Deep underground space is the circulatory system of human society, covering multiple aspects such as coal mining, subway transportation, and subsea transportation, etc., the structural safety of which is the primary consideration for the system to play fundamental role. In this work, innovative framework for mobile construction of 3D point cloud model and structural deformation measurement in deep underground space is proposed under GNSS denial and extreme working conditions. The construction method of point cloud model for coal mine roadways based on LiDAR and IMU fusion has been proposed to solve the problem of no satellite signal assistance for mapping. Further, regional measurement of structural deformation is carried out through a series of data processing including point cloud denoising, simplification, registration and deformation analysis. Experiments were conducted in simulated and actual deep coal mine to validate the performance of the proposed framework. Results show that the proposed method is effective and accurate in measuring regional deformation in different types of deep coal mine roadways, achieving an average relative measurement accuracy of 96.36 % and an average absolute measurement error of 2.68 mm in on-site environment of underground space with coal dust and water mist.

Coal mine roadways provide necessary underground space for personnel and equipment transportation, and mining operation, the reliability of which is critical for safe and efficient daily of operation of coal mine. Due to the impact of rock burst and coal mining movement, the deformation of roadway surrounding rock is more and more common in the process of coal mining, and has gradually become a high-risk factor that affects and restrains the development of high-quality coal mines. Given this, this work proposed an innovative regional measurement method of surface displacement of coal mine roadway based on D laser scanning technique. In the proposed method, point clouds of roadway from different phases are obtained in combination with the proposed approach of denoising and simplification, and the global deformation analysis of the point cloud is further carried out. Finally, the regional surface displacement distribution is acquired to realize the quantification of surrounding rock deformation. The proposed method has been applied in the actual environment of coal mine in northwest China, and reaches ideal measurement performance in different working conditions, including mining roadway, preparation roadway, and permeant roadway. Results show that the absolute errors of surface displacement measurement in three kinds of roadway environment are 0.2363 cm, 0.21 cm, and 0.4138 cm, respectively. The proposed measurement method also exhibits a resolution of better than 2 cm, which is sufficient for warning of roadway surrounding rock deformation.

Large Language Models (LLMs) have demonstrated remarkable capabilities in capturing complex conceptual representations from textual data for a wide range of real-world applications. However, in Intelligent Fault Diagnosis (IFD), leveraging sensor data such as vibration signals is essential but remains a challenge due to the modality gap between time series and LLMs’ inputs. Existing efforts to bridge this gap often treat LLMs merely as classifiers, overlooking their potential for understanding and reasoning over vibration-based data. In this paper, we propose a novel LLM-based fault diagnosis framework (FD-LLM) that aligns vibration signals with LLMs by encoding the signals into textual representations. FD-LLM introduces a classification-oriented approach, which formulates fault diagnosis as a multi-class classification task for benchmarking LLMs’ performance, and a context-aware spectrum language modeling approach that enables explainable, reasoning-driven fault analysis. We evaluate four open-source LLMs using FD-LLM across multiple datasets and noise conditions, assessing their validity, adaptability, and robustness. The results demonstrate that models such as LLaMA models achieve robust diagnostic performance, strong zero-shot adaptability across operating conditions, and effective generalization in cross-dataset scenarios with few-shot learning. The results further indicate that explainable fault diagnosis can be achieved in LLMs.

Stray current corrosion is the main and most serious form of corrosion suffered by subway shield tunnels,1 which is induced only when the environmental Cl- concentration of reinforced concrete structure (RCS) reaches a threshold value. Hence, accurate monitoring of ambient Cl- concentration of RCS is crucial for early warning of stray current corrosion in shield tunnels. In view of this, an artificial intelligence-assisted predictive method was proposed to conduct the monitoring of Cl- concentration based on electrochemical noise (EN) measurement. The proposed model effectively constructs the regression relationship between Cl- concentration and EN signal characteristics by means of key features extraction. This work mainly highlights the practicality and interpretability of the XGBoost-based model in environmental Cl- concentration prediction. The implementation of the explainable AI technique also helps to reveal the EN signal features that have the most obvious impact on Cl- concentration. Results show that EN signal of RCS under stray current interference with different Cl- concentration show significant differences in both time domain and frequency domain. Key features, including mean value, skew, kurt, white noise level (WL), total energy of D4 ~ D6 layers (E2), and total energy of D7 ~ D9 layers (E3), were extracted as the input variables of the XGBoost-based model by means of training contribution rates. Compared with other concentration prediction algorithm, the proposed model exhibits superior performance advantages in terms of prediction accuracy and stability, and can reach an average prediction accuracy of 92.95 %, an RMSE evaluation index value of 0.046, and a prediction resolution of at least 0.1 mol/L, showing a potential application value for health monitoring and reliability assurance of RCS.

With the advancement of new energy technologies and the demand for intelligentization, hydraulic cylinders in crane booms are gradually being replaced by electric cylinders. However, electric cylinders may induce an initial retraction phenomenon during the crane startup phase when the electromagnetic torque has not yet matched the lifting load. To address this problem, this paper proposes a load-aware position switching control method for an electric crane boom with an integrated brake mechanism. In the braking phase, a fixed-time convergence controller combined with a radial basis function neural network observer is used to build up the holding torque of the permanent-magnet synchronous motor (PMSM) rapidly under uncertain load torque. After the safety current threshold is satisfied and the brake is released, the control law switches to a feedforward Proportional-Integral (PI) controller for position tracking. The switching threshold is derived from the estimated load torque and the PMSM torque constant, and a hysteresis/dwell-time condition is introduced to avoid chatter near the switching boundary. Simulation and experimental results show that the method effectively reduces the initial retraction of the boom.

This study presents a novel electrochemical noise (EN)-based data mining approach for non-invasive measurement of environmental chloride ion (Cl−) concentration in reinforced concrete structures (RCS) exposed to stray current interference. A custom experimental system captures EN signals from mortar-embedded steel rebars under varying Cl− concentrations (0.05–0.9 mol/L) and stray current densities (0.05–0.1 A/cm²). Time-domain statistical features and frequency-domain wavelet-decomposed energy parameters are extracted from EN signals as regression inputs. To overcome the complexity of signal-environment relationships, an intelligent algorithm (WOA-XGBoost-Attention) is proposed, integrating Whale Optimization Algorithm (WOA) for hyperparameter training, XGBoost for regression, and an attention mechanism to weight critical features dynamically. Validation shows the model achieves 95.33% average accuracy and a 0.9929 correlation coefficient (R2) for Cl− prediction, significantly outperforming benchmark methods (XGBoost, Random Forest, etc.). The framework enables early warning of stray current corrosion by detecting critical Cl− thresholds, offering a robust solution for monitoring subway shield tunnel durability where traditional methods are impractical.

With the extension of urban rail transit systems, more and more buried gas pipelines will suffer from stray current interference due to contact with adjacent metro systems. Stray current interference will generate a hydrogen evolution reaction through cathodic current flow, which leads to severe hydrogen embrittlement hazards on pipeline steel. In this work, we used scanning electron microscopy to obtain the morphology of fracture surfaces, which is ascribed to hydrogen embrittlement for Q235 pipeline steel under stray current interference, considering varying defect area and environmental pH value. The fracture surfaces were then quantitatively characterized by multifractal methods, including multifractal spectrum and generalized fractal dimension. A novel method was proposed to accurately describe the surface morphology of hydrogen-induced fracture by combining spectral width A alpha and generalized dimension threshold width ADq . The results indicate that HE fracture surfaces of Q235 pipeline steels under stray current interference show obvious self-similar and multifractal characteristics. Moreover, multifractal spectrum f ( alpha)-alpha and generalized fractal dimension D ( q )-q exhibit high-accuracy performance in depicting the roughness and homogeneity of complex fracture surfaces, considering varying defect area and environmental pH value. Spectral width A alpha and generalized dimension threshold width ADq show a good negative correlation with hydrogen embrittlement susceptibility. The underlying mechanism was analyzed through the energy-absorbing effect during the transition process from ductile to brittle fracture. The findings of this research offer a new perspective that describes the fracture surface through probability distribution. The proposed approach provides a potential application for hydrogen embrittlement susceptibility evaluation through an on-site screening technique. (c) 2025 Published by Elsevier Ltd on behalf of The editorial office of Journal of Materials Science & Technology.

Voxel-based LiDAR-inertial odometry (LIO) is accurate and efficient but can suffer from geometric inconsistencies when single-Gaussian voxel models indiscriminately merge observations from conflicting viewpoints. To address this limitation, we propose Azimuth-LIO, a robust voxel-based LIO framework that leverages azimuth-aware voxelization and probabilistic fusion. Instead of using a single distribution per voxel, we discretize each voxel into azimuth-sectorized substructures, each modeled by an anisotropic 3D Gaussian to preserve viewpoint-specific spatial features and uncertainties. We further introduce a direction-weighted distribution-to-distribution registration metric to adaptively quantify the contributions of different azimuth sectors, followed by a Bayesian fusion framework that exploits these confidence weights to ensure azimuth-consistent map updates. The performance and efficiency of the proposed method are evaluated on public benchmarks including the M2DGR, MCD, and SubT-MRS datasets, demonstrating superior accuracy and robustness compared to existing voxel-based algorithms.

Secondary collapse poses a severe threat to rescue operations, as subtle structural deformation may precede abrupt failure. Mobile LiDAR SLAM enables scene reconstruction, but it does not directly provide reliable measurement of local structural deformation under moving sensing and pose uncertainty. This paper presents ALERT (Adaptive LiDAR Early Warning for Structural Response Tracking), an online measurement and warning framework for structural deformation monitoring in environments with secondary collapse risk. Given pose estimates and point clouds from SLAM, ALERT formulates local deformation measurement as an uncertainty constrained state estimation problem. It integrates observation modeling with pose uncertainty propagation, Interacting Multiple Model state estimation constrained by local geometry, and multiscale spatiotemporal evidence verification. This design estimates deformation position, magnitude, direction, and confidence while suppressing false measurements caused by pose drift, sparse nonrepetitive scanning, and background bias. The controlled experiments show that ALERT consistently reports confirmed regions of persistent geometric-deformation evidence, confirming 81% of the moving structural targets at a track precision of 0.977 with about one false track per minute, while maintaining an average processing time of 299.0 ms against a 500 ms cycle budget. Preliminary alerts preceded high-confidence confirmations by 3.3 to 5.8 s on average across motion conditions, and the smallest observed confirmation displacement was 3.19 mm. Deployment on a quadruped robot in a constructed confined space further verifies its onboard deformation measurement and warning capability under realistic sensing conditions. These results indicate that ALERT provides an onboard measurement layer for tracking local structural deformation and supporting early warning during robotic rescue operations.

Mobile laser scanning (MLS) has become an effective technique for deformation monitoring in subway shield tunnels. Among various deformation characteristics, segment dislocation is an important indicator of tunnel structural health because it reflects the relative deformation between adjacent segments and may affect the mechanical behavior and waterproof performance of segmental joints. However, existing MLS-based methods for dislocation detection still suffer from facility interference, inaccurate seam localization, and limited automation in quantitative analysis. To address these challenges, this study proposes an automated method for shield tunnel segment dislocation detection based on MLS point cloud processing. The proposed framework consists of three main steps. First, a point cloud filtering strategy integrating offset features and semantic segmentation is developed to remove facility-related noise while preserving tunnel wall information. Second, a tunnel segment segmentation method combining bolt hole extraction and moving template matching is introduced to achieve accurate localization of both horizontal and longitudinal seams, where bolt holes are identified using normal vector and distance constraints. Finally, automated segment dislocation analysis is performed based on the filtering and segmentation results. Experimental results demonstrate that the proposed filtering method improves accuracy by 8.7% and 5.6% compared with conventional ellipse fitting and cylinder fitting methods, respectively. Using manually interpreted reference values derived from the same MLS dataset as the evaluation reference, the proposed method achieves less than 2 mm deviation in both seam localization and dislocation analysis, demonstrating high consistency with manual interpretation. Compared with existing automatic approaches, the proposed method provides more accurate and reliable automated dislocation analysis, significantly reducing the need for manual inspection. The proposed method enhances the automation, consistency, and reliability of shield tunnel deformation assessment and provides an effective solution for structural health monitoring.

The support bearing of permanent magnet synchronous motor (PMSM) is a key component of permanent magnet direct-drive system (PMDDS), and its reliability plays a critical role in the operational efficiency and safety of PMDDS. Existing PMSM bearing studies typically rely on static assumptions, neglecting time-varying reliability and degradation modeling. This study comprehensively considers the effects of electromechanical coupling and hybrid eccentricity of the PMSM rotor, investigates the evolution of bearing dynamic loads under varying system parameters and operating conditions, establishes a reliability assessment model for PMSM bearings, and proposes a dynamic reliability evaluation and sensitivity analysis method based on the active learning kriging (ALK) method. The results show that the proposed method greatly reduces calls to the actual performance function and provides highly accurate prediction results. Furthermore, the simulation results reveal the effects of various system parameters on the dynamic reliability of PMSM bearing. This study can provide a reference for the time-varying reliability prediction of PMSM bearings and the optimized design of PMDDS.

Aero-engine rotors, assembled from multi-material components fastened by bolts, primarily operate under supercritical conditions and are prone to excessive vibration near critical speeds. To mitigate this issue, elastic support structures are widely employed in engineering practice to reduce critical speeds and modulate vibration amplitudes. However, the piecewise-linear stiffness characteristics of bolted joints and substructure resonance in elastic support structures introduce localized nonlinear stiffness, potentially inducing vibration instability. To address these challenges, this study investigates the dynamic characteristics of rotor systems under the coupled effects of bolted joints and elastic support structures. First, a 10 DOFs bolted joint element is established to characterize the piecewise-linear stiffness behavior of bolted connections. Finite element simulations are employed to obtain the frequency-dependent stiffness of the squirrel-cage support, enabling the development of a rotor system dynamic model that incorporates both the piecewise-linear stiffness of bolted joints and the frequency-dependent dynamic stiffness of the squirrel-cage support. Numerical simulations and experimental results demonstrate that the squirrel-cage support effectively reduces both the critical speed and vibration amplitude of the rotor system while mitigating the stiffness-softening effect of bolted joints. Parametric investigations explore the relationship between the squirrel-cage support dynamic stiffness and the rotor critical speed and vibration amplitude through the variation of structural parameters. The findings elucidate the coupled dynamic mechanisms of bolted joints and elastic supports, providing theoretical foundations and design guidelines for vibration control and elastic support optimization in bolted rotor systems.

Vibration monitoring technology has been widely applied in industrial, transportation, and other environments. Traditional wired and battery-powered methods face challenges such as complex wiring, high energy consumption, and environmental pollution. Here, we have developed a selfpowered wireless acceleration monitoring system (SWAMS) for monitoring the operating condition of bearings in a weak vibration environment. A vibration triboelectric nanogenerator array (WEB-TENGs) is designed to efficiently harvest energy from micro-amplitude vibration, and supplies power to SWAMS. The WEB-TENGs achieves a notable peak power density of 4.3 mu W/ cm3 under an acceleration of 0.07 g (10 mu m excitation amplitude at 42.1 Hz). The WEB-TENGs can efficiently charge lithium batteries by a power management module. The SWAMS can transmit vibration data within each hour without additional power consumption, using a timing circuit module. The collected acceleration data from the bearing housing is wirelessly transmitted to a PC via a transmission module. Fault identification of bearings is performed using a hybrid CNN-LSTM model, achieving a classification accuracy of 98 % on the test set. The SWAMS enables self-powered vibration monitoring of industrial equipment in unattended environments, providing a sustainable and maintenance-free solution for fault diagnosis.

To address the difficulty of obtaining real fault data from elevator traction drive systems and the potential overestimation caused by random window splitting, this study uses two public motor-drive datasets to examine how validation protocols, signal modalities, and noise conditions affect diagnostic evaluation, rather than to claim direct validation of field performance in actual elevator systems. A PMSM inverter-drive fault diagnosis dataset is used as the main dataset to evaluate multiclass classification performance based on electrical, thermal, and derived features. A multimodal MOTOR dataset is used as an independent secondary dataset to analyze the effects of validation protocols, signal modalities, and noise disturbance on model performance. The results show that Random Forest achieves a Macro-F1 of 0.9901 under random splitting on the PMSM dataset. On the MOTOR dataset, the Macro-F1 reaches 0.9682 under random splitting but decreases to 0.5856 under strict block-split validation, indicating that random window splitting may substantially overestimate generalization performance for continuous signal data. The modality ablation results show that the vibration-only modality performs best under strict block-split validation, with a Macro-F1 of 0.6420, whereas the noise analysis indicates that this modality is sensitive to disturbance. The results show that public motor-drive data can provide a reproducible methodological test bed for studying evaluation bias and signal reliability, but they should not be interpreted as direct evidence of diagnostic performance in actual elevator systems.

Firefighting robots play a critical role in fire suppression. Ensuring the water stream precisely hits the target during autonomous fire extinguishing is of paramount importance. By visually detecting the landing point of the water jet, closed-loop control of the extinguishing process can be achieved. However, achieving accurate jet landing point localization in complex environments, such as changes in ambient lighting, jet end divergence, and jet breakup, presents a challenging task. To address this, we propose a novel CIA-YOLOX (Channel Interaction Attention-You Only Look Once) model for the precise identification of water jet landing points in firefighting robots using unmanned aerial vehicle (UAV) visual information. First, the model introduces the Triplet Attention (TA) mechanism to capture feature dependencies across different dimensions, enriching feature information. Second, a module named Coordinate Attention Transformer (CA-Trans) is designed to establish long-range dependencies between directional feature vectors, enabling the extraction of precise positional information critical for accurate impact point prediction. Additionally, a Dual-branch Channel Interactive Attention Fusion (DCIAF) module is proposed to enhance feature representation capabilities by facilitating feature complementation through semantic modeling of channel interactions. Experimental results indicate that the proposed model surpasses current state-of-the-art methods in performance while maintaining low computational costs, confirming its efficacy. This approach enhances the robot's ability to perceive complex environments, providing valuable insights for implementing firefighting actions in real-world scenarios.

In current computer-aided process planning (CAPP) systems, the quality of the typical process routes employed directly influences the overall quality of subsequent process planning. With the advent of the big data era, automated analysis and discovery of typical process routes using advanced artificial intelligence (AI) techniques have become a critical issue to address. Current research primarily focuses on linear/simple process routes, with relatively limited exploration of networked process routes. Therefore, considering the characteristics of networked process routes, this paper proposes a novel approach for discovering typical networked process routes based on networked sequence similarity and intelligent clustering. Specifically, by thoroughly analyzing the information requirements of networked process routes and integrating five embedded process information types, a multi-dimensional process information fusion-based comprehensive similarity measure is constructed using the Kuhn-Munkres (KM) algorithm and principal component analysis (PCA). Furthermore, to ensure the clustering effectiveness of the discovered typical networked process routes, quantity and radius soft constraints are introduced into the traditional typical process route discovery problem. Two nutcracker optimization algorithm (NOA)-optimized affinity propagation (AP) algorithms (i.e., NOA-OAP and NOA-IAP) are proposed to address this problem, aiming to enhance clustering performance and identify more suitable and practical typical networked process routes for CAPP. Finally, numerical illustrations validate that the proposed similarity measure can effectively distinguish subtle differences among various networked process routes, and the two proposed clustering algorithms can discover more representative and effective typical process routes.

Purpose Stray current generated from rail transit system poses great integrity threat on surrounding buried metal pipelines, especially for oil and gas pipelines. Apart from corrosion damage, hydrogen damage will also occur in widely distributed cathode area on the pipeline surface due to locomotive operating conditions and location. However, limited studies focus on the hydrogen permeability characteristics of Q235 pipeline steel under stray current interference. The purpose of this paper is to study hydrogen permeation behavior of Q235 steel under stray current interference from urban rail transit system. Design/methodology/approach Since hydrogen permeation is prerequisite for inducing hydrogen damage, this study deals with hydrogen permeation behavior of Q235 pipeline steel under direct stray current interference. Besides, the impact of surface defect and pH value on the hydrogen permeation behavior was also studied with electrochemical techniques. Nonlinear fitting was conducted with high accuracy to explore the quantitative relationship between hydrogen permeation process and influencing factors. Findings Experimental results show that stray current flowing into metal substrate could obviously promote hydrogen permeation process, and surface defect and decreasing pH value could also enhance hydrogen permeation effect. Hydrogen penetration kinetic parameters: hydrogen flux J and diffusion coefficient D were found to be positively correlated with increasing amplitude of stray current density and area of surface defect, and negatively correlated with the increasing environmental pH value. Originality/value This study clarifies main influencing factors of external environment on the hydrogen permeation under stray current interference. Experimental results of this study are of great significance for daily maintenance of buried gas pipelines adjacent to rail transit system and prevention of hydrogen-induced cracking from the perspective of influencing factors.

Table structure recognition, as the most important subtask in image-based table recognition, recognizes the logical relationship between adjacent cells and represents the structured essence of the table. However, in reality, tables often appear in many different styles, such as those containing many blank cells, large row–column spans, missing separator lines, or even no separator lines at all. Different styles present great challenges to table structure recognition. In response to the above situation, we carefully analyzed the structural characteristics of the table and believed that rows and columns are their inherent characteristics. However, row–column annotations are rarely encountered in existing publicly available datasets, and cells with row–column spans are often ignored. Thus, we propose an innovative table structure annotation scheme, where the objects to be annotated include rows, columns, and cells with row–column spans. Furthermore, we released a challenging dataset named Row–Column Segmentation for Table Structure Recognition (RCSTSR), which contains more than 12,000 table images of different styles, each of which is notated with the corresponding mask. Then, on the basis of this dataset, we construct an effective semantic segmentation-based solution for table structure recognition. It consists of two main parts: an improved masked-attention mask transformer model, named TableStructureFormer, and corresponding postprocessing. Among them, the former is responsible for predicting the masks of objects in the table image, and the latter is used to generate the table structure on the basis of the predicted masks. Considering that long-distance feature maps are more useful than local feature maps for row–column segmentation, we propose the dual-path adaptive weighted attention module to aggregate the multilevel long-distance feature maps and adaptively select more input information by introducing enhanced strip pooling and learnable weighted parameters, thereby improving the segmentation performance. In addition, to address the difficulty of perfectly segmenting the details of rows and columns, we propose the deep detail supervision module to guide the segmentation model to learn the detailed feature maps about objects, thereby further correcting their masks. The experimental results show that for row–column segmentation, the mean intersection over union (mIoU) values of TableStructureFormer are 92.15