基于BP神经网络的FAST馈源舱融合测量预测研究OA北大核心CSTPCD

When the Five-hundred-meter Aperture Spherical Radio Telescope(FAST)per-forms the tracking observation task,for cooperating with this task,the feed has got spatial motion.The fine-tuning positioning of the feed is realized by the feed cabin,so the high-precision measurement of the position of the feed cabin is great significance.However,when the total station equipment fails,it is unable to correct the GPS/IMU fusion measurements with the Kalman algorithm,it causes the accuracy of the feed cabin measurements decreas-ing.In order to solve this problem,this paper designs a prediction model based on BP neural network,which is composed of three parts,the data preprocessing,the model design and the model training validation.And the model training data is the real measurement data of FAST with a data volume of about 40 GB.In order to verify the generalization ability of the model,three kinds of motion trajectory data are selected to test the model prediction accuracy,and the results show that the accuracy meets the 15 mm requirement under three kinds of motion trajectories.