Automated system for interpreting technical condition from dynamograms based on machine learning tools

UDK: 681.518:622.276.58

DOI: 10.24887/0028-2448-2021-4-102-105

Key words: oil production, sucker rod pumps, dynamogram, machine learning, diagnostics

Authors: M.G. Volkov (RN-BashNIPIneft LLC, RF, Ufa), D.V. Silnov (RN-BashNIPIneft LLC, RF, Ufa), A.S. Topolnikov (RN-BashNIPIneft LLC, RF, Ufa), B.M. Latypov (RN-BashNIPIneft LLC, RF, Ufa), A.V. Katermin (Bashneft PJSOC, RF, Ufa), R.M. Enikeev (Bashneft PJSOC, RF, Ufa)

The article presents the results of work on the development of an automated system for interpreting deviations from dynamograms based on machine learning tools. The work contains the results of factor analysis of the reasons affecting the accuracy of the dynamometer recording of the sucker rod pump and the reasons affecting the accuracy of the dynamogram interpretation models and the principle of the implementation of the tool for recognizing deviations in work dynamometer sucker rod pump. It has been shown that the accuracy of a dynamogram is influenced by many factors, such as: the state of the polished rod (dimensions) due to deviations caused by abrasion and wear, the deviation of the elastic modulus of the steel grade of the polished rod from the calculated value, the deviation of the Poisson coefficient, and the error from temperature drift by the device itself. It is shown that the quality of the implemented machine learning model will be affected by: the quality of the training sample and the test sample (the number of erroneous interpretations in the samples); prognostic ability of the model itself. The scheme of operation of the system for interpreting deviations from dynamograms and the results of assessing the quality of the developed models are presented. For the model of binary classification of dynamograms, the Fisher metric was 97%, for the multiclass model - 82%, for the multilable model - 87%. The developed automated system for interpreting deviations from dynamograms based on machine learning tools is integrated into the decision support system implemented as part of R&D project “Operational Service” Bashneft PJSOC. The system allows you to quickly identify simultaneously several types of deviations in the dynamogram.
References
1. Bakhtizin R.N., Urazakov K.R., Latypov B.M. et al., The influence of regular microrelief forms on fluid leakage through plunger pair of sucker rod pump (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2017, no. 4, pp. 113–116.
2. Urazakov K.R., Latypov B.M., Ishmukhametov B.K., Experimental studies of the influence of configuration of regular microrelief of plunger surface on sucker-rod pump delivery, Chemical and Petroleum Engineering, 2018, V. 54, no. 3–4, pp. 172–176.
3. Urazakov K.R., Latypov B.M., Ishmukhametov B.Kh., Study of the influence of form of regular microrelief of the plunger on the output flow of sucker rod pump (In Russ.), Khimicheskoe i neftegazovoe mashinostroenie, 2018, no. 3, pp. 23–25.
4. Yamaliev V.U., Ishemguzhin I.E., Latypov B.M., Friction assessment plunger to barrel of sucker rod pump in design rod string (In Russ.), Izvestiya Samarskogo nauchnogo tsentra RAN, 2017, V. 19, no. 1, pp. 70–75.
5. Mansafov R.Yu., A new approach to the diagnosis of sucker rod pumps work on the dynamometer card (In Russ.), Inzhenernaya praktika, 2010, no. 9, pp. 82–89.
6. Urazakov K.R., Latypov B.M., Komkov A.G., Davletshin F.F., Calculation of the theoretical dynamogram of a differential sucker-rod pump for the production of high-viscosity oil (In Russ.), Oborudovanie i tekhnologii dlya neftegazovogo kompleksa, 2017, no. 4, pp. 41–47.
7. Deng L., The MNIST database of handwritten digit images for machine learning research, IEEE Signal Processing Magazine, 2012, V. 29, no. 6, pp. 141–142.
8. Wu B. et al., Multi-label learning with missing labels for image annotation and facial action unit recognition, Pattern Recognition, 2015, V. 48, no. 7, pp. 2279–2289.
9. Madjarov G. et al., An extensive experimental comparison of methods for multi-label learning, Pattern recognition, 2012, V. 45, no. 9, pp. 3084–3104.
10. Aliev T.A., Rzayev A.H., Guluyev G.A. et al., Robust technology and system for management of sucker rod pumping units in oil wells, Mechanical Systems and Signal Processing, 2018, V. 99, pp. 47–56.
11. Mikhaylov A.G., Shubin S.S., Alferov A.V. et al., Improvement of efficiency of diagnostics of rod pumps with use of deep neural networks (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2018, no. 9, pp. 122–126.
12. Bakhtizin R.N., Urazakov K.R., Timashev E.O., Belov A.E., A new approach of quantifying the technical condition of rod units with the solution of inverse dynamic problems by multidimensional optimization methods (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2019, no. 7, pp. 118–122.
13. Li K., Xianwen G., Zhongda T., Zhixue Q., Using the curve moment and the PSO-SVM method to diagnose downhole conditions of a sucker rod pumping unit, Petroleum Science, 2013, V. 10, pp. 73–80.
14. Li K., Xianwen G., Zhou H.B., Han Y., Fault diagnosis for down-hole conditions of sucker rod pumping systems based on the FBH-SC method, Journal of Petroleum Science and Engineering, 2015, V. 12, pp. 135–147.