基于支持向量机时栅数控转台时序预测研究

时栅代替光栅等传统位移传感器运用到全闭环数控转台做角位移检测部件,需采用时空变换算法将时栅的时域信息转换到空域信息。运用时间序列分析出时栅数控转台的测试数据依存特性,采用支持向量机建立起未来测试数据和历史样本的映射关系,从而得到测试数据中隐含的规律。依据过去相关测量值采用支持向量机回归预测下一采样时刻角位移,将原本等时采样的绝对式角位移转换为全闭合数控系统需要的等空间增量式连续脉冲。并且在误差控制方面,采用当前预测值对上一次预测误差进行实时修正,消除累计误差保证测量精度。实验证明支持向量回归的时间序列预测算法能有效保证动态数控角位移测量误差控制在±2.5″以内,实现精密全闭环角位移测量。

Research on forecast method for time grating CNC rotary table based on SVR and time series

In order to apply time grating in full closed-loop computerized numerical control(CNC) rotary table,it is necessary to transform temporal information into spatial information with time-space transformation algorithm and use time grating sensors as angle detectors to replace traditional displacement sensors such as optical grating.According to the correlation of measurement data for time grating CNC rotary table analyzed with time series,the mapping relation between future measurement value and past measurement value can be obtained with support vector machine.So the next measurement value can be forecasted with support vector regression(SVR) based on a series of past relative measurement values of the displacement.As a result,the original absolute displacement signal sampled in equal time interval can be converted into continuous incremental pulses required by the full closed-loop CNC system.Moreover,the last predicted error is always timely revised by current measurement value so as to eliminate the accumulated error for realizing precision measurement.Experiment results prove that the dynamic displacement measurement error is within ±2.5″ with SVR and time series prediction method.