Damage Detection in Simply Supported Beams Using 1D Convolutional Neural Networks with Vibration Signals

Abstract

Structural health monitoring (SHM) is critical for ensuring the safety and serviceability of civil infrastructures. Traditional vibration-based damage detection methods often rely on manually extracted features and expert knowledge, which can be time-consuming and subjective. This paper proposes a novel, data-driven approach for damage detection in simply supported beam structures utilizing a one-dimensional Convolutional Neural Network (1D-CNN). The proposed method automatically learns discriminative features directly from raw acceleration response signals under ambient excitation, eliminating the need for manual feature engineering. A numerical model of a simply supported beam is established to generate acceleration data for various damage scenarios, including different locations and severity levels of cracks. The collected time-domain signals are used to train and validate the designed 1D-CNN model. Experimental results demonstrate that the proposed CNN model achieves high accuracy in identifying the presence, location, and severity of damage. The model exhibits strong robustness to noise and outperforms traditional methods that rely on modal parameters (e.g., natural frequencies and mode shapes). This study confirms the feasibility and effectiveness of using deep learning, specifically 1D-CNN, as a powerful and efficient tool for automated damage diagnosis in beam-like structures, offering significant potential for real-world SHM applications.