基于鲸鱼算法优化LSSVM的铣刀磨损监测

Milling cutter wear monitoring based on whale algorithm optimized LSSVM

为解决铣刀磨损状态监测问题,提出一种改进的鲸鱼算法优化最小二乘支持向量机的状态识别方法.首先,采用变分模态分解处理铣削过程中的振动信号,分解得到的固有模态分量进行特征提取;然后,针对鲸鱼算法易陷入局部最优解、收敛精度低的问题,引入混合反向学习算法和非线性收敛因子进行改进,并采用基准测试函数验证改进后的鲸鱼算法的有效性;最后,将改进的鲸鱼算法优化LSSVM模型应用于铣刀磨损状态识别仿真实验.实验结果表明,相较于粒子群算法与传统鲸鱼算法,改进的鲸鱼算法优化LSSVM具有更高的识别精度.

In order to solve the problem of monitoring the wear status of milling cutters,an improved whale algorithm is proposed to optimize the state recognition method of least squares support vector machine.Firstly,variational modal decomposition is used to process the vibration signal in the milling process,and the characteristics of the inherent modal functions obtained by decomposition is extracted;then,to tackle the problem that whale algorithm is easy to fall into local optimal solution and low convergence accuracy,a hybrid reverse learning algorithm and the nonlinear convergence factor is introduced,and benchmark functions are used to verify the effectiveness of the improved whale algorithm;finally,the improved whale algorithm optimized LSSVM model is applied to the simulation experiment of milling cutter wear status recognition.The experimental results show that,compared with particle swarm algorithm and traditional whale algorithm,the improved whale algorithm optimized LSSVM has higher recognition accuracy.