基于IPSO优化LS-SVM的铣削刀具磨损状态监测方法研究

Monitoring method of milling tool wear status based on IPSO optimized LS-SVM

刀具是机械加工中的重要组成部分,刀具磨损会影响加工精度和加工效率,准确掌握加工时刀具磨损状态至关重要,提出了一种改进粒子群优化(improved particle swarm optimization, IPSO)算法优化最小二乘支持向量机(least square support vector machine, LS-SVM)的刀具磨损状态监测方法。采集铣削时的切削力信号,分别利用经验模态分解(empirical mode decomposition, EMD)和主成分分析(principal component analysis, PCA)法进行特征提取和降维,IPSO算法改进了粒子速度、位置的更新策略和权重搜索方法,IPSO算法通过计算适应度函数对LS-SVM识别模型的惩罚因子和核参数迭代寻优。结果表明,降维后的特征可分性较强,IPSO算法寻优能力强于传统PSO和LdPSO算法,将降维后的特征当作IPSO-LS-SVM模型的输入,模型的识别精度和效率优于PSO和LdPSO优化的LS-SVM模型。

Tool wear is an important part of machining. Tool wear can affect machining accuracy and efficiency, so it is very important to accurately grasp the tool wear status during machining. A tool wear status monitoring method was proposed based on an improved particle swarm optimization(IPSO) algorithm optimized least square support vector machine(LS-SVM). The cutting force signals during milling were collected, and the feature extraction and dimension reduction were carried out by the empirical mode decomposition(EMD) and principal component analysis(PCA) respectively. The IPSO algorithm was modified by improving the strategy of updating particle velocity and position as well as by reforming the method of weight search. Based on the IPSO algorithm, the penalty factor and kernel parameters of the LS-SVM recognition model were iteratively optimized by calculating the fitness function. The results show that the features with dimensionality reduction are highly separable, and the optimization ability of the IPSO algorithm is stronger than that of traditional PSO and LdPSO algorithms. When the features with dimensionality reduction are taken as the input of the IPSO-LS-SVM model, the recognition accuracy and efficiency of the model are better improved than those of the LS-SVM model optimized by PSO and LdPSO.