
Accurate time synchronization among spatially distributed sensors is a prerequisite for reliable seismic structural health monitoring (S 2 HM), particularly for widely used lightweight S 2 HM systems that rely on inter-story drift ratios as the key damage-sensitive feature. In real-world S 2 HM applications, however, sensor measurements are inevitably affected by time delays arising from unreliable hardware- or protocol-induced synchronization. Such delays can lead to non-negligible errors in assessing inter-story drift outputs, even when the monitoring system operates under nominal conditions. This study proposes a progressive approximation time synchronization algorithm for practical lightweight S 2 HM systems under noisy measurement environments. Time synchronization is formulated as a sequentially constrained estimation problem, in which time delays are progressively estimated through cascading steps from coarse to fine alignment. The proposed algorithm is validated using shaking table tests of a full-scale ten-story reinforced concrete structure, where various time-delay scenarios representative of real-world monitoring conditions are conceived. The results show that the proposed algorithm can robustly reduce arbitrary inter-story time delays to within 5 ms, and constrain the relative error of peak inter-story drift estimations to within 1% from values exceeding 100%, even when the S 2 HM system is operated with the low-cost network time protocol synchronization.
Vibration signals of rotating machinery collected in industrial scenarios are usually corrupted by heavy noise, and time-varying operating conditions further degrade the performance of intelligent fault diagnosis methods. To address these issues, this article proposes an interpretable semi-supervised wavelet graph convolutional network (SW-GCN) for noise-robust fault diagnosis under varying operating conditions. First, a multi-wavelet graph convolution kernel with a scale-smoothing property is designed, and an interpretable wavelet graph convolution layer is constructed to improve denoising capability and model interpretability. Second, an entropy-weighted semi-supervised network architecture is developed to excavate node correlation within graph data and enhance the capability of capturing fault-sensitive features under variable working conditions. Finally, feature extraction and denoising mechanisms are visualized to interpret the internal logic of SW-GCN and increase the trustworthiness of diagnosis results. Two rotating machinery datasets under different noise intensities and time-varying speeds are utilized for validation. The results demonstrate that the proposed SW-GCN can effectively extract fault-sensitive features under strong noise interference, achieving competitive diagnostic accuracy and prominent noise robustness.
To assess the stress state of contact interfaces in bridge anchorage systems, a theoretical model integrating contact mechanics and nonlinear acoustics is developed. Based on one-dimensional longitudinal wave scattering theory, an analytical expression for the reflection nonlinear coefficient γ R is derived, incorporating first- and second-order interface stiffnesses K 1 and K 2 . An ellipsoidal asperity multi-stage contact model is introduced as a physically based complement to the power-law model, which is concise but relies on empirical stiffness–pressure fitting. Laboratory graded tensioning and ultrasonic tests on a single-strand anchorage system, together with field tensioning and ultrasonic tests on a multi-strand anchorage system, validate the model. Results show that reflection nonlinear coefficient γ R increases monotonically with contact stress P , reflecting interface compaction and stiffness enhancement. The power-law model is convenient within a calibrated pressure range, but its accuracy depends on fitted parameters C and m and does not explicitly encode surface topography or material properties, limiting extrapolation. The proposed model derives the stiffness–stress relationship from rough-surface topography and material parameters, reducing empirical fitting. In the laboratory single-strand anchorage test, the acoustic characterization model based on the contact model gives a coefficient of determination of 0.9460 for reflection nonlinear coefficient γ R . In the field multi-strand anchorage test, the relative error between the predicted and measured contact stresses is 8.49%. These results confirm the field applicability and engineering reliability of the model for quantitative stress evaluation of anchorage interfaces.
Structural health monitoring (SHM) systems depend on reliable sensor measurements, rendering sensor fault diagnosis essential for maintaining data quality and supporting decision-making. Deep learning enables automated sensor fault diagnosis, while explainable artificial intelligence (XAI) improves the transparency of model predictions. In parallel, edge computing processes monitoring data closer to sensor networks and reduces dependence on centralized systems. However, most edge-based SHM studies focus on prediction performance and computational feasibility, while agreement between predictions and explanations remains insufficiently examined. This study proposes an integrated edge-based XAI-enabled evaluation methodology for multisensor fault diagnosis in SHM. Sensor faults are formulated as classes combining the faulty sensor and fault type (sensor-fault classes), enabling sensor fault detection, isolation, and identification within a single classification task. An established gated recurrent unit-based sequence-to-label classifier is used to classify multisensor time-series windows into “no-fault” or “sensor-fault” classes. Gradient × input generates time-step-level attributions, which are aggregated into sensor-level attributions. The main contribution is a task-specific quantitative evaluation of whether the dominant attributed sensor agrees with the sensor encoded in the predicted class. Prediction performance, explanation faithfulness, and sensor-level prediction–explanation agreement are evaluated for an independently trained edge deployment and compared with a reference centralized deployment under controlled implementation conditions. Local explanation stability, attribution-method dependence, and the influence of chronological response portions are additionally examined for the centralized deployment, whereas computational feasibility is assessed only for the edge deployment. The results show broadly comparable prediction performance and aggregate sensor-level attribution behavior between the two deployments, although neither achieves full prediction–explanation agreement. Supplementary analyses indicate moderate local faithfulness and high stability of the sensor-attribution distributions, while dominant-sensor selection varies among attribution methods and classification performance is sensitive to chronological response portions for specific fault classes. The measured mean runtime indicates that, on average, prediction and explanation generation are completed on the next unit of computing (NUC)-based edge-computing node within the selected window update interval. The findings provide proof-of-concept evidence for the investigated edge-based XAI-enabled multisensor fault diagnosis, establishing a basis for further development of decentralized SHM systems with larger sensor networks, different structures, and varying operating and environmental conditions.
Reliable condition monitoring of multi-sensor mechanical systems requires the joint assessment of the fault state and its temporal progression. However, fault diagnosis and prognosis are commonly treated as separate tasks, whereas existing multitask approaches model their interaction implicitly through shared representations. It does not explicitly account for the relationship between diagnosis and prognosis, where information about the developing fault type can provide useful context for estimating time remaining before fault onset. To address the gap, this study introduces a directed cross-task attention within a multitask Transformer, in which the fault-diagnosis representation informs the time-to-onset prediction, rather than relying implicitly on a shared encoder. Learnable task tokens are used to derive task-specific representations through an attention mechanism, whereas an uncertainty-based auto-weighted loss adaptively balances the learning objectives during training. The framework is validated on multivariate sensor data from a rail engine air compressor comprising 14,712 air and 218,846 oil leak readings. The proposed framework achieves 95.2% classification accuracy and time-to-onset errors of 9.3 and 3.7 min for air and oil leaks, respectively. It achieves strong overall performance against deep learning baselines, with the component-wise ablation confirming the substantial contribution of both the directed cross-task attention and auto-weighted loss. Furthermore, the model detects air leak 96 min and oil leaks more than 2 days ahead of the built-in alarm system. The model also demonstrates robustness to missing data, noise, and drift while remaining sensitive to sparse outlier spikes. The results demonstrate the effectiveness of modeling dependency between fault diagnosis and time-to-onset prediction for integrated condition monitoring of multi-sensor mechanical systems.