Noise Suppression Of Dynamic Measurement Of Aviation Gravity Gradiometer

Due to its high signal resolution and high accuracy,gravity gradient measurement makes it play a very powerful role in military and civilian use.Military aspects such as submarine detection,underground military target detection,etc.;civilian aspects can be used for geological survey investigation,natural disaster prevention,exploration of various minerals and groundwater resources.Aviation gravity gradient measurement refers to the placement of gravity gradient instruments on aircraft,airships,and other carriers;due to the carrier's motion capabilities,this measurement method can be measured under harsh terrain,such as mountains,lakes,swamps,and unmanned areas,etc.This greatly improves exploration efficiency and is therefore a very important technology.But precisely because of its convenience and high precision,the data of the instrument needs to undergo complex processing(such as post mission compensation)to finally obtain the signal we want.The manufacturing of the gradiometer and the process of data processing are extremely sensitive technologies,so Western countries have always imposed a technical blockade on China.China urgently needs to develop its own gradient data compensation technology.This article's work is based on this reason,summarizing experience from the few foreign publications and production reports of commercial companies,and many domestic studies on the error analysis of gravity gradiometers,and find ways to improve it.A gradient signal noise suppression method based on deep learning is proposed.This method takes the main factors causing the error,that is,the residual linear velocity,angular velocity,and angular acceleration of the carrier relative to the gradiometer as feature quantities,which are input into the neural network.Through the training of the neural network,it can grasp the complicated internal relationship between the factors that cause the error,so as to predict the error generated by the instrument in the dynamic measurement,and use it to correct the output signal of the instrument.In this way,a method of data compensation was found This method eliminates the tediousness of manual modeling when it is coupled with multiple factors,and is not limited by certain specific numerical models,but is driven by data,allowing the network to find an appropriate model by itself.In order to accomplish the above goals,the main work of this article is as followsOverview of the current status of gradient data processing methods at home and abroad,showing the current gradient data processing flow of commercial companies,focusing on the analysis of a post mission compensation(PMC)method proposed by Bell Aerospace Inspired,the main factors causing the instrument error(the residual linear velocity,angular velocity,and angular acceleration of the carrier relative to the gradiometer)were selectedStarting from the measurement equation of the gradiometer and combining with the actual situation,the output formula of the gradiometer due to the mismatch of scale factors is derived.Based on this,combined with the frequency of the prototype,the forward output obtained the theoretical output data of the instrumentSeveral underground spherical models were designed,and combined with the actual situation,the parameters such as the flight height and speed of the aircraft were given,and the gradient component under ideal conditions was obtained through forward simulation.The output signal of the gradiometer is demodulated using a quadrature amplitude demodulator and low-pass filtering to obtain the required gradient component signalBuild a neural network,and make the data obtained from the forward performance into a data set that meets the needs,input it to the neural network to train the parameters,and use the trained network to predict the noise.The predicted noise is used to correct the output signal of the gradiometer,so that the purpose of noise suppression is achieved...