为监测刀具的磨损状态,该文建立了一个基于声发射的刀具磨损状态监测系统。在刀具磨损状态监测实验中,采集加工过程中的声发射信号,提取方根幅值、绝对值均值、均方根、最大值作为反映刀具磨损的时域特征值。针对人工神经网络容易陷入局部极小值、结构难以确定、学习收敛速度慢等缺点,提出最小二乘支持向量机(least square support vector machine,LS-SVM)的刀具磨损状态识别方法。针对LS-SVM性能依赖于惩罚因子和核参数,利用粒子群优化(particle swarm optimization,PSO)算法对LSSVM参数进行自动寻优,建立PSO优化LS-SVM模型进行刀具磨损状态识别。结果表明:与LS-SVM识别模型相比,优化后的LS-SVM模型具有更高的识别率。 In order to monitor the wear status of the tool, an acoustic emission-based tool wear monitoring system was established. In the tool wear state monitoring experiment, the acoustic emission signals during processing were collected and the square root mean square value, root mean square value and root mean square value were extracted as the time-domain eigenvalues ​​reflecting tool wear. Aimed at the disadvantages of artificial neural network, such as local minima, easy to determine structure and slow learning convergence, a method of tool wear identification based on least square support vector machine (LS-SVM) is proposed. The LS-SVM performance is dependent on the penalty factor and kernel parameters. The particle swarm optimization (PSO) algorithm is used to optimize the LSSVM parameters automatically. The PSO-optimized LS-SVM model is established to identify the tool wear state. The results show that compared with the LS-SVM recognition model, the optimized LS-SVM model has a higher recognition rate.