基于BP神经网络的时栅时序预测测量研究

Study of Time Series Predictive Measurement of Time Grating Based on BP Neural Network

为了将时栅应用于全闭环数控系统,需完成时栅信号由时域到空域的转换。通过BP神经网络预测模型找出实测数据中的隐含规律进而建立起样本和未来实测数据的映射关系,预测出下一个周期内时栅的测量角度值,实现时栅绝对式角度值与光栅数控系统所需的增量式连续脉冲的转换;为了保证测量精度,利用当前测量值对上一次的预测误差进行校正。实验表明:基于BP神经网络预测算法的时栅系统可以实现时域信号向空域信号的转换,且误差精度为±2″,满足了数控系统对测量精度的要求。

In order to apply the time grating to the full-closed CNC system,it is necessary to complete the conversion of the time grating signal from time domain to space domain.BP neural network prediction model was used to find out the implicit law in the measured data to establish the mapping relationship between historical samples and future measured data,the measured anglevalue of the next measurement period of the time grating was predicted,and conversion between absolute angle of time grid and incremental continuous pulse required by optical grating CNC system was realized.In order to ensure the measurement accuracy,the current measurement was used to correct the last prediction error.The experiment shows that the time grating system based on BP neural network prediction algorithm can realize the conversion from time domain signal to space domain signal,and the error accuracy is±2″,satisfying the measurement accuracy requirements of CNC system.