Search Results (1,152)

Currently, data-driven fault diagnosis methods have achieved remarkable progress. However, in industrial scenarios, acquiring a sufficient amount of fault data poses a challenge, thereby leading to the issue of imbalanced data in intelligent fault diagnosis. Furthermore, manual recording and instrument measurement errors will introduce label noise, which significantly impacts diagnosis performance. To address these problems, this paper proposes a safe-domain generative adversarial network with Swin Transformer (SDGAN-ST). A safe domain selection method is utilized to eliminate noisy samples and construct a pure dataset that poses no risk to the GAN training process. Consequently, GAN can generate high-quality minority samples to rebalance the original dataset. Additionally, the Swin Transformer is employed as a classifier to capture global information for each fault sample, thereby achieving high diagnostic accuracy. Experiments on the CWRU dataset and a real-world oxygen compressor bearing dataset demonstrate the effectiveness of the proposed method. On the CWRU dataset, SDGAN-ST achieves accuracies of 98.88%, 97.63%, and 97.50% under imbalance ratios of 1:10, 1:20, and 1:30, respectively. On the real-world dataset, SDGAN-ST achieves 100% accuracy under all three imbalance ratios. Additional experiments under noise ratios of 20%, 30%, and 40% show that SDGAN-ST maintains stable diagnostic performance and is more robust to label noise than ordinary WGAN-GP-based methods. Full article

Deepfakes threaten sensor-based authentication systems, including biometric sensors, surveillance cameras, and IoT edge devices. Unimodal detectors remain vulnerable to modality-specific attacks. We propose a multimodal deepfake detection framework optimized for resource-constrained edge devices, featuring a novel cross-modal attention fusion mechanism with adaptive gating. The architecture combines enhanced Res2Net for audio, temporal 3D CNN with SE attention for video, and bidirectional cross-modal attention with quality-based gates. On our benchmark (5472 audio + 1842 video samples), the fusion model achieves 96.7% accuracy, 96.6% F1-score, 0.988 AUC-ROC, and 3.3% EER. Adversarial testing shows 92.3% accuracy under the Fast Gradient Sign Method (FGSM) attack. The model has a 30.3 MB footprint and runs at 20 FPS on edge hardware. Modality contribution analysis reveals adaptive weighting (72% audio for TTS forgery, 78% video for lip-synced attacks). Cross-dataset evaluation on FakeAVCeleb achieves 92.3% overall accuracy, confirming generalization. Full article

Large-scale renewable energy integration introduces random power fluctuations into microgrids, increasing the difficulty of frequency regulation. To improve regulation stability and training efficiency, this article proposes sample selection generative adversarial networks (SSGANs) based on sample selection networks (SSNs), conditional generative adversarial networks (CGANs), and the actor–critic framework. First, the SSNs are trained to evaluate sample information values and prioritize informative samples for model training. Second, the CGANs learn the conditional mapping between microgrid operating states and control actions, and the pretrained generator is transferred into the actor–critic framework as the actor. Third, the actor–critic framework further optimizes the control policy online to generate real-time frequency regulation commands. The proposed method is tested on a standard two-area system and further validated on a complex four-area system. Case studies show that SSGANs achieve faster convergence and better frequency regulation performance than typical control algorithms. Full article

This study addresses the challenge of modeling the complex non-linear relationship between process parameters and relative density in selective laser melting (SLM) of IN718 nickel-based superalloy under small-sample conditions. A data-driven prediction framework integrating data augmentation, physics-informed feature engineering, machine learning, and model interpretability analysis was developed and systematically validated. Fourteen sets of experimental data covering both vertical and horizontal building directions were collected by varying laser power (P), scan speed (v), and hatch spacing (h). To overcome the small-sample limitation, three augmentation strategies—radial basis function (RBF) interpolation, generative adversarial network (GAN), and K-nearest neighbors (KNN)—were systematically compared under unified physical constraints combining local perturbation and volumetric energy density (E_vol) filtering, with Pearson correlation coefficient consistency used to select the optimal strategy. Eight physically meaningful input features were constructed, including E_vol and line energy density (E_line), explicitly embedding SLM process physics into the learning framework. Support vector regression (SVR), random forest (RF), and artificial neural network (ANN) models were trained and their hyperparameters were systematically optimized via exhaustive grid search combined with leave-one-out cross-validation (LOO-CV), ensuring robust model selection under small-sample constraints. A physics-based baseline model (E_vol quadratic fitting, LOO-CV average R² = 0.2534) was established to quantify the gain of machine learning over empirical formulas. LOO-CV results show that ANN achieves the highest average R² of 0.9269, followed by SVR (0.9148) and RF (0.8393), all of which substantially outperform the physical baseline. Feature importance analysis reveals that E_vol accounts for 51.58% of the predictive power, and ablation experiments confirm that introducing physics-derived features improves the average R² by 0.0246 compared with raw process parameters alone. To further elucidate the predictive mechanism of the optimal ANN model, Partial Dependence Plot (PDP) analysis was conducted for all eight input features, visualizing their marginal effects on predicted density and confirming physical consistency with SLM mechanisms. This framework provides a reliable, interpretable, data-driven solution for intelligent SLM process optimization with limited experimental data. Full article

To address fault diagnosis under limited sample conditions, this paper proposes a small-sample diagnosis framework integrating multibody dynamic modeling and a GAN–CNN fusion strategy. First, a rigid–flexible coupled multibody dynamic model of the reducer is established to simulate vibration responses under typical fault modes, including broken gear tooth, gear wear, and bearing outer ring fault, thereby generating representative simulation samples. Second, to reduce the distribution discrepancy between simulated and measured data, the simulated samples are introduced into a generative adversarial learning framework for feature enhancement, with limited measured samples used as references. Cosine similarity is employed to evaluate the consistency between the enhanced simulated data and the measured data in the feature space. Finally, the enhanced simulated samples are fused with measured samples to construct a hybrid dataset for convolutional neural network training and fault classification. Experimental results show that the proposed framework improves the similarity between simulated and measured data, with cosine similarity increasing from below 0.65 to above 0.80. Under small-sample conditions, the mean diagnosis accuracy reaches 83.81%, which is 17.33 percentage points higher than that obtained using the original small-sample dataset. The proposed framework provides an effective modeling and algorithmic approach for reducer fault diagnosis under data-scarce conditions. Full article

With the rapid development of the Internet of Things (IoT) and the increase in the frequency of cyberattacks, accurate identification of IoT end devices is critical to their security. Existing identification methods are based on raw, statistical, and deep features of network traffic, each with their own advantages and disadvantages. Raw feature-based methods have difficulty performing feature extraction and insufficient information. As such, the recognition accuracy of statistical feature-based methods is limited by the distinguishment machine learning classifiers, and the deep feature-based methods do not take into account the problem of large differences in traffic samples, which leads to low recognition accuracy in some devices. For this reason, this paper proposes the IoTDI-ImbS method. The method selects the network traffic payload information as the original features and converts them into grayscale images; uses a generative adversarial network-based IoT terminal devices traffic generation (NTGAN) algorithm to generate traffic samples for devices with fewer samples through generative adversarial network to solve the sample imbalance problem; and constructs a ResNet18-BiLSTM model, mining spatial features with ResNet18 and extracting temporal features with BiLSTM to improve recognition accuracy. The experimental results on different sizes of IoT terminal device datasets show that IoTDI-ImbS has performance advantages over other methods in recognition accuracy, better leverages the sample imbalance problem in the dataset, and provides a more effective solution for IoT device recognition. Experimental results on the UNSW and IoT Sentinel dataset demonstrate that IoTDI-ImbS significantly outperforms baseline methods. Specifically, on the UNSW dataset, our method achieves an overall accuracy of 99.1% and an F1-score of 0.985. After integrating the NTGAN module, the identification accuracy for minority classes improved by approximately 3.5%. On the IoT Sentinel dataset, the model maintains a high precision of 98.7%, proving its robustness in diverse IoT environments. Full article

This study investigates class imbalance in motor fault diagnosis. Fault samples, especially those at different severity levels, are often much fewer than healthy samples. To address this issue, a self-attention guided Wasserstein conditional GAN with gradient normalization (SWGAN) is proposed. The method is based on synchronized three-phase current and vibration measurements. It separately generates label-conditioned current spectra and vibration spectra to supplement minority fault classes. Self-attention is used to capture long-range spectral dependencies. Gradient normalization is introduced to improve adversarial training stability. The generated current and vibration spectra are then fused at the feature level and fed into a stacked autoencoder (SAE)-based multi-modal classifier. Experiments were conducted on a PMSM stator fault dataset and a variable-speed three-phase asynchronous motor dataset. On the PMSM dataset, SWGAN achieved highest accuracies of 98.90% and 97.81% under two fault-category imbalance settings. On the variable-speed motor dataset, the proposed method achieved accuracies of 98.10% and 97.65%, respectively. These results show that SWGAN can provide effective supplementary samples for minority fault classes. They also indicate that the proposed method improves diagnostic performance under both fixed-speed and variable-speed conditions. Full article

Synthetic aperture radar (SAR) is of great value in intelligent image interpretation. However, the acquisition of real SAR data is costly, and manual annotation heavily relies on expert experience. These factors severely restrict the development of SAR intelligent interpretation algorithms. Meanwhile, the high-frequency details of SAR images contain rich target information. Traditional generation methods cannot effectively capture these key features. To address the above issues, this paper proposes a dual-branch feature decoupling generative adversarial network (GAN) with wavelet constraint designed to achieve high-quality and parameter-controllable SAR image generation. The framework leverages discrete wavelet transform (DWT) to separate spatial structure from high-frequency details, which are independently modeled by a structure branch and a detail branch, respectively. A wavelet consistency loss function is introduced to constrain the distribution of generated and real images in high-frequency subbands, thereby enhancing the model’s capability to model scattering details. To fuse features from the two branches, a cross-attention fusion module is adopted to realize the adaptive compensation of structural features with texture details. Furthermore, to achieve joint control over the semantic attributes and azimuth of generated samples, the framework further integrates auxiliary classification and azimuth regression tasks. A multi-task learning mechanism is constructed to realize precise control over the target’s semantic category and azimuth. For the continuous variable of azimuth, an angle-aware hypernetwork transform module is introduced to perform dynamic convolution modulation on the structure branch at the feature map scale, which improves the model’s fine control capability over continuous azimuth variations. Experimental results on the MSTAR dataset demonstrate that the proposed model can significantly improve the semantic consistency and visual fidelity of the generated samples. The generated samples exhibit high statistical alignment with real data distributions, confirming the model’s effectiveness in characterizing the feature space of SAR imagery and enabling controllable SAR data simulation, thereby augmenting datasets for image interpretation tasks. Full article

Under the Active Vision hypothesis, eye movements are not passive responses to visual stimuli but are actively guided by task demands and internal goals. In driving, scanpaths may therefore reflect an ongoing process of information sampling for risk assessment. However, current computational models often isolate scanpath prediction from risk assessment, overlooking their intrinsic cognitive coupling. In this study, we investigate whether driver scanpaths and traffic risk perception can be jointly modeled within a unified framework. We propose a computational approach based on the introduced Adversarial Inverse Reinforcement Learning (AIRL), where gaze behavior is interpreted as a policy that maximizes a latent safety-related reward. By employing a generator to simulate human-like sequences of fixations and saccades, and a discriminator to approximate the internal reward signal, our framework ensures that generated scanpaths synergistically inform downstream risk perception. To facilitate this research, we constructed the BDDA dataset, aggregating over 13,000 spatio-temporal gaze points with explicit risk annotations to study this joint mechanism. Experimental results indicate that simultaneously modeling the “where” (scanpath dynamics) and the “why” (risk perception) significantly outperforms the compared baseline methods on the proposed BDDA dataset. These findings provide computational evidence for a functional coupling between visual attention and risk perception, supporting the view that eye movements serve as an active mechanism for acquiring task-relevant information in safety-critical environments. Full article

This paper proposes a spectrum-aware generative learning framework for intelligent motor fault diagnosis under small-sample conditions. To address the challenges of insufficient labeled fault data and imbalanced distributions in motor systems, a hybrid model integrating a generative adversarial network (GAN) with an attention-enhanced deep neural network is developed. First, vibration signals of the motor are transformed into time–frequency representations to capture discriminative spectral features. Then, the GAN is employed to augment minority classes and improve data diversity, while the SE (squeeze-and-excitation) mechanism enhances feature extraction by emphasizing critical fault-related components. Finally, a deep classifier is trained on the augmented dataset for fault identification. Experimental results on benchmark datasets demonstrate that the proposed method achieves superior diagnostic accuracy and robustness compared with several state-of-the-art approaches, especially under severe data scarcity and imbalance scenarios. The results indicate that the proposed framework effectively improves generalization performance and provides a reliable solution for intelligent motor fault diagnosis in practical industrial applications. Full article

Graph neural networks (GNNs) are highly effective on relational data, yet their performance degrades sharply when graph topology is poisoned before training. Existing defenses usually assume a fixed attack pattern and a fixed graph structure, which makes them brittle when the poisoned graph changes across attacks, perturbation budgets, or deployment conditions. We propose GADD, a game-inspired adversarial distillation framework for robust graph defense. GADD first constructs multiple positive and negative graph views through a homophily-aware graph sampling scheme, allowing the model to learn from both purified and high-risk subgraphs. It then trains a heterogeneous group of student GNNs online, where each student receives global class-distribution knowledge from its peers and local structural knowledge through an adversarial cyclic distillation objective. Finally, GADD replaces uniform ensembling with an entropy-regularized adaptive aggregation rule that assigns graph-adaptive weights according to confidence and inter-model agreement. On Cora, CiteSeer, and PubMed, GADD consistently improves robustness against both Meta and Nettack attacks while preserving clean accuracy. Under the strongest Meta and Nettack settings in the main benchmark, GADD improves the best competing baseline by up to 2.99 and 3.42 percentage points, respectively. Additional ablations show that graph sampling, adversarial distillation, and adaptive aggregation all contribute materially to the final robustness gains. Full article

This study aims to explore a leakage-aware and explainable machine learning framework for predicting tunnel water inrush quantity (WIQ) under small-sample and high-heterogeneity geological conditions. A project-level dataset was compiled at a fixed spatial granularity of 30 m per excavation segment by integrating forward prospecting outputs, construction-face observations, and geological reports, and six hydrogeological–structural indicators were used to predict the water inflow rate in cubic meters per hour. To overcome data scarcity and improve generalization, a tabular generative adversarial network (GAN) was introduced to augment the training distribution while preserving marginal statistics and inter-variable dependence, and a swarm-intelligence optimizer was employed to tune a Categorical Boosting (CatBoost) regressor for stable performance. In addition, six mainstream tree-based learners were benchmarked under a unified protocol, and model transparency was ensured through a multi-level interpretability suite combining SHapley Additive exPlanations (SHAP) attribution, partial dependence with individual conditional expectation (ICE) diagnostics, and interaction surfaces. Results show that, under the present fixed split, training-set augmentation was associated with improved performance for the evaluated baseline learners, and the proposed hybrid model achieved encouraging hold-out accuracy. However, because the dataset contains only 55 real samples and the test set contains only 11 real samples, the reported performance should be interpreted as an initial project-specific indication rather than robust evidence of generalizable reliability. Interpretability analyses further identify lithologic and reflector-related factors as dominant drivers, and reveal nonlinear response patterns and interaction-sensitive high-risk regions. Overall, the proposed framework shows potential to improve predictive performance and engineering interpretability for the studied project, and may provide a useful reference for drainage and reinforcement planning. Further confirmation through repeated data splitting, additional samples, and external validation is still needed before broader application. Full article

Industrial Internet of Things (IIoT) intrusion detection requires compact, latency-efficient models whose behavior remains assessable under adversarial stress, yet compression can alter the feature-attribution structure learned by a full-precision model. This paper presents X-GATE (eXplanation-Guided Adversarial Training Engine), an attribution-aware training framework for compressed Edge-IIoT intrusion detection. X-GATE combines Explanation-Consistency Distillation (ECD), which aligns Teacher–Student feature-attribution rankings with a differentiable soft-rank Spearman penalty, and Explanation-Guided Adversarial Training (EGAT), which hardens the Student on Teacher-salient feature coordinates. On the full Edge-IIoTset 2022 benchmark, the latest three-seed ablation gives Full X-GATE 89.30 ± 3.89% F1-Macro with 0.617 M parameters, within approximately 0.6 percentage points of the full-precision Teacher; a Random Forest model remains a stronger clean-F1 reference, so X-GATE is not framed as the clean-accuracy optimum. In a separate deployment-subset rerun, X-GATE obtains 78.83 ± 5.83% float F1-Macro and 79.11 ± 5.47% INT8 F1-Macro, reduces the adversarial false-positive rate from 0.46 ± 0.08% for KD-only to 0.16 ± 0.09% under the evaluated single-step white-box explanation-evasion protocol, and reduces CPU latency from 4.16 to 1.25 ms/sample. Component ablation further shows that ECD reduces Logical Drift by 17.24%, while EGAT improves adversarial F1 by 10.57 percentage points. Taken together, these benchmark- and protocol-bounded results position X-GATE as a compact neural operating point for the Edge-IIoT setting studied here, balancing attribution consistency, targeted hardening, and CPU-side efficiency. Full article

Aiming at the problems of extremely scarce measured samples and significant domain shift between simulated and measured data in automatic target recognition (ATR) of air targets for spaceborne radar, this paper proposes an inverse synthetic aperture radar (ISAR) image recognition method for air targets combining physics-driven data augmentation guided by detection prior information with domain adversarial transfer learning. First, the mapping relationship between scattering point projection and ISAR images is established by using the target 3D point cloud and radar observation geometric priors, and a 2D sinc kernel function is introduced for energy distribution rendering. Then, under the unsupervised transfer learning paradigm, aiming at the distribution inconsistency between augmented data (source domain) and unlabeled simulated data (target domain), this paper designs a cross-domain recognition task experiment including six types of typical aircraft targets, and compares the cross-domain recognition performance of three transfer learning methods (model fine-tuning, deep domain confusion (DDC) and domain-adversarial neural networks (DANN)) on the target domain. Meanwhile, t-distributed stochastic neighbor embedding (t-SNE) visualization is used to analyze the feature distribution alignment ability of the models. Simulation experiments show that the DANN model with a dynamic inversion coefficient introduced in the gradient reversal layer (GRL) achieves a recognition accuracy of 99.5% on the unlabeled target domain, which is significantly superior to the model fine-tuning and DDC methods. Moreover, it makes the feature distributions of source and target domain samples highly overlapping, and maintains a strong inter-class discriminability while eliminating the domain shift. The proposed scheme provides a physically interpretable and robust technical path for few-shot radar target image recognition. Full article

The reinforcement and retrofit of damaged steel buildings has emerged as a primary focus in civil engineering. It should be noted that completing the reasonable strengthening design for avoiding the sudden collapse of a structure in extreme engineering conditions was an urgent task, while the existing method required a long time which significantly influenced the reinforcing practice. In the present study, an improved explainable machine learning (ML) framework was developed for the rapid assessment of the bending property of repaired laminated channel beams. Firstly, a comprehensive database of 192 samples combining experimental and finite element data was established. The Mahalanobis distance analysis and Pearson correlation analysis were sequentially performed to evaluate the singularity of the samples and the dependencies between the variables. Secondly, the adversarial tests were conducted on the randomly selected 10 pairs of training and testing sets to determine the database with the best distribution consistency. Then, three machine-learning models of artificial neural networks (ANN), random forest (RF), and extreme gradient boosting tree (XGBoost) were respectively trained and validated. Finally, the explainability analysis of the XGBoost model was carried out in the global and local perspectives based on the SHAP method. The prediction accuracy (R²) of all ML models exceeded 90%, demonstrating good accuracy and providing a useful reference within the current database for the reinforcement design of damaged steel beams in emergency situations. In addition, the XGBoost model achieved superior prediction accuracy (R² = 97.98%) and stability (CoV = 0.82%) compared to ANN and RF. The explainability analysis revealed that boundary conditions and load type had the most significant influence on bending capacity. The proposed ML approach enabled efficient and reliable bending capacity estimation, supporting rapid decision-making in emergency reinforcement scenarios for damaged steel structures. Full article