The utilization of well log data to provide a lithologic description of complex formation is well-evidenced. Historically, however, most approaches assumed fixed, known, and distinct lithologies for analyst-specified zones. The proposed approach attempts to alleviate this restriction by estimating the "probability of a model" for the most likely models suggested by the reservoir geology. Lithologic variables are then simultaneously estimated from response equations for each model and combined in accordance with the probability of each respective model.

The initial application of the proposed approach has been the recognition of clays in the presence of calcite, quartz, dolomite, feldspar, anhydrite, or salt. This application has been realized through utilization of natural gamma-ray spectra, photoelectric effect, bulk density, and neutron porosity information.

For each specified model, response equations and parameter selections are obtained through utilization of the thorium versus potassium crossplot and the apparent matrix density versus apparent volumetric photoelectric cross-section crossplot. The thorium and potassium response equations are used to estimate the volumes of clay and feldspar. The apparent matrix density and volumetric cross-section response equations can then readily be corrected for the presence of clay and feldspar. However, a test is applied to ensure that the clay correction lies within limits consistent with the assumed lithology model. If inconsistency is detected, either the estimate of clay volume or the lithology model, or both, are changed.

A computer program has been written to test the proposed approach. Initial field testing in the U.S.A. and Canada has been completed and the program appears to recognize, with minimal analyst intervention, illite, chlorite, and a third clay which is considered to be a kaolinite-montmorillonite mixture. Results from the field are presented for an assortment of wells with varying lithologies.

End_of_Article - Last_Page 1446------------