基于RBF神经网络的数字闭环光纤陀螺温度误差补偿

为了消除数字闭环光纤陀螺温度误差，设计了基于径向基函数（RBF）神经网络的温度误差补偿方案，对该方案所采用的标度因数误差模型和偏置误差模型进行了研究。首先，根据光纤陀螺的温度误差分布情况设计了标度因数误差和偏置误差联合补偿的方案。接着，将基于多尺度分析的噪声和趋势项分离算法应用于建模数据预处理，以提高建模数据准确性。然后，建立了RBF神经网络模型，并改进模型的学习方法以防止网络的过拟合。最后，讨论了模型输入向量对神经网络规模的影响。温度补偿的结果表明：标度因数误差模型的残差均方（RMS）达到0.73 ，偏置误差模型的RMS达到0.051 。该建模方法可以满足中、高精度光纤陀螺实时温度补偿的要求。

Temperature Errors Compensation for Digital Closed-Loop Fiber Optic Gyroscope Using RBF Neural Networks

In order to supress temperature errors of digital closed-loop fiber optic gyroscope (FOG), a scheme based on radial basis function (RBF) neural networks was designed for temperature errors compensation and its applied scale factor error model and bias error model were investigated. First, based on the distribution of FOG’s temperature errors, a scheme, which combined scale factor error compensation and bias error compensation, was designed for temperature errors compensation. A multiscale analysis algorithm of signal feature extraction was used to preprocess original testing data for higher modeling accuracy. Then two RBF neural network models were developed and their learning algorithm was improved to avoid over-fitting. Finally, the effects of the models’ input vectors on the models’ scale are discussed as well. Analysis of simulation results indicate that residaul mean square (RMS) of the scale factor error model is 0.73 and the RMS of the bias error model is 0.051 . It can satisfy the requirements of real-time temperature compensation for middle and high precision FOG.