--> --> Application of an Assisted, Analogical Reasoner to Exploration and Production, by Anthony Gary, Arun K. Majumdar, William Full, and John Sowa; #90052 (2006)

Datapages, Inc.Print this page

Application of an Assisted, Analogical Reasoner to Exploration and Production

Anthony Gary1, Arun K. Majumdar2, William Full3, and John Sowa2
1 University of Utah, Salt lake City, UT
2 VivoMind Intelligence, Inc, Alexandria, VA
3 Tramontane Inc, Salt Lake City, UT

Reasoning by analogy is a cornerstone of problem solving. This is acutely so in petroleum geology where field work is routinely conducted to understand the depositional analogues of new prospects and where a decade or more of discipline-specific training/expertise is needed. Drilling decisions are made based on applying this experience to new problem situations. There has been a significant net loss of geoscientists that are experienced in E&P, and many of the knowledge management (KM) practices presently being employed offer only partial solutions. They focus on the preservation and transfer of existing knowledge elements, such as reports, charts and data but not on the processes critical to creating new knowledge. In E&P new insights and knowledge are often derived from analogies with past experience and performed by human assets. As these assets become scarce, it is likely that knowledge creation will diminish in E&P. We apply a conceptual-graph based analogy finder, developed by one of the authors, in the discovery of new E&P analogues. These graphs are constructed based on a variety of sources (e.g., text, graphs, data bases) common to E&P and that are the basic elements of many KM strategies being currently employed. The conceptual graphs can then be characterized by their knowledge fingerprints in order to find analogies by structure mapping these fingerprints. Mapping, however, produces multiple analogies (i.e., by matching labels, matching sub-graphs and combining nodes and fingerprints), so a weight of evidence is computed using heuristics that estimate the closeness of the matches. This methodology can be used to explore for analogies (unsupervised) or for case-based reasoning (supervised), and is suitable for a wide range of E&P applications.