Using Computer Modeling To Provide Missing Information For Interpreting Resistivity Logs

Computer modeling is playing an increasingly important role in the integration of logging and non-logging sources of petrophysical data by providing a means for testing tool response in formation models built from multiple sources of information. This paper details three case studies where computer modeling is used to validate the correctness of information obtained from sources other than the resistivity logs themselves. In laminated formations, there are situations where SFL reads as much as a decade above the induction tools, with ID and IM overlaying. MSFL logs, when available, indicate that formations where this effect occurs are laminated. Computer models of SFL and induction response in laminated formations duplicate the field log. Additional modeling demonstrates that the resistivity average as measured by SFL does not equal the reciprocal of the average conductivity as measured by ID or IM, (i.e., 1/IL not equal to Rsfl). Therefore we must be particularly careful when attempting to equate small-scale electrical features to bulk electrical properties when comparing induction and laterolog logs. In the second case, periodic sine wave like oscillations occur on a high resistivity section of a North Sea Dual Induction log. The caliper shows that the borehole contains grooves about a quarter inch deep and 2 ft. long that appear to result from stabilizers on the drill-string. Modeling done using a range of mud, mudcake, and formation resistivities demonstrates by a process of elimination that under the right conditions, small grooves on the borehole wall can indeed generate periodic oscillations on induction logs. In addition, Fourier analysis verifies that the only possible way for IM or ID to resolve features 2 ft. thick or less on logs is when they are caused by anomalies located with 1 ft of the borehole wall. This conclusion can be applied in interpreting small anomalies on induction logs in general. In extremely resistive beds, induction logs recorded prior to the use of modern signal processing methods such as Phasor* processing are difficult to interpret because of extreme shoulder effect. Accurate interpretation of these old logs is still important for the ongoing analysis of existing reservoirs. In the final case, a Dual Induction log from Oklahoma exhibits strange horns as the tool passes from conductive shoulders into a very resistive bed. Potential causes of the horns, such as shoulder effect, dip, and deconvolution are systematically simulated by modeling. The horns are shown to be normal occurrences in an approximately rectangular bed that are accentuated for ID by deconvolution, and are made to appear even more pronounced by sonde error adjustments of the center-bed readings.