时栅动态测量误差建模与补偿技术研究

Research on Modeling and Compensation Technology for Time-grating Sensor's Dynamic Measurement Errors

为提高时栅传感器动态测量精度,针对时栅的误差特点,提出对动态测量误差的周期性成分和随机性成分分别建模的思想。采用傅里叶级数逼近的方法对误差中的周期性成分进行建模,利用最小二乘方法对逼近模型参数进行寻优,选取比重较大的谐波参数对误差的周期性成分进行分离,对于分离后残留的随机性成分采用支持向量回归(Support vector regression,SVR)模型进行预测,利用交叉验证的方法对回归预测模型进行参数寻优和细化,选取最优的核函数参数g和惩罚因子C,使得残差均方达到最小。研发误差补偿系统,对时栅动态测量误差进行补偿。试验结果表明,运用该建模方法和模型,时栅传感器动态测量误差的峰峰值由38.2″降至3″,有效地降低了测量误差,大幅度提高了传感器的测量精度。

In order to improve time-grating sensor＇s measurement accuracy, according to the characters of errors, targeted modeling thought is proposed for the periodic part and the random part of dynamic measurement errors. Fourier series approach model is presented to establish the periodic error compensation model. By using least square method, parameters of approach model can be determined, the major of which are chosen to realize error separation. In order to model the random part of errors, support vector regression（SVR） model is proposed. The global optimal parameters of the model SVR which are the best kernel function parameter g and penalty factor C are evaluated by using cross validation method. A synthesis error compensation system is investigated, and realize error compensation. The accuracy analysis certificates that time-grating sensor＇s dynamic measurement error is reduced from38.2″to 3″by applying the error compensation method and compensation system. Sensor＇s accuracy is effectively improved by using the compensation system.