FAULT DETECTION OF A PLANETARY GEAR SYSTEM BASED ON NON-LINEAR DYNAMIC MODELING AND VIBRATION SIGNALS VIA NON-STATIONARY TIME SERIES MODELS

Date

2023-3-31

Author

Hosseiniaghdam, Behrang

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

115
views

24
downloads

Cite This

The current study aims to investigate the fault diagnosis problem in a gear system using non-stationary time series models and non-linear dynamic modeling. Fault diagnosis of gearboxes is an ongoing and significant research topic in the context of condition monitoring. Various gears and gearboxes are used in machinery found in different industries and vehicles. During manufacturing, detecting gearbox faults is also an important task. Faults such as tooth profile error, helix angle error (of helical gears), and assembly errors are of the faults as mentioned earlier. Some other faults occur when the machinery is operating and if they are not prognosed in advance they can finally result in catastrophic failures. Instances of such faults include gear tooth crack, surface pitting, and spalling. For fault detection, first, a non-linear dynamic model, including tooth root crack, is developed. Then the dynamic model with fault is verified by using the results given in the available literature. In the next step, fault detection using experimental data is carried out. The process starts with the analysis of the vibration signals measured from a test setup to identify the fault features in the frequency spectrum via Fast Fourier Transform (FFT). Before FFT analysis, the signals are averaged via Time Synchronous Averaging (TSA) method. Subsequently, the TSA signals are modeled via a non-stationary time series model called Functional Time Series Time Dependent Autoregressive Moving Average (FS-TARMA) and its another form called FS-TAR. The developed method of fault detection, utilizes the identified models for vibration analysis and the estimation of Power Spectral Densities to evaluate fault effects in the time-frequency domain. Subsequently, a fault detection and localization algorithm is developed by comparing models associated with healthy and faulty gearboxes. Finally, the experimental results, as well as theoretical results, are analyzed by the use of the developed method to demonstrate its applicability and effectiveness.

Subject Keywords

Gear Dynamics; Non-linear vibration; Vibration signal analysis; Fault detection; Time series models

URI

https://hdl.handle.net/11511/103095

Collections

Graduate School of Natural and Applied Sciences, Thesis