Yuan Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Yuan Yang has a total of 199 co-authors. Here are the top co-authors by number of shared publications:
- Fengjuan Wang
Shaanxi University Key Laboratory of Photonic Power Devices and Discharge Regulation, Xian University of Technology, Xian, China
18 shared publications Ningmei Yu
Xian University of Technology,Department of Electronic Engineering,Xian,China
16 shared publications- Ningmei Yu
Xi’an University of Technology
14 shared publications - Xiangkun Yin
Xidian University
14 shared publications - Wei Xiang
La Trobe University
13 shared publications - Yang Wen
Xian University of Technology, Xian, China
13 shared publications
Latest Publications
- 19 November 2026
- Published by IOP Publishing in Physiological Measurement
Adaptive observation–history fusion for EEG-based tracking of PERCLOS-anchored drowsiness
Objective. This study evaluates explicit adaptive observation–history fusion for continuous EEG-based tracking of PERCLOS-anchored drowsiness and tests whether reward-driven proximal policy optimisation (PPO) provides predictive benefit over supervised fusion and temporal alternatives. Approach. SEED-VIG data from 21 participants and 23 recording sessions were evaluated using strict 21-fold subject-wise leave-one-subject-out (LOSO) validation. In each fold, an EEG encoder generated a PERCLOS-supervised observation proxy, which was recursively combined with the previous estimate through an adaptive observation–history fusion coefficient. Eleven primary methods spanned raw-proxy prediction, conventional filters, supervised and PPO-based fusion, and neural temporal models. Percentage of eyelid closure over time (PERCLOS) provided the ocular-behavioural reference for supervision, PPO reward construction, model selection, and evaluation. Main results. Among methods available in all 21 folds, S4D had the lowest descriptive mean root-mean-square error (RMSE; 0.240746), followed by Gate-7D (0.241490). PPO-7D was available in 20/21 folds. In the prespecified matched comparison (n = 20), the mean RMSE difference (Gate-7D minus PPO-7D) was -0.0022941 (95% bootstrap CI -0.00813595 to 0.00361267; Wilcoxon p = 0.348810), which did not establish a predictive advantage for PPO-7D. None of the 18 broader prespecified PPO-versus-alternative RMSE comparisons met its rejection criterion after Holm correction. Validation-side controlled perturbation altered the EEG-derived proxy, but the prespecified stronger reliability-response criterion was not satisfied. Significance. Explicit observation–history fusion makes the relative contributions of the current EEG proxy and recursive history inspectable. PPO yielded descriptively slower, more persistent fusion-coefficient dynamics than Gate-7D, but its predictive advantage and necessity were not established. The coefficient describes recursive weighting, not calibrated observation reliability.
- 19 November 2026
- Published by IOP Publishing in Physiological Measurement
Adaptive observation–history fusion for EEG-based tracking of PERCLOS-anchored drowsiness
Objective. This study evaluates explicit adaptive observation–history fusion for continuous EEG-based tracking of PERCLOS-anchored drowsiness and tests whether reward-driven proximal policy optimisation (PPO) provides predictive benefit over supervised fusion and temporal alternatives. Approach. SEED-VIG data from 21 participants and 23 recording sessions were evaluated using strict 21-fold subject-wise leave-one-subject-out (LOSO) validation. In each fold, an EEG encoder generated a PERCLOS-supervised observation proxy, which was recursively combined with the previous estimate through an adaptive observation–history fusion coefficient. Eleven primary methods spanned raw-proxy prediction, conventional filters, supervised and PPO-based fusion, and neural temporal models. Percentage of eyelid closure over time (PERCLOS) provided the ocular-behavioural reference for supervision, PPO reward construction, model selection, and evaluation. Main results. Among methods available in all 21 folds, S4D had the lowest descriptive mean root-mean-square error (RMSE; 0.240746), followed by Gate-7D (0.241490). PPO-7D was available in 20/21 folds. In the prespecified matched comparison (n = 20), the mean RMSE difference (Gate-7D minus PPO-7D) was -0.0022941 (95% bootstrap CI -0.00813595 to 0.00361267; Wilcoxon p = 0.348810), which did not establish a predictive advantage for PPO-7D. None of the 18 broader prespecified PPO-versus-alternative RMSE comparisons met its rejection criterion after Holm correction. Validation-side controlled perturbation altered the EEG-derived proxy, but the prespecified stronger reliability-response criterion was not satisfied. Significance. Explicit observation–history fusion makes the relative contributions of the current EEG proxy and recursive history inspectable. PPO yielded descriptively slower, more persistent fusion-coefficient dynamics than Gate-7D, but its predictive advantage and necessity were not established. The coefficient describes recursive weighting, not calibrated observation reliability.
- 10 November 2026
- Published by IOP Publishing in Biomedical Physics & Engineering Express
- Vol. 12 (6) , 065013
ST-M 2 EDA: spatio-temporal metric-guided multi-stream expert domain adaptation network for fatigue monitoring
In the human-machine collaboration paradigms of Industry 5.0, the intrinsic non-stationarity and domain drift of electroencephalogram signals pose significant challenges to fatigue monitoring. Consequently, achieving accurate and robust cross-subject fatigue detection is paramount for workplace safety. This paper proposes a Spatio-Temporal Metric-Guided Multi-Stream Expert Domain Adaptation Network. The framework initially employs Singular Value Decomposition-based signal reconstruction to attenuate non-stationary noise, followed by Mixup data augmentation to enhance and smooth the feature space distribution. Furthermore, a hybrid architecture integrating Domain-Specific Batch Normalization with a ResNet-Transformer backbone is constructed to precisely decouple statistical discrepancies between domains. Finally, Multi-Kernel Maximum Mean Discrepancy is introduced as a dynamic routing criterion to facilitate fine-grained, adaptive decision-making for target samples. The baseline experiments on the SEED-VIG dataset yield a subject-wise mean accuracy of 90.54% across all 23 subjects, including S15. Visual analysis of brain topography confirms that the model successfully aligns distributional shifts caused by individual heterogeneity—specifically correcting for specific frontotemporal lobe activation observed in subject S7—and establishes distinct decision boundaries. These findings indicate that synergizing signal reconstruction with Spatio-Temporal metric-guided multi-stream experts effectively mitigates negative transfer, ensuring reliable fatigue detection in complex cross-subject tasks.
- 10 November 2026
- Published by IOP Publishing in Biomedical Physics & Engineering Express
- Vol. 12 (6) , 065013
ST-M 2 EDA: spatio-temporal metric-guided multi-stream expert domain adaptation network for fatigue monitoring
In the human-machine collaboration paradigms of Industry 5.0, the intrinsic non-stationarity and domain drift of electroencephalogram signals pose significant challenges to fatigue monitoring. Consequently, achieving accurate and robust cross-subject fatigue detection is paramount for workplace safety. This paper proposes a Spatio-Temporal Metric-Guided Multi-Stream Expert Domain Adaptation Network. The framework initially employs Singular Value Decomposition-based signal reconstruction to attenuate non-stationary noise, followed by Mixup data augmentation to enhance and smooth the feature space distribution. Furthermore, a hybrid architecture integrating Domain-Specific Batch Normalization with a ResNet-Transformer backbone is constructed to precisely decouple statistical discrepancies between domains. Finally, Multi-Kernel Maximum Mean Discrepancy is introduced as a dynamic routing criterion to facilitate fine-grained, adaptive decision-making for target samples. The baseline experiments on the SEED-VIG dataset yield a subject-wise mean accuracy of 90.54% across all 23 subjects, including S15. Visual analysis of brain topography confirms that the model successfully aligns distributional shifts caused by individual heterogeneity—specifically correcting for specific frontotemporal lobe activation observed in subject S7—and establishes distinct decision boundaries. These findings indicate that synergizing signal reconstruction with Spatio-Temporal metric-guided multi-stream experts effectively mitigates negative transfer, ensuring reliable fatigue detection in complex cross-subject tasks.
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (22) , 55403-55416
A High-Robustness Surrogate Modeling Method for On-Chip Microwave Passive Components in Chiplet-Based IoT Integrated Microsystem
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (22) , 55403-55416
A High-Robustness Surrogate Modeling Method for On-Chip Microwave Passive Components in Chiplet-Based IoT Integrated Microsystem
- 26 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. PP (02780046) , 1-14
Physics-Guided Deep Learning for Online Bond Wire Lift-Off Monitoring in IGBTs via Acoustic Emission
Bond wire lift-off is one of the dominant failure mechanisms in insulated gate bipolar transistor (IGBT) power modules. This article presents a physics-guided acoustic emission (AE) monitoring framework that couples mechanism modeling with a lightweight deep network (AE–STFT–R18-Lite-SE) for real-time bond wire health assessment. A coupled electromagnetic-force and constrained-thermal-expansion analysis identifies two degradation-sensitive AE frequency bands and derives closed-form amplitude scaling laws versus the number of lifted wires. These physics priors are systematically embedded into short-time Fourier transform (STFT) parameterization, squeeze-and-excitation attention allocation, channel pruning, and 8-bit integer (INT8) quantization, yielding a 1.9 M-parameter model with 3.2 ms inference latency on an edge multiprocessor system-on-chip (MPSoC). Comprehensive experimental validation on a multimodule platform covering eight degradation states grouped into four health levels achieves 98.6% classification accuracy. Accuracy above 95.9% is retained under noise, sensor placement variation, solder voiding, cross-package transfer, ambient temperature, H-bridge PWM drive, and full-lifecycle power cycling to natural failure. These results demonstrate that physics-informed AE monitoring offers an early, noninvasive, operating-point-robust approach to power module condition assessment.
- 26 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. PP (02780046) , 1-14
Physics-Guided Deep Learning for Online Bond Wire Lift-Off Monitoring in IGBTs via Acoustic Emission
Bond wire lift-off is one of the dominant failure mechanisms in insulated gate bipolar transistor (IGBT) power modules. This article presents a physics-guided acoustic emission (AE) monitoring framework that couples mechanism modeling with a lightweight deep network (AE–STFT–R18-Lite-SE) for real-time bond wire health assessment. A coupled electromagnetic-force and constrained-thermal-expansion analysis identifies two degradation-sensitive AE frequency bands and derives closed-form amplitude scaling laws versus the number of lifted wires. These physics priors are systematically embedded into short-time Fourier transform (STFT) parameterization, squeeze-and-excitation attention allocation, channel pruning, and 8-bit integer (INT8) quantization, yielding a 1.9 M-parameter model with 3.2 ms inference latency on an edge multiprocessor system-on-chip (MPSoC). Comprehensive experimental validation on a multimodule platform covering eight degradation states grouped into four health levels achieves 98.6% classification accuracy. Accuracy above 95.9% is retained under noise, sensor placement variation, solder voiding, cross-package transfer, ambient temperature, H-bridge PWM drive, and full-lifecycle power cycling to natural failure. These results demonstrate that physics-informed AE monitoring offers an early, noninvasive, operating-point-robust approach to power module condition assessment.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 19, 1-18
PND-Net: A Unified Framework with Attention Fusion and Dynamic Feature Mining for Visible-Infrared Object Detection
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 19, 1-18
Open AccessPND-Net: A Unified Framework with Attention Fusion and Dynamic Feature Mining for Visible-Infrared Object Detection
- 30 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 73 (5) , 7937-7948
Junction Temperature Estimation Method for Power Modules Based on Gate Voltage Undershoot Amplitude-Phase 2-D Features WithLoad Current Independence
Junction temperature ($T_{\mathrm{j}}$) monitoring plays a crucial role in predicting the lifetime and assessing the reliability of power semiconductor modules. However, existing $T_{\mathrm{j}}$ estimation methods face three major challenges: insufficient measurement accuracy, strong load current dependence, and vulnerability to bond wire degradation. To address these issues, this article proposes a novel dynamic junction temperature monitoring method based on 2-D amplitude-phase feature fusion derived from gate voltage undershoot signals. The core innovations are twofold: 1)fusion of time-domain amplitude and phase characteristics constructs a 2-D temperature-sensitive feature, leveraging multidimensional complementarity to improve $T_{\mathrm{j}}$ estimating accuracy and enhance anti-interference capabilities; and 2) integration of Gramian angular field (GAF) encoding with convolutional neural networks (CNNs) enables load-current-independent junction temperature estimation. Moreover, the proposed method demonstrates inherent immunity to bond wire degradation. Its effectiveness is validated through experimental double-pulse tests, which show that the maximum $T_{\mathrm{j}}$ estimation error remains within $\pm 2\ ^{\circ}$C, even under varying load conditions.
- 30 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Electronics
- Vol. 73 (5) , 7937-7948
Junction Temperature Estimation Method for Power Modules Based on Gate Voltage Undershoot Amplitude-Phase 2-D Features WithLoad Current Independence
Junction temperature ($T_{\mathrm{j}}$) monitoring plays a crucial role in predicting the lifetime and assessing the reliability of power semiconductor modules. However, existing $T_{\mathrm{j}}$ estimation methods face three major challenges: insufficient measurement accuracy, strong load current dependence, and vulnerability to bond wire degradation. To address these issues, this article proposes a novel dynamic junction temperature monitoring method based on 2-D amplitude-phase feature fusion derived from gate voltage undershoot signals. The core innovations are twofold: 1)fusion of time-domain amplitude and phase characteristics constructs a 2-D temperature-sensitive feature, leveraging multidimensional complementarity to improve $T_{\mathrm{j}}$ estimating accuracy and enhance anti-interference capabilities; and 2) integration of Gramian angular field (GAF) encoding with convolutional neural networks (CNNs) enables load-current-independent junction temperature estimation. Moreover, the proposed method demonstrates inherent immunity to bond wire degradation. Its effectiveness is validated through experimental double-pulse tests, which show that the maximum $T_{\mathrm{j}}$ estimation error remains within $\pm 2\ ^{\circ}$C, even under varying load conditions.