Key Takeaways & Executive Findings
- •• • MCOAN replaces the fixed empirical threshold of 0.5 used in conventional open-set back-propagation (OSBP) adversarial loss with per-sample dynamic thresholds derived from dual K-way and K+1-way classifier similarity scores, eliminating the subjective prior threshold setting that causes negative transfer when unknown-class proportions vary across wind farm sites. • • The non-adversarial domain classifier stabilizes dynamic weight computation, preventing the weight oscillation that degrades known-class alignment when adversarial gradients from unknown samples dominate the shared feature space under variable speed and load conditions. • • Validation across two datasets confirms high-precision shared-class feature distribution alignment and unknown-class recognition, with robustness maintained under the distribution shift regime characteristic of wind turbine drivetrains operating at variable rotational speeds and torque loads. • • The architecture addresses the closed-set assumption failure documented in domain adversarial neural networks (DANN) and statistical moment matching methods, which misclassify unknown bearing fault modes as known classes and induce negative transfer when the entire target domain is blindly aligned to the source domain.
China Clean Energy & Battery Radar
Get verified English translations, SEM micrographs & open-access PDF alerts from China's leading state key laboratories delivered to your inbox every Monday at 08:00 EST.
Abstract
Rolling bearings in wind turbine generator systems operate under variable speed and load conditions that induce significant data distribution shifts between training and field data, while unknown fault modes absent from the source domain are frequently misclassified as known classes. This study proposes a multi-classifier open adversarial network (MCOAN) for open-set fault diagnosis. Within an adversarial domain adaptation framework, a K-way classifier and an additional K+1-way one-vs-all classifier independently estimate target-sample similarity to the source domain. These similarity scores drive a dynamic weighting mechanism that adaptively reweights target samples during open-set adversarial training and supplies per-sample dynamic thresholds for known/unknown discrimination, thereby promoting cross-domain alignment of shared-class features while suppressing negative transfer from unknown samples. A non-adversarial domain classifier is introduced to stabilize dynamic weight estimation. Validation on two datasets demonstrates high-precision shared-class distribution alignment and unknown-class recognition with favorable robustness. The method removes reliance on empirically preset thresholds that plague conventional open-set back-propagation approaches, where a fixed threshold of 0.5 in the binary cross-entropy adversarial loss provides no per-sample adaptivity. By coupling K-way and K+1-way similarity estimates, MCOAN achieves simultaneous known-class alignment and unknown-class separation without prior knowledge of the unknown-class cardinality, addressing a persistent bottleneck in wind turbine drivetrain condition monitoring where unanticipated bearing failure modes emerge under field conditions.
1. Introduction
Data-driven rolling bearing fault diagnosis for wind turbine generator systems has advanced rapidly through deep learning, yet commercial deployment remains constrained by a fundamental mismatch: models trained on source-domain data collected under controlled conditions degrade sharply when applied to target-domain data acquired under variable rotational speed and load. Domain adaptation techniques, including domain adversarial neural networks and statistical moment matching, partially mitigate this distribution shift but operate under a closed-set assumption that the source and target label spaces are identical. In field operations, target domains routinely contain fault modes absent from the source domain, and closed-set models forcibly assign these unknown samples to known classes, producing false negatives that can escalate to unplanned turbine downtime.
Open-set domain adaptation methods attempt to resolve this by introducing a K+1-dimensional classifier with an empirical threshold to separate unknown samples. However, the threshold is typically fixed at 0.5, a value that lacks per-sample adaptivity and performs poorly when unknown-class proportions fluctuate across operating regimes. Blind alignment of the entire target domain to the source domain under such fixed thresholds induces negative transfer, corrupting the shared-class feature space. MCOAN addresses this bottleneck by computing independent similarity estimates from a K-way one-vs-all classifier and a K+1-way classifier, converting their outputs into dynamic per-sample weights and thresholds. This mechanism enables adaptive reweighting during adversarial training, aligning known-class features across domains while isolating unknown samples, and a non-adversarial domain classifier ensures the weight computation remains stable under adversarial gradient pressure.
Loading authentic research manuscript (Pages 1–5)...
HU Qinyi, DENG Aidong, ZHOU Zhongzhi, XIAO Kaiwen, SHEN Yang, WU Yifan (2026). Multi-Classifier Open Adversarial Network for Rolling Bearing Fault Diagnosis in Wind Turbine Generator Systems. Acta Energiae Solaris Sinica. https://doi.org/10.19912/j.0254-0096.tynxb.202608_9727
Research & Educational Purpose Only: The translations, structured abstracts, analytical annotations, and data reports provided by SinoGreenTechare intended exclusively for academic research, internal corporate R&D, and educational benchmarking. They do not constitute formal engineering, chemical safety, legal, or professional advice.
Copyright & Intellectual Property Notice: Original copyright of the underlying source articles and experimental data remains with the respective authors, institutions, and original publishing journals. SinoGreenTech claims intellectual property only over its proprietary translations, analytical syntheses, and AEO structured enhancements in accordance with international fair use and academic citation principles.
Frequently Asked Questions
How does MCOAN prevent negative transfer when unknown-class samples constitute a large fraction of the target domain, a scenario where fixed-threshold OSBP fails?
MCOAN computes per-sample dynamic weights from the agreement between a K-way one-vs-all classifier and a K+1-way classifier. Samples with low similarity to all known source classes receive low weights during adversarial alignment, effectively excluding them from the shared-class feature alignment. This contrasts with OSBP, which applies a fixed threshold of 0.5 to the K+1-dimensional softmax output and cannot adapt when the unknown-class proportion shifts. The non-adversarial domain classifier further stabilizes weight estimation by decoupling it from adversarial gradient interference.
What is the operational cost of deploying MCOAN on wind turbine condition monitoring hardware compared to existing DANN-based diagnostic pipelines?
MCOAN adds one K-way classifier and one non-adversarial domain classifier to the standard DANN generator-discriminator backbone. The additional inference cost is limited to forward passes through these two lightweight classification heads, which operate on the same feature vector extracted by the shared generator. No additional sensor channels or data acquisition hardware are required, preserving compatibility with existing vibration monitoring installations on wind turbine drivetrains.
How does the dynamic threshold mechanism perform when the source domain contains only a subset of the fault modes present in the target domain, and the unknown-class cardinality is not known a priori?
The K+1-way classifier assigns the K+1-th dimension as the unknown-class probability without requiring knowledge of how many unknown classes exist. The dynamic threshold is computed per sample from the similarity scores, so the method does not assume a fixed unknown-class count. This is critical for wind turbine applications where new bearing failure modes may emerge from field conditions not represented in the source domain, and the number of such modes is inherently unpredictable.
What failure mechanisms in rolling bearings under variable wind turbine operating conditions are most likely to produce unknown fault modes that MCOAN must identify?
Variable speed and load conditions in wind turbine drivetrains generate fault signatures that differ from those in constant-speed laboratory datasets. Unknown modes may include compound faults where inner-race and outer-race defects coexist, lubrication starvation-induced wear patterns, and electrical discharge damage from generator bearing currents. These modes are absent from source-domain training data and would be misclassified by closed-set domain adaptation methods, whereas MCOAN isolates them through the K+1-way classifier and dynamic weighting.
How does the non-adversarial domain classifier improve dynamic weight accuracy compared to using only the adversarial domain discriminator?
The adversarial domain discriminator is trained to confuse source and target features, which introduces gradient noise into the feature space that corrupts similarity estimation. The non-adversarial domain classifier operates without gradient reversal, providing a stable domain-invariance signal that is not subject to the oscillatory dynamics of adversarial training. This separation allows the dynamic weights to reflect true similarity between target samples and source known classes rather than artifacts of the adversarial game, improving both known-class alignment and unknown-class recognition precision.
Related Chinese Research & Cross-Citations
Wind Turbine Gearbox Fault Diagnosis Method Based on Improved CNN-XGBoost Fusion Model Under Gramian Angular Difference Field
Gearbox failures account for 20–30% of total wind turbine faults and incur maintenance costs equivalent to 10–15% of overall turbine value. Conventional vibration diagnostic pipelines—complementary ensemble empirical mode decomposition with singular value energy spectrum, time-varying filtering empirical mode decomposition, and Teager energy spectrum analysis—remain bounded below 90% accuracy and depend on expert-driven feature engineering that is sensitive to non-stationary operating conditions and noise. This study proposes an intelligent diagnostic architecture that converts one-dimensional gearbox vibration signals into two-dimensional images via Gramian angular difference field (GADF) transformation, preserving intrinsic temporal correlation and time-frequency structure while exploiting matrix sparsity to suppress interference. An improved convolutional neural network (CNN) extracts multi-dimensional features: a convolutional block attention module (CBAM) is embedded in the convolutional layers to weight critical channels and focus on fault-sensitive spatial regions, and a modified βc-ACONC activation function replaces ReLU to mitigate neuron necrosis and enable selective activation. The extracted composite features are then fed into an XGBoost network whose hyperparameters are optimized by an improved sparrow search algorithm (ISSA). Validation on a laboratory wind turbine gearbox dataset yields diagnostic accuracy exceeding 99%, demonstrating robust fault identification capability under complex operating conditions.
Joint Forecasting of Wind and Photovoltaic Power Considering Complementarity
The inherent spatiotemporal complementarity between wind and solar resources offers a theoretical basis for improving renewable power forecasting accuracy. This study proposes a joint wind-photovoltaic (PV) power forecasting strategy that explicitly exploits this complementarity. A bidirectional long short-term memory (BiLSTM) neural network serves as the baseline forecasting model, and a novel sorting and comparative optimization (SCO) algorithm is developed to optimize the model's hyperparameters. The SCO algorithm ranks individuals in ascending order and compares adjacent fitness values to escape local optima, a known deficiency in conventional metaheuristics such as genetic algorithms and particle swarm optimization. For wind farms and PV plants exhibiting significant complementarity, the joint forecasting strategy first aggregates their power outputs, normalizes the combined signal, and then feeds it into the optimized BiLSTM model. Experimental results demonstrate that the proposed SCO-BiLSTM model reduces the eMAPE by 10.313% compared with PSO-BiLSTM. Furthermore, joint forecasting under SCO-BiLSTM lowers the eRMSE by 27.443% relative to standalone PV power forecasting. The study also establishes that forecasting accuracy improves with stronger wind-solar complementarity but degrades as the forecasting horizon extends. These findings confirm that exploiting complementarity in joint forecasting substantially enhances predictive performance for renewable energy integration.
Ultra-Short-Term Wind Power Forecasting Based on Fluctuation Continuation Scenario Identification
Existing ultra-short-term wind power forecasting methods exhibit limited performance due to insufficient extraction of fluctuation information and inadequate analysis of evolution patterns. This paper proposes an ultra-short-term wind power forecasting method based on fluctuation continuation scenario identification. First, the coupling mechanism of wind power fluctuations under multiple turbulence processes is investigated, and historical wind power dynamics are decoupled into a combination of nonlinear and linear fluctuation components. A fluctuation continuation concept is introduced, and the future continuation scale of wind power fluctuations is derived from nonlinear and linear decoupling parameters, thereby classifying fluctuation continuation scenarios. A sparse neural network (SNN) oriented to high-dimensional sparse features is constructed to identify historical fluctuation continuation scenarios, and ultra-short-term power forecasting is conducted separately for each scenario. Validation using measured wind speed and power data from three wind farms shows that, compared with baseline models, the proposed model improves root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) by at least 1.46%, 2.44%, and 14.67%, respectively, demonstrating superior accuracy and stability. The method addresses the limitations of signal decomposition techniques that lack physical interpretability and are sensitive to hyperparameters, and overcomes the high-dimensional sparsity challenges faced by traditional scenario identification models.
Unsupervised Automated Identification Method for Abnormal States of Wind Turbine Gearboxes
Addressing the scarcity of labeled data for training classification models in wind turbine planetary gearbox anomaly identification, this study proposes an unsupervised automated detection method. Log Mel-band energy features are extracted from raw vibration signals and fed into an unsupervised anomaly recognition model centered on a U-net autoencoder. A health-state threshold is established based on reconstruction error between model input and output, enabling anomaly identification. The method is validated using factory gearbox test data and operational data from a wind farm in Yangtouya, Shanxi. For factory gearboxes, dual validation is performed using a spectrum amplitude modulation-based signal processing method. Results demonstrate that the proposed method achieves 93.34% recognition accuracy on both factory and wind farm test sets, confirming its capability to automatically and correctly separate abnormal wind turbine gearboxes. The approach eliminates reliance on labeled fault data, offering a scalable solution for full-lifecycle health monitoring, from factory acceptance testing to in-service early anomaly detection, adaptable across different operating conditions and turbine models.
Improved Adaptive Super-Twisting Sliding Mode Control for Permanent Magnet Synchronous Motors
Adaptive super-twisting sliding mode control (ASTSMC) for permanent magnet synchronous motors (PMSM) suffers from prolonged convergence and insufficient disturbance rejection under complex operating conditions. This paper proposes an improved ASTSMC incorporating a fixed-time disturbance observer (FTDO). A power term is introduced into the adaptive super-twisting controller to accelerate convergence far from the origin, while the discontinuous sign function is replaced by a continuous h(s) function to mitigate chattering. The FTDO ensures disturbance estimation converges within a fixed time independent of initial states, overcoming the limitations of traditional and finite-time observers. The estimated disturbance is fed forward to the sliding mode controller for compensation, enhancing robustness. The FTDO design is based on an auxiliary state variable z and its derivative, with error dynamics analyzed via Lyapunov stability. Comparative simulations against conventional disturbance observers and sliding mode controllers validate the proposed strategy. The results demonstrate shorter convergence time, improved dynamic performance, and superior disturbance rejection, making the approach suitable for high-performance servo systems. The method addresses the critical need for robust, fast-response control in electric vehicles, rail transit, and aerospace applications where PMSM drives face significant uncertainties and external disturbances.
Data-Model Jointly Driven Fault Diagnosis for Wind Turbine Planetary Gearboxes
Fault diagnosis of wind turbine planetary gearboxes is severely constrained by the scarcity of high-quality fault data, as gearboxes operate predominantly in healthy states and automatic shutdowns prevent fault progression. This paper proposes a data-model jointly driven diagnosis method to address low diagnostic accuracy under limited fault samples. A high-fidelity lumped-parameter dynamic model of the planetary gearbox is constructed to generate pseudo-fault data, supplementing the training set. A domain-shared residual network feature extractor incorporating convolutional block attention modules extracts key physical features from both pseudo and measured data. Local maximum mean discrepancy aligns feature distributions at the fault-category level between pseudo and real fault data. A Kolmogorov-Arnold network module enhances the model's capacity to learn complex data relationships, enabling classification and identification of different fault types. Validation on a wind turbine planetary gearbox fault diagnosis test rig demonstrates that the proposed method achieves superior diagnostic performance under fault sample scarcity compared to classical methods. The framework offers an effective solution for known fault types, though identification of unknown and atypical faults remains a challenge for future work via open-set domain generalization.