A hierarchy of pattern recognition algorithms for the diagnosis of sucker rod pumped wells

A useful pattern recognition method for sucker rod pumping diagnosis is presented in this paper. It has been developed in a hierarchical manner using a collection of very simple algorithms. To accurately and quickly diagnose the conditions at the sucker rod string, subsurface pump, or both, a precomputed library of downhole dynamometer cards is used. The techniques used incorporate new ideas such as separated upstroke and downstroke diagnosis. They allow us to detect simple and multiple problem cases with various degrees of severity and occurring at different locations. A graphical system called AGE provides a visual display of the diagnosis on the computer screen. The proposed algorithms can be easily incorporated into a more general automated system.<<ETX>>