An Approach to Detailed Dip Determination Using Correlation by Pattern Recognition

Publisher Website
Google Scholar
Abstract
This paper describes a new method for processing dipmeter logs. Each correlation curve is decomposed into a depth-ordered sequence of ranked elements (geological events reflected as curve features). These curve elements are correlated from curve to curve, with higher-quality correlations accepted first. This method has much greater resolution for definition of thin-bed boundaries than conventional techniques. Introduction Correlation of the variations in continuous recordings of a physical property measured in two nearby wells is basic to describe the subsurface of an oil field. If the rock sequence was perfectly stratified, correlation would be simple and immediate. A significant feature may exist on both records, yet one record may contain details that do not occur on the other record. The features on the two recordings also may have different amplitudes or different thicknesses. Thus, correlations can have different degrees of ambiguity or doubt. Geologists devote much knowledge and experience to correlating such records. Yet, correlation is time consuming and can become tedious. This is particularly true of the analysis of high-resolution dipmeter logs, where, instead of records in two nearby wells, we seek to correlate four records of microresistivity made at four azimuths around the same borehole. Even in such a reduced space, correlations are as difficult as in the case of different wells. Because stratification is complex, abundant, and variable, and because interpretations can vary significantly, ways have been sought to correlate dipmeter curves automatically by using computers. At present, machine computations of dipmeter results use "fixed-interval" correlations. The information provided is useful, but it is deficient in three ways: provided is useful, but it is deficient in three ways:the correlation interval is preset and is not adapted continuously to the geological context;the basic cause of the correlation (the bedding) is not described; andonly similar features of practically identical thickness can produce a high-quality correlation. Geodip processing overcomes those three objections by applying methods of pattern recognition to process the dipmeter surveys automatically. With this technique, each dipmeter curve to be correlated is decomposed mathematically into a depth-ordered sequence of ranked elements. Correlations of higher-ranking elements are made first. Subsequent correlations of lower-ranking elements are not allowed to cross correlations of higher-ranking events. The higher-ranking correlations guide the search for the lower-ranking ones, until all possible correlations are made. possible correlations are made.The resulting dips are recorded on a log display that includes the original dipmeter curves, arrow plots of the hole-deviation information and of the dip results, and, at periodic intervals, azimuth-frequency diagrams. Because periodic intervals, azimuth-frequency diagrams. Because of the increased number of dips available, this log has a depth scale of 1/40 [30 in./100 ft (1 m/40 m)] or 1/24 [50 in./100 ft (1 m/24 m)], as compared with the more usual 1/240 or 1/200 arrow-plot scales. Detailed information also is furnished on listings. Dipmeter Technique The downhole equipment of today's high resolution dipmeter tool (HDT) has two perpendicular, extendable pairs of arms. Each arm carries a pad, which is pairs of arms. Each arm carries a pad, which is hydraulically pressed against the wall of the hole during logging. JPT P. 232

This publication has 0 references indexed in Scilit: