为了解决弹载MEMS陀螺实时输出预测问题,提高陀螺输出预测精度并减小计算量,通过分析弹载陀螺信号时间序列特性,结合小波分析和神经网络两种方法的优势,提出了一种基于小波神经网络的MEMS陀螺输出预测方法.选取并优化神经网络结构参数,利用实测陀螺数据对建立的小波神经网络进行训练,并预测出未来一段较短时间的陀螺输出值.与陀螺实测值和ARMA模型预测结果的对比发现,小波神经网络方法有效地提高了MEMS陀螺输出预测的精度、减小了计算量,从而验证了该方法的有效性和精度.%To solve the problem of prediction for MEMS gyros real-time output, improve the forecast accuracy of gyro output and reduce calculation, the time series of gyro output and the superiority of the wavelet and the neural network were analyzed and a method of predicting MEMS gyro output based on wavelet neural network was presented. Selecting and optimizing the network structure parameters and using the measured data aim at training the network and predicting gyro output in a future period of time. It could be found from the comparison with gyro measured data and the prediction by use of the ARMA model that the method of wavelet neural net work improved the prediction accuracy and reduced the amount of calculation, which verified the effectiveness and the precision of the method.
展开▼
