This paper investigates the predefined-time tracking control problem for a type of strict-feedback stochastic nonlinear systems, in which the output variable is required to be constrained within an asymmetric interval and the nonlinearities are unknown. First, an improved practical predefined-time stability (PPTS) Lyapunov criterion for stochastic nonlinear systems is presented. Secondly, an asymmetric integral barrier Lyapunov function is constructed to handle the output constraint issue. On this basis, an adaptive predefined-time adaptive fuzzy tracking control algorithm is developed via the backstepping framework, where fuzzy logic systems (FLSs) are utilized to approximate unknown system dynamics. Through rigorous mathematical analysis and numerical simulations, it can be concluded that the proposed control scheme can not only drive all system variables to converge to steady states within a prescribed time in probability, but also make the output track the desired signal without violating the output constraint.
The damping ratio is the modal parameter that is hardest to identify reliably and exhibits the largest scatter in structural health monitoring and operational modal analysis; a trustworthy baseline relies on multi-estimator cross-validation and uncertainty quantification. Measured damping data for high-rise reinforced-concrete (RC) frame–shear-wall buildings under low-amplitude ambient vibration are scarce. For a 26-storey RC frame–shear-wall residential building, an operational modal analysis pipeline built only on open-source Python libraries is developed from three-component ambient-vibration records of eight measured floors and a continuous top-floor reference point. Three independent estimators—the random decrement technique, covariance-driven stochastic subspace identification, and the half-power bandwidth method—are applied and benchmarked against one another to estimate the damping ratios of the horizontal and vertical modes and to quantify their uncertainty. The measured ambient modal damping ratios are about 0.26%–0.6% in the horizontal direction and about 1.0%–1.6% in the vertical, the vertical damping being markedly higher than the horizontal. For spectrally isolated modes, the random decrement technique and the stochastic subspace identification yield highly consistent estimates, with per-mode differences within about 0.2%; for the fundamental horizontal mode, whose structural peak adjoins another spectral peak, the parametric subspace estimate is adopted, whereas the half-power bandwidth method is systematically biased high and serves only as a coarse reference. A trustworthy ambient-damping baseline with quantified uncertainty, together with an identification workflow built on open-source libraries (no commercial software), is thus provided for this class of high-rise buildings, supporting subsequent SHM baseline establishment and long-term condition tracking.
Attitude maneuvers of dumbbell-shaped flexible spacecraft are bottlenecked by rigid-flexible cross-coupling and dense, extremely low-frequency structural modes. In this paper, consider the case of large flexible spacecraft with low flexural modes and coupling with the bandwidth of the control system. To circumvent the frequency aliasing and dimensionality explosion inherent in complex online adaptive estimators, this paper proposes a passivity-based, dual-loop robust control framework. At the trajectory level, a sinusoidal spectral-notch path planning algorithm is formulated to analytically enforce source energy isolation precisely at the structural fundamental frequency. At the execution level, an outer-loop robust PD controller accommodates lumped uncertainties, while an inner-loop, collocated Positive Position Feedback (PPF) controller redistributes robustness through a wide-notch damping envelope ( ξ ci ∈ [ 0.5 , 1 ] ) to tolerate extreme parameter mismatches. Simulation results demonstrate that the proposed strategy improves steady-state pointing accuracy by two orders of magnitude (from 2 × 10 − 4 ∘ / s to 8 × 10 − 6 ∘ / s ) and suppresses flexible disturbance torque by 88.3%. Furthermore, the attitude stabilization time is reduced by 36% (23.7 s), and global asymptotic stability is strictly maintained even under severe, synchronous perturbations (e.g., ± 20 % inertia and ± 15 % resonance frequency shifts).
To address the limitation that a single sensor is insufficient for comprehensively extracting deep fault features in strong industrial noise environments, which constrains bearing diagnosis accuracy, this paper proposes an acoustic-vibration collaborative fusion network. First, an Adaptive Gated Residual Block (AGRB) is designed and combined with a Twin-Gated Residual Block (TGRB) architecture to effectively extract highly robust deep local acoustic and vibration features amidst strong background noise. Second, a Bidirectional Attention Sensing Module (BASM) is constructed to perform deep interaction and complementary calibration of heterogeneous acoustic-vibration features in the global semantic dimension, breaking through the limitations of traditional shallow concatenation of multimodal features. To verify the effectiveness of the proposed model, an experimental study was conducted on a 6205 deep groove ball bearing using a non-contact acoustic-vibration synchronous acquisition system with a 25 cm acoustic monitoring distance and a 5096 Hz sampling rate. The dataset contains nine diagnostic categories, including one healthy state and eight fault states.Experimental results indicate that this method can achieve deep dynamic alignment of heterogeneous data. The VA-DFN demonstrates exceptional noise-resistant robustness under varying signal-to-noise ratio (SNR) conditions from −6 dB to 2 dB, achieving a maximum diagnostic accuracy of 99.55%, which is significantly superior to existing single-modality and conventional deep learning baseline models.
The 3D scanning probe is the “eye” of precision measuring instruments, coupling errors among its axes remain an urgent problem; accurate decoupling is essential for improving probe data accuracy. This paper presents an improved algorithm that integrates Particle Swarm Optimization (PSO)、Latin Hypercube Sampling (LHS), and an Opposition-Based Learning (OBL) strategy. The inertia weight w is dynamically adjusted using a cycloidal schedule, ensuring both rapid initial convergence and refined exploration in later stages. Additionally, the cognitive and social coefficients c 1 and c 2 are linearly decreased to minimize the risk of the swarm becoming trapped in local optima. Compared with the standard PSO algorithm, the improved algorithm can obtain the global optimal solution, while the standard algorithm only achieves the local optimal solution and thus cannot be applied to probe decoupling. Compared with the LSM algorithm, the coupling matrices obtained by the two methods are basically consistent under the condition of outlier-free data. Nevertheless, when outliers exist in the dataset, the LO-PSO algorithm achieves superior robustness and decoupling accuracy relative to LSM. Tests were conducted on the original dataset, and the results show that decoupling improves the measurement accuracy of test data in the X, Y and Z directions. Compared with the pre-decoupling results, the four X-group datasets showed a maximum type-I error of 0.747%, corresponding to an average gain of 8.02%; the Y-group datasets showed a maximum type-I error of −0.956%, corresponding to an average gain of 10.82%; and the Z-group datasets showed a maximum type-I error of 1.658%, corresponding to an average gain of 8.86%. The results show that the improved algorithm significantly enhances the data accuracy of the 3D scanning probe. Under laboratory calibration test conditions, the LO-PSO algorithm achieves superior decoupling accuracy and outlier resistance robustness compared with the least squares method, and it has the engineering potential to be integrated into precision measurement software for on-site decoupling.
Aiming at the problems of long-range dependency modeling difficulty, weak fault feature extraction challenge, and severe class imbalance with limited minority fault samples in long time-series data of 25 Hz phase-sensitive track circuits, this paper proposes a fault diagnosis model called WT-Transformer by integrating discrete wavelet transform and Transformer encoder. Firstly, the original track circuit signal is decomposed by multi-scale discrete wavelet transform to extract global long-term trend features and local abrupt change features caused by faults. A soft-threshold denoising method based on the Minimax rule is adopted to suppress noise interference while retaining critical fault information. Secondly, the wavelet-enhanced signal and the original signal are combined by sample concatenation to enrich feature diversity and improve the identifiability of minority-class faults. Sine-cosine position encoding is integrated to provide temporal structure of the track circuit signal. Finally, a Transformer encoder with multi-head self-attention, feedforward network, and residual connection is constructed to capture long-range temporal dependencies and enhance feature representation ability. Experimental results on a simulated fault dataset show that the proposed model achieves superior performance in track circuit fault diagnosis, especially in the classification of rare faults with few samples. The effectiveness of wavelet transformation and time-frequency feature enhancement is verified, which provides a feasible and effective solution for intelligent fault diagnosis of long sequence data in track circuits.
Containment control for multi-agent systems (MAS) under disturbances is an important research topic. Although numerous containment control methods have been proposed, learning-based optimal control approaches for MAS still suffer from the drawbacks of low policy interpretability and poor convergence efficiency. In this paper, a receding-horizon neuro-fuzzy reinforcement learning approach with experience replay is proposed for the containment control of MAS. In the proposed method, the optimal control problem of MAS is formulated as a sequence of finite-horizon forward-receding optimization problems. Then, a value iteration framework that integrates policy improvement and policy optimization is adopted. Moreover, a distributed actor-critic policy framework based on fuzzy neural networks is designed and combined with a weight update mechanism using experience replay, thus improving online learning efficiency and enhancing policy interpretability. Finally, numerical simulations are conducted, and the results demonstrate that the proposed approach outperforms the receding-horizon reinforcement learning and neural network-based heuristic dynamic programming methods in terms of performance cost and mean error norm.
Automated defect detection plays a crucial role in maintaining product quality and improving production efficiency in modern manufacturing. This study established a reproducible benchmark to compare the performance of recent YOLO architectures for casting defect detection under standardized experimental conditions and to investigate the trade-off between detection accuracy and computational efficiency. A publicly available casting defect dataset from Kaggle was used to evaluate four object detection models, namely YOLOv8x, YOLOv10s, YOLO11s, and YOLOv12s. All models were trained and evaluated using identical preprocessing procedures, training configurations, and evaluation protocols. Performance was assessed using precision, recall, mAP50, and mAP50–95, while computational efficiency was evaluated in terms of model size, parameter count, FLOPs, inference latency, and frames per second (FPS). Experimental results showed that YOLOv12s achieved the best overall performance, attaining an mAP50 of 70.91% and an mAP50–95 of 33.37% on the testing dataset while maintaining a lightweight architecture with 9.23 million parameters, 21.5 GFLOPs, and a model size of 18.06 MB. Although YOLOv10s achieved the highest precision (70.78%) and YOLOv8x achieved the highest recall (69.54%), YOLOv12s provided the most balanced performance in terms of detection accuracy, localization robustness, and computational efficiency. The proposed benchmark provides a fair and reproducible comparison of recent YOLO architectures and demonstrates that YOLOv12s offers the most favorable trade-off between detection performance and computational cost, making it a promising solution for automated visual inspection in resource-constrained industrial manufacturing environments.
The non-stationarity of high-speed train axle-box bearing vibration signals, combined with distribution discrepancies between laboratory and field data, makes it challenging to directly apply fault diagnosis methods trained on experimental data to actual operating conditions. This paper proposes a four-branch multi-modal unsupervised domain-adaptive fault diagnosis framework, termed CRG-DA Net, based on ConvNeXt and ResNet1D, aimed at enabling cross-condition fault diagnosis under unlabeled target data. Multi-scale and multi-domain fault features are extracted from raw vibration signals using envelope spectrum analysis, short-time Fourier transform, and wavelet transform. A four-branch parallel network is then constructed, employing ConvNeXt-Tiny for modeling the time-frequency representations and ResNet1D for learning the time-domain characteristics of raw signals, with each branch generating high-dimensional feature representations and corresponding fault prediction logits. To account for the varying discriminative contributions of different modalities, a gated dynamic fusion mechanism is introduced, which computes sample-specific fusion weights from the concatenated branch features and integrates the individual branch predictions into a final fused output. In addition, adversarial domain adaptation combined with pseudo-label self-training is employed to align the source and target domain feature distributions, while target samples are classified following the same gated fusion and prediction procedure. Extensive experiments on multi-condition laboratory datasets and real-world operating data demonstrate that the proposed CRG-DA Net achieves outstanding fault diagnosis performance, meeting the expected experimental performance and exhibiting strong generalization across different operating conditions.
Inter-turn short-circuit (ITSC) faults are among the most frequent stator winding faults in induction motors, often leading to irreversible damage. In this context, this work proposes a Machine Learning (ML) based framework for classifying ITSC faults using experimental stator current data comprising 13 categories. The framework employs Direct-Quadrature (dq) transformation, signal windowing, and feature extraction for data processing, followed by the training of multiple ML classifiers, including Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forests, XGBoost, and LightGBM. For benchmarking, a Convolutional Neural Network (CNN) was also trained directly on raw signals. The hyperparameters of all the models were tuned using Particle Swarm Optimization (PSO), Optuna, and random search. The results show that the proposed framework achieves high classification performance, with XGBoost tuned using random search reaching up to 99.60% accuracy across 13 classes. The CNN, relying on end-to-end learning, achieved lower performance compared to the developed ML classifiers, highlighting the importance of data representation for accurate fault classification under limited data. For hyperparameter tuning, random search achieved performance comparable to complex methods with a lower processing burden, making it a viable option for hyperparameter optimization. These findings confirm that careful data engineering is as important as model complexity and is key to achieving efficient ITSC fault diagnosis.
In statistical process control, the performance of control charts is commonly evaluated using the average run length (ARL), which represents the expected number of observations required to signal a change in the monitored process. In many practical applications, quality characteristics are observed as time series data, motivating the incorporation of time series models into control chart design. The principal innovation of this study is the development of a new explicit analytical formula for the ARL of the modified exponentially weighted moving average (EWMA) control chart for autoregressive integrated moving average (ARIMA) processes subject to exponentially distributed white noise. The proposed explicit solution enables direct and instantaneous computation of ARL. Its accuracy and consistency are validated against the numerical integral equation (NIE) method using a conformity measure, demonstrating nearly identical results. Comparative analysis demonstrates that the explicit formulation provides equivalent accuracy while substantially reducing computational time. The detection performance of the modified EWMA chart is further investigated under various design parameters and shift sizes revealing improved sensitivity to small shifts compared with the standard EWMA chart. The practical applicability of the proposed approach is illustrated using real disaster datasets, where the modified EWMA chart detects structural changes more rapidly than the EWMA chart. These findings confirm the accuracy, computational efficiency, and practical effectiveness of the proposed explicit ARL formulation for monitoring autocorrelated processes.
This paper addresses the state estimation problem for fractional-order systems (FOS)–systems described by differential equations involving non-integer (fractional) derivatives–with nonlinear dynamics under communication constraints. To reduce communication load while effectively utilizing network resources, the FlexRay protocol (FRP), an automotive network protocol that combines time-triggered (static segment) and event-triggered (dynamic segment) communications, is incorporated into the estimator design, where its hybrid communication mechanism is explicitly exploited. A piecewise state estimation scheme is constructed using distinct scheduling rules for the static and dynamic segments. The system nonlinearity is handled under the Lipschitz condition (a mathematical condition restricting how rapidly the system’s nonlinear part can change), and the stability of the resulting estimation error system is analyzed using Lyapunov theory (a standard technique in stability analysis) and Mittag-Leffler stability criteria (a form of stability specific to fractional-order systems). By formulating the design conditions as linear matrix inequalities (LMIs, which are constraints involving matrices that must be positive definite), sufficient conditions for stability are derived, and the estimator gains are obtained analytically via the Schur complement (a method for simplifying matrix inequalities). Finally, simulation results based on an autonomous guided vehicle (AGV) tracking scenario demonstrate the effectiveness and convergence of the proposed method.
Accurate remaining useful life (RUL) prediction is critical for optimizing the performance and safety of lithium-ion battery systems across various applications. To enhance prediction performance, a novel fusion framework is proposed. This framework uniquely integrates the interacting multiple-model unscented Kalman Filter (IMM-UKF) with a wavelet neural network (WAVENN), leveraging their complementary strengths. In this framework, the IMM-UKF is employed to integrate various existing mathematical models that describe the capacity degradation of lithium-ion batteries, while simultaneously incorporating predictions generated by the WAVENN. The method features automatic identification and switching between different models through the online updating of weight coefficients and model probabilities. Consequently, an optimal capacity estimate is derived from a weighted aggregation of the forecasts. This iterative process ultimately yields the RUL prediction. The proposed multi-model fusion method is rigorously evaluated against traditional single-model benchmarks. Experimental results confirm its superior predictive accuracy.
To address the problem that pre-set unmanned aerial vehicle (UAV) inspection routes become invalid due to barge movement, resulting in the inability to capture inspection images of key components of floating cranes, this paper proposes a flight path correction method based on rigid-body kinematics. This method treats the barge and its associated waypoints as a single rigid body. By utilizing real-time kinematic (RTK) positioning technology onboard the UAV to precisely measure the three-dimensional (3D) pose variation of the barge after displacement, including latitude and longitude offsets, elevation changes, and heading angle deviations, this method employs forward and inverse transformations between geodetic coordinates and spatial Cartesian coordinates to map the barge's pose variation onto each waypoint, synchronously correcting both position coordinates and heading angles. Compared with the iterative optimization strategies of meta-heuristic algorithms, this method does not require the definition of an objective function or search space, has a computational complexity of O(n), and strictly preserves the relative positions of the original waypoints and the inspection coverage logic. Three independent field tests demonstrated that the mean feature point matching rates between the corrected re-flight images and the reference images were 94.61%, 95.64%, and 95.23%, respectively, all consistently above 90%. This performance significantly outperforms the 1.46% achieved with uncorrected flight paths, validating the feasibility and robustness of the proposed method and providing a technical reference for UAV inspections in non-fixed reference scenarios.
Controller workload remains the principal functional constraint on air traffic management system capacity, yet its measurement in operational settings is complicated by substantial inter-individual variability that aggregate traffic-based models fail to capture. This case study investigates whether physiological workload differs significantly between certified air traffic controllers (ATCs) exposed to identical operational conditions, a question with direct implications for both safety management and dynamic capacity planning. Three licensed tower (TWR) controllers were monitored during a standardised 50-min heavy-load simulation exercise. During the experiment, physiological stress indicators were monitored, specifically LF/HF, SDNN, and mean RR, in 5-min intervals measured by a single-lead ECG monitor with a sampling frequency of 1000 Hz, supplemented by a 3D accelerator for actigraphy. Simultaneously, photoplethysmographic (PPG) recording was performed in synchronization with a second single-lead ECG for control purposes. Despite identical traffic scenarios, statistically significant inter-individual differences were confirmed for all three parameters ( p ≤ 0.002). Median LF/HF values differed by up to 74% between controllers, SDNN by up to 45%, and mean RR by 5.83%. These differences exceed the magnitudes typically reported between low- and high-workload conditions in within individual aviation studies, demonstrating that individual physiological reactivity, rather than traffic complexity alone, is a primary determinant of operational workload. The findings challenge the current practice of treating airport capacity as a fixed threshold and support the concept of dynamic, real-time capacity management informed by continuous individual physiological monitoring. Practically, the approach enables identification of stress-susceptible controllers, supports personalised shift scheduling, and provides a foundation for deploying operationally compatible wearable ECG monitoring during actual ATC operations.
Marine transportation carries about 80% of global cargo, making it vital to global logistics. Growing environmental and fuel-efficiency demands drive the need for cleaner, low-emission engines. Homogeneous charge compression ignition (HCCI) offers high thermal efficiency and minimal emissions but faces challenges in maintaining combustion stability under varying conditions. Conventional in-cylinder pressure-based control works well within narrow phasing ranges but lacks responsiveness during large deviations.This study presents a novel vibration-based framework for detecting combustion states in an HCCI single-cylinder engine. Vibration signals, recorded from the research unit, were processed to extract time-domain and frequency-domain features, which were then fed into a supervised classification model trained on combustion-phasing data (CA50). Combustion states—Normal, Late, and Very Late—were classified using K-nearest neighbors (KNN), support vector machines (SVM), and artificial neural networks (ANN). All models showed high accuracy, with F1-scores above 98% for most classes. The ANN achieved the best performance (98.3% accuracy) and demonstrates strong potential for robust, real-time combustion monitoring and control applications.
This paper presents a new terminal non-singular sliding mode (NSTS) controller proposed for a second-order four-degree-of-freedom (4-DOF) robotic manipulator. This system is designed with a new exponent for the terminal sliding surface. The latter is designed to solve the singularity problem in the first derivative of the sliding surface. Such a problem is associated with the conventional linear hyperplane terminal sliding mode control. Error convergence time from any initial state is guaranteed to be a finite time. Lyapunov stability analysis and Simulation results are presented in this paper. Furthermore, a comparison is implemented with the conventional linear sliding mode in the simulation part. The obtained results prove improved performance in trajectory tracking.
Pavement icing induced by complex temperature fluctuations during winter weather poses a significant safety hazard. Given that existing icing detection methods are predominantly passive, this study designs an active pavement temperature simulation device based on a thermoelectric cooler (TEC). Using the predicted pavement temperature as the thermal control target, the device achieves dynamic temperature tracking and simulates both the pavement temperature and ice-snow status through active cooling. A temperature control model for the device is established by incorporating heat transfer theory and applying small-signal analysis methods. To address the system's inherent nonlinear and coupled characteristics, a fuzzy adaptive PID decoupling control strategy is proposed. This strategy enables accurate pavement temperature tracking and facilitates active early warnings for runway icing. The results demonstrate that under typical input signals, the proposed control method exhibits excellent steady-state performance, high tracking accuracy, and robust anti-interference capability. Specifically, the active simulation device tracks the target temperature with a steady-state error not exceeding 0.08 degrees C. The tracking error follows a normal distribution, with 99.7% of the error samples falling within +/- 0.2 degrees C. This performance improvement is statistically significant compared to traditional PID control (p < 0.05), thereby effectively simulating the thermal evolution of the runway surface. Ultimately, this research provides a novel approach for active runway icing early warnings, aligning with the Global Reporting Format (GRF).
Convex Combination Filtered-x Least Mean Square (C-FxLMS) algorithms face an inherent conflict between achieving rapid convergence speed and smaller steady-state error. To address this limitation, this paper proposes a multi-agent Game Theory Adaptive Weighting (GTAW) based C-FxLMS control strategy for active vibration suppression. The framework integrates two different sub-controllers within a strategic competition process, specifically an efficiency-oriented agent utilizing a variable step size and an accuracy-oriented agent employing a small fixed step size, a state-dependent Nash equilibrium strategy is designed to dynamically modulate the control weights. This strategy allows the system to balance the conflicting objectives of rapid convergence speed and smaller steady-state error. The convergence and stability of the Active Vibration Control (AVC) system are rigorously verified through the construction of a Lyapunov function and boundary analysis. Finally, quantitative results across four distinct scenarios demonstrate that the proposed method achieves an average reduction of 23.9% in steady-state error and a 41.7% decrease in control energy consumption compared to benchmark methods.
This study aims to address the challenge of detecting wood surface defects with high accuracy in real-world manufacturing environments, which are affected by various conditions. The goal is to build a hybrid deep learning model that can operate effectively in real time, serving automated quality control in wood processing. The research team proposes a hybrid architecture that combines multiple models: You Only Look Once - version 10 and CenterNet for accurate and fast localization of defect areas, Graph Attention Network to exploit spatial relationships between features, and Multilayer Perceptron for the final classification stage. At the same time, the Battle Royale Optimization algorithm is applied to select the optimal feature. The model is trained and tested on a large-scale dataset of 20,275 high-resolution wood surface images, covering 10 different types of defects. Experiments show that the model achieves a classification accuracy of 92.7% and a processing speed of 40 fps, meeting the requirements of real-time industrial systems. The model also demonstrates good generalization ability when effectively detecting both obvious defects and subtle abnormalities. The proposed hybrid modeling framework not only improves the efficiency of wood surface defect detection but also provides an automated, robust, and highly scalable solution for quality control in smart manufacturing. This contributes to optimizing the use of raw materials, minimizing waste, and enhancing the competitiveness of the wood processing industry.
