基于遗传RBF神经网络的高速电主轴热误差建模

High-speed motorized spindle thermal error modeling based on genetic RBF neural network

针对高速电主轴热误差建模,对HMC80加工中心电主轴单元进行了热误差测量实验,综合利用模糊聚类法和灰色关联度分析法对测温点进行优化,使测温点数量从8个减少到3个,该方法同时考虑了温度变量之间的复共线性和测温点温度与热误差之间的相关性.以优化后的温度变量为输入,热误差为输出,建立基于遗传算法径向基函数（RBF）神经网络预测模型,并与其他方法进行比较.分析结果表明：相比于传统RBF神经网络法和多元线性回归法,遗传RBF神经网络建立的热误差预测模型精度更高、鲁棒性更强.

The experiment of HMC80 electric spindle was carried out for high-speed motorized spindle thermal error modeling.Fuzzy clustering and grey correlation analysis was conducted in temperature classification,and the number of temperature measuring sensor was reduced from 8 to 3.This method considered the multicollinearity between temperature variables,and the correlation between temperature and thermal error concurrently.Radial basis function neural network modeling was established based on the genetic algorithm,and the optimized three temperature candidates were used as the input while the thermal error was used as the output.Compared with other approaches,analysis results show that the genetic algorithm radial basis function neural network model performs better than the traditional radial basis function neural network model and multiple linear regression model in accuracy and robustness.