文中提出了一种基于3层 BP 人工神经网络新的光纤陀螺捷联系统动态误差模型的建模方法。在Matlab 软件平台上,利用正交三轴转台速率实验产生的大量数据样本对网络结构模型进行了学习训练,并使用实测数据对得到的模型进行了验证研究。实...文中提出了一种基于3层 BP 人工神经网络新的光纤陀螺捷联系统动态误差模型的建模方法。在Matlab 软件平台上,利用正交三轴转台速率实验产生的大量数据样本对网络结构模型进行了学习训练,并使用实测数据对得到的模型进行了验证研究。实验结果表明,训练后的3层 BP 网络模型辨识精度高,可以做光纤陀螺捷联惯导系统的动态误差模型直接使用,为系统的动态误差补偿提供了数学基础。展开更多
The dynamic errors of gyros are the important error sources of a strapdown inertial navigation system. In order to identify the dynamic error model coefficients accurately, the static error model coefficients which la...The dynamic errors of gyros are the important error sources of a strapdown inertial navigation system. In order to identify the dynamic error model coefficients accurately, the static error model coefficients which lay a foundation for compensating while identifying the dynamic error model are identified in the gravity acceleration fields by using angular position function of the three-axis turntable. The angular acceleration and angular velocity are excited on the input, output and spin axis of the gyros when the outer axis and the middle axis of a three-axis turntable are in the uniform angular velocity state simultaneously, while the inner axis of the turntable is in different static angular positions. 8 groups of data are sampled when the inner axis is in 8 different angular positions. These data are the function of the middle axis positions and the inner axis positions. For these data, harmonic analysis method is applied two times versus the middle axis positions and inner axis positions respectively so that the dynamic error model coefficients are finally identified through the least square method. In the meantime the optimal angular velocity of the outer axis and the middle axis are selected by computing the determination value of the information matrix.展开更多
文摘文中提出了一种基于3层 BP 人工神经网络新的光纤陀螺捷联系统动态误差模型的建模方法。在Matlab 软件平台上,利用正交三轴转台速率实验产生的大量数据样本对网络结构模型进行了学习训练,并使用实测数据对得到的模型进行了验证研究。实验结果表明,训练后的3层 BP 网络模型辨识精度高,可以做光纤陀螺捷联惯导系统的动态误差模型直接使用,为系统的动态误差补偿提供了数学基础。
文摘The dynamic errors of gyros are the important error sources of a strapdown inertial navigation system. In order to identify the dynamic error model coefficients accurately, the static error model coefficients which lay a foundation for compensating while identifying the dynamic error model are identified in the gravity acceleration fields by using angular position function of the three-axis turntable. The angular acceleration and angular velocity are excited on the input, output and spin axis of the gyros when the outer axis and the middle axis of a three-axis turntable are in the uniform angular velocity state simultaneously, while the inner axis of the turntable is in different static angular positions. 8 groups of data are sampled when the inner axis is in 8 different angular positions. These data are the function of the middle axis positions and the inner axis positions. For these data, harmonic analysis method is applied two times versus the middle axis positions and inner axis positions respectively so that the dynamic error model coefficients are finally identified through the least square method. In the meantime the optimal angular velocity of the outer axis and the middle axis are selected by computing the determination value of the information matrix.