论文部分内容阅读
硅微陀螺漂移具有混沌特性,可以通过相空间重构对漂移进行预测.计算出硅微陀螺漂移序列的Lyapunov指数为0.000 529,估计出随机漂移可预测的时间尺度为1 890,求出相空间重构所需的延迟时间、关联维数和嵌入维数分别为57,7.042和15.以相空间重构后的漂移序列为输入变量,提出利用RBF神经网络和陀螺阵列技术,对陀螺静态测试和动态测试时的随机漂移序列进行预测.预测结果表明:基于相空间重构的陀螺静态和动态测试情况下的预测精度可分别提高5.39和2.65倍,优于常用的时序法和未经相空间重构的神经网络法.
The drift of silicon micro-gyroscope is chaotic, and the drift can be predicted by phase space reconstruction.The calculated Lyapunov exponent of the micro-gyroscope drift sequence is 0.000529, and the predictable time scale of random drift is estimated to be 1890. The phase space The delay time, correlation dimension and embedded dimension needed for reconstruction are 57, 7.042 and 15. Based on the phase space reconstructed drift sequence as the input variables, the RBF neural network and gyro array technology are proposed to test the gyro static test And the random drift sequence of the dynamic test are predicted.The prediction results show that the prediction accuracy of the gyro static and dynamic test based on phase space reconstruction can be increased by 5.39 and 2.65 times, respectively, which is better than the common sequential method and non-phase space Reconstruction of neural network method.