Data mining of the Dynamometry of oil Production Sucker Rod Pumping Unit

This paper proposes an approach to diagnostics of the state of submersible pumping equipment. The use of a two-dimensional wavelet transform of a dynamogram based on the Morlet wavelet for generating features and subsequent construction of a classifier is described. The main steps of dynamometer data processing are considered in order to identify features and prepare training and test sets. A convolutional neural network is used as a classifier, which makes it possible to achieve an estimate of the f1 score = 0.91 on a test set. Neural network analysis of technological process parameters will allow to extract and interpret the features of each of the classes of equipment states. On this basis it is possible to build a database of early features and precursors of abnormal conditions, which will provide significant decision support in predictive analytics tasks.