Intelligent Structural Damage Detection with MEMS-Like Sensors Noisy Data

In the recent period, in the civil engineering field, innovative structural health monitoring (SHM) techniques are under the spotlight, promoted by the latest developments in artificial intelligence (AI). Even just focusing on the first SHM level, i.e., the damage identification task, ground-breaking non-destructive evaluations based on output-only vibration signals have been developed lately. The current work presents two alternative methods to address an AI-based multi-class damage classification on a numerical benchmark beam problem. The main goal is to investigate the noise impacts on the simulated monitored system of micro-electro-mechanical system (MEMS) accelerometer sensors with three reasonable signal-to-noise ratio (SNR) levels. A multi-layer perceptron (MLP) neural architecture has been employed for the current task. The numerical results essentially showed that the analyzed deep learning (DL) model presents a fairly good noise immunity performance for damage detection needs with proper damage-sensitive features.