刀具磨损的人工神经网络估计

难以对刀具磨损进行较准确的监测与界定是目前国内外自动化加工中一个急待解决的课题。本文利用优化理论对ＢＰ网络的缺陷进行了分析，提出动态步长法等优化方法，并结合选择了对刀具状态较敏感而对加工条件变化稳定性的相对切削力比值作为特征量，将其方向用于刀具磨损量的估计，实验证明，采用的方法是正确的、有效的、可行的，可以广泛应用于工程技术领域。

EVALUATIONS OF TOOL WEAR IN ARTIFICIAL NEURAL NETWORK

It is difficult to survey and define tool wear exactly,and it is an improtant problem in automation production to be urgently solved inside and outside the nation at present.In this thesis,the limitations of BP network were analyzed using optimization theory,the optimization method of dynamic step algorithm was advanced.And selected the relative cutting force ratio as the character variable which was sensitive to the tool state but was steady to machining condition,and then using the direction to estimate tool wear.Tool wear experiments and computer simulations show that the method and the means are correct,effective and valid.And it can be widely used to engineering and technology fields.