Yuyang Song

The grape and wine industry faces prolonged barrel aging, microbial spoilage, protein haze, chemical additives, and inefficient by‐product valorization. These challenges motivate nonthermal technologies that reduce chemical inputs. Ultrasound operates in two mechanistically distinct regimes: sensing, based on acoustic propagation, reflection, and echo detection, and processing, driven mainly by acoustic cavitation and its physical and chemical effects, including microjets, shock waves, localized hotspots, and reactive radicals. These complementary modes provide a versatile physical platform. This review covers ultrasound across the grape‐wine continuum: acoustic imaging for precision viticulture; postharvest preservation with green activated water; green extraction enhanced by natural deep eutectic solvents; grape protein modification for ingredient assembly; modulation of wine color, aroma, and colloidal stability; selected aging‐associated reactions involving 1‐hydroxyethyl radicals; selective microbial inactivation and fermentation modulation; and scale‐up based on energy characterization, continuous‐flow reactors, and data‐driven modeling. Current evidence indicates that ultrasound is evolving from an empirically optimized unit operation toward a mechanistically informed and quantitatively characterized physical‐field platform. Reported effects include accelerated aging‐related transformations, modified protein structure and stability, and selective microbial responses, although they depend on treatment conditions and matrix properties. Absorbed energy density supports quantitative comparison of treatments, but reliable scale‐up also requires acoustic‐field distribution, residence time, temperature, and reactor geometry. Future progress will require standardized energy reporting, pilot‐ and continuous‐flow validation, physics‐informed data‐driven modeling, and integration of process analytical technologies with closed‐loop control. These advances would support more reproducible, scalable, and sustainable ultrasound processing in the grape and wine sector.

Microbiomes influence tea quality, safety, and health from production to consumption, yet most studies and reviews examine individual stages. This review uses the Tea Plant‐Fermentation‐Gut Continuum to assess cross‐stage microbial and metabolic connectivity, not as an established transmission chain. It separates microbial origin and persistence from processing‐associated chemical transformation and host responses. At the leaf‐to‐processing interface, phyllosphere and endophytic communities may contribute to the processing microbial pool, but taxonomic overlap does not demonstrate directional transmission, and strain‐level persistence remains unproven. During processing, plant enzymes and microbial activities shape tea chemistry differently across tea types, affecting pigment and aroma development. At the processing‐to‐gut interface, current evidence mainly supports chemical and metabolic continuity: tea‐ and fermentation‐derived compounds reach the gut and are metabolized by resident microbiota, whereas evidence that fermentation‐associated microorganisms survive, persist, or colonize the gut remains insufficient. These metabolites are linked to FXR‐FGF15, AMPK, and AhR signaling and to metabolic, anti‐inflammatory, and neuroprotective effects, mostly in preclinical models; human evidence remains limited. Together, the evidence does not support an uninterrupted microbial transmission pathway across these stages. Key gaps remain in microbial origin, persistence, activity, causality, and host relevance. Shared taxa may reflect persistence, environmental acquisition, or compositional overlap rather than direct transmission; tea type, cultivar, processing, dose, and host microbiota limit generalization. Strain‐resolved tracking, controlled experiments, functional omics, and metabolic tracing are needed to distinguish association from persistence, activity, and causal function. Addressing these gaps will support evidence‐based tea processing, precision fermentation, and microbiome‐informed nutrition.

Machine learning is transforming the grape and wine industry by shifting traditional experience‐driven practices toward data‐driven and intelligent decision making across the entire value chain. This review provides a conceptually driven synthesis of machine learning‐enabled technologies spanning vineyard sensing, precision viticulture, fermentation monitoring, and winemaking optimization. An integrated analytical framework is proposed, linking multisource data acquisition, preprocessing and representation learning, model development, and intelligent decision making into a unified end‐to‐end pipeline. Building on this framework, the review critically examines advances related to five key scientific challenges: multimodal data integration, model interpretability, cross‐domain generalization, whole‐chain decision coordination, and scalable industrial deployment. Particular emphasis is placed on the mechanisms and trade‐offs of multiscale sensing technologies (including spectroscopy, chromatography–mass spectrometry, imaging, and electronic sensing), as well as data preprocessing, feature engineering, and multimodal fusion strategies. A task‐oriented and data‐structured perspective on model selection is highlighted, covering linear models, kernel methods, ensemble learning, deep neural networks, and probabilistic frameworks, alongside evaluation protocols and interpretability approaches. In contrast to fragmented task‐specific studies, this review highlights the importance of cross‐stage integration and closed‐loop decision systems linking vineyard management with downstream vinification and quality evaluation. Despite rapid progress, key challenges remain, including data scarcity and heterogeneity, limited model transferability, and high implementation costs. Emerging directions such as knowledge‐guided machine learning, causal inference, small‐sample learning, and human–AI collaboration are expected to enhance robustness and interpretability. Overall, this review provides a structured roadmap for advancing intelligent and sustainable practices in the grape and wine industry.

Fengxiang Taibai baijiu (FTB) fermentation involves complex microbial interactions that critically influence aroma formation; however, the efficient selection of high ester‐producing yeasts and the prediction of their fermentation performance remain challenging. In this study, yeasts were systematically isolated from different stages of FTB fermentation and evaluated using a combination of fermentation experiments and machine learning approaches. A total of 156 strains were obtained, among which seven Wickerhamomyces anomalus strains exhibited high ester‐producing capacity (> 1.15 g/L). Among them, strain F12D‐38 demonstrated the strongest tolerance and aroma‐producing potential under ethanol–acetic acid co‐stress conditions. An XGBoost model achieved high predictive accuracy ( R ² = 0.89) and was applied to guide strain screening and process optimization, identifying an optimal inoculation range of 14%–16% (v/v). Fermentation validation showed that five strains outperformed the traditional control, with F12D‐38 achieving an approximately 15.0% increase in total volatile compounds. Overall, this study demonstrates that integrating machine learning with experimental validation enables efficient strain selection and fermentation optimization, providing a practical strategy for improving aroma quality and process control in FTB production.

The flavor complexity of mead is influenced by various factors, including the geographic origin of honey, botanical sources, fermentation processes, and the addition of fruits and herbs. Chemical characteristics of mead are regulated by factors such as honey origin, yeast metabolism, latitude, climate, light intensity, and fermentation parameters, which collectively affect the formation of volatile and nonvolatile compounds. This review examines the chemical profiling of mead, with a particular focus on detecting volatile compounds. Advanced analytical techniques like HS–SPME–GC–MS, RP–UHPLC–ESI–MS, and GC × GC–HRMS are employed to detect and quantify these compounds. Quantitative methods, such as the internal standard method and standard addition method, effectively mitigate interference from high‐sugar matrices in mead, ensuring accurate results. The flavor characteristics of mead, especially pure mead, are primarily determined by the choice of honey, typically exhibiting floral, fruity, and sweet honey flavors. European meads are marked by phenolic compounds and fruity esters, Asian meads are rich in terpenes and phenolic compounds, North American and South American meads are characterized by methoxy‐aromatic compounds and lactones, while African meads contain unique ester and sulfur compounds. Despite challenges like the lack of standardized methods and regional imbalances in research, standardizing methods and integrating “omics” technologies will help uncover regional characteristics in mead production and promote its global recognition and development.