Eventos Anais de eventos
COBEM 2017
24th ABCM International Congress of Mechanical Engineering
DAMAGE DETECTION ON ALUMINUM BEAMS USING VIBRATION-BASED METHOD AND ARTIFICIAL NEURAL NETWORKS.
Submission Author:
Luísa Völtz , SC
Co-Authors:
Luísa Völtz, Eduardo Lenz Cardoso, Ricardo De Medeiros
Presenter: Luísa Völtz
doi://10.26678/ABCM.COBEM2017.COB17-0583
Abstract
Damage detection is a critical engineering area that allows corrective measures to be applied in order to ensure structural safety. Significant efforts have been devoted to developing Non-Destructive Techniques (NDT) for damage identification and prediction in structures. In this work, an Artificial Neural Network (ANN) is proposed to detect damage in metallic beams. An experimental setup is designed in both, intact and damaged, free-free beam boundary condition, excited by an impact hammer, with the response measured by an accelerometer attached to the beam. A vibration-based method using Frequency Response Functions (FRF) and modal parameters are studied and used to train the ANN. A multilayered feedforward neural networks architecture with a learning algorithm is proposed. In addition, experimental and modeling results are performed and compared. The main idea is training and simulating with different setups for the hyperparameters and topology to reach an optimized ANN architecture for damage detection. Therefore, the strategy presented can be helpful in the study of damage detection systems that uses ANN as part of its process.
Keywords
Artificial neural networks, Damage Detection, Vibration-Based Methods, experimental analysis, metallic beam
