基于小波分析与LSSVM的陀螺仪随机漂移建模

Modeling method of gyroscope's random drift based on wavelet analysis and LSSVM

为了提高陀螺仪的使用精度,以陀螺仪随机漂移时间序列为研究对象,建立了基于小波分析和最小二乘支持向量机(LSSVM)的陀螺仪随机漂移模型。陀螺仪作为高精度敏感器件,其随机漂移信号具有非线性、弱平稳性等特点,难以补偿。为了提高补偿精度,这里采用小波分析对陀螺仪随机漂移信号进行多尺度分解,利用最小二乘支持向量机方法对重构后的近似序列和细节序列建立非线性子模型,最后将各子模型输出融合作为组合模型输出。最后将该算法用于动调陀螺仪的随机漂移建模,实验结果表明基于该组合算法的非线性模型能够有效地反映陀螺仪的随机漂移特性,建模效果明显优于直接采用LSSVM和ANN建立的模型。

For improving the gyro＇s operational accuracy, the random drift error model based on Wavelet Analysis and LSSVM（Least Square Support Vector Machine） was established taking the time series of gyro＇s random drift as study object. As a highly precise instrument, gyro＇s random drift is a nonlinear and weak-stability random process, and is difficult to compensate. To solve this problem, multi-scale decomposition was made on the random drift series by wavelet analysis. Then the reconstructed approximate series and detail series were regressed and modeled respectively by using LSSVM as the sub-model. At last, all sub-models＇ outputs were fused as the output of random drift error model. At last, this method was used to DTG＇ random drift error modeling. The experiment result indicates that this model can effectively reflect the characteristics of gyro＇s random drift, and is much superior than the model by LSSVM and ANN.