【摘要】 以正交表和有限元软件 ANSYS的 APDL为手段建立样本模型 ,用 BP神经网络构建滑轮结构设计变量与第一阶固有频率及质量的非线性映射关系 ,实现了以提高第一阶固有频率为主要目标的多目标优化 .可以在较少的有限元分析次数下获得较好的工程上实用的整体可行优化解 ,降低了人为因素的干扰 ,得到了较满意的结果更多还原

【Abstract】 A non-linear mapping function from multiple input design variable of pulley to output the basic structure frequency and mass is constructed within BP neural networks whose samples are got by mean of APDL of ANSYS and orthogonal table .In order to set the basic frequency to the largest, the main-object optimal solution is carried out, which relieves the subjective interference and obtains the satisfactory result and indicates that a better optimum solution can be obtained using less finite element analyses.更多还原