论文部分内容阅读
为了有效抑制二频机抖激光陀螺(RLG)输出数据中的随机漂移,提出了一种结合谐波小波预滤波进行网络训练的径向基函数(RBF)神经网络滤波方法。原始的陀螺输出经过谐波小波预滤波消噪,消噪后的数据作为RBF神经网络训练样本的期望输出,对神经网络进行训练。以实测的长时间陀螺输出对完成训练的RBF神经网络的有效性进行验证,并利用Allan方差法对网络滤波前后的陀螺数据进行分析。结果表明:该方法有效降低了机抖激光陀螺的各项随机误差,提高了其使用精度。
In order to effectively suppress the random drift in the output data of the two-frequency gyroscope (RLG), a Radial Basis Function (RBF) neural network filtering method based on harmonic wavelet pre-filtering for network training is proposed. The original gyro output is filtered through harmonic wavelet pre-filtering. The denoised data is used as the expected output of the RBF neural network training samples to train the neural network. The validity of the trained RBF neural network is verified by the measured long-term gyro output, and the gyro data before and after the network filtering are analyzed by Allan variance method. The results show that this method effectively reduces the random errors of the LMGR and improves its using precision.