Sangoh Kim

Kimchi fermentation involves dynamic physicochemical and microbial changes; however, conventional monitoring methods are generally dependent on intermittent measurements, resulting in limitations in the real-time detection of abnormal fermentation. In this study, a Wireless Sensor Network (WSN)-based Fermentation Monitoring System (WFMS) and a Long Short-Term Memory (LSTM)-based Anomaly Detection System (LADS) were developed to continuously monitor internal pressure changes during kimchi fermentation. Kimchi samples were prepared under normal fermentation conditions (CON) and glucose-added conditions (GLU-6). Pressure data were collected at 10 min intervals using 15 psi and 30 psi pressure sensors connected to an Arduino Nano 33 IoT board and were transmitted to the ThingSpeak platform. During the fermentation period, pressure data were collected stably, while the external temperature was maintained at approximately 25 °C. Both CON and GLU-6 samples exhibited a rapid increase in internal pressure during the early fermentation stage, followed by a gradual decrease. However, relatively larger pressure fluctuations were observed in the middle and late fermentation stages of the GLU-6 samples. An LSTM autoencoder model trained using CON data established a reconstruction error-based threshold of 0.0025 and successfully detected anomalies in the GLU-6 samples. Anomalies were mainly identified during the initial fermentation stage and between fermentation days 2 and 4. These results demonstrate that pressure-based real-time monitoring combined with LSTM autoencoder analysis can be effectively applied for the non-destructive tracking of kimchi fermentation and the early detection of abnormal fermentation patterns.

This study examined cognitive style-related differences (analytic vs. holistic) in consumer liking, sensory perception, and ideal sensory profiles across three evaluation contexts (real café, sensory booth, and mixed reality). A total of 77 participants were divided into an analytic group (N = 34) and a holistic group (N = 43) based on the Analysis–Holism Scale. They evaluated six café latte samples varying in sugar concentration (0, 2.5, 5%) and espresso-to-milk ratio (1:2 and 1:3) for three environments using a within-subject design. Consumer evaluation comprised overall liking and sensory perception assessed using CATA (25 attributes) and Ideal CATA, with descriptive analysis (DA) conducted in parallel by eight trained panelists. The results showed no significant differences between cognitive styles in overall liking, but differences appeared in sensory perception and ideal product mapping between the booth and real café. The analytic group focused on dominant attributes with little variation for environments, whereas the holistic group integrated contextual cues, showing more context-dependent patterns. Compared with the other two environments, the MR environment showed high similarity to the DA results in terms of attribute profiles (RV = 0.88). This study indicates that cognitive style is a key factor in consumer sensory evaluation and should be considered to improve sensory evaluation methodology.

Teaching-based robotic systems offer an accessible alternative to complex inverse kinematics programming for food processing automation. Traditional model-based approaches require precise system identification and analytical solutions that are challenging for custom-built robots with manufacturing tolerances and mechanical uncertainties. This study developed a custom six-degree-of-freedom robotic arm using modular brushless motors controlled via Controller Area Network communication and Robot Operating System 2, a teaching mode where users manually demonstrate trajectories that are recorded at 100 Hz. Forty-five demonstration trajectories were collected across three geometric patterns (rectangle, triangle, circle) and augmented to 270 samples. A bidirectional Long Short-Term Memory network with attention mechanism was trained to classify patterns, achieving 83.33% test accuracy and outperforming baseline deep learning models (1D-CNN: 77.78%, TCN: 66.67%, GRU: 44.44%), while being marginally exceeded by Random Forest (86.11%). Rectangle patterns showed strongest recognition (78.57% F1-score), while circle patterns achieved highest performance (91.67% F1-score). However, severe overfitting was observed, with validation accuracy peaking at 85.19% at epoch 24 before degradation, indicating insufficient dataset size despite five-fold augmentation. The results demonstrate proof-of-concept feasibility for pattern recognition from limited teaching demonstrations, providing a pathway for robotic food processing without extensive programming expertise, though larger datasets and robust feedback control strategies are required for production deployment.

The convergence of blockchain, IoT, and cloud technologies has created new opportunities for secure, transparent, and autonomous data management in smart manufacturing systems. In this study, we developed a Cloud-Blockchain Integrated Fermentation Device (CBIFD) that enables real-time monitoring and distributed logging of environmental data during fermentation. The system employs a modular sensor configuration to measure pressure, temperature, humidity, internal temperature, CO₂, and pH, and is integrated with an Arduino Nano 33 IoT. Its cloud-native, self-contained architecture eliminates the need for a local server, ensuring high mobility and scalability. System performance was validated in a laboratory setting, demonstrating seamless data acquisition, reliable cloud synchronization, and consistent execution of blockchain transactions. Moreover, comparative analysis with existing IoT–blockchain architectures highlights the superior autonomy and infrastructure flexibility of the CBIFD. This work provides a robust foundation for future research aimed at enhancing process transparency and data-driven control in intelligent fermentation environments.

Extended Reality (XR) technologies—including Virtual Reality, Augmented Reality, and Mixed Reality—are increasingly applied in the food industry to simulate sensory environments, support education, and influence consumer behavior, while robotics addresses labor shortages, hygiene, and efficiency in production. This review uniquely synthesizes their convergence through digital twin frameworks, combining XR’s immersive simulations with robotics’ precision and scalability. A systematic literature review and keyword co-occurrence analysis of over 800 titles revealed research clusters around consumer behavior, nutrition education, sensory experience, and system design. In parallel, robotics has expanded beyond traditional pick-and-place tasks into areas such as precision cleaning, chaotic mixing, and digital gastronomy. The integration of XR and robotics offers synergies including risk-free training, predictive task validation, and enhanced human–robot interaction but faces hurdles such as high hardware costs, motion sickness, and usability constraints. Future research should prioritize interoperability, ergonomic design, and cross-disciplinary collaboration to ensure that XR–robotics systems evolve not merely as tools, but as a paradigm shift in redefining the human–food–environment relationship.

Despite significant advancements in sensor technologies, real-time monitoring and prediction of fermentation dynamics remain challenging due to the complexity and nonlinearity of environmental variables. This study presents an integrated framework that combines deep learning techniques with blockchain-enabled data logging to enhance the reliability and transparency of fermentation monitoring. A Long Short-Term Memory (LSTM)-based Fermentation Process Prediction Model (FPPM) was developed to predict Fermentation Percent (FP) and cumulative Fermentation Quantification (FQ) using multivariate time-series data obtained from modular sensor units (PBSU, GBSU, and FQSU). Fermentation conditions were systematically varied under controlled environments, and all data were securely transmitted to a Fermentation–Blockchain–Cloud System (FBCS) to ensure data integrity and traceability. The LSTM models trained on AAG1–3 datasets demonstrated high predictive accuracy, with coefficients of determination (R²) between 0.8547 and 0.9437, and the estimated FQ values showed strong concordance with actual measurements. These results underscore the feasibility of integrating AI-driven prediction models with decentralized data infrastructure for robust and scalable bioprocess control.

Accurate demand forecasting in the restaurant industry is critical for optimizing inventory management, minimizing food waste, and enhancing operational efficiency. This study developed an AI-based system that predicts menu-specific daily sales using historical sales and meteorological data collected from 2021 to 2023. Approximately 384 menu items were individually modeled using deep neural networks configured for multi-class classification. The system achieved strong predictive performance with a mean Pearson correlation coefficient of 0.7945. Additionally, flexible visualization options were implemented to sort predictions by expected or actual sales volumes. The results demonstrate the feasibility of AI-driven demand prediction systems and their potential to transform food service operations toward greater sustainability and efficiency.

Pressure drop processes, such as dissolved inorganic carbon and gun-puffing, have shown utility in the food industry, but their reliance on heat remains a limiting factor. This study involved the development of a processor capable of performing nonthermal pressure drop treatment, which minimizes thermal changes in food. In addition, its effects on the structure and soaking efficiency of adzuki beans were analyzed. Two improved pressure drop processes were tested: PDA, which applied 1 kgf/cm² of pressure before release, and PDB, which applied a higher pressure and gradually decreased it in steps of 1 kgf/cm². Both the PDA and PDB pretreatments enhanced soaking more effectively than heat treatments at 60 °C and 100 °C, whereas no significant effect was observed at 25 °C, indicating a minimal heat requirement for moisture and gas release. Notably, repeated PDB application (more than 40 times) further increased the moisture absorption without thermal influence. Scanning electron microscopy revealed that the PDA, PDB, and heat treatments caused cracks in the hilum region and increased surface wrinkling and mesh structure deformation. These findings demonstrate the potential of pressure drop treatment to improve soaking efficiency through structural modification, supporting its use as an effective nonthermal pretreatment method.

In this study, a Multi-Sensor IoT Device (MSID) is developed that is designed to collect various environmental data and interconnect with the cloud and blockchain to ensure reliable data management. The MSID is designed with a flexible, modular structure that supports a variety of sensor configurations and is easily expandable with 3D-printed components. The system performance was monitored in real-time, with a high cloud upload success rate of 98.35% and an average transmission delay of only 0.64 s, confirming stable data collection every minute. Blockchain-based sensor data storage ensured data integrity and tamper-proofness, with all transactions successfully recorded and verified via smart contract. The proposed Blockchain-based Mobile IoT System (BMIS) has shown strong potential for use in environmental monitoring, industrial asset management, and other areas that require reliable data collection and long-term preservation.

The Food Process Robot Intelligent System (FPRIS) integrates a 3D-printed six-axis robotic arm with Artificial Intelligence (AI) and Computer Vision (CV) to optimize and automate the coffee roasting process. As an application of FPRIS coffee roasting, this system uses a Convolutional Neural Network (CNN) to classify coffee beans inside the roaster and control the roaster in real time, avoiding obstacles and empty spaces. This study demonstrates FPRIS’s capability to precisely control the Degree of Roasting (DoR) by combining gas and image sensor data to assess coffee bean quality. A comparative analysis between the Preliminary Coffee Sample (PCS) and Validation Coffee Sample (VCS) revealed that increasing roast intensity resulted in consistent trends for both samples, including an increase in weight loss and Gas sensor Initial Difference (GID) and a decrease in Sum of Pixel Grayscale Values (SPGVs). This study underscores the potential of FPRIS to enhance precision and efficiency in coffee roasting. Future studies will expand on these findings by testing FPRIS across various food processes, potentially establishing a universal automation system for the food industry.