基于动态模糊神经网络的机床时变定位误差补偿

Time-varying Position Error Compensation of Machine Tools Based on Dynamic Fuzzy Neural Networks

为提高数控机床的定位精度,提出基于动态模糊神经网络进行数控机床时变定位误差补偿的方法。针对数控机床定位误差影响因素复杂、模糊规则难于获取的情况,改进动态模糊神经网络,使其能够应用于多输入多输出系统,并实现模糊规则的自动在线辨识与生成。通过测量机床温度和定位精度,应用改进后的动态模糊神经网络建立机床时变定位误差预测模型。运用该模型对数控机床进行定位误差补偿试验,并与径向基神经网络模型补偿的效果进行比较,结果显示,基于动态模糊神经网络的数控机床时变定位误差预测模型精度高、泛化能力强、鲁棒性优,适用于对数控机床定位误差的长时间、高精度的实时补偿。

A new time-varying position error compensation method for machine tools based on dynamic fuzzy neural networks（D-FNN） is presented to improve the positioning accuracy of numerical control（NC） machine tools.In view of the complexity of influencing factors of NC machine tool positioning accuracy and the difficulty to obtain fuzzy rules,the D-FNN is improved to fit for multiple-input multiple-output system,and also to automatically online identify and generate fuzzy rules.Through measuring the temperature and positioning accuracy of NC machine tool,a NC machine tool time-varying position error prediction model is build on the basis of the improved D-FNN.Then this model is used to compensate the NC machine tool＇s positioning error,and its effect is compared with the compensation effect of a radial basis function（RBF） neural network model,which shows that the D-FNN model features high accuracy,strong generalization ability and excellent robustness,thus being more suitable for long-time,high-precision real time compensation of NC machine tools.