基于前向线性预测算法的光纤陀螺零漂的神经网络建模

在详细分析光纤陀螺零漂的基础上,提出了先用滤波算法对光纤陀螺信号进行预处理,然后采用RBF神经网络对滤波后的信号进行建模的方法。针对光纤陀螺信号特点分别采用FLP算法、小波滤波算法、解相关变步长LMS自适应滤波算法对其进行了预处理,比较三种滤波方法,小波滤波算法效果优于其它两种预处理方法,但针对基于预处理后的陀螺信号采用RBF神经网络进行建模时,小波滤波预处理后的信号在建模精度上却是最差的,而对FLP算法滤波后的信号进行RBF建模,建模精度提高了两个数量级。结果表明:基于FLP算法的RBF神经网络在光纤陀螺中的建模是有效的,可大大提高建模的精度。

Neural network modeling for FOG zero point drift based on forward linear prediction algorithm

By analyzing the zero point drift of FOG,a method for preprocessing the FOG signal using filtering algorithm and then modeling it with radial basis function ( RBF) neural network (NN) was presented. The FOG signal was preprocessed by Forward Linear Prediction (FLP) filter,wavelet filter and De-correlation variable-step Least Mean Square (LMS) adapting filter respectively. The results show that the effect of wavelet filter is the best among three filtering algorithms,but the modeling precision using RBF neural network for the FOG signal after preprocessing is the worst. By preprocessing the signal of FOG using FLP filter,the precision of RBFNN modeling can be improved two orders higher than that using wavelet filter. Simulation results show that by using FLP algorithm for preprocessing,and then modeling for the Zero Point Drift of FOG by RBF neural network,the precision of modeling can be greatly improved.