基于混合粒子群算法优化BP神经网络的机床热误差建模

Optimization of thermal error modeling of machine tool in BP neural network based on hybrid particle swarm optimization

为了降低机床热误差对主轴加工精度的影响,采用了混合粒子群算法优化BP神经网络结构,并对优化结果进行实验验证.引用了粒子群算法耦合遗传算法,给出BP神经网络结构简图,通过混合粒子群算法优化BP神经网络结构.构造机床热误差优化目标函数,采用混合粒子群算法优化目标函数,给出了混合粒子群算法优化BP神经网络流程图.建立BP神经网络热误差预测模型和BP神经网络热误差优化模型,采用三轴立式铣床对两种预测结果进行实验验证.实验结果表明:采用BP神经网络热误差预测模型,机床y轴、z轴预测结果与实验结果偏差最大值分别为6.9μm和6.7μm;采用BP神经网络热误差优化模型,机床y轴、z轴预测结果与实验结果偏差最大值分别为3.3μm和3.5μm.采用混合粒子群算法优化BP神经网络结构,能够提高机床热误差预测精度.

In order to reduce the influence of the thermal error of the machine tool on the machining accuracy of the spindle,a hybrid particle swarm optimization algorithm was adopted to optimize the BP neural network structure,and the optimization results were verified experimentally.In this paper,the genetic algorithm of particle swarm optimization（PSO）is introduced,and a schematic diagram of BP neural network is presented,and the BP neural network structure is optimized by the hybrid particle swarm optimization algorithm.The optimization objective function of the thermal error of the machine tool is constructed.The hybrid particle swarm optimization algorithm is used to optimize the objective function,and the flow chart of BP neural network is optimized by the hybrid particle swarm optimization algorithm.The BP neural network thermal error prediction model and BP neural network thermal error optimization model were established,and the two prediction results were verified by the three-axis vertical milling machine.The experimental results show that the thermal error by using BP neural network prediction model,y and z axis machine tool predicted results with the experimental results a maximum deviation were 6.9μm and 6.7μm,thermal error by using BP neural network optimization model,y and z axis machine tool predicted results with the experimental results a maximum deviation were3.3μm and 3.5μm.A hybrid particle swarm optimization algorithm is used to optimize BP neural network structure,which can improve the accuracy of thermal error prediction.