Eventos Anais de eventos
DINAME 2023
XIX International Symposium on Dynamic Problems of Mechanics
Bolt loosening Detection Based in Data-Driven of Bolted Beam Connections by Support Vector Machine Method
Submission Author:
Jefferson Coelho , AM
Co-Authors:
Jefferson Coelho, Amanda Aryda Silva Rodrigues de Sousa, Marcela Machado
Presenter: Jefferson Coelho
doi://10.26678/ABCM.DINAME2023.DIN2023-0111
Abstract
Structures are commonly jointed using fasteners such as rivets or bolts arranged in various configurations depending on the required performance. Bolts are widely employed because of their numerous advantages, avoiding possible movement and ensuring the stability and security of the bolted joints. However, one of the main disadvantages of fasteners is the loosening that occurs under various causes, such as shock, vibration, and others that can cause serious damage and lead to structural failure. The application of machine learning (ML) techniques to bolt joint verification are still limited. This work investigates the application of the supported vector machine (SVM), a machine learning method, to bolt loosening detection based on a data-driven bolt-joint structure. A damage index was calculated using the system’s frequency response to classify the state of the bolted connection in binary form. It will show some advantages of SVM used to monitor the bolted structure, which also can be applied to nonlinear classification problems using kernel functions. Results show the use of SVM to track the jointed structure data-driven collected under different conditions. Discussion about the challenges of usage, performance and implementation of the technique are presented.
Keywords
Bolt-loosening, SVM Machine Learning, Damage diagnosis, Data-Driven
