基于RBF神经网络的井眼方位角误差建模及补偿

对于陀螺钻井测斜技术而言,由于惯性器件本身误差及井下恶劣条件带来的干扰均具有复杂性,误差补偿是影响测量精度的关键因素。在阐述陀螺测斜原理、分析测斜系统可能产生的各种误差的基础上,进行了测斜仪转台实验,并采用神经网络中的径向基函数理论建立了井眼方位角误差模型,将其建模性能指标与双线形插值建模指标进行综合对比,结果表明RBF网络建模时间短、拟合性能好、预测能力强、补偿后方位角误差得到有效抑制、补偿精度高,各项指标均优于双线形插值方法。

Azimuth Error Modeling and Compensation of Drilling Trajectory Using Radial Basis Function Neural Network

As for the gyroscopic survey technology, because of the complicacy of the inertia components themselves and the environment underground, error compensation is the main influence factor of the surveying accuracy. After analyzing the surveying principle and kinds of error factors, the experiment process based on the gyro calibration system was proposed and the radial basis function (RBF) neural network was used to model and compensate the azimuth error of the system. The modeling performances of RBF approach and the bilinear interpolation were synthetically compared. The test result shows that RBF network model makes more accurate predictions and compensation with less modeling time, and the azimuth error is restrained more effectively than the bilinear model.