Automated Interpretation For Lwd Propagation Resistivity Tools Through Integrated Model Selection

Petrophysicists often have difficulty interpreting logs from today's multispacing, multifrequency logging-while-drilling (LWD) propagation resistivity tools. Which of the many resistivity curves represents the true formation resistivity? The logs may be affected to varying degrees by Borehole effect, tool eccentering, shoulder-bed effects, fractures, invasion, anisotropy and/or dielectric effects. These effects may occur individually, or multiple effects may be present in the same zone. Identifying these effects and correcting for them is challenging, especially when conclusions are needed quickly. Much of the information required to answer these questions is contained in the array measurements themselves. This paper presents a general automated scheme that can help analysts to identify environmental effects and to select the appropriate environmentally corrected formation resistivity. It answers the key question of how to select the correct model amongst many candidates, based primarily on inversion of the tool response and possibly other additional information. The key idea is to invert different formation models that may apply and select the one most consistent with the measurements of the tool and with auxiliary data from user inputs and/or other logs. When a dominant effect is identified, the correction is applied automatically, the relevant environmentally corrected data are generated, and confidence of interpretation is assigned. When none of the models result in an adequate fit, the data are flagged to indicate that an automatic interpretation could not be made because of more complicated or compounded environmental effects. In addition, petrophysicists can accept or reject certain models and impose petrophysical constraints to improve the interpretation. The program based on this approach has been used to process field logs with diverse formation characteristics, and it has been evaluated by experienced petrophysicists. The program includes borehole, invasion, dielectric, and anisotropy models in its model-base. Results show that the algorithm successfully identifies most environmental effects and highlights zones that need further analysis because of complex or compounded effects. The benefits of such an integrated-interpretation-through-model-selection approach will be demonstrated through four field log examples. Availability of these results at the wellsite is expected to improve both the timeliness and the quality of decisions made based on resistivity data.