Understanding causal heterogeneity is crucial for building robust and interpretable learning systems that operate reliably under environmental shifts. However, existing methods lack causal awareness, with insufficient modeling of heterogeneity, confounding, and observational constraints, leading to poor interpretability and difficulty distinguishing true causal heterogeneity from spurious associations. We propose an unsupervised framework, HCL (Interpretable Causal Mechanism-Aware Clustering with Self-Improving Adaptive Heterogeneous Causal Structure Learning), that jointly infers latent clusters and their associated causal structures from mixed-type observational data without requiring temporal ordering, environment labels, interventions or other prior knowledge. HCL relaxes the homogeneity and sufficiency assumptions by introducing an equivalent representation that encodes both structural heterogeneity and confounding. It further develops a bi-directional iterative strategy to alternately refine causal clustering and structure learning, along with a self-supervised regularization that balances cross-cluster universal and specific mechanisms under shifts. Together, these components enable convergence toward interpretable, heterogeneous causal patterns. Theoretically, we show identifiability of heterogeneous causal structures under mild conditions. Empirically, HCL achieves superior performance in both clustering and structure learning tasks, and recovers biologically meaningful mechanisms in real-world single-cell perturbation and clinical intervention data, demonstrating its utility for discovering interpretable, mechanism-level causal heterogeneity.

Information Extraction (IE) stands as a cornerstone in natural language processing, traditionally segmented into distinct sub-tasks. The advent of Large Language Models (LLMs) heralds a paradigm shift, suggesting the feasibility of a singular model addressing multiple IE subtasks. However, the efficacy of employing LLMs directly trained on chat-based data for IE tasks is considerably subpar when juxtaposed with conventional methods employed in prior studies. In order to address this limitation and harness the robust generalization capabilities inherent in LLMs, we propose the General Information Extraction Large Language Model (GIELLM). GIELLM seamlessly integrates various tasks, including Text Classification, Sentiment Analysis, Named Entity Recognition, Relation Extraction, and Event Extraction, employing a unified input-output schema. This innovation marks the first instance of a model simultaneously handling such a diverse array of IE subtasks. Notably, the GIELLM leverages the Mutual Reinforcement Effect (MRE), enhancing performance in integrated tasks compared to their isolated counterparts. Our experiments demonstrate State-of-the-Art (SOTA) results in five out of six Japanese mixed datasets, significantly surpassing GPT-3.5-Turbo. Further, an independent evaluation using the novel Text Classification Relation and Event Extraction(TCREE) dataset corroborates the synergistic advantages of MRE in text and word classification. This breakthrough paves the way for most IE subtasks to be subsumed under a singular LLM framework. Specialized fine-tune task-specific models are no longer needed.

The unique tilt-servo mechanism of the tilt-rotor unmanned aerial vehicle (UAV) facilitates seamless transitions between multi-rotor and fixed-wing modes, enhancing both flexibility and maneuverability. However, traditional modeling methods, which treat each flight mode independently, fail to provide a unified dynamic representation, limiting the accurate description of aerobatic maneuvers during mode transitions. This paper introduces a novel modeling approach based on transient Computational Fluid Dynamics (CFD) to capture the aerodynamics of the transition mode, resulting in a multimodal, consistent dynamics model. This model simplifies the mathematical representation for specific tilt angles, ensuring compatibility with both multi-rotor and fixed-wing dynamics, and accurately describes aerobatic maneuvers. An autonomous feedback motion planning method, utilizing third-order Bézier curves for angular velocity planning, is applied, along with a modal switching strategy to address the limitations of traditional fixed-wing UAVs. The feasibility of this method was validated through numerical simulations, hardware-in-the-loop simulations, and outdoor flight experiments of a tilt-rotor UAV performing the Cobra maneuver in transition mode.

Early aging diagnosis of winding insulation is crucial for preventing insulation failures. However, early aging suffers from unclear mechanisms and weak changes of electrical parameters, which makes it difficult for mechanism and intelligent diagnosis approaches to achieve effective modeling. To address this issue, this article proposes a mechanism-guided intelligent enhancement method for early aging information based on the common-mode impedance spectrum. First, an effective mechanism knowledge focusing strategy is proposed, which can effectively alleviate coupling of aging mechanism features under different aging types, so as to provide the pure mechanism features for the dynamic knowledge matching. Then, a dynamic knowledge matching strategy is presented, by which aging mechanism features and early aging information are reliably matched, thereby supporting the strong generalization of the mechanism-guided reconstruction. Finally, a mechanism-guided reconstruction network is constructed to enhance the insulation early aging information, so as to achieve high accuracy and strong-generalization diagnosis modeling of insulation early aging. Notably, our method provides a promising solution for the diagnosis modeling of early aging. The experimental results on a 4-kW permanent magnet synchronous motor drive system demonstrate that our work outperforms the existing methods, which achieves the 60.74% reduction in the early aging diagnosis error of winding insulation.

The thermal sensitive electrical parameter (TSEP) junction temperature estimating method is an important way to enhance the reliability of SiC MOSFETs, with the temperature-dependent model (TDM) serving as the fundamental basis. The drain-source turn-on current changing rate σds,on is a crucial TSEP, as it governs the turn-on characteristics and other associated transients TSEPs. However, the existing TDMs for σds,on are inaccurate and lack analytical rigor, thereby impeding the calibration process and limiting method generalization. To address this issue, we have developed an accurate nonlinear temperature-dependent analytical model (TDAM) for σds,on based on the gate turn-on transient model proposed in our previous study. First, the mechanism of carrier scattering is analyzed to determine the theoretical relationship between the mobility and the junction temperature. Furthermore, the relationship between the gate threshold voltage and the junction temperature is examined based on the principles of energy band theory. Third, the TDAM is established according to the obtained theoretical results. The proposed model is verified by the dual pulse experiment. Compared with existing models, it is more accurate and analytical, displaying the nonlinearity characteristics. The proposed model can pave the way for advancing transient TSEP methods and switching loss optimization strategies.

Causal discovery based on observational data is important for deciphering the causal mechanism behind complex systems. However, the effectiveness of existing causal discovery methods is limited due to inferior prior knowledge, domain inconsistencies, and the challenges of high-dimensional datasets with small sample sizes. To address this gap, we propose a novel weakly supervised fuzzy knowledge and data co-driven causal discovery method named KEEL. KEEL introduces a fuzzy causal knowledge schema to encapsulate diverse types of fuzzy knowledge, and forms corresponding weakened constraints. This schema not only lessens the dependency on expertise but also allows various types of limited and error-prone fuzzy knowledge to guide causal discovery. It can enhance the generalization and robustness of causal discovery, especially in high-dimensional and small-sample scenarios. In addition, we integrate the extended linear causal model into KEEL for dealing with the multi-distribution and incomplete data. Extensive experiments with different datasets demonstrate the superiority of KEEL over several state-of-the-art methods in accuracy, robustness and efficiency. The effectiveness of KEEL is also verified in limited real protein signal transduction process data, with the better performance than benchmark methods. In summary, KEEL is effective to tackle the causal discovery tasks with higher accuracy while alleviating the requirement for extensive domain expertise.

SiC power devices have been increasingly used in industrial applications. Junction temperature estimating is the basis of high reliability operation for SiC devices, since thermal stress is one of the dominating failure inducing factors. Thermal sensitive electrical parameter (TSEP) methods have been widely used for junction temperature estimating for the advantage in low invasiveness and fast response speed. However, conventional TSEP methods merely focus on a single semiconductor device instead of a converter, which results in high cost and difficulty in integration. Low sensitivity is another limitation. To solve the problems, a high-sensitivity converter-level online junction temperature estimating method is proposed in this paper based on the current oscillation (Δi OS ) of DC and AC sides in a single-phase inverter. Theoretical analysis is provided to illustrate why Δi OS can serve as a converter-level junction temperature (T j ) indicator. The relationship between Δi OS and T j is obtained by experiment for calibration based on a dual pulse test. Online experiment for T j estimating based on a single-phase inverter is provided to verify the effectiveness of the proposed method.

Anomaly detection is a critical technique that ensures the reliability of WSNs. However, most existing anomaly detection methods only consider the case of single modal data flow anomaly detection for each node or multiple modal time series data flow anomaly detection for a single node and do not consider the case of multiple nodes and multiple time series data flow simultaneously,and it limited the ability of anomaly detection. In this paper, a novel anomaly detection model is proposed for multimodal WSN data flows. First, the temporal features and modal correlation features extracted from each sensor node are fused into one vector representation, then it is further aggregated with the spatial features represented the spatial position relationship of the nodes; finally,the current time-series data of WSN nodes are predicted, and abnormal states are identified according to the fusion features. The simulation results obtained on a public dataset show that the proposed approach can significantly improve upon existing methods interms of robustness, and its F1 score reaches 0.90, which is 14.2% higher than that of the graph convolution network (GCN) with longshort-term memory (LSTM).

On-line aging monitoring is the basis for high reliability operation of converters because many converter failures are related to aging defects. Conventional aging monitoring methods mainly focus on a single semiconductor device or module, and their industrial applications are limited. First, several aging monitoring circuits are required in a converter, which raises the cost. Second, the converter structure has to be modified, which is impractical. To solve this problem, a converter-level aging monitoring method is proposed in this paper, which is based on the on-line extraction of DC-link bus leakage current. The monitored aging mode of the proposed method is Si material degradation, which is one of the chip-related aging defects. First, the DC bus leakage current is analyzed theoretically as an indicator of the Si material degradation. A corresponding description parameter dSiconis proposed. Second, an on-line measuring method is proposed for the DC bus leakage current, which is called the state-selection method. Third, a converter-level Si material degradation monitoring strategy is proposed. On-line experiments based on a constant-voltage-constant-temperature (CVCT) accelerated aging test verify the effectiveness of the proposed method.

The permanent magnet synchronous generator (PMSG) connected to pulse width modulation (PWM) rectifier is widely used in the renewable energy applications, such as direct–drive wind turbines and marine current turbines. The reliability of the PMSG is crucial due to the high maintenance cost of the turbines. The stator winding insulation is one of the main reasons for the PMSG failures. However, the current technologies can only detect the insulation failure when the short-circuit failure has already happened, which cannot avoid the corresponding downtime losses. Therefore, the online monitoring method for stator winding insulation is performed on the PMSG-PWM rectifier system in this paper for the first time. The factors affecting the voltage harmonics on the stator winding insulation are studied and their effects on the monitoring results are analyzed. Based on the flow path of the insulation leakage current and the phase-to-ground voltage, the insulation impedance can be measured online. The effects of the rotor position and the working condition are also analyzed. The effectiveness of the online condition monitoring method for PMSG-PWM rectifier system under various working conditions is verified using offline and online tests.