摘要
采用振动传感器采集刀具车削时的信号,对振动信号进行短时傅里叶变换,将频谱集中区域(0~6250)Hz内的频率幅值直接输入到BP神经网络中进行训练,使神经网络建立振动信号频谱与刀具磨损量之间的映射关系,从而实现刀具磨损监测。人工提取的特征值一般数量较少,往往不能全面细致地刻画信号的特点,而该方法则充分发掘了神经网络强大的学习能力,具有方法简单、识别精度高、稳定性好的优点。实验结果表明,该方法可以快速准确地预测刀具磨损量。
Using vibration sensor to acquire the signal in tool turning, execute short-time Fourier transform to the vibration signal, then inputs the frequency amplitude in frequency spectrum concentrated area （0-6250）Hz into BP neural network directly to training, make the neural network to establish the .mapping relationship between vibration signal frequency spectrum and tool wear, so as to realize tool wear monitoring. The number of characteristic value that extracted artifwially is generally less, often can not depiction the signal＇s characteristics fully and detailedly. While this method makes a good use of the neural network＇s powerful learning ability. This method has the advantages of simple, high accuracy and good stability. The experimental results show that, this method can predict tool wear quantity quickly and accurately.
出处
《机械设计与制造》
北大核心
2017年第10期113-116,共4页
Machinery Design & Manufacture
基金
国家科技重大专项课题-高档数控机床与基础制造装备(2012ZX04005-021)
关键词
刀具磨损
振动
频谱
BP神经网络
Tool Wear
Vibration
Frequency Spectrum
BP Neural Network
作者简介
库祥臣，（1968-），男，河南人，博士研究生，副教授，主要研究方向：数控技术、工业自动化技术
