Structural damage detection is still a challenging problem owing to the difficulty of extracting damage-sensitive and noise-robust features from structure response. This article presents a novel damage detection approach to automatically extract features from low-level sensor data through deep learning. A deep convolutional neural network is designed to learn features and identify damage locations, leading to an excellent localization accuracy on both noise-free and noisy data set, in contrast to another detector using wavelet packet component energy as the input feature. Visualization of the features learned by hidden layers in the network is implemented to get a physical insight into how the network works. It is found the learned features evolve with the depth from rough filters to the concept of vibration mode, implying the good performance results from its ability to learn essential characteristics behind the data.

Structural damage detection (SDD) still suffers from environmental uncertainties or modeling errors, causing a gap between the numerical model and the real structure. It results in performance degradation in the application of many model‐based methods, which are usually designed on a numerical model and needed to be applied to a real structure. Such a situation is defined as a cross‐domain SDD problem in this work. This paper aims to address the cross‐domain SDD problem by designing a feature‐extractor to generate both damage‐sensitive and domain‐invariant features, instead of trying to reduce the gap, as the traditional methods do. A domain adaptation (DA) neural network is designed and trained on the data from both the numerical model and the real structure at the same time. In addition, no damage label of the real structure is needed. Both numerical and laboratory experiments show that the proposed method has excellent performance and outperforms the baseline model, a traditional convolutional neural network (CNN). This paper provides a new methodology to solve the cross‐domain SDD problem, that is, to learn better features instead of just trying to reduce the gap.

This article presents a novel data-driven structural damage detection method named moving embedded principal component analysis to monitor the bridge condition and detect the damage occurrence using only one sensor. A fixed moving window is used to cut out the time series of the recorded data for the analysis. The data set inside the window is embedded to be a multidimensional state space using time delay method. The matrix of the state space is analyzed using the standard principal component analysis method, and a novel damage index R_j defined with the eigenvalue is proposed to identify structural damage occurrence. The window length is determined by a new approach through examining the convergent spectrum of the contribution ratio of the first principal component of the embedded state space. The time delay is determined by the autocorrelation function of the response, and the embedding dimension is obtained by the cumulative contribution ratio of the state space. The windowed damage index can be calculated continuously by moving the window along the recorded vibration data. To demonstrate the performance of the proposed method, responses of a beam bridge model subjected to stochastic loads obtained with numerical simulations and experimental tests are analyzed to monitor the structural conditions. The results demonstrate that the proposed method can accurately identify the occurrence of damage and the abnormal behavior of the structure. The recorded data on a large suspension bridge are also analyzed. The analysis successfully identified an incident on this bridge when it was slightly scraped by the mast of a sand ship. This further verifies the effectiveness of the proposed method.

A novel structural damage detection (SDD) method is proposed in this work, which is based on adversarial‐based transfer learning to achieve the cross‐domain information transfer of damage locations between numerical simulations and real bridge structures. Although the advancement of numerical modeling technology makes it accessible for relatively accurate finite element (FE) models, it is still difficult to meet the needs of practical engineering. The idea of adversarial training is introduced to enable the traditional feature extraction network to obtain the domain independent features between the numerical simulation and the real bridge structure. The dynamic response data from the numerical simulations are labeled with damage, while those from the real structure are unlabeled. To verify the effectiveness of the proposed method, we established a FE model of a simply support beam and regarded it as the benchmark model, and the target model with discrepancies from the benchmark model is obtained by quantitatively increasing the uncertainties. The results of the simulation show that the proposed method can overcome the discrepancy caused by uncertainty to a certain extent compared with the traditional method and obtain a high damage localization accuracy on the target model. In the laboratory experiment, the proposed method still achieves promising results. The primary contributions of this work are twofold: first, it delves deeper into the effectiveness of adversarial training for extracting domain‐invariant features, which are crucial for structural damage identification. Second, it provides a quantitative assessment of the performance degradation of traditional methods due to modeling errors and uncertainty. Additionally, it demonstrates the significant performance enhancement achieved by the proposed method.

Structural health monitoring (SHM) is vital for ensuring structural integrity by continuously evaluating conditions through sensor data. However, sensor anomalies caused by external disturbances can severely compromise the effectiveness of SHM systems. Traditional anomaly detection methods face significant challenges due to reliance on large labeled datasets, difficulties in handling long-term dependencies, and issues stemming from class imbalance. To address these limitations, this study introduces a hierarchical attention Transformer (HAT)-based method specifically designed for sensor anomaly detection in SHM applications. HAT leverages hierarchical temporal modeling with local and global Transformer encoders to effectively capture complex, multi-scale anomaly patterns. Evaluated on a real-world dataset from a large cable-stayed bridge, HAT achieves superior accuracy (96.3%) and robustness even with limited labeled data (20%), significantly outperforming traditional models like CNN, LSTM, and RNN. Additionally, this study visualizes the convergence process of the model, demonstrating its fast convergence and strong generalization capabilities. Thus, the proposed HAT method provides a practical and effective solution for anomaly detection in complex SHM scenarios.

A deterministic algorithm for VLSI block placement was developed through human's accumulated experience in solving “packing” problem. Rectangle packing problem is just a simplified case of the polygon-shape stone plate packing problem that the ancient masons needed to face. Several “packing” principles derived from the so-called “less flexibility first” experience of the masons. A k-d tree data structure is used for manipulating the packed rectangles under the derived packing principles. Experiment results demonstrate that the algorithm is effective and promising in building block layout application