The proliferation of Internet of Things (IoT) devices in industrial environments generates vast heterogeneous data streams, yet extracting actionable process insights remains challenging due to the semantic gap between low-level sensor measurements and high-level business activities. Traditional process mining techniques assume the availability of structured event logs, whilst IoT environments produce raw multimodal data requiring extensive manual annotation to bridge this gap. Existing approaches suffer from three critical limitations: reliance on single modalities, poor semantic richness with minimal contextual attributes, and lack of spatio-temporal consistency validation. This paper introduces MEGAEL (Multimodal Event Generation using Adaptive Ensemble Learning), a comprehensive framework that orchestrates Large Language Models, Vision-Language Models, Retrieval-Augmented Generation, and hybrid Transformer-Graph Neural Network architectures for automated event log generation from heterogeneous IoT data. The framework operates through five synergistic phases: multimodal ingestion with temporal synchronisation, parallel LLM-based semantic extraction and vision-based contextual enrichment, knowledge-grounded augmentation using organisational documentation, and dual-branch spatio-temporal validation ensuring both temporal coherence and spatial consistency. Extensive evaluation on a real-world smart manufacturing dataset comprising 2.5 million sensor readings, video streams, and system logs demonstrates MEGAEL’s superiority over existing approaches. The framework achieves 0.91 F1-score, generates semantically rich event logs with 12+ attributes per event (versus 3-5 for baselines), reduces temporal violations by 74
Abstract Wind turbine reliability depends on timely identification of electromechanical faults, especially in generator-related subsystems under variable mechanical loads. This study presents a simulation-based, multi-signal, physically interpretable diagnostic workflow for wind turbine electrical systems. It combines multiphysics simulation, FFT feature extraction, and explainable machine learning, emphasising the integration of existing methods rather than new AI models. A COMSOL Multiphysics (2D electromagnetic with 3D multibody dynamics) model of an induction machine simulated both healthy and imbalanced operating conditions with increasing stator phase-A current imbalance (parameter ϵ ). The verified fault mechanism was incorporated into the model. From these simulations, a multisignal dataset was built using electromagnetic torque, rotor speed, electromagnetic force, and foundation force responses across 9 configurations, resulting in 54 samples (each with 100 features) classified into healthy, minor, and major imbalance groups. We tested support vector machine (SVM), multilayer perceptron (MLP), and random forest (RF) algorithms. Repeated stratified cross-validation showed RF performed best, with an average accuracy of 92.3% (±4.1%) and macro- F 1 of 0.764 (±0.146). A leave-one-configuration-out test, where no data from the same configuration appears in both training and testing, produced more conservative results: 46.3% accuracy and 0.317 macro- F 1, with no healthy-condition samples correctly classified, because only one independent healthy configuration was available. SHAP analysis identified foundation-force spectral energy in the 50–150 Hz range as the most important predictor, suggesting imbalance severity at the configuration level. Since foundation-force features are fixed within each configuration, this indicator should be seen as a configuration-level marker rather than an individual sample marker. Overall, the sample-level results are promising, indicating that the multi-signal, physics-based feature set and interpretability are useful. However, the configuration-level results suggest that the current 9-configuration simulation setup is not yet sufficient for definitive diagnostic accuracy. Future steps include increasing the number of simulations, performing mesh convergence studies, and validating through experiments or analytical methods.
In this study, the characterization, production, and application of antifungal metabolites obtained from the Levilactobacillus (L. brevis S27) and two Lactiplantibacillus (L. pentosus S42 and L. plantarum S62) strains were evaluated. The lactic acid bacteria (LAB) cells showed antifungal activity against molds. The cell-free supernatant (CFS) of Levilactobacillus and Lactiplantibacillus presented antimicrobial activity against fungi and foodborne pathogenic bacteria, and the antifungal activity was significantly (p < 0.05) higher than that of the antibacterial activity. Moreover, the antifungal metabolites were characterized as proteinaceous compounds that remained stable under both high and low temperatures and demonstrated activity across a broad pH range. Additionally, metabolite production was significantly higher (p < 0.05) at an initial pH of 5 when incubated at either 25 °C or 37 °C. All strains and their CFSs exhibited strong bio-preservative effects against Penicillium digitatum in yogurt and against Aspergillus niger on orange fruit. Consequently, these Lactobacilli strains and their antifungal metabolites represent a novel approach to biocontrol in the agri-food industry and agricultural products.
In this paper, we will study ve dierential equations describing two-strain COVID-19 infection dynamics with vaccination strategy. The variables of our model will represent the susceptible, the two strain infected sub-populations and the vaccinated individuals. First, we will study the well-posedness of our model. Next, we will give the dierent equilibria of our model. After that, we will study the global stability of each equilibrium. Finally, we will give dierent numerical simulations in order to illustrate the convergence of the solutions toward the equilibria. In addition, the comparison between the numerical tests and COVID-19 clinical data is conducted.
This study compares the biofilm-forming potential of Escherichia coli and Staphylococcus aureus acclimatized to two different conditions: natural orange juice and orange juice reconstituted and standardized from natural concentrate. The adhesion results predicted by the XDLVO model were compared with the experimental. Strains acclimatized to natural orange juice show significantly increased adhesion, unlike orange juice from the food industry, which considerably reduces adhesion. FTIR analyses reveal an increase in polysaccharides in strains exposed to natural juice. The results also showed the limitations of the predictive XDLVO approach, as experimental adhesion remains relatively high even when the ∆GTotal values are relatively larger and positive, which theoretically indicates weak adhesion. These results highlight the impact of bacterial acclimatization and the need to integrate biological interactions into the study of biofilm formation. They also require evaluating the link between the food environment and the behaviour of gut microbiota bacteria, with a view to preventing cross-contamination and improving food safety.
