ABSTRACT
Structural performance degradation can be identified by analyzing changes in response behavior patterns. This study employs artificial neural networks to evaluate structural aging and deterioration through behavior pattern recognition, enabling effective data-driven assessment. Using a cable-stayed bridge model subjected to simulated dynamic loads from multiple vehicles, we applied the Long Short-Term Memory (LSTM) algorithm to recognize and predict response patterns. Trained on time-series data, the algorithm assessed structural condition by analyzing prediction errors, which reflect changes in properties due to incremental performance degradation across multiple regions. The proposed methodology incorporates factors such as response type, measurement location, and timing to generate a multivariate error dataset that captures stiffness degradation stages. Building on this dataset, we developed a deep learning model to quantify structural state changes. Validation results demonstrate the model’s effectiveness in accurately representing the extent of structural performance degradation.
