Pattern Recognition As A Means Of Formation Evaluation

Techniques for interpretation of patterns obtained from the possible cross-plots of log responses are described. It is shown that the pattern recognition approach described hereencompasses the more traditional interpretation methods (calculation of water saturation, Rwa, MOS, Rt/Rxo, Fr/Fs, etc.), while offering significant advantages over these methods. The basic advantage of the pattern recognition approach is that resistivity index (a measure of water saturation) can be derived without having a calibration for the porosity device (i.e., grain density, matrix delta t, etc.), without a knowledge of m, Rw, or Rmf, and in the presence of many types of logging "calibration"-type errors. The basic limitations are that a statistically significant number of zones are required and that unknown parameters (m, Rw, etc.) be reasonably constant for the zones used in the analysis. However, effects of changes or of presence of more than one value of a parameter or parameters can be predicted and accounted for. Under favorable conditions the patterns can be used to obtain values of m, can be distinguished matrix at and grain densities: rock types and calibration errors in the measurements can be detected and corrected. Uncertainties in the evaluation parameters (i.e., water saturation, resistivity index, etc.) can be estimated. It is concluded that pattern recognition is one of the most powerful interpretation approaches yet devised. Examples of the points discussed above are presented.