Eventos Anais de eventos
DINAME 2023
XIX International Symposium on Dynamic Problems of Mechanics
The influence of proper orthogonal decomposition and smooth orthogonal decomposition on rotor unbalance fault classification
Submission Author:
Leonardo Valero Pereira , RJ
Co-Authors:
Leonardo Valero Pereira, Thiago Ritto
Presenter: Leonardo Valero Pereira
doi://10.26678/ABCM.DINAME2023.DIN2023-0033
Abstract
This work aims to classify different unbalance intensities located at the tip of gas turbine blades. The objective is to verify how the inclusion of proper (POD) and smooth orthogonal decomposition (SOD), as features of the machine learning problem, impact the accuracy of classifying the intensity of unbalance presented at the gas turbine. First, a full gas turbine model with previously identified bearings is set as a reference for test and training data set data generation. These rotordynamic models provided features related to displacement and velocity measured in both bearing nodes, which compose the base for the classification model under different unbalance conditions. Then, different groups of features were proposed to compare the influence of POD and SOD in classification results, which presented a gain of 3% compared with the models without these pair of features. Results demonstrated that both orthogonal decompositions provide different aspects of the physical phenomena, thus improving the classification of rotordynamic severity unbalance.
Keywords
Rotordynamics, Fault classification, Smooth Orthogonal Decomposition, Proper orthogonal decomposition (POD)
