A Fast and Transparent Bayesian Log Interpretation Method

Abstract

This paper presents a fast and transparent Bayesian-based computational method that can be used for the interpretation of well logs. Rather than single values for formation properties, the results are probability distributions. The method is fast because the responses of logging tools, for a large set of formation realizations, are calculated only once and then stored with the realizations in a database. Thereafter, only this database is used for the interpretation of real well logs. The method is transparent because it entirely relies on selecting formation realizations with calculated log responses that are, within a given error margin, equal to the real logging tool responses. Therefore, only internally consistent interpretations can be found. The size of the database is realizable and the log evaluation computation time sufficiently short. The method has been tested with good results on more than 100 wells. Some of the well results are presented and discussed. The presented evaluations use databases of about 300 MB containing about 25 million realizations. Database construction and whole well evaluation each require less than a few seconds.