基于粒子群优化的螺纹刀具磨损状态监测模型

Research on Wear Condition Monitoring Model for Thread Cutting Tools Based on Particle Swarm Optimization

螺纹刀具的状态监测是制造加工中非常重要的问题。由于刀具振动信号具有复杂非线性、强耦合等关系,常用的基于支持向量机(SVM)的刀具监测模型由于参数的设置极其依赖人为经验,设置不当会导致监测的识别率不高,在刀具磨损状态判别中收到了限制。针对此难题,依据螺纹刀具的振动特性,结合改进的粒子群算法(PSO),采用异步更新学习因子策略实现刀具状态监测模型优化。结果表明,优化后的PSO-SVM刀具状态监测模型能够有效对SVM的关键参数进行寻优,异步更新学习因子也可加强模型在迭代后期的寻优能力,从而提高刀具状态监测识别的精度。

Condition monitoring of thread cutting tools is vital in manufacturing.Due to the complex nonlinear and strong coupling relationship of the tools’vibration signal,the commonly used model for tools monitoring,which is primarily based on a Support Vector Machine(SVM),relies heavily on human experience,where improper settings can lead to a low recognition rate.Considering the vibration characteristics of thread cutting tools,the combination algorithm of Particle Swarm Optimization and SVM,together with the strategy of asynchronously updating learning factors,is adopted to optimize the process of conventional tools’condition monitoring model.Results show that the improved PSO-SVM model for condition monitoring can effectively optimize the key parameters of the SVM,while the asynchronously updating learning factor can enhance the optimization ability of the model at later stages of iteration,and thus the accuracy of tool condition monitoring is improved.