Fatigue failure is a common mode of failure in ferromagnetic materials, yet traditional electromagnetic detection techniques and assessment models struggle to accurately estimate fatigue levels. To address this, a reliability method for evaluating the fatigue degree of ferromagnetic materials is proposed in this paper, integrating multi-electromagnetic features with a multi-model stacking ensemble learning approach. Four types of electromagnetic parameters, namely magnetic Barkhausen noise, incremental permeability, tangential magnetic field, and hysteresis loops, are collected as feature parameters. Subsequently, a stacking ensemble learning framework is constructed. Least squares support vector machine models serve as base learners with a multivariate linear regression model serving as the meta-learner. Experimental results on Q345 steel samples demonstrate that the proposed method achieves an average error rate of 7.3% in fatigue degree assessment, enabling early fatigue detection and lifespan prediction for materials.

As photovoltaic (PV) penetration increases, reduced system inertia intensifies frequency stability challenges in interconnected power grids. Conventional proportional integral derivative–based and single-loop fractional-order load frequency control schemes often exhibit limited disturbance rejection and robustness under stochastic PV fluctuations and parameter uncertainties. This paper proposes a cascaded tilted fractional-order integral–derivative–(PDN+1) load frequency control framework for a two-area PV–storage-integrated system, enhanced by a semantics-guided intelligent optimization strategy. A modified multi-objective function combining integral squared time absolute error and squared control effort embeds control semantics related to disturbance persistence and energy limitation. The lemur optimizer is employed for parameter tuning. Simulation results demonstrate faster response, smaller frequency deviations, and improved robustness compared with Proportional–Integral–(Proportional–Derivative+1), functional-order–proportional integral derivative, and single-objective integral squared time absolute error–based controllers, while maintaining engineering feasibility for future low-inertia power systems.

Short-term traffic flow prediction is essential for intelligent transportation systems, enabling signal control, route planning, and congestion mitigation. However, existing methods often overlook flow fluctuations and underutilize temporal positional information, limiting spatiotemporal modeling performance. Therefore, this study proposed a dynamic spatiotemporal feature extraction (DSTFE) model that integrated temporal information and fluctuation features (TIFF) for enhanced traffic prediction—DSTFE-TIFF. Specifically, DSTFE-TIFF employed trigonometric time encoding to capture multiscale patterns and an exponentially weighted sliding window to emphasize recent, abrupt changes. These features were then integrated through a time-weighted attention framework, which modeled both local and long-range dependencies to form comprehensive spatiotemporal representations. Experiments on three data sets—the traffic-speed data set in the Los Angeles County road network, traffic-flow data set in the San Francisco Bay area, and traffic-flow data set in California—showed that DSTFE-TIFF achieved state-of-the-art performance, reducing key error metrics (mean absolute error, root mean square error, mean absolute percentage error) by over 22% on average versus baselines, demonstrating its effectiveness and robustness.

This study explores whether generative artificial intelligence (AI) can exhibit decision-making behavior aligned with that of human experts. A total of 200 startup projects were assessed across four key dimensions. Each project received parallel evaluations from investors and generative AI models. AI models aligned with human evaluators in overall score levels and moderately predicted human ratings yet differed substantially in their score distributions, and the contrasts between the top 20% and the bottom 80% segments across all three models further revealed a distinctly two-tier alignment structure. Two indicators showed the practical impact: human labor time decreased by 94–99.6%, and monetary cost per report dropped by 350–550 times. The results reveal general logic but missed expert-level nuance in bounded, gradual, and stratified alignment with expert evaluators. Bounded alignment reflects AI's structural limits, gradual alignment describes dimension-specific convergence with human judgment, and stratified alignment captures tiered patterns of human–AI co-evaluation.

This paper proposes a novel hybrid authentication framework that combines offline and online verification to enhance security and performance in IoMT networks. In the offline phase, user and device identities are verified directly within the local area network (LAN), reducing latency and dependence on continuous server communication. The online phase employs a centralized authentication server to complete system-wide verification. The proposed method integrates Elliptic Curve Cryptography (ECC) for lightweight and secure key exchange, XOR operations for efficient credential combination, and one-way hash functions for ensuring data integrity. To further strengthen the authentication process, the Improved Lotus Effect Optimization Algorithm (ILEOA) is introduced to select the optimal cryptographic key, minimizing latency and improving resistance to attacks. The architecture is specifically designed to meet the constraints of IoMT devices while achieving multiple security objectives, including protection against replay, impersonation, and man-in-the-middle attacks.

With the rise of blockchain and decentralized technologies, doubts about traditional financial institutions' efficiency have increased. Meanwhile, Web3 offers transparency, security, and autonomy. However, the existing literature overlooks role the role of doubt as a push factor while focusing on the positive effects of trust. Moreover, the role of crypto wallets as a mooring factor remains underexplored. This study applies push-pull-mooring theory to examine Web3 literacy, trust in machines, doubt in institutions, and switching costs. Data were collected from 165 survey respondents. The results indicate that Web3 literacy increases doubt in traditional institutions but does not significantly affect trust in Web3. Additionally, switching costs moderate the relationship between Web3 literacy and doubt. When switching costs are low, doubt rises significantly. This study provides a new perspective on Web3 adoption, showing doubt's push effect and the role of push-pull mooring in migration, thus addressing gaps in the literature. Furthermore, the findings highlight how decentralized finance's trust mechanism is evolving, offering insights for Web3 adoption.

Large Language Models (LLMs) show strong performance in natural language tasks but are prone to hallucinations, limiting reliability in knowledge-intensive fields such as cultural heritage. This paper presents an Ontology-Based Retrieval-Augmented Generation (OB-RAG) framework that embeds subject–predicate–object triples from domain ontologies into a vector space, retrieving relevant knowledge via semantic search to ground LLM outputs. Unlike traditional RAG using unstructured text, the framework integrates manually and semiautomatically generated ontologies for explicit contextual grounding. A cultural heritage case study illustrates implementation and evaluation. Performance is assessed with quantitative metrics (Faithfulness and Answer Relevancy) and expert validation. Results show the OB prototype outperforms baseline LLMs, reducing hallucinations and improving factual accuracy and contextual alignment. The study offers both an architectural framework and empirical evidence that ontology-based RAG strengthens trustworthiness and user acceptance of LLMs in specialized domains.

Cloud computing is a complex and articulated semantic structure designed to facilitate the discovery, composition, and integration of cloud services offered by different providers. The cloud services reference ontology, in its current version, represents an articulated and complex semantic structure that facilitates the interoperability and portability between different cloud platforms. This work builds on previous research that has explored the use of semantic representations in cloud computing to improve the portability, interoperability, and automatic discovery of cloud systems. This paper proposes a series of improvements to simplify the class hierarchy, making it more readable and intuitive, and to enrich it with new classes, object properties, and data properties in order to expand the descriptive capacity and improve the accuracy and detail of the semantic description. The work presented here is based on a critical review of existing ontologies, accompanied by expansion to meet emerging challenges in the cloud computing.

Autism Spectrum Disorder (ASD) is a lifelong condition that affects communication, social interaction, and behavior. Artificial intelligence (AI) shows promise for early detection, but many models struggle with accuracy, scalability, and interpretability, limiting clinical use. To address these gaps, this paper proposes a semantic web–enabled explainable AI (XAI) framework for accurate and interoperable ASD diagnosis. The framework has three parts: (1) a semantic data integration layer that harmonizes heterogeneous datasets, (2) a scalable feature engineering process using MapReduce with the Binary Capuchin Search Algorithm (BCSA), and (3) interpretable classifiers enriched with SHAP for transparent predictions. Experiments on ASD datasets achieved about 87% accuracy, outperforming baselines by 7–10% and federated methods by 5%. Precision and F1 improved by 6–8%, while semantic integration enhanced interpretability and trust. By uniting semantic technologies with explainable ML, the framework ensures scalability and offers a reliable, transparent pathway toward clinically useful AI.