刀具磨损早期故障智能诊断研究

Early Fault Intelligent Diagnosis of Tool Wear

针对刀具的早期故障监测中因存在强烈的背景噪声而难以提取故障特征的问题,提出了基于二次采样随机共振消噪和B样条神经网络智能识别的故障诊断方法。首先利用在随机共振过程中,噪声增强振动信号的信噪比特性,将刀具振动信号进行随机共振输出,提取有效特征,再输入到B样条神经网络进行智能识别,进而获得刀具的磨损值。同时,为了得到与输入信号最佳匹配的随机共振参数,提出了基于遗传算法的多参数同步优化的自适应随机共振算法,克服了传统随机共振系统只实现单参数优化的缺点。实验结果表明,该方法能实现弱信号检测,能有效地应用于刀具磨损故障诊断中。

In view of the difficulties of fault feature extraction from strong background noise in tool wear early fault diagnosis ,a method was proposed based on twice sampling SR and B-spline neural net-work .First ,SR was employed to remove noise in tool wear vibration signals because of its benefits for enhancing the signal-to-noise ratio ,then ,tool wears with the good fault features were identified by B-spline neural network .In order to improve the deficiency of a single parameter be optimized in the tra-ditional SR and achieve the best SR parameters ,an adaptive SR was proposed based on genetic algo-rithm ,which realized multi-parameter synchronous optimization .The experimental results show that this method can realize the weak signal detection and apply to tool fault diagnosis effectively .