Bài báo khoa học

In civil engineering, monitoring the structural damage becomes critically important to ensure safety and avoid sudden failures of structures. Therefore, improving the accuracy of methods for Structural Health Monitoring problems remains a priority. This paper proposes a new framework that combines the Burg Autoregressive (BAR) and Stacked Autoencoder-based Deep Neural Network (SAE-DNN) for the damage detection of steel frames using time-series data. Firstly, features of the time-series data are extracted using the BAR method. Then, the Autoencoder (AE) network is employed to reduce the dimension and learn sensitive features. Finally, the AE and Softmax layers are stacked and trained in a supervised manner of DNN for structural damage detection. The experimental data of two steel frame benchmarks are adopted to verify the performance of the proposed framework. The results show that the proposed framework could achieve high accuracy (97.8 and 99%) in the damage identification of steel frames.