论文部分内容阅读
监测刀具磨损是自动化加工的一个很重要的研究课题，人工神经网络的出现，为刀具磨损的在线监测开创了广阔的前景，但在应用时出现了易被忽略的问题，例如常用的ＢＰ网络存在着收敛速度慢、易陷入局部极小点等缺陷。该文利用优化理论对ＢＰ网络缺陷进行分析，提出相应的改进措施，即神经网络分解算法，并用于刀具磨损量的估计，实验证明采用的方法是正确的。
Monitoring tool wear is an important research topic in automated machining. The emergence of artificial neural network has opened up a broad prospect for on-line monitoring of tool wear. However, there are some problems that are neglected when applied, such as the existence of common BP network Slow convergence, easy to fall into the local minimum defects. In this paper, the optimization theory is used to analyze the defects of BP network, and the corresponding improvement measures, ie, the neural network decomposition algorithm, are proposed and used to estimate the tool wear. Experiments prove that the method adopted is correct.