Fast inversion of approximate resistivity for oil-based mud resistivity imager based on open-short calibration Authors: Fengfeng Li, Zhou Feng, Hongliang Wu, Hao Zhang, Han Tian, Peng Liu, Lixin Yuan Published: 2026 ABSTRACT AbstractThis paper proposes a new method for mud-effect correction and inversion of approximate apparent resistivity for oil-based mud microresistivity imaging logging. To correct the interference of the mud layer with electrode measurement signals, the Open-Short Calibration method is combined with a three-layer impedance model, and the open- and short-circuit states of the tool are simulated by the finite element method. By treating the electrode and the pad standoff as an integrated system, the method independently computes and explicitly removes the mud impedance, thereby extracting the formation impedance signal. To rapidly estimate formation parameters, a resistivity-consistency iterative inversion method is further proposed. After the formation impedance is extracted, the consistency residual among multifrequency approximate apparent resistivities is taken as the objective function, and nonlinear optimization is used to efficiently solve for formation resistivity, relative permittivity, and pad standoff. Numerical-model validation and field-well application results show that the proposed method can separate mud-signal interference over the full measurement range and enable quantitative characterization of formation apparent resistivity and related parameters.Keywords: Oil-based mud, High-definition electrical imaging logging, Microresistivity imaging, Standoff effectSubject terms: Energy science and technology, Engineering, Physics Introduction With the continued expansion of oil and gas exploration into complex domains such as deep formations and unconventional reservoirs, oil-based mud (OBM) has been widely adopted because of its excellent lubricity, stability, and tolerance to high-temperature and high-pressure conditions. However, the nonconductive nature of OBM makes conventional electrode-type resistivity logging methods difficult to apply directly1,2. Compared with water-based-mud electrical imaging logging, OBM electrical imaging logging based on capacitive coupling uses MHz-level high-frequency signals to drive current through the nonconductive mud and back to the return electrodes. Combined with multifrequency measurements, this markedly improves the resolution of formation resistivity and permittivity3–5. The raw measurement signals of OBM imaging tools are readily affected by mud resistivity, permittivity, and mud layer standoff, and therefore cannot directly and accurately represent the true formation properties. At present, orthogonal projection, optimization inversion, machine learning, and related techniques are commonly used to decouple the real and imaginary parts of the raw complex impedance and to construct resistivity-dominated images, permittivity-dominated images, and standoff images, thereby separating mud effects from formation properties6–9. In orthogonal projection, formation properties are estimated by decomposing the total impedance vector: in low- and medium-resistivity formations, the Z90 component perpendicular to the mud-impedance vector is extracted as an approximation of the formation impedance; in high-resistivity formations, the formation impedance is obtained by subtracting an estimate of the mud impedance from the total impedance magnitude10. Because this method is based mainly on a simplified parallel RC equivalent model, an unavoidable deviation remains between the estimated equivalent mud impedance and the actual mud impedance. In addition, the apparent resistivity is generated by stitching together measurement modes from different frequency bands after composite processing, and its accuracy depends strongly on both the switching thresholds and the calculated phase angle of the mud impedance. In practice, threshold effects can lead to resistivity reversals or stitching artifacts, so this method is generally suitable only for qualitative interpretation of formation properties. Optimization-based inversion and neural-network-driven quantitative inversion are important approaches for achieving higher-accuracy formation-parameter estimation. For a new-generation oil-based mud resistivity imager, Chen et al. established a high-fidelity forward model of the tool, simulated the impedance and phase responses under different mud and rock conditions, and proposed a multistep inversion workflow6. In that workflow, mud properties are first iteratively extracted over a short depth interval, after which these parameters are used to jointly invert pad standoff, formation resistivity, and permittivity over a longer logged interval. Although this inversion algorithm effectively corrects for mud-standoff and formation-dielectric effects and avoids resistivity reversal or artifacts in composite processing, the inversion is intrinsically highly underdetermined: at a single measurement point, four impedance inputs are used to solve for eight unknowns, including mud electrical properties (dual-frequency permittivity and conductivity), formation parameters (resistivity and dual-frequency permittivity), and the standoff. This severe underdetermination makes the optimization inversion highly sensitive to the inverted mud parameters, prone to local minima and convergence difficulties, and relatively inefficient, making it hard to satisfy the demand for rapid wellsite processing. To improve the computational efficiency of multiparameter numerical inversion, Bayraktar et al. introduced cascade feedforward neural networks (NNs) as an alternative11. By stepwise training and prediction, this method improves inversion speed while maintaining image resolution comparable to that of conventional inversion. Furthermore, machine learning and advanced data-driven algorithms have recently demonstrated outstanding capabilities in array signal processing, multi-parameter inversion, and feature localization across various engineering disciplines12,13. However, as a purely data-driven approach, its accuracy and generalization ability depend strongly on how broadly the training set covers mud and formation properties, which leads to reduced accuracy under complex formation conditions and weak controllability of the prediction process14,15. To alleviate the underdetermination of multiparameter joint inversion, reduce dependence on large training datasets, and rapidly obtain clear and reliable formation-parameter images without composite artifacts, this paper proposes a mud-effect correction and inversion method based on Open-Short Calibration (OSC) and multifrequency resistivity-consistency constraints. The proposed method first introduces Open-Short Calibration (OSC) into imaging-logging modeling and uses finite-element analysis (FEM) to strip the mud-impedance contribution from the total measured signal. On this basis, exploiting the physical fact that the resistivity of the same formation should remain relatively consistent at different measurement frequencies, this paper constructs an iterative inversion algorithm that uses the consistency residual of multifrequency apparent resistivity as the objective function to solve for formation parameters and mud standoff. Open and short-circuit calibration of mud impedance Calibration method Open-Short Calibration (OSC) is an impedance-calibration technique based on measurements at two extreme states. By acquiring the responses of the system under the two ideal boundary conditions of complete open circuit and complete short circuit, an error model is established to eliminate the bias introduced by the measurement system itself16 To accurately separate and calibrate the influence of mud on the measured impedance signal, this paper extends the technique to correct the interference of the mud layer with electrode measurements. A three-layer impedance model consisting of electrode, mud layer, and formation is established, in which the measuring electrode and mud layer are treated as one integrated system component. The impedance response of this system under open- and short-circuit states is simulated by the finite element method, and the mud impedance contribution is then separated from the measured total impedance. On this basis, an efficient approximate impedance-calculation model is built through polynomial fitting, and the open-short calibration formula is ultimately applied to remove mud layer interference and accurately recover the formation impedance. The pad configuration of the imaging tool studied here is shown in Fig. 1. It includes 24 button electrodes, one annular focusing electrode, and two return electrodes, and operates at two frequencies, F1 (low) and F2 (high). The imaging pad structure of the OBM imager. As shown in Fig. 2, the calibration model of the imaging tool consists of three layers: the outermost layer is an ideal conductor (short-circuit state) or air (open-circuit state); the middle layer is the mud layer to be calibrated; and the inner layer is the measurement pad. In the numerical simulation, the perfect electric conductor for the short-circuit state is modeled with a resistivity of , while the air for the open-circuit state is modeled with a resistivity of . Let the total measured impedance be , the mud impedance be , the pad impedance be denoted by , and the external-medium impedance be denoted by . The calculation model of OSC. In the short-circuit state, the external medium is an ideal conductor whose impedance approaches zero, and the total impedance of the measurement system is approximately: In the open-circuit state, the external medium is air whose impedance approaches infinity, and the total impedance of the measurement system is approximately: The open-circuit state provides the maximum external impedance reference. Taking the open-circuit state as the “zero-signal” reference, the effective admittance change is defined as The external medium impedance can be obtained through short-circuit calibration. However, when the formation resistivity is relatively high, retaining the second term for differential calibration can further improve the calculation accuracy. At this time, the formation impedance after open-short calibration is expressed as The formation impedance can be written in complex form as The apparent resistivity and apparent relative permittivity can then be expressed as The measured impedances in the open- and short-circuit states are simulated by the finite element method, and the computational model is shown in Fig. 2. It includes the pad, the mud layer, and the external medium. Figure 3 compares the magnitude and phase of the impedance computed in the short-circuit state with those of the equivalent mud impedance7. The equivalent mud impedance is calculated as follows: where is the standoff in m, S is the area of the button electrode in m, is the mud conductivity in S/m, and is the relative permittivity of the mud layer. Comparison of mud impedance amplitude and phase. As shown in Fig. 3, the equivalent mud impedance differs substantially in magnitude from the mud impedance obtained by the finite element calculation, whereas their phase angles are consistent. A fourth-order polynomial is used to piecewise fit the magnitude of the equivalent mud impedance to the magnitude of the FEM-computed mud impedance. As shown in Fig. 4, The relative error between the fitted results and the true values is less than 0.5%. The fitting results of mud impedance amplitude. A fourth-order polynomial is also used to fit the measured impedance under open-circuit conditions. The relative errors of the fitted impedance magnitude and phase are shown in Fig. 5, and the maximum relative error does not exceed 2.5%. The three fitting results together with the phase angle of the equivalent mud impedance are then used for efficient impedance calculation under open- and short-circuit conditions: where A is the magnitude of the equivalent mud impedance , – are fitting coefficients, and is the mud-layer standoff. The relative error of impedance and phase in the open state. Numerical validation To verify the effectiveness of the open-short correction method in separating mud effects, approximate apparent-resistivity responses were calculated and analyzed at excitation frequencies of 1 MHz and 30 MHz under different mud and formation conditions (Figs. 6, 7, 8 and 9). OSC corrected and vertical/parallel coupling results ( = 10000 , = 8, stdf = 1 mm, = 20). Approximate apparent resistivity calculated for different mud resistivities ( mm, , ). (a) Open-Short calibration at 1 MHz. (b) Open-Short calibration at 30 MHz. Approximate apparent resistivity calculated for different standoffs (, , ). (a) Open-Short calibration at 1 MHz. (b) Open-Short calibration at 30 MHz. Approximate apparent resistivity calculated for different formation permittivities (, , mm). (a) Open-Short calibration at 1 MHz. (b) Open-Short calibration at 30 MHz. Figure 6 compares the calculated apparent resistivity responses of the OSC method and the equivalent-circuit methods (vertical and parallel coupling7) under specific borehole environmental parameters (mud resistivity of , mud relative permittivity of 8, mud layer thickness of 1 mm, and formation relative permittivity of 20). The results indicate that the OSC-calibrated apparent resistivity exhibits excellent continuity across the entire dynamic range. Although the equivalent-circuit methods can present a similar overall trend after ideal segmented stitching, the OSC method eliminates the dependence on threshold-based stitching, thereby avoiding resistivity reversals and stitching artifacts caused by improper threshold selection in practical data processing. Considering the electrical dispersion of mud, Fig. 7 shows the OSC-corrected results for different mud resistivities (). The change in mud resistivity has very little influence on the corrected results, and the apparent-resistivity curves for all cases nearly overlap. In double-logarithmic coordinates, the apparent resistivity is essentially linearly related to the true formation resistivity with an approximately 1:1 relationship. Figure 8 verifies the adaptability of the algorithm to different mud standoffs. As can be seen, standoff variations have only a weak influence on the apparent resistivity extracted by the OSC method. The method effectively removes the effect of the mud layer over a range of standoff values. Figure 9 shows the corrected apparent-resistivity results for different formation permittivities. When the formation relative permittivity is less than 100, the apparent resistivity computed by the OSC method varies approximately linearly with the true resistivity. When the relative permittivity exceeds 100, the dominant role of high-frequency displacement current gradually increases in the high-resistivity interval, causing the apparent-resistivity growth curve to flatten and resulting in a nonlinear deviation from the true formation resistivity. Overall, the OSC method effectively separates mud-layer impedance interference and, without relying on threshold-based stitching in composite processing, provides a complete apparent-resistivity curve over an extremely wide dynamic range, thereby significantly reducing the inherent uncertainty of composite-processing methods. Consistency inversion For the same formation resistivity, the impedance measured by the imaging tool at different frequencies shows significant frequency dependence because of borehole-environment effects, permittivity, and other factors. This difference mainly arises from dispersion in formation relative permittivity, whereas the formation resistivity itself is nondispersive and does not vary with frequency. Based on this physical principle, inversion of formation resistivity using impedance data at multiple frequencies should theoretically yield a unique solution. Accordingly, this paper proposes an optimization inversion method constrained by apparent-resistivity consistency. By minimizing the residual between apparent resistivities at different frequencies, the inversion is transformed into a constrained optimization problem that is solved iteratively to obtain the optimal estimate of the formation apparent resistivity. Formation resistivity consistency residual Based on the above physical consistency principle, the objective function for multifrequency apparent-resistivity consistency is constructed as The core of the inversion is to find the optimal parameter vector that minimizes the objective function: During iterative inversion, the parameter vector contains five independent components: where the parameters denote the mud resistivities at the two measurement frequencies, the mud relative permittivities at the corresponding frequencies, and the pad standoff. The specific inversion procedure is as follows: Parameter initialization and bound setting: assign initial values to mud resistivity, mud permittivity, and standoff, and define the allowable range of each parameter. Parameter initialization and bound setting: assign initial values to mud resistivity, mud permittivity, and standoff, and define the allowable range of each parameter. Impedance calculation and mud-signal separation: based on the current parameter values, use the OSC model to calculate the mud-layer impedance and strip it from the total measured impedance to obtain the formation impedance. Impedance calculation and mud-signal separation: based on the current parameter values, use the OSC model to calculate the mud-layer impedance and strip it from the total measured impedance to obtain the formation impedance. Apparent-resistivity conversion: using the extracted formation impedance, compute the corresponding approximate apparent resistivities of the formation at the two measurement frequencies. Apparent-resistivity conversion: using the extracted formation impedance, compute the corresponding approximate apparent resistivities of the formation at the two measurement frequencies. Consistency-residual evaluation: substitute the above dual-frequency approximate apparent resistivities into (10) to evaluate the multifrequency resistivity-consistency residual. Consistency-residual evaluation: substitute the above dual-frequency approximate apparent resistivities into (10) to evaluate the multifrequency resistivity-consistency residual. Iterative optimization and convergence: repeatedly iterate with the optimization algorithm until the calculated consistency residual falls below the specified accuracy threshold, at which point the algorithm stops and outputs the optimal formation parameters. Iterative optimization and convergence: repeatedly iterate with the optimization algorithm until the calculated consistency residual falls below the specified accuracy threshold, at which point the algorithm stops and outputs the optimal formation parameters. Numerical validation To validate the inversion algorithm, a horizontal layered model with variable thickness was established using the finite element method, as shown in Fig. 10. The thickness of each formation layer decreases stepwise from 100 to 3 mm, and the formation resistivity alternates between and . Under dual-frequency operation, the mud resistivities are and , and the corresponding mud relative permittivities are 12 and 10. Dispersion of formation permittivity is also considered: for the low-resistivity layers, the dual-frequency relative permittivities are 27 and 15, respectively, whereas for the high-resistivity layers they are 18 and 12, respectively. Resistivity-consistency inversion results for the layered formation model. Figure 10 shows the measured impedance magnitude and the consistency-inversion results for standoffs of 1, 3, and 6 mm. The optimization process is implemented using the generalized Gauss–Newton inversion framework with bound constraints proposed17. The inverted apparent-resistivity curves () nearly overlap over the different standoff ranges and can still accurately characterize millimeter-scale thin interbeds at the bottom of the model. Spikes appear in the inverted standoff curves at formation interfaces, reflecting the nonlinear interfacial polarization effect of the high-frequency electromagnetic field, and the inversion algorithm compensates for the residual by increasing the standoff. The apparent relative permittivity is a derived parameter calculated from a parallel RC equivalent-circuit model and does not directly participate in the consistency-inversion optimization; it is more accurate at small standoff and exhibits numerical deviation at large standoff. Figures 11 and 12 show the standoff-inversion process and residual-decay process for 1-mm and 6-mm standoffs, respectively. Under the 1-mm standoff condition, the algorithm starts from an initial value of 3.0 mm, which deviates from the true value. The inverted standoff curve decreases rapidly, smoothly, and monotonically, and converges at the ninth iteration. Under the large 6-mm standoff condition, the tool detection depth is only about 5 mm, so the measurement signal is strongly attenuated by the mud and the system nonlinearity becomes much stronger. With the initial value set to 1 mm, the residual decreases relatively slowly at the beginning; after about 18 iterations, it starts to drop sharply and finally converges to the true standoff of 6.0 mm at the 24th iteration. These two convergence histories show that the consistency-inversion algorithm converges rapidly under strong formation signals and still converges stably under weak formation signals. Iterative consistency-inversion curve for a 1-mm standoff. Iterative consistency-inversion curve for a 6 mm standoff. The consistency inversion algorithm was employed to resolve a numerical formation model containing three fractures filled with different materials, with the baseline configuration referenced from Chen et al.8 (Fig. 13). The simulation parameters were set as follows: the pad standoff was 1.54 mm; at frequency 1, the mud resistivity was 8400 and the relative permittivity was 12; at frequency 2, the mud resistivity was 339 and the relative permittivity was 10. The background formation resistivity was 30 , and the corresponding relative permittivities at the two operation frequencies were 38 and 19, respectively. The spatial spacing between adjacent fractures was set to 20 mm, with a uniform fracture aperture of 6.35 mm, and the internal fracture resistivities increased sequentially. From top to bottom, these three fractures represent a sealed conductive fracture, a sealed resistive fracture, and an open fracture, respectively. The resulting multi-parameter inversion profiles are presented in Fig. 14, which displays the reconstructed curves of apparent resistivity, pad standoff, and high-frequency relative permittivity. It can be observed from the inversion results that all three fractures are clearly captured and well-reflected within the continuous inversion curves. Notably, the inverted standoff curve exhibits a significantly more pronounced and sensitive response to the open fracture, which is in excellent agreement with the findings and conclusions reported in Chen et al.8. The fractured formation models. Resistivity consistency inversion results for fractured formations. Field example The measured logging data from Well X were processed using three methods: the consistency inversion proposed in this paper, conventional Z90 composite processing, and quantitative-parameter inversion6. Figure 15 presents the integrated inversion results for this interval. From left to right, the tracks show the caliper log, the center-electrode resistivity logs extracted by the different methods, and the corresponding dynamic resistivity images. In addition, the figure shows the standoff image and the high-frequency relative-permittivity image simultaneously output by the consistency inversion and the quantitative inversion. Comparison of inversion results for Well X. As shown in Fig. 15, the resistivity curves extracted by the consistency inversion and the quantitative inversion are generally of the same order of magnitude as the array induction log (with a 10-in investigation depth). Although the two inversion results do not coincide exactly because of differences in algorithmic mechanisms, they are highly consistent in their portrayal of formation-property variations and, compared with the Z90 method, characterize high-resistivity formations more accurately. By contrast, the Z90-processed image exhibits obvious numerical artifacts and polarity reversal caused by nonlinear effects. In sharp contrast, the resistivity image produced by the consistency inversion has high vertical resolution and can delineate formation textures finely and accurately ; following the consistency inversion method, the sinusoidal continuous bedding structures or fracture texture features become extremely clear, exhibiting excellent lateral and vertical continuity, which directly enhances the recognizability of these geological features. In addition, the simultaneously provided standoff and permittivity images offer multidimensional physical evidence for real-time wellsite evaluation. The consistency inversion algorithm was further applied to process the field logging data from another borehole, Well Y. As illustrated in Fig. 16, the comprehensive processing suite comprises the conventional Z90 static and dynamic images, the inverted static and dynamic resistivity images, as well as the simultaneously output pad standoff and relative permittivity profiles. Notably, two open fractures are well-developed at XX75 m and XX79 m. Within the reconstructed profiles, compared with the blurred boundaries and structural obscurities in the traditional Z90 tracks, the consistency inversion results–particularly the pad standoff image–exhibit an exceptionally pronounced, sharp, and sensitive response to these open fractures. The processing results demonstrate that the inverted multi-parameter profiles can provide a more reliable and direct physical basis, thereby significantly enhancing the capabilities of wellsite fracture interpretation and reservoir characterization. Comparison of inversion results for Well Y. Discussion and conclusions It should be noted that while the proposed framework significantly enhances image continuity and successfully suppresses conventional composite processing artifacts, it possesses several inherent physical and algorithmic limitations. Fundamentally, the proposed method remains a qualitative to semi-quantitative imaging interpretation approach rather than a perfectly quantitative parameter evaluation tool. Particularly under large pad standoff conditions or within strongly dispersive formations, the dominant high-frequency non-linear effects and displacement currents cause the OSC-calibrated apparent resistivity to exhibit a non-linear deviation from the absolute true formation parameters. Furthermore, regarding the optimization layer, the employed Gauss-Newton iterative algorithm exhibits a relatively strong dependence on the selection of initial values and the accuracy of the preceding mud parameter calculations, which is inherently tied to the underdetermined nature of localized multi-parameter optimization. Consequently, the primary practical value of this method lies in providing a brand-new inversion approach for efficiently stripping mud-layer interference. In conclusion, this paper introduces the open-short calibration method into the measurement signal processing of the oil-based mud resistivity imagers, and proposes a new mud impedance calculation method. The measurement system is modeled as a three-layer model including the impedances of the electrode, mud layer, and external medium. The finite element method is used to simulate the impedance responses under open-circuit and short-circuit states, and the calibration formula is applied to eliminate the signals from the system and mud, thus effectively separating the formation signal. Validations on multiple models show that this method is superior to the vertical/parallel coupling method. It can realize full-range resistivity calculation from low-resistivity to high-resistivity formations and avoid the uncertainty caused by resistivity threshold splicing. Taking the multi-frequency resistivity consistency residual as the objective function, we propose an approximate apparent resistivity consistency iterative inversion method. Only a small number of datasets need to be pre-constructed for this method, which can complete the optimal solution of 5 parameters (mud layer parameters) through 4 measured values, realizing the inversion of formation apparent resistivity, permittivity, interval and other parameters. The calculation efficiency is 2 3 times higher than that of quantitative inversion. The analysis and verification of the defined objective function show that when the mud layer parameters are reasonably selected, the objective function attains the minimum value, indicating that the resistivity consistency residual can be used to invert the formation resistivity. The theoretical verification on the complex fractured formation model and the application test on the case well show that the consistency inversion can effectively calculate the formation characteristics. Although it is fundamentally an approximate qualitative to semi-quantitative method, its inversion results are consistent with the quantitative inversion results , improving the lateral and vertical continuity of sinusoidal bedding and fracture textures. Author contributions Fengfeng Li: Methodology, Writing - original draft, Conceptualization, Visualization, Writing - review and editing, Investigation. Zhou Feng: Conceptualization, Writing - review and editing, Project administration, Investigation. Hongliang Wu: Funding acquisition, Project administration. Hao Zhang: Software. Han Tian: Resources. Peng Liu: Resources. Lixin Yuan: Writing - review and editing, Formal analysis, Validation. Funding This work was supported in part by the Science and Technology Project of China National Petroleum Corporation under Grant No. 2024ZG45; and in part the National Science and Technology Major Project of China under Grant No. 2025ZD1402106. Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request. Restrictions apply to the availability of some field data, which were obtained from PetroChina and are therefore not publicly available. Declarations Competing interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared toinfluence the work reported in this paper. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Fengfeng Li, Email: cj21701@163.com. Zhou Feng, Email: fz@petrochina.com.cn. References Associated Data Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request. Restrictions apply to the availability of some field data, which were obtained from PetroChina and are therefore not publicly available.