Abstract:After analysis of real nuclear magnetic resonance logging data,it is found that some data error might not follow a normal Gauss distribution.Therefore,a regression M-estimation (RME) was applied to relaxation spectra inversion,overcoming the effect of non-Gaussian distribution data error on the relaxation spectra resulted from ordinary regularization inversion method.The regression M-estimation suppressed the effect of large data errors on total errors by taking small right in large off-group data errors and large right in small data errors and this influence can be further overcome by an iterative regression process.In the inversion of nuclear magnetic resonance logging data,first step is using RME to calculate error.data set being pre-liminary modified,then using ordinary regularization method for relaxation spectra inversion in modified error data.The results of real data calculation show that combining RME with ordinary regularization inversion can improve the precision of relaxation spectra and the effect of calculated porosity.
翁爱华, 李舟波, 王雪秋, 陆敬安, 莫修文. 基于回归M-估计的核磁测井数据反演[J]. 石油地球物理勘探, 2002, 37(3): 258-261.
Weng Aihua, Li Zhoubo, Wang Xueqiu, Lu Jingan, Mo Xiuwen. Inversion of nuclear mag-netic resonance logging data based on regression M-estimation. OGP, 2002, 37(3): 258-261.