Weighted processing for microresistivity imaging logging in oil-based mud using a support vector regression model

Abstract

We have developed a weighted processing method for wireline microresistivity imaging logging in oil-based mud to estimate not only the low but also the high formation resistivity quantitatively using a support vector regression (SVR) model. Furthermore, the standoff between the logging tool and the formation is also optimized out quite well. The general vertical coupling process is an unweighted method that is only suitable for low-resistivity formation and causes a puzzling reversal phenomenon in high-resistivity formation. Therefore, we have first determined the necessity of introducing a weighted coefficient, and then we developed an improved coupling processing method, i.e., a weighted processing model. We have implemented a sensitivity analysis to determine the change regularity and range of the weighted coefficient. We have developed an SVR model with four main controlling parameters: frequency, real part of measured impedance, imaginary part of measured impedance, and equivalent resistivity to optimize out the weighted coefficient, thus to estimate the formation resistivity accurately. Furthermore, a similar SVR model is also developed to obtain the unknown standoff. We have determined the effectiveness and advantage of the weighted processing in estimating the formation resistivity with two references, respectively, the imaging results in water-based mud and from the general unweighted processing, using three simulated cases. Meanwhile, the optimization of the standoff is also verified.