的非线性映射关系,对移动硬盘的关键零件参数设计具有理论指导意义。

In view of inner structure and complying with quality equivalent principle, a 2.5 inch mobile hard disk was modeled based on ANSYS/LS-DYNA. The non-linear mapping relation between structural parameters of the key components and drop impact response was built using FEM （finite element method） and experimental design. Drop impact processes of different head actuator arm thickness, cantilever thickness and pivot bearing stiffness were simulated in order to obtain training specimen for the ANN （artificial neutral network）. The BP ANN prediction model was established. Multiple objective optimization algorithms for crashworthiness of mobile hard disk were analyzed based on ANN and GA （genetic algorithm）. The results indicate that the method is effective and may provide theory gist of preferences in key part parameter design.