Quantitative Demonstration of a High-Fidelity Oil-Based Mud Resistivity Imager Using a Controlled Experiment

Abstract

The objective of this paper is to describe and validate a new approach for acquiring images that provides both qualitative and quantitative information on the formation electrical properties using a high-resolution, oil-based mud imager (HROBMI) tool. This new multifrequency imaging tool is able to function at high frequencies (in the MHz range) in oil-based muds. To allow for the quantitative estimation of formation and mud properties from the HROBMI data, a hybrid machine-learning/inversion approach was implemented. In this hybrid approach, machine-learning models corresponding to different candidate mud properties are trained, and the resulting regression functions are stored. For a given measurement data set, predictions of these different models are used to quickly identify an optimum mud candidate. This information is then fed into an inversion algorithm that provides accurate quantitative information on the logging environment of the HROBMI. The accuracy of this algorithm has been verified using a test fixture that enables the change of formation properties in different mud environments. The measurements from the HROBMI are a function of the formation properties: resistivity and permittivity, frequency, and mud properties. The hybrid algorithm can untangle HROBMI data from multiple frequencies to obtain true formation resistivity images independent of the other parameters that affect the tool measurements. In addition, the algorithm provides formation permittivity images as well as a standoff image. The results have been provided from both the controlled experiments in the test fixture and from field logs.