基于径向基函数网络的刀具磨损识别

Recognizing Tool Wear Based on Radical Basic Function Network

提出了一种基于灰色关联度优化网络神经元数目和径向基函数网络用于刀具磨损量预测的方法.以选取合理的涵盖影响刀具磨损的有关因素,采用不同切削条件下铣削加工过程刀具后刀面磨损的多组实验数据对网络模型进行训练以及对刀具磨损量进行估计和预测,预测结果与实际基本吻合.结果表明,该方法克服了用一个多元线性公式描述由切削条件和切削带来的后刀面磨损量的变化的刀具磨损高度非线性模型方法的缺陷,对于与刀具磨损量相关因素的非线性本质较易准确表达,所建立的刀具磨损网络模型可以较满意地计算出不同切削条件下刀具后刀面的磨损量.

A method that optimizes the neural number of radial basic function network （RBFN） based on the grey relating degree analysis is put forward to forecast tool wear extent. The net model is trained by selecting reasonable factors related to tool wear, adopting several groups of data of tool wear in different cutting conditions. The tool wear can be forecasted by the network. The forecast results can accord with the practical data. Experimental results indicate that the disadvantage of the high nonlinear model of tool wear, which depicts cutting conditions by a multi-linear equation, is overcome, the nonlinearity essence of correlative factors is exactly explained and the tool wear extent in different cutting conditions can be calculated by using the net model of cutting tool wear.