An online neural network compensating algorithm for wing distortion influence on transfer alignment

An online neural network compensating algorithm for wing distortion influence on transfer alignment is proposed in the paper, which avoids augmenting measurement noise and system state in conventional methods. The wing distortion is modeled as the multi-order colored measurement noise firstly. Then the modified Kalman filter for solving the problem is derived. To compensate the noise and carry out the modified filtering process, online neural network is designed. The neural network can not only train the parameters of multi-order noise, but also adjust the Kalman filter with the plant noise. Meanwhile, the gain of Kalman filter is substituted with the neural network. The algorithm is efficient for rapidly and accurately estimating the misalignment in transfer alignment under complicate air environment without knowing the noise statistics. Simulations are done to compare the algorithm with conventional methods. Results of the simulations show that the algorithm outperforms other methods and attains good filtering performance in rapidity and accuracy.