基于动态树理论的刀具磨损监测技术

TOOL WEAR MONITORING BASED ON DYNAMIC TREE

提出了基于动态树理论的刀具磨损监测方法,通过相关系数法提取传感器信号与刀具磨损最相关的几组特征,并采用具有局部记忆的B样条模糊神经网络建立刀具磨损量与声发射信号、切削力信号和振动信号特征之间的非线性映射关系,构造了任意加工条件下的刀具磨损监测系统,刀具磨损的识别结果由集成神经网络输出。试验结果表明,基于此方法构建的刀具磨损监测系统具有精度高、可靠度强、增殖性好和在线识别速度快等优点,值得工业推广。

A new methodology of tool wear classification based on dynamic tree is proposed. The correlation coefficients approach is utilized to extract several features with a close relation to tool wear. B-spline neural networks charactered by local memory is introduced to establish the nonlinearity relation between tool wear amounts and monitoring features extracted from acoustic emission, dynamometer and vibration sensors. Tool wear monitoring systems is so built under arbitrary machining conditions, and the integrated neural networks give the final classifying results of tool wear. The experimental results indicate that the tool wear monitoring system founded on the methodology is provided with high precision, high reliability, good multiplication and rapid recognizing speed, so it is good for popularization in industry.