Improvement of the on-board vibration control system of an aviation gas turbine engine based on multi-level processing of vibration signals

Abstract

This paper proposes and substantiates the improvement of the on-board vibration control system of aircraft engines to expand the system’s functional capabilities and provide multi-class diagnostics based on multi-level processing of vibration signals. The possibility of diagnosing initial crack-like damage to the rotor shaft and impeller blade is considered, as well as the possibility of detecting such operational violations of the engine's regular modes, such as minor imbalance, ingress of small and medium-sized foreign objects, and grinding of the impeller blades. Such damages and disturbances do not lead to increased vibration of the rotary harmonics and are not detected by traditional analysis methods. For multi-class diagnostics, multi-level processing of vibration signals based on various combinations and/or sequential application of time-frequency analysis, multi-spectral analysis, and fractal analysis methods were used. This work proposes the structure of a subsystem of multi-class diagnostics as part of the standard on-board engine vibration control system in steady-state and transient operating modes, as well as algorithms for detecting initial crack-like damages of rotating elements and disturbances in standard operating modes. As diagnostic signs, the following are used: the Hurst index of the vibration signal in the region of subharmonic resonance in the transient mode (for the diagnosis of the initial transverse crack of the rotor shaft); dimension of Minkowski time-frequency spectra of vibration signals in stationary and transient modes (for diagnosing the initial crack of the impeller blade); simultaneous use of the Hurst index of samples of vibration signals and dimension of Minkowski estimates of the module of the bispectrum of vibration signals (for diagnosing violations of standard operating modes). The developed software algorithms use received and pre-processed signals from vibration and engine speed sensors, provide the determination of the mode of vibration disturbance, perform time-frequency, bispectral, and fractal analysis in the sequence determined for each diagnostic task, calculate the value of the diagnostic feature and compare it with the reference value, based on the results of the comparison, establish the appropriate diagnosis and submit it to the decision-making system. The proposed algorithms are intended for use in the corresponding blocks of the multi-class engine diagnostic subsystem.