Early fault detection of single-point rub in gas turbines with accelerometers on the casing based on continuous wavelet transform

• Published in: Journal of sound and vibration Vol. 487; p. 115628
• Main Authors: Silva, Alejandro, Zarzo, Alejandro, Machuca González, Jacobo Manuel, Munoz-Guijosa, Juan Manuel
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier Ltd 24.11.2020; Elsevier Science Ltd
• Subjects: Accelerometers; Algorithms; Continuous wavelet transform; Damage prevention; Early fault diagnosis; Fault detection; Fault diagnosis; Fourier transforms; Gas turbines; Industrial applications; Machinery; Machinery condition monitoring; Malfunctions; Post-processing; Process parameters; Rotating machinery; Rotor-casing rub; Rotor-dynamics; Rotors; Signal processing; Wavelet transforms; Accelerometers; Early fault diagnosis; Rotor-dynamics; Rotor-casing rub; Continuous wavelet transform
• ISSN: 0022-460X; 1095-8568
• DOI: 10.1016/j.jsv.2020.115628

The diagnosis of rotor-casing rub, one of the most common malfunctions in rotating machinery, is usually straightforward in machines equipped with proximitors measuring relative rotor-casing displacement. However, constructive reasons, aeroderivative gas turbines are frequently equipped with accelerometers instead of proximitors. Since these turbines are increasingly used in many industrial applications, this poses substantial maintenance and safety problems. In this work, we apply a properly designed method based on Continuous Wavelet Transform (CWT) to the detection of early single–point rub in gas turbines by means of accelerometers on the machine casing. This procedure is validated on an experimental rotor rig with a flexible cylindrical casing. With this method we expect to prevent machine damage or failure by supplying a very clear picture of rotor-casing rub from the very onset. This is monitored under diverse conditions with accelerometers on casing points and with proximitors assembled between rotor shaft and casing. The acceleration signals collected are then post-processed and analyzed using the Discrete Fourier Transform (DFT) and the proposed CWT-based method. The signal post-processing parameters of the signal post-processing were adjusted with the help of a computational model of the experimental rig. We prove that an adequately designed CWT-based algorithm outlines the signal features that better characterize single-point rub, and unlike the DFT it is sensitive to an outbreak of rub within a time series.