Damage Detection of Plate Structure with 1D Convolutional Neural Network (1D-CNN) Using Modal Data

Abstract

Health monitoring of structures and detecting structural damages are crucial for maintaining structural integrity and assuring desirable service life. Traditionally, damage detection has been carried out through visual inspection, but this method has several limitations. Therefore, Vibration-based damage detection has become a more effective substitute for traditional techniques. The core idea is that the existence of damage or weakening in a structure leads to alterations in its dynamic characteristics. This method can be applied using changes in different dynamic properties like natural frequency, mode shapes, and their derivatives. Artificial Intelligence (AI) is being extensively used in this field to detect damage with greater accuracy because of its useful features. This study proposes a deep learning (DL) based damage detection technique for damage analysis of plate-like structures with modal data. The plate structure is simulated with the FE analysis software ABAQUS, and modal data is collected. Noisy synthetic data is generated by adding different quantities of Gaussian random noise to the modal data. The 1D-CNN algorithm is utilized here to perform damage localization and quantification using noisy synthetic data as input. The proposed damage detection method has proved to be highly effective in detecting damage location and severity of plate structure.