Machine Learning Introduction to Classify the Erosive Wear Degree of the Compressor Stage Blades

Authors: Blinov V.L., Deryabin G.A., Zubkov I.S. Published: 13.01.2024
Published in issue: #4(147)/2023  

DOI: 10.18698/0236-3941-2023-4-88-105

Category: Power Engineering | Chapter: Turbomachines and Combination Turbine Plants  
Keywords: erosive wear, technical condition, axial compressor, blade apparatus, machine learning, numerical simulation, gas turbine unit


The problem of predicting erosion level of an axial compressor stage based on its operating parameters was solved using the machine learning methods. Current state of erosion research using the machine learning methods was reviewed, and the approach to solving the problem was proposed. Program code was developed in the Python 3 language to study the models and features applicability. The stage operation parameters obtained as a result of the numerical experiment were used as the initial data. Five erosion degrees were studied, they were determined by changing the flow pass geometry in accordance with the known laws of the erosion wear distribution. Principles are proposed for formation of the parameter-features sets explaining the erosion degree. To determine the erosion degree, simple machine learning models were used to solve the classification problems presented in the Scikit-learn Python library. Best accuracy of the study results when using pressure and temperature ratios, as well as mass flow through the axial compressor, was 0.82 (maximum one unit). When using the parameters that were actually measured in real operation, the accuracy dropped to 0.76. The random forest model showed the best results. Study results could be introduced in design and development of the compressor diagnostic systems

This work was supported by the RSF (project no. 22-79-00169)

Please cite this article in English as:

Blinov V.L., Deryabin G.A., Zubkov I.S. Machine learning introduction to classify the erosive wear degree of the compressor stage blades. Herald of the Bauman Moscow State Technical University, Series Mechanical Engineering, 2023, no. 4 (147), pp. 88--105 (in Russ.). DOI: https://doi.org/10.18698/0236-3941-2023-4-88-105


[1] Burnes D., Kurz R. Performance degradation effects in modern industrial gas turbines. Proc. Zurich Global Power and Propulsion Forum, 2018, pp. 3--9.

[2] Sallee G.P. Performance deterioration based on existing (historical) data. NASA report NASA-CR-135448. NASA Lewis Research Center, 1978.

[3] Blinov V.L., Zubkov I.S., Iskortsev E.Yu., et al. Special aspects of numerical simulation of a two-stage axial-flow compressor with defective blades. Vestnik Samarskogo universiteta. Aerokosmicheskaya tekhnika, tekhnologii i mashinostroenie [Vestnik of Samara University. Aerospace and Mechanical Engineering], 2021, vol. 20, no. 4, pp. 7--19 (in Russ.). DOI: https://doi.org/10.18287/2541-7533-2021-20-4-7-19

[4] Gumerov A.V., Akmaletdinov R.G. Compressor blade erosion modeling. Vestnik Samarskogo gosudarstvennogo aerokosmicheskogo universiteta im. Akademika S.P. Koroleva [Vestnik of Samara State Aerospace University Named After Academician S.P. Korolev], 2011, no. 3-2, pp. 233--239 (in Russ.).

[5] Ghenaiet A., Tan S.C., Elder R.L. Experimental investigation of axial fan erosion and performance degradation. Proc. Inst. Mech. Eng. A, 2004, vol. 218, no. 6, pp. 437--450. DOI: https://doi.org/10.1243/0957650041761900

[6] Dvirnik Ya.V., Pavlenko D.V. Evaluation limit state of helicopter GTE compressor blades by frequency of natural vibrations. Aviatsionno-kosmicheskaya tekhnika i tekhnologiya, 2016, no. 7, pp. 79--83 (in Russ.).

[7] Wilcox M., Kurz R., Brun K. Technology review of modern gas turbine inlet filtration systems. Int. J. Rotating Mach., 2012, vol. 2012, art. 128134. DOI: https://doi.org/10.1155/2012/128134

[8] Efanov V.S., Prokopenko A.N., Ovchinnikov A.V., et al. Erosion resistance of helicopter GTE compressor blades with different types of coatings. Vestnik dvigatelestroeniya, 2017, no. 1, pp. 120--123 (in Russ.).

[9] Nochovnaya N.A., Nikitin Ya.Yu. The current state of the question in the field of cleaning of the compressor GTE from operational pollution (review). Trudy VIAM [Proceedings of VIAM], 2017, no. 3 (in Russ.). DOI: https://dx.doi.org/10.18577/2307-6046-2017-0-3-5-5

[10] Liu Y., Ravichandran R., Chen K., et. al. Application of machine learning to solid particle erosion of aps-TBC and EB-PVD TBC at elevated temperatures. Coatings, 2021, vol. 11, no. 7, art. 845. DOI: https://doi.org/10.3390/coatings11070845

[11] Bahrainian S., Bakhshesh M., Hajidavalloo E., et al. A novel approach for solid particle erosion prediction based on Gaussian Process Regression. Wear, 2020, vol. 466--467, art. 203549. DOI: https://doi.org/10.1016/j.wear.2020.203549

[12] Zahedi P., Parvandeh S., Asgharpour A., et al. Random forest regression prediction of solid particle Erosion in elbows. Powder Technol., 2018, vol. 338, pp. 983--992. DOI: https://doi.org/10.1016/j.powtec.2018.07.055

[13] Giorgi M.G., Campilongo S., Ficarella A. A diagnostics tool for aero-engines health monitoring using machine learning technique. Energy Procedia, 2018, vol. 148, pp. 860--867. DOI: https://doi.org/10.1016/j.egypro.2018.08.109

[14] Chai T., Draxler R.R. Root mean square error (RMSE) or mean absolute error (MAE)? --- arguments against avoiding RMSE in the literature. Geosci. Model Dev., 2014, vol. 7, no. 3, pp. 1247--1250. DOI: https://doi.org/10.5194/gmd-7-1247-2014

[15] Joshuva A., Sugumaran V. A machine learning approach for condition monitoring of wind turbine blade using autoregressive moving average (ARMA) features through vibration signals: a comparative study. Prog. Ind. Ecol., 2018, vol. 12, no. 1-2, pp. 14--34. DOI: https://doi.org/10.1504/PIE.2018.095867

[16] Enriquez Zarate J., Gomez Lopez M.A., Carmona Troyo J.A. Analysis and detection of erosion in wind turbine blades. Math. Comput. Appl., 2022, vol. 27, no. 1, art. 5. DOI: https://doi.org/10.3390/mca27010005

[17] Reid L., Moore R.D. Design and overall performance of four highly loaded, high-speed inlet stages for an advanced high-pressure-ratio core compressor. NASA report NASA-TP-1337. NASA Lewis Research Center, 1978.

[18] Pedregosa F., Varoquaux G., Gramfort A., et al. Scikit-learn: machine learning in Python. J. Mach. Learn. Res., 2011, vol. 12, pp. 2825--2830.

[19] Denton J.D. Lessons from rotor 37. J. of Therm. Sc., 1997, vol. 6, no. 1, pp. 1--13. DOI: https://doi.org/10.1007/s11630-997-0010-9

[20] Blinov V.L., Deryabin G.A., Zubkov I.S. Classification of the gas path erosion level of the insulated stage of the axial compressor. Energeticheskie sistemy [Energy Systems], 2022, vol. 7, no. 1, pp. 8--18 (in Russ.).

[21] Dvirnyk Y., Pavlenko D., Przysowa R. Determination of serviceability limits of a turboshaft engine by the criterion of blade natural frequency and stall margin. Aerospace, 2019, vol. 6, no. 12. DOI: https://doi.org/10.3390/aerospace6120132

[22] Vogel A., Durant A., Cassiani M., et al. Simulation of volcanic ash ingestion into a large aero engine: particle--fan interactions. J. Turbomach., 2019, vol. 141, no. 1, art. 011010. DOI: https://doi.org/10.1115/1.4041464

[23] Potapov V.A., Sanko A.A. Performance simulation of multi-stage axial-flow compressor of turbo-shaft engine with account for erosive wear nonlinearity of its blades. Nauchnyy Vestnik MGTU GA [Civil Aviation High Technologies], 2020, vol. 23, no. 5, pp. 39--53 (in Russ.). DOI: https://doi.org/10.26467/2079-0619-2020-23-5-39-53

[24] Lever J., Krzywinski M., Altman N. Classification evaluation. Nat. Methods, 2016, vol. 13, no. 8, pp. 603--604. DOI: https://doi.org/10.1038/nmeth.3945

[25] Han H., Guo X., Yu H. Variable selection using mean decrease accuracy and mean decrease gini based on random forest. IEEE ICSESS, 2016, pp. 219--224. DOI: https://doi.org/10.1109/ICSESS.2016.7883053

[26] Lix L.M., Keselman J.C., Keselman H.J. Consequences of assumption violations revisited: a quantitative review of alternatives to the one-way analysis of variance F test. Rev. Educ. Res., 1996, vol. 66, no. 4, pp. 579--619. DOI: https://doi.org/10.3102/00346543066004579