基于遗传算法优化Elman神经网络的机床热误差建模

Thermal Error Modeling for Machine Tool Based on Genetic Algorithm Optimization Elman Neural Network

为了提高数控机床热误差模型的预测精度,以某型号立式加工中心为实验对象,采用模糊聚类与灰色综合关联度相结合的方法对机床测温点进行优化,将测温点从8个减少到3个。利用遗传算法(GA)优化的Elman神经网络建立了主轴热漂移误差预测模型,通过实例比较了GA-Elman神经网络模型与普通Elman神经网络模型的预测效果。结果表明,与普通Elman神经网络所建的预测模型相比,GA-Elman神经网络模型对主轴轴向热漂移误差的预测精度较高,残差较小,网络的泛化能力较好。

In order to improve the prediction accuracy of CNC machine thermal error model,a vertical machining center was taken as a research object. The temperature measuring points of machine are optimized by using a combined method of fuzzy clustering and grey comprehensive relationship degree. The temperature measuring points were reduced from 8 to 3. The prediction model of spindle thermal drift error was established based on genetic algorithm optimization Elman neural network. The predictive effect of GA-Elman neural network model and the common Elman neural network models were compared through the example.Compared with the prediction model built by the ordinary Elman neural network,the results showthat the Elman neural network has higher fitting accuracy,smaller residual error and better generalization capacity on spindle axial thermal drift error.