论文部分内容阅读
复杂运行工况下的电主轴温度检测缺乏全面性与实时性。同时,电主轴的温升过程有较强的非线性、时变性和滞后特性,采用常规的线性化模型存在算法难以在线实施,预测精度不能保证等问题。针对高速电主轴表面温度预测问题,采用遗传算法和神经网络相结合,通过遗传算法优化BP神经网络,建立电主轴温度预测模型,提高预测精度。仿真研究表明,经遗传算法优化后的BP神经网络温升预测最大相对误差百分比减小了0.97%,平均相对误差百分比减小了0.24%,即遗传算法优化的BP神经网络预测电主轴表面温度精度高、稳定性强。
Under the complex operating conditions, the spindle temperature detection lacks comprehensiveness and real-time performance. At the same time, the temperature rise process of the spindle has strong nonlinear, time-varying and hysteresis characteristics. The conventional linearized model is difficult to be implemented online and the prediction accuracy can not be guaranteed. According to the prediction of surface temperature of high-speed spindle, genetic algorithm and neural network are combined to optimize BP neural network through genetic algorithm to establish the prediction model of spindle temperature and improve prediction accuracy. The simulation results show that the maximum relative error of BP neural network optimized by genetic algorithm is reduced by 0.97% and the average percentage of relative error is reduced by 0.24%, that is, the BP neural network optimized by genetic algorithm predicts the temperature accuracy High, stable.