基于AIC-RBF的油气管柱挤压形变估计方法

油气管柱长期受到地层运动的影响会发生挤压变形,且挤压程度难以度量。利用脉冲涡流的油气管柱挤压形变估计反演算法,提出了一种基于AIC-RBF的油气管柱挤压形变估计方法,该方法包括基于赤池信息量准则(Akaike Information Criterion,AIC)的油气管柱形变多项式拟合优化算法和基于径向基函数(Radial Basis Function,RBF)神经网络的多项式参数估计模型。对管柱不同挤压段的脉冲涡流信号进行测试,获得对应的形变多项式函数,对挤压段的最小臂长进行量化,以估计其形变程度。实验结果表明:与传统RBF神经网络算法、BP神经网络算法相比,AIC-RBF算法的量化误差更小、稳定性更强、量化速度更快,满足油气管柱挤压程度无损量化的需求。

Extrusion deformation estimation method of oil and gas string basedon AIC-RBF

It is difficult to measure the deformation of oil and gas strings affected by the ground movement for a long time.In order to solve this problem,an inversion algorithm for extrusion deformation estimation of oil and gas strings based on pulsed eddy current was studied,and an extrusion deformation estimation method of oil and gas strings based on AICRBF was put forward.In the method,the AIC-based polynomial fitting optimization algorithm for deformation of oil and gas strings and RBF-based polynomial parameter estimation model were included.The pulsed eddy current signals of the different extrusion sections of the string were tested,the deformation polynomial function was obtained,and the minimum arm length of the extrusion section was quantified to estimate the degree of deformation.As shown by the experimental results,the AIC-RBF algorithm has smaller quantization error,better stability and faster quantization speed than the traditional RBF neural network algorithm and BP neural network algorithm,capable of satisfying the requirement of accurate quantization of the extrusion degree of oil and gas strings.