The goal of single-domain generalization (SDG) fault diagnosis is to train a model that can adapt to unseen target conditions using only one source condition. Yet the task is extremely challenging due to large domain shift and insufficient data diversity, existing SDG methods rely heavily on data augmentation and consistency regularization, techniques that still struggle to capture the complexity of real-world operating-condition variations. To address this, a SDG framework for trustworthy fault diagnosis is proposed, i.e., Adaptively Guided Balance of Diversity and Discrimination. First, the low-frequency statistics of source-domain data are modeled as a Gaussian distribution to simulate potential domain shift, yielding auxiliary domains that are both diverse and semantically consistent. Next, the cross-domain mutual information estimation is dynamically optimized by measuring the distribution differences between the source domain and the auxiliary domain, guiding the encoder to learn domain-invariant features. Finally, a reliability assessment framework is developed, which transforms the diagnosis objective from accuracy maximization to decision risk minimization through uncertainty quantification and clustering-guided adaptive revision. Empirical evaluations on the three datasets of bearings and gearboxes demonstrate significant performance improvements in cross-condition fault diagnosis, thereby confirming the effectiveness of our approach.
Interconnecting power grids through HVDC links requires an efficient cost–benefit analysis to fully realize the advantages of such interconnections, including the mitigation of widespread cascading risks. This paper proposes a resilience-aware planning framework based on dynamic simulation studies and introduces a Cascading-aware Cost-Benefit Analysis (C2BA) for evaluating alternative Points-of-Connection (POCs) for a voltage-source-converter multi-terminal DC (VSC-MTDC) system. Candidate POCs are identified through a combination of cascading-vulnerability assessment, geographical and engineering considerations, and are subsequently evaluated through a resilience-focused techno-economic analysis incorporating a historically informed initiating-event occurrence model. The framework employs Integrated Cost-Benefit Metrics (ICBMs) that combine resilience benefits, quantified as the monetized value of cascading-failure mitigation through avoided interruption costs, with the corresponding MTDC investment and topology-dependent supporting AC-side reinforcement costs. Unlike conventional approaches that rely primarily on static analyses, including security-constrained power flow methods, the proposed C2BA incorporates detailed time-domain RMS simulations of cascading mechanisms, explicitly capturing the dynamic interactions among controllers and protection devices, particularly those associated with converter-interfaced technologies. The methodology is applied to the islanded power grid of Cyprus, a real-world case study, and its effectiveness in identifying an efficient MTDC interconnection topology is demonstrated by achieving a 64% improvement in the system's supplied load.
The inherent individual variability and multi-state uncertainty within the degradation process of mechanical equipment frequently make a single degradation model incapable of fully capturing the entire degradation trajectory. In addition, the performance decline of mechanical equipment often involves long-range dependencies (LRD) and non-stationarity characteristics. Traditional Markov process models fail to capture such degradation relationships. To address these issues, the present paper develops a novel framework for predicting remaining useful life (RUL) of mechanical systems. Firstly, a multi-path improved Bi-fractional Brownian Motion (MP-BFBM) RUL prediction model is introduced. MP-BFBM leverages multiple parallel degradation paths and effectively captures the LRD characteristics and non-stationarity of the degradation process. Subsequently, an enhanced weighted optimization particle filter (WOPF) algorithm is employed to estimate the states associated with the performance degradation model. Finally, a trajectory similarity matching method based on surface weighted warping distance (SWWD) is developed. This method effectively measures the non-linearity of damage propagation and the time lags in trajectory alignment during the system's degradation process, leading to more accurate RUL predictions. The advancement of the MP-BFBM model and SWWD are validated through extensive testing on heavy truck axle bridge lifetime data and the PHM2012 benchmark bearing dataset.
Accurate mission reliability evaluation provides an essential basis for assessing whether multistation manufacturing systems can continuously accomplish production missions along the functional chain of processing equipment, work-in-process (WIP) states, and production tasks. Inherent operational uncertainties propagate along this functional chain and evolve into interdependent and competing functional failure processes. Existing expectation-based decoupled models usually rely on averaged functional states, which may overlook propagation-induced amplification and the competition among multi-layer functional failure processes in triggering system failure, leading to biased mission reliability estimates. To address this issue, this study proposes an operational uncertainty-oriented integrated mission reliability evaluation approach considering competing functional failures. First, an operational-uncertainty-propagation-driven mechanism for functional failure process evolution is formulated to explain how cross-layer interactions transmit, transform, and amplify operational uncertainties. Second, a WIP-state stochastic flow manufacturing network (WSFMN)-based reliability modeling method is then developed to construct a threshold-driven hierarchical reliability model for degradation-related, quality-related, and production-related competing functional failure processes, with WIP states as the core carriers of uncertainty propagation. Third, these multi-layer functional failure processes are further integrated into a path-wise mission reliability evaluation framework to capture system mission failure governed by the first threshold-crossing process. Finally, a case study of an aerospace fastener manufacturing system demonstrates the effectiveness of the proposed method.
Accurate remaining useful life (RUL) prediction under cross-domain conditions is essential for reliability-oriented maintenance of industrial assets. In practice, target domain degradation trajectories are typically incomplete and unlabeled due to operational constraints that prevent assets from running to failure, while distribution discrepancies across domains further degrade prediction performance. Existing domain adaptation (DA) methods mainly rely on global feature alignment and therefore often suffer from semantic misalignment and negative transfer when only partial degradation trajectories are available. To address this challenge, this paper proposes a semantic-level segment-wise domain adaptation (SSD-Adapt) framework for cross-domain RUL prediction under incomplete target degradation trajectories. An adaptive degradation trajectory segmentation module is first developed to automatically identify multi stage degradation structures without requiring manual stage annotations. Based on the segmented trajectories, a multi-level semantic alignment strategy jointly performs global feature alignment together with degradation stage-level and healthy stage-level alignment, while a contrastive learning-based feature preservation module mitigates excessive alignment and preserves target-specific degradation characteristics. Extensive experiments on two turbofan degradation datasets demonstrate that the proposed framework consistently outperforms state-of-the-art DA methods under multiple cross-domain prediction scenarios. The framework is further validated on an electric bus battery dataset collected from two real metropolitan transit systems, demonstrating its effectiveness for fleet-level cross-city RUL prediction under realistic operating conditions with incomplete degradation observations. These results indicate that the pro posed framework provides an effective solution for cross-domain RUL prediction with incomplete degradation observations and offers strong potential for reliability-oriented prognostics in real world industrial applications.
