VBAKF-based Correction Algorithm for Pure Inertial Solving of Attitude Angles

One of the key issues to solve the problem of train derailment detection during motion requires accurate information about the pitch angle of the train. In order to solve the problem of inaccurate measurement of acceleration carrier attitude using traditional Kalman filter and complementary filter. In this paper, we propose a method to correct the dispersed pitch angle obtained by pure inertial solution by taking the pitch angle obtained by variational Bayesian adaptive Kalman filter(VBAKF) solution as a reference, and introduce variational Bayesian adaptive filtering method to correct the attitude weight calculated by the accelerometer by designing the acceleration threshold to adapt to the acceleration change; Then ,the data are stored in a queue of size 100s, and the queue is updated every moment to ensure the continuity of the data; Finally, the relative change in pitch angle is obtained by making a difference between the data in the queue for the 0.3s intervals within the queue, aiming at detecting derailment by measuring short-term changes in the train's direction. The algorithm is experimented and the effectiveness of the algorithm is verified by the actual data. The root mean square error from the baseline is within the acceptable range of 0.05°.