【摘要】 提出了采用多水平正交表选取 BP神经网络训练样本的方法，在保证足够建模精度的条件下，采用此法大大减少了神经网络的建模工作量，因而有很大的实用价值。基于此法建立了磨床主轴系统的 BP神经网络模型并进行了结构修正与优化计算。利用 BP神经网络模型进行大型复杂结构的分析优化具有简单、高效的优点。更多还原

【Abstract】 The method selecting BP neural networks training samples with many levels table of orthogonal arrays is presented. That the method has great value in practical is because it can decrease enormously neural network modeling workload on condition that the modeling precision is contented. BP neural networks model of the main spindle system of the grinder is built and the structure model updating and optimization calculation are carried through based on the method. The analysis and optimization of the big complicated structures is very simple and efficient based on BP neural networks.更多还原