基于 MSVM 的多品种小批量动态过程在线质量智能诊断

摘要

An on-line quality intelligent diagnosis method was proposed based on MSVM.In off-line training,the feature data was extracted using wavelet reconstruction of quality patterns,the MSVM recognition model and estimation model then were trained and tested.In online diagnosis,the abnormal pattern and the parameter of data flow in “monitoring window”were recognized and esti-mated by the trained recognition model and estimation model,then the dynamic process was diag-nosed by the construed diagnosis library.Finally,the proposed method is applied to online diagnosis the precision-axis machining process,and the results of application example show its effectiveness.%提出了基于多分类支持向量机(MSVM)的多品种、小批量动态过程在线质量智能诊断方法。离线训练时，提取异常模式仿真数据的小波重构特征，对 MSVM 识别和估计模型进行训练和测试，同时建立异常因素诊断库；在线诊断时，对“监控窗口”数据特征的过程模式及参数进行识别与估计，而后利用异常因素诊断库实现对多品种、小批量动态过程实时在线智能诊断。某精密轴加工过程实例验证了该智能诊断方法的有效性。