(50-4) 12 * << * >> * Russian * English * Content * All Issues

Neural network model for classifying dynamometer cards of sucker-rod pumping units
A.M. Vulfin1, K.F. Tagirova1, A.D. Kirillova1, A.E. Sulavko2, P.S. Lozhnikov2

1Ufa University of Science and Technology, Zaki Validi Str. 32, Ufa, 450076, Russia;
2Omsk State Technical University, Prospekt Mira 11, Omsk, 644050, Russia

  Full text (PDF)

DOI: 10.18287/COJ1940

Article ID: 1940

Language: Russian

Abstract:
The work is devoted to the problem of increasing the efficiency of recognizing characteristic signs of complex faults in images of dynamometer charts of submersible pumping equipment under conditions of limited expert marking, pronounced class imbalance and the presence of measurement noise. A set of 20164 practical dynamometer charts, divided into 10 classes, is created. A systematic comparison of six neural network architectures (ResNet-18, ResNet-50, EfficientNet-B0, Vision Transformer, ConvNeXt) and four methods of representing dynamometer charts as images (contour representation, continuous wavelet transform scalogram, signed transform of distances to the contour, and their combination) is performed. The basic approach used is based on manual feature construction (Freeman chain code, approximation by geometric primitives) with classification using the CatBoost gradient boosting method. Model training and evaluation are performed using a 5-fold stratified cross-validation protocol with imbalance compensation via loss function weighting. It is shown that representations with a high proportion of informative pixels (scalogram, signed distance map) provide more stable learning and outperform sparse contour representation, especially for architectures with attention mechanisms. The best quality is achieved by combining the ResNet-18 model and scalogram (F1-macro = 0.831 with Accuracy = 0.899). The scientific novelty lies in the quantitative assessment of the influence of the type of dynamometer chart representation on the efficiency of different families of models under a single training protocol and a pronounced class imbalance. The practical significance lies in increasing the recognition accuracy of dynamometer chart images with complex faults and reducing the labor intensity of manual marking.

Keywords:
dynamometer card, image classification, convolutional neural networks, wavelet transform, class imbalance.

Acknowledgements:
The research was financially supported by the Ministry of Science and Higher Education of the Russian Federation under project FSGF-2023-0004.

Citation:
Vulfin AM, Tagirova KF, Kirillova AD, Sulavko AE, Lozhnikov PS. Neural network model for classifying dynamometer cards of sucker-rod pumping units. Computer Optics 2026; 50(4): 1940. doi: 10.18287/COJ1940.

References:

  1. Li K, Gao X-W, Yang W-B, Dai Y-L, Tian Z-D. Multiple fault diagnosis of down-hole conditions of sucker-rod pumping wells based on Freeman chain code and DCA. Pet Sci 2013; 10: 347-360. DOI:10.1007/s12182-013-0283-4.
  2. Li K, Gao X-W, Tian Z, Qiu Z. Using the curve moment and the PSO-SVM method to diagnose downhole conditions of a sucker rod pumping unit. Pet Sci 2013; 10: 73-80. DOI:10.1007/s12182-013-0252-y.
  3. Abdalla R, El Ela MA, El-Banbi A. Identification of Downhole Conditions in Sucker Rod Pumped Wells Using Deep Neural Networks and Genetic Algorithms. SPE Prod Oper 2020; 35: 435-447. DOI:10.2118/200494-PA.
  4. Cheng H, Yu H, Zeng P, Osipov E, Li S, Vyatkin V. Automatic Recognition of Sucker-Rod Pumping System Working Conditions Using Dynamometer Cards with Transfer Learning and SVM. Sensors 2020; 20: 5659. DOI:10.3390/s20195659.
  5. Dong G, Li W, Dong Z, Wang C, Qian S, Zhang T, Ma X, Zou L, Lin K, Liu Z. Enhancing Dynagraph Card Classification in Pumping Systems Using Transfer Learning and the Swin Transformer Model. Appl Sci 2024; 14: 1657. doi:10.3390/app14041657.
  6. Tan C, Chen P, Feng Z, Ai X, Lu M, Zhou Q, Feng G. Multi-Scale Normalization Method Combined With a Deep CNN Diagnosis Model of Dynamometer Card in SRP Well. Front Earth Sci 2022; 10: 852633. DOI:10.3389/feart.2022.852633.
  7. Yuan C, Wu W, Li X. Dynamometer card generation for pumping units based on CNN and electrical parameters. Sci Rep 2024; 14: 18657. DOI:10.1038/s41598-024-69516-y.
  8. Tan X, Wang S, Wu H, Chen F. Enhancing pumping unit diagnosis with similarity splicing data augmentation and wavelet denoising. Sci Rep 2025; 15: 37045. DOI:10.1038/s41598-025-20819-8.
  9. Mihajlov AG, Shubin SS, Alferov AV, Imashev RN, Yamaliev VU. Improvement of efficiency of diagnostics of rod pumps with use of deep neural networks. Neftyanoye khozyaystvo 2018; 9: 122-126. DOI:10.24887/0028-2448-2018-9-122-126.
  10. Volkov MG, Silnov DV, Topolnikov AS, Latypov BM, Katermin AV, Enikeev RM. Automated system for interpreting technical condition from dynamograms based on machine learning tools. Neftyanoye khozyaystvo 2021; 4: 102-105. DOI:10.24887/0028-2448-2021-4-102-105.
  11. Martinović A, Bijanić M, Danilović D, Petrović A, Delibasić B. Unveiling deep learning insights: a specialized analysis of sucker rod pump dynamographs, emphasizing visualizations and human insight. Mathematics 2023; 11: 4782. DOI:10.3390/math11234782.
  12. Sun Y, Wang H, Zhang X, Yang J, Wang C, Shao J, Li Y. Comparative Analysis and Suggestions for Sucker Rod Pump Working Condition Diagnosis Using Machine Learning Techniques. SPE Asia Pacific Oil and Gas Conf and Exhib. SPE 2024; D021S012R008.
  13. Gumerova VI, Korobkov GE. Assessment of the technical condition of oil industrial equipment using digital technologies. Obshchestvo 2023; 3(30): 12-16.
  14. Danilov SO. Identification of complications and malfunctions of submersible equipment of sucker rod pumping units using neural networks [In Russian]. Molodoy uchenyy 2019; 15(253): 17-22.
  15. Sharaf SA. Beam pump dynamometer card prediction using artificial neural networks. KnE Eng 2018; 198-212.
  16. Wang X, He Y, Li F, Wang Z, Dou X, Xu H, Fu L. A Working Condition Diagnosis Model of Sucker Rod Pumping wells Based on Deep Learning. SPE Prod Operations 2021; 36(2): 317-326. DOI:10.2118/205015-PA.
  17. Prokhorenkova L, Gusev G, Vorobev A, Dorogush AV, Gulin A. CatBoost: unbiased boosting with categorical features. Advances in Neural Information Processing Systems 2018; 31: 6638-6648.

151, Molodogvardeiskaya str., Samara, 443001, Russia; E-mail: journal@computeroptics.ru; Tel: +7 (846) 242-41-24 (Executive secretary), +7 (846) 332-56-22 (Issuing editor), Fax: +7 (846) 332-56-20