To address the fundamental challenge of translating probabilistic health assessments into reliable maintenance decisions for elevator mechatronic systems, this paper proposes the first uncertainty-aware dual-threshold optimization framework that systematically bridges Bayesian perception with optimal decision-making. Leveraging posterior health distributions (mean and variance) output by the Bayesian Fusion for Elevator Health Assessment, the method constructs supervised learning pairs from multi-year historical records. A three-objective composite loss function-integrating matching accuracy, uncertainty coverage, and decision separability-is optimized via Bayesian optimization to calibrate optimal thresholds. Critically, the 95% confidence interval is explicitly embedded as a constraint, ensuring high-uncertainty equipment remains safely distant from decision boundaries. Validated on one-year operational data from 100 elevators, the framework achieves 84.0% matching accuracy (17.0% improvement), 0.71 maintenance cost index, and 0.12 safety risk index, delivering Pareto-optimal trade-offs among decision accuracy, maintenance economy, and operational safety. The offline learning and online deployment paradigm requires no additional hardware and integrates directly into existing supervision platforms.

The inverse kinematics (IK) of anthropomorphic (humanoid) upper-limb manipulators presents two main challenges: the inherent non-uniqueness of solutions and inter-frame discontinuities near singularities or joint limits. Under real-time constraints, the trade-off among end-effector accuracy, near-singularity robustness, and motion smoothness becomes critical. To address these challenges, this paper proposes an on-demand global-local hybrid solver that minimizes end-effector position error, incorporates a manipulability-based robustness term, and penalizes inter-frame joint changes under a unified cost function. The solver combines damped least squares (DLS) for efficient local refinement with gated particle swarm optimization (PSO), which is activated only when increasing frame difficulty is indicated by rising error, declining manipulability, or emerging feasibility risks. At the trajectory level, smoothness and feasibility are maintained through previous-frame initialization, early stopping, and joint-limit projection. Under a unified metric suite, the proposed method is systematically compared with Always PSO-DLS, DLS-only, classical Jacobian pseudoinverse, and quadratic programming-based IK baselines. The results show that the proposed method achieves a favorable balance among tracking accuracy, continuity, robustness, and computational cost: it preserves behavior close to always PSO-DLS while substantially reducing computation, improves difficult-frame stability and reliability relative to purely local refinement, and maintains low runtime on low-complexity frames by avoiding the fixed overhead of always-on global search. Ablation and sensitivity analyses further support the effectiveness and stability of the on-demand mechanism and its key parameter settings. Overall, the proposed framework provides a promising IK core for rehabilitation robots, upper-limb exoskeletons, and anthropomorphic manipulators in continuous tracking tasks with real-time requirements.

In recent years, deep learning-based intelligent fault diagnosis methods have made significant progress. However, most of these approaches rely heavily on large amounts of training data and involve substantial computational costs, which can adversely affect both model performance and real-time applicability. To achieve high accuracy with low computational overhead in endto-end fault diagnosis, this article directly utilizes 1-D vibration signals as model inputs and introduces frequency domain analysis into the Swin-Transformer framework, proposing a novel frequency-enhanced Transformer method. Specifically, we design a frequency-enhanced block that performs frequency domain transformation and data augmentation on 1-D vibration signals. In this process, the Fourier-transformed vectors are reorganized into a 2-D matrix, providing a richer and more informative input representation. Additionally, a new frequency-enhanced attention mechanism is developed to replace the traditional attention mechanism during model training. Building upon these frequency-domain data augmentation and attention enhancements, the network architecture is further optimized. Experimental results demonstrate that, compared with the standard Swin Transformer using time-domain signals, the proposed method reduces computational cost by approximately 9% while achieving 99.225% accuracy. Moreover, our approach exhibits faster convergence and superior generalization capabilities across cross-domain tasks. Finally, when applied to real-world industrial motor data, the method achieves 89.2% accuracy, highlighting its practical value for industrial fault diagnosis.

With the rapid development of modern industrial systems, production equipment has become increasingly complex, leading to high diversity in equipment types, fault modes, and operating conditions. Equipment failures not only degrade production efficiency but may also pose serious threats to operational safety, which places higher demands on the accuracy and reliability of fault diagnosis methods. To address these challenges, this paper proposes a multi-scale feature fusion fault diagnosis method based on a cross-attention mechanism. Built upon the Vision Transformer (ViT) framework, the proposed method employs two feature extraction branches operating at different scales to model and fuse multi-scale characteristics of vibration signals, thereby enabling more comprehensive feature representation. Specifically, the raw one-dimensional vibration signals are transformed into two-dimensional frequency-domain representations with an input size of 32×32, which are then fed into a dual-branch Vision Transformer architecture. Furthermore, a cross-attention-based feature interaction module is introduced to facilitate effective information exchange and complementary fusion between different scale branches, enhancing the discriminative capability of fault features. Experimental results demonstrate that the proposed method achieves a fault diagnosis accuracy of 98.942%, outperforming several state-of-the-art approaches, including CrossViT, ViT, ResNet, VGG, and WDCNN.

Existing elevator maintenance relies on fixed cycles or pure data-driven assessments, facing two bottlenecks: lack of physical failure perspectives and dependence on large historical failure samples. This paper proposes FT-DG-Physics, a mechanism-data fusion framework for elevator operational reliability assessment enabling transition from fixed to on-demand maintenance. The framework integrates three-source probability fusion at inspection:maintenance:physical degradation = 5:3:2, fault tree to directed graph transformation rules with CPTs, and differentiated common cause failure analysis by failure mechanism types. Unlike purely data-driven Bayesian networks learning both structure and parameters from data, the FT-DG transformation preserves logical gate semantics while enabling physics-based probability injection through CPTs. Validation on 10 degraded elevators shows 44% MAPE reduction over baselines and 40% cost reduction via interval adjustment.

Accurately assessing the health status of elevators is key to transitioning from scheduled maintenance to on-demand maintenance. At present, significant progress has been made in the health assessment of electromechanical devices, particularly in assessment methods that lack sufficient data and integrated expert knowledge. However, the existing methods exhibit static characteristics when balancing data and knowledge: they fail to fully utilize the continuously accumulating data, or cannot ensure the continuity of the model in the engineering. In order to address the shortcomings, a health evaluation system for elevators based on a Bayesian framework is proposed. The model takes fuzzy expert knowledge as prior information and data-driven fault prediction results as likelihood evidence, and realizes dynamic health score assessment through the Bayesian update mechanism. A dataset consisting of 100 elevator samples is used for health evaluation, experimental results show that the proposed BayFuse-EHA framework exhibits excellent robustness in handling sparse data and niche operating condition samples; meanwhile, by providing confidence intervals of health scores, the framework quantifies assessment uncertainty and provides more reliable decision support for elevator safety operation and maintenance.

Constrained close-proximity service scenarios, exemplified by bedside rehabilitation, require collaborative manipulators to operate safely in cluttered human-robot shared spaces with dense obstacles and narrow passages. To address the low sampling efficiency and slow convergence of multi-DoF manipulator path planning in such constrained environments, this paper proposes a bidirectional algorithm named Progress-driven Multi-Channel RRT* (PMC-RRT*). This method integrates a progress-driven sampling approach with a multi-channel extension technique. First, a progress-driven sampling scheme is proposed to adaptively shift the sampling center and shrink the sampling domain to balance global exploration and local exploitation as the search progresses. Second, a multi-channel extension strategy is developed to combine direct steering, spherical shell preferential detouring, and potential field-based tangential sliding to enhance search guidance and obstacle avoidance flexibility. We also create a whole-body collision checking model based on the manipulator’s forward kinematics, enforcing its joint limits and a global minimum clearance constraint as a unified feasibility criteria throughout sampling, extension, and rewiring. To validate the performance of the PMC-RRT*, we conducted comparative experiments against six baseline algorithms—RRT-Connect, RRT*, Bi-RRT*, GB-RRT*, BAI-RRT*, and Bi-APF-RRT*—in terms of planning time, number of nodes, and iteration, path length, mean clearance, and mean turning angle. The results show that in most scenarios, PMC-RRT* achieved significantly lower planning time, number of nodes, and number of iterations than the baseline algorithms while maintaining comparable path length. Furthermore, PMC-RRT* also obtained larger mean clearance and smaller mean turning angle, and performed exceptionally well in cluttered and narrow bottleneck environments. Simulation and physical prototype experiments further confirm that the planned paths satisfy joint limits and safety clearance constraints. The manipulator achieves stable tracking without collision throughout the entire motion, which demonstrates robustness and physical executability in clearance-critical cluttered scenes, supporting its potential for bedside rehabilitation and other close-proximity applications.

Robotic therapy has demonstrated significant potential in upper limb rehabilitation, particularly for patients with unilateral motor function disorders resulting from stroke. Dual-robot systems introduce innovative approaches to enhance rehabilitation efficiency, yet optimizing their spatial layout to maximize collaborative workspace poses a considerable challenge. This paper presents a robust, computer-assisted optimization framework for the spatial layout of a dual-robot system aimed at unilateral upper limb rehabilitation, integrating advanced collision detection mechanisms to ensure safe and effective operation. We developed a mathematical model consisting of two six-degree-of-freedom robots designed to assist in limb movement. A genetic algorithm was employed to maximize the collaborative workspace while preventing collisions, using a penalty-based mechanism to enforce safety constraints. Through extensive simulations, optimal layout parameters were determined as [Formula: see text]. Stability tests revealed convergence of the fitness function at 6.18e7[Formula: see text]mm 3 after 274 generations, confirming the algorithm’s reliability. Furthermore, we validated the accuracy of the genetic algorithm by uniformly sampling layout parameters, achieving a relative error of no more than 1.57% from the maximum volume. Visualization techniques affirmed that the optimal parameters resided within the global optimum region, further validating the algorithm’s effectiveness and robustness. This research lays a solid methodological foundation for optimizing spatial layouts in dual-robot-assisted unilateral upper limb rehabilitation, paving the way for future investigations into collaborative control and real-time adaptability, ensuring improved patient outcomes and clinical applicability.

Task-oriented upper-limb rehabilitation requires trajectories that are accurate, smooth, and biomechanically plausible, yet the 7-DoF redundancy of the human arm leads to non-unique inverse-kinematics (IK) solutions and potential discontinuities. This paper presents an on-demand hybrid trajectory-planning framework that discretizes the task path and optimizes a unified objective combining end-effector position error, joint-angle variation, and angular-velocity variation under joint-limit constraints. Most samples are solved efficiently by a warm-started gradient-based local refinement; a small-budget genetic algorithm is invoked only at automatically detected geometric key transitions or when tracking error or inter-frame joint change exceeds a threshold, and its output is briefly polished locally. In simulation on a representative disk-stacking task, the planned path closely tracks the target with a peak position error below 1 mm, while joint-angle and angular-velocity profiles remain smooth without abrupt spikes, evidencing the intended penalties on variation terms. Global search is used sparingly (approximate to 6% of samples), yielding an average computation time of 0.053 s per sample, indicative of practical real-time, interactive applicability. The approach improves trajectory smoothness and efficiency and is well suited to broader task-oriented training scenarios.

Electric motors are common devices widely used in the industrial sector, making the study of motor fault diagnosis highly representative. In particular, for rail grinding vehicles, which play a significant role in the preventive maintenance and periodic upkeep of railway tracks, ensuring optimal train operation is of paramount importance. However, due to harsh operating conditions, the grinding motors on rail grinding vehicles frequently experience failures. Typically, these motors are periodically inspected and repaired by railway workers, which often leads to delayed handling of faulty motors, thereby compromising the efficiency of rail grinding operations and increasing the maintenance costs associated with motor repairs. Consequently, there is a need to investigate a fault diagnosis model for grinding motors and establish a system for remote fault diagnosis of these motors. To address this issue, the first step involves analyzing the maintenance records of grinding motors to identify common failure locations and types, and subsequently collecting corresponding vibration data. Next, a fault diagnosis model is developed based on the specific failure characteristics of grinding motors. This model is trained and optimized using a data set of vibration data from grinding motors to determine a suitable fault diagnosis model for this specific application. Finally, the developed fault diagnosis model for grinding motors is applied to diagnose faults in these motors, thereby validating the practical effectiveness of the model. By conducting this research, it is anticipated that a comprehensive understanding of the fault diagnosis process for grinding motors can be achieved, leading to the implementation of a remote fault diagnosis system for these motors. Ultimately, this will contribute to improved operational efficiency and reduced maintenance costs in rail grinding operations.

Rational parameters of TBM (Tunnel Boring Machine) are the key to ensuring efficient and safe tunnel construction. Machine learning (ML) has become the main method for predicting operating parameters. Grid Search and optimization algorithms, such as Particle Swarm Optimization (PSO), are often used to find the hyper parameters of ML models but suffer from excessive time and low accuracy. In order to efficiently construct ML models and enhance the accuracy of predicting models, a BPSO (Beetle antennae search Particle Swarm Optimization) algorithm is proposed. Based on the PSO algorithm, the concept of BAS (Beetle Antennae Search) is integrated into the updating process of an individual particle, which improves the random search capability. The convergence of the BPSO algorithm is discussed in terms of inhomogeneous recursive equations and characteristic roots. Then, based on the proposed BPSO prototype, a hybrid ML model BPSO-XGBoost (eXtreme Gradient Boosting) is proposed. We applied the model to the Hangzhou Central Park tunnel project for the prediction of screw conveyer rotational speed. Finally, our model is compared with existing methods. The experimental results show that the BPSO-based model outperforms other traditional ML methods. The BPSO-XGBoost is more accurate than PSO-XGBoost and BPSO-RandomForest for predicting the speed. Also, it is verified that the hyper parameters optimized by the BPSO are better than those optimized by the original PSO. The comprehensive prediction performance ranking of models is as follows: BPSO-XGBoost > PSO-XGBoost > BPSO-RF > PSO-RF. Our models have preferable engineering application value.

Abstract The cooperative planning in intermodal transport networks can obtain the global optimal decision under the premise of ensuring the data privacy of each role in the cooperation and avoiding massive data transmission. For the control of container flow in intermodal transport networks, the distributed model predictive control (DMPC) method can effectively realize cooperative planning, but the convergence speed of the existing DMPC methods is slow. Therefore, this study attempts to construct faster DMPC methods for cooperative planning. The Jacobi proximal distributed model predictive control (JP‐DMPC) and dual consensus distributed model predictive control (DC‐DMPC) methods are constructed for container flow control based on two variants of alternating direction method of multipliers (ADMM). The simulation experiments prove that the convergence speed of JP‐DMPC and DC‐DMPC methods is higher than that of the state‐of‐the‐art method on the premise that the time cost and interaction data volume of each iteration do not change much, and the DC‐DMPC method improves planning speed particularly significantly. This study provides new methods for intermodal transport cooperative planning and has significance for the development of synchromodal transport.

In storage systems, there are heat differences between data. Traditional algorithms such as LRU are limited by specific data structures. These methods cannot be well applied to industrial big data storage systems. Methods based on "temperature" are usually limited by static parameters and it is unable to adapt to dynamic load. Based on Newton's law of cooling, we proposed an identification model called AdjustDTM, with adjustable parameters. Our method identifies the hot and cold by assigning the attribute "temperature" to the data. Then, the model can dynamically adjust the parameters according to accessing interval and frequency. Our model can also preheat the correlative data. Finally, The experimental results showed that the hit rate of AdjustDTM is higher than other strategies.

针对自动扶梯原有故障检测装置的老化以及缺乏远程监测功能的问题,设计了一套无线数据采集系统.利用ESP-WIFI-MESH网络所具备的低成本、广覆盖、可扩展等特点,可以在具备较低部署成本的情况下,以无线的方式在系统的各个节点之间实现数据传输.数据采集系统由一个中心节点和若干边缘节点构成,边缘节点实现数据采集功能,中心节点实现数据缓存及上传功能.该系统通过对扶梯运行数据的实时采集与上传,实现对于自动扶梯运行状态的监测,保障自动扶梯的安全运行.

Wrist rehabilitation robots have been proved to be an effective training solution to patients who lose some motor function of wrist joint. During human-robot interaction, axis misalignment between human wrist joint and robot joint may bring some discomfort and even injuries to patients. Many researches mainly focused on the analysis and validation of misalignment compensation, while the quantitative assessment of axis misalignment only attracted little attention. This paper introduced a self-aligning mechanism (SAM) on the previously developed wrist rehabilitation device and proposed a robot-assisted assessment method of axis misalignment for wrist joint by analytic geometry on the foundation of kinematic model of human-robot interaction. To validate the feasibility and accuracy of the proposed assessment method, 9 healthy subjects were invited to participate the experiments to measure rotation angle in wrist flexion/extension and radial/ulnar deviation, along with the values of self-compensation of SAM. Surface Electromyography was implemented to keep the consistency of passive motion. Experimental results showed high reliability of the axis misalignment of each subject with intraclass correlation coefficient more than 0.75, standard errors of measurement of axis misalignment less than 2.75 mm. The experiment results are consistent with previous studies well, which proves the feasibility of proposed assessment method of axis misalignment.

在桥式起重机运行环境双目视觉三维建图任务中,为减少因反光耀斑所造成的建图误差,文中对双目相机采集的图像使用偏振滤镜对偏振光进行过滤,使用搭建的实验台进行了实验,比较了偏振滤镜对三维地图精度的影响.结果表明:偏振滤镜可以有效过滤物体表面反光,减小建图误差,提高建图结果的鲁棒性.

Bridge cranes must be able to sense their working environment to achieve autonomous operation. An active visual mapping system is presented in this research to adapt the measurement range based on the crane's operational state. The rotation angles of the two servos are computed based on the crane's speed so that the binocular camera is deflected and a greater field of view is gained in front, improving the crane's safety. An experimental platform is developed to simulate the operation process of the active vision 3D mapping system, and experiments are carried out using the approach proposed in this paper. The experimental results indicate that the technology can successfully enhance the mapping scope of the bridge cranes' digital operating environment and improve crane operation safety.

Aiming at the problem of obtaining internal parameters of binocular camera-lidar, a joint calibration method is proposed. Firstly, the corner points in the left and right images are detected and sub-pixel precision is performed, and the corner points in the left and right images are matched through epipolar constraints. Then, the internal parameters of the binocular camera are obtained according to the three-dimensional space coordinate relationship and the corner point matching relationship. Through the coordinate relationship between the feature points of the lidar and the feature points of the binocular camera, the binocular camera and the lidar are jointly calibrated to obtain the external parameters of the binocular camera-lidar system. Finally, an experiment was designed and carried out for the method proposed in this paper. The experimental results show that the calibration method proposed in this paper is effective.

铁路轨道打磨车打磨电动机工作环境复杂恶劣,故障发生率相对较高,并且打磨电动机作为铁路轨道打磨车的关键部件,其运行状态直接关系到轨道打磨的效率与质量,故可靠的故障诊断技术是提高打磨效率与打磨质量的关键技术之一.文中研究基于数据驱动的方式,通过传感器采集不同打磨电动机数据构建数据集,采用不同的机器学习算法构建电动机智能故障诊断模型,对打磨电动机运行状态进行诊断,并通过对比不同算法的诊断准确率,探索更加适用于打磨电动机的故障诊断算法.结果表明,所研究的智能故障诊断技术准确度较高,对提高打磨电动机运行可靠性以及提高其工业智能化程度具有较大意义.