Russian Federation
Russian Federation
Russian Federation
Purpose: To develop an objective approach to determine the most effective wavelets for specific types of diesel engine vibration signals based on quantitative criteria, with the possible further application in locomotive diagnostic systems. The research aims to create a universal methodology for comparative evaluation of different wavelets, which will be an important step in the development of modern vibration diagnostic methods. Methods: Experimentally obtained vibration signals using a three-position sensor on a diesel engine followed by data processing with wavelet packet transformation in the Python programming environment. The effectiveness of 35 different wavelets from the Daubechies, Coiflets, Biorthogonal, Reverse Biorthogonal, Symlets and Haar families was evaluated according to two objective criteria: minimum mean square error (MSE) of the reconstruction and maximum spectral energy of the signal. Comparative analysis was performed using a specially developed algorithm for ranking and selecting optimal wavelets based on parallel evaluation against both criteria. Results: A universal method for selecting the optimal mother wavelet function for the analysis of diesel engine vibration signals has been presented. For the studied signal, the wavelets bior3.3, bior3.5 and bior3.7 were found to be optimal according to both criteria, offering the best balance between reconstruction accuracy and preservation of energy characteristics. The results obtained allow a quantitative justification of the choice of wavelet for a specific type of signal, increasing the reliability of the diagnosis. Practical significance: The developed methodology allows the objective selection of optimal mother wavelets for wavelet packet transformation of vibration signals, which will increase the efficiency of vibration diagnostics of locomotive diesel engines and will enable the improvement of existing intelligent systems for technical condition monitoring with the possibility of early fault detection and prediction of remainin g useful life.
Diesel engine, vibration diagnostics, wavelet transformation, wavelet packet, mean squared error, spectral energy, signal decomposition
1. Malla S. Veyvlety v obrabotke signalov. — M.: Mir, 2005. — 671 s.
2. Dobeshi I. Desyat' lekciy po veyvletam. — Izhevsk: RHD, 2001. — 464 s.
3. Lee G. R. PyWavelets: A Python package for wavelet analysis / G. R. Lee, R. Gommers, F. Wasilewski, K. Wohlfahrt et al. // Journal of Open Source Software. — 2019. — Vol. 4(36). — P. 1237. — DOI:https://doi.org/10.21105/joss.01237.
4. Vorob'ev V. I. Teoriya i praktika veyvlet-preobrazovaniya / V. I. Vorob'ev, V. G. Gribunin. — SPb.: VUS, 1999. — 204 s.
5. Mozharov G. P. Sravnitel'nyy analiz adaptivnyh algoritmov veyvlet-paketov / G. P. Mozharov // Vestnik Moskovskogo gosudarstvennogo tehnicheskogo universiteta im. N. E. Baumana. Seriya «Priborostroenie». — 2016. — № 1(106). — S. 75–88.
6. Kewei O. Adaptive Multi-Scale Wavelet Neural Network for Time Series Classification / O. Kewei, H. Yi, Z. Shilin, Z. Ye // Information an International Interdisciplinary Journal. — 2021. — T. 12. — Iss. 6. — P. 252. — DOI:https://doi.org/10.3390/INFO12060252.
7. Yu H. AdaWaveNet: Adaptive Wavelet Network for Time Series Analysis / H. Yu, G. Peikun, S. Akane // arXiv preprint arXiv:2405.11124. — 2024. — DOI: 10.48550/ arXiv.2405.11124.
8. Wolter M. Adaptive wavelet pooling for convolutional neural networks / M. Wolter, J. Garcke // Proceedings of Machine Learning Research. — 2021. — Vol. 130. — Pp. 1–12.
9. Oppenheim A. V. Discrete-Time Signal Processing / A. V. Oppenheim, R. W. Schafer. — 3rd ed. — Prentice Hall, 2009. — 1144 p.
10. Antonini M. Image coding using wavelet transform / M. Antonini, M. Barlaud, P. Mathieu, I. Daubechies // IEEE Transactions on Image Processing. — 1992. — Vol. 1. — Iss. 2. — Pp. 205–220.
11. Oppenheim A. V. Discrete-Time Signal Processing / A. V. Oppenheim, R. W. Schafer. — 3rd ed. — Prentice Hall, 2009. — 1144 p.
12. Grachev V. V. Diagnostirovanie elektronnogo oborudovaniya lokomotiva s ispol'zovaniem intellektual'nyh klassifikatorov / V. V. Grachev, M. V. Fedotov, A. V. Grischenko i dr. // Eltrans 2023: sbornik trudov XI Mezhdunarodnogo simpoziuma, Sankt-Peterburg, 31 maya — 2 iyunya 2023 goda / Peterburgskiy gosudarstvennyy universitet putey soobscheniya Imperatora Aleksandra I. — Sankt-Peterburg: OOO «IPK «NP-Print», 2023. — S. 134–145.
