METHODOLOGY FOR AUTOMATIC FAULT RECOGNITION OF ROD-WELL PUMPING UNITS

Abstract

One of the most effective ways of monitoring is dynamometry. Dynamograms are used to diagnose and monitor the operation of the recognition of rod-well pumping units  (RDPU). A new approach to detecting and detecting failures, malfunctions in the operation of the RDPU system, excluding the human factor. The use of Fourier descriptors for the recognition of dynamometric maps and machine learning techniques for the classification of various pump conditions is proposed. With the help of the Fourier descriptor, it is  possible to predict and diagnose malfunctions of the downhole pump. These descriptors simplify, normalize, and describe 
each map well. The proposed method is trained using data from real dynamometric maps.