Research on Filtering Algorithm of Attitude Signal While Drilling in Directional Drilling Rig

In the process of coal seam gas extraction and geological parameter acquisition, it is very important to control the drilling trajectory of directional drill pipe accurately in real time to ensure the life safety of underground workers. To achieve this goal, the premise is to obtain high-precision drilling tool motion attitude information. Measurement While Drilling (MWD) technology collects the acceleration, angular velocity and magnetic field strength of the drill tool in real time through the three-axis accelerometer, three-axis gyroscope and three-axis magnetometer in the measurement unit, and then obtains the attitude Angle of the drill tool through the attitude solution method. However, due to the complexity of the working environment and the error of the measuring instrument, the accuracy of the attitude calculation result may be affected. In order to improve the accuracy of MWD attitude calculation, this paper combines Bayesian Kalman filter (BSKF) with long short-term memory (LSTM) neural network, and simulates and compares the algorithm with Bayesian Kalman filter and back propagation (BP) neural network. The simulation results show that the algorithm based on LSTM neural network can effectively reduce the error of attitude solution and improve the stability of attitude solution. Compared with BSKF filter and BSKF and BP neural network fusion algorithm, the proposed algorithm has smaller relative error, the smallest error parameters (standard deviation, mean absolute value error and mean square error), and the solution accuracy is better than other algorithms. In summary, the BSKF algorithm integrated with LSTM neural network shows higher accuracy and stability in MDD attitude calculation, and provides reliable technical support for coal seam gas extraction and geological parameter acquisition.