Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
Detecting and localizing damage in an active automotive suspension using machine learning methods and a filter bank
Submission Author:
Bruno Santos , SP
Co-Authors:
Bruno Santos, Fabíola Martins Campos de Oliveira, Murilo Loiola, Helói Genari
Presenter: Bruno Santos
doi://10.26678/ABCM.COBEM2023.COB2023-1473
Abstract
The structural health monitoring area has received increasing attention from the industry and the scientific community in recent decades. Several techniques applied to damage detection and localization have been proposed, including recent developments based on artificial intelligence. In this context, this paper investigates a framework that combines a machine learning method with a residue generator to detect and localize multiple damaged regions. The residue signal is the output of a filter bank, built with Luenberger observers. This type of estimation allows a direct relation between the residue signals and the changes in dynamics caused by damage. However, the correlation between the residue signal and the damage localization is not straight. For this purpose, this paper analyzes the performance of three classifiers: k-nearest neighbors, a decision tree, and a support vector machine to identify damage occurrence and its respective localization using the temporal residue data. An active automotive suspension model is used as a case study structure, generating the temporal data to build the classifiers. Simulated results show that the support vector machine method is the best tested classifier to identify changes in the suspension components, achieving a classification accuracy between 79.6% and 100%.
Keywords
Structural Health Monitoring (SHM), Artificial Intelligence, Active suspension, filter bank
