Eventos Anais de eventos
DINAME 2023
XIX International Symposium on Dynamic Problems of Mechanics
Malfunction Parameters Determination using Bayesian Neural Networks applied to a Multi-Fault Rotor
Submission Author:
Olympio Belli , SP
Co-Authors:
Olympio Belli, Helio Fiori de Castro
Presenter: Olympio Belli
doi://10.26678/ABCM.DINAME2023.DIN2023-0087
Abstract
In addition to the qualitative diagnosis of rotating machinery, it is of great importance to determine its severity and characteristics in order to provide sufficient information for its correction. In this sense, the present work aims to use Bayesian neural networks to develop meta-models for regression of these parameters. The methodology used simulated data from a multi-fault rotor. The displacement signals as a function of time in the four bearings were measured. After this, a full spectrum analysis was applied to extract the amplitude values of the six positive and negative harmonics. From these data, models were formulated for the regression of the following parameters: transverse shaft crack size, crack location, angle and parallel distance misalignment of the coupling. From the Bayesian model the uncertainties related to the predictions were extracted. Finally, the results were satisfactory given the expected physical behavior of the model.
Keywords
Bayesian neural networks, Parameter Regression, Angular and Paralell Misalignment, Crack Shaft, Multi-Fault Rotor
