采用量子粒子群算法耦合差分进化算法优化BP神经网络的铣床热误差预测研究

Using the quantum particle swarm algorithm coupled differential evolution algorithm to optimize the prediction of the thermal error of milling machines of BP neural network

针对铣床主轴运行产生的热误差问题,采用改进BP神经网络预测模型,并对预测结果进行验证。融合量子粒子群算法和差分进化算法的各自优点,给出混合算法寻优操作流程。分析BP神经网络结构,给出改进BP神经网络优化流程图,构造铣床热误差适应度函数,采用混合算法优化BP神经网络预测模型。通过具体实例对铣床热误差进行实验验证,预测结果显示:BP神经网络预测偏差值较大,在Y轴、Z轴方向预测产生的偏差最大值分别为7.3μm和7.5μm,改进BP神经网络预测偏差值较小,在Y轴、Z轴方向预测产生的偏差最大值分别为2.8μm和2.9μm。同时,改进BP神经网络预测铣床热误差与实际偏差值波动较小。采用改进BP神经网络预测铣床热误差精度较高,可以提高主轴加工工件的精度。

Aiming at the thermal error of the milling machine spindle,the prediction model of BP neural network is improved and the prediction results are verified. Combining the advantages of the quantum particle swarm optimization algorithm and the differential evolution algorithm,the optimal operation flow of the hybrid algorithm is given. The BP neural network structure is analyzed,the optimized flow chart of BP neural network is improved,the thermal error fitness function of the milling machine is constructed,and the BP neural network prediction model is optimized by the hybrid algorithm. Through specific examples of milling machine experiment verification,thermal error prediction results showed that the BP neural network prediction deviation is bigger,in the Y and Z axis direction to predict the maximum deviation were 7. 3 μm and 7. 5 μm,smaller values of the improved BP neural network prediction deviation,in the Y and Z axis direction to predict the maximum deviation were 2. 8 μm and 2. 9 μm. At the same time,the improved BP neural network prediction of the thermal error of the milling machine and the actual deviation is less. It can improve the precision of machining parts by improving BP neural network.