Eventos Anais de eventos
DINAME 2023
XIX International Symposium on Dynamic Problems of Mechanics
Assessment of Damage in a Reinforced Beam by K-Nearest Neighbor Algorithm
Submission Author:
Amanda Aryda Silva Rodrigues de Sousa , DF
Co-Authors:
Amanda Aryda Silva Rodrigues de Sousa, Jefferson Coelho, Marcela Machado, Maciej Klosak
Presenter: Amanda Aryda Silva Rodrigues de Sousa
doi://10.26678/ABCM.DINAME2023.DIN2023-0081
Abstract
Structural damage is considered any change in a structure’s local flexibility or mass that creates undesirable displacements and vibrations. Mass loss in the structure can increase sensitivity to local damage and potentiate the effects of discontinuities in the dynamic response, which can be used in damage identification and location. Therefore, this work aims to investigate the application of k-nearest neighbor (k-NN) machine learning for damage detection in a cantilevered beam with reinforcement. Under undamaged and damaged conditions, a damage index (DI) is assumed to build the dataset from the frequency response function. Structure reinforcement mass loss directly influences the system’s vibration; hence, the DI is applied to detect damage and quantify its severity. Numerical results demonstrate that the Euclidean, Manhattan, and Braycurtis metrics are considered robust for both datasets with and without noise, and k values lower than 75, are considered fast and accurate in the detection problem and estimation of damage in beam structures.
Keywords
Damage Identification, machine learning, k-Nearest Neighbor, beam with additional auxiliary mass
