基于EMD-SVR的光纤陀螺随机误差预测

为了提升光纤陀螺随机误差建模的准确性及补偿结果,提出了一种基于经验模态分解与支持向量机结合的随机误差预测方法。鉴于随机误差的非线性及不稳定性,直接进行预测时精度不高,采用经验模态分解对原始数据进行分解以降低时间序列的复杂程度;然后根据经验模态分解得到的各本征模态函数及趋势序列,构建基于支持向量机的预测模型;再将所得的各分量的预测结果综合以得到光纤陀螺随机误差的预测结果。以光纤陀螺随机误差数据作为验证,结果表明,相较于传统的预测方法,均方根误差与平均绝对误差分别降低了78.4%和75.5%,有效提高了回归精度。

Fiber Optical Gyro Random Errors Prediction Based on EMD-SVR

In order to improve the accuracy and compensation result of fiber optic gyro random error modeling, a random error prediction method based on empirical model decomposition and support vector machine is proposed. In view of the non-linearity and non-stability of random errors, as well as the low accuracy of direct prediction, the empirical model decomposition is used to decompose the original data to reduce the complexity of time series. The intrinsic mode function and trend sequence are constructed with the empirical model decomposition, and the prediction model based on support vector machine is constructed. Then, the prediction results of the obtained components are combined to obtain the prediction result of the fiber optic gyro random error. Based on the random error data of fiber optic gyro, the results show that compared with the traditional prediction method, RMSE and MAE are reduced by 78.4% and 75.5%, respectively, which effectively improves the regression accuracy.