Search Results (505)

Existing reviews on AI in tea production are either agriculture-generic or limited to isolated tasks. This review thoroughly compares vision technologies (RGB, hyperspectral, near-infrared, thermal, Light Detection and Ranging (LiDAR), Unmanned Aerial Vehicle (UAV)) and establishes a task-oriented algorithm selection framework for the tea industry. For small-sample or near-linear problems, traditional machine learning (ML) (support vector machine (SVM); partial least squares regression (PLSR)) remains effective. For unstructured field tasks, deep learning achieves superior performance: pest detection accuracy exceeds 97%, tea bud detection reaches 96.8% with RGB images, and hyperspectral imaging predicts nitrogen content with R² > 0.90 and tea polyphenols with R² up to 0.925. Algorithm choice further differentiates by task granularity: lightweight convolutional neural networks (CNNs) balance speed and accuracy for edge deployment at 16 fps; You Only Look Once (YOLO) series detectors enable real-time localization on mobile platforms at 93.1% accuracy, 24 ms per target. No single algorithm dominates all tea tasks; selection is a trade-off among accuracy, speed, data availability, and computational constraints. These findings outline a structured analysis of the challenges and pathways for transitioning computer vision (CV) from laboratory research toward field-deployable tools. Full article

Frozen seafood plays an important role in the global food supply, but maintaining its quality during frozen storage and cold-chain distribution remains a significant challenge. Although freezing effectively slows microbial growth and enzymatic activity, it cannot completely prevent quality deterioration. During frozen storage, seafood undergoes a series of interconnected physicochemical changes, including ice crystal growth, protein denaturation and oxidation, lipid oxidation, water redistribution, and texture deterioration. These changes gradually reduce sensory quality, nutritional value, and overall commercial acceptability. Conventional quality assessment methods, including destructive laboratory analyses and sensory evaluation, are still widely used. However, they are often labor-intensive, time-consuming, and unsuitable for rapid or real-time monitoring in modern cold-chain systems. As a result, increasing attention has been given to non-destructive sensing technologies that can evaluate seafood quality quickly and objectively. This review summarizes the major mechanisms responsible for quality deterioration in frozen seafood, together with recent advances in sensing technologies used to monitor these changes. The sensing approaches discussed include near-infrared (NIR) and Raman spectroscopy, hyperspectral and fluorescence imaging, low-field nuclear magnetic resonance (LF-NMR), electronic nose (E-nose), electronic tongue (E-tongue), colorimetric sensor arrays (CSAs), and biosensors. This review also discusses the growing role of artificial intelligence in frozen seafood quality assessment, including chemometrics, machine learning, deep learning, and multi-sensor data fusion. Particular attention is given to their applications in quality prediction, industrial implementation, and decision support. Finally, current challenges and future research needs are highlighted, with emphasis on the development of interpretable, transferable, and real-time monitoring systems that can support more reliable quality assurance throughout the frozen seafood supply chain. Full article

Time-series analysis is important for various scientific and industrial fields, such as remote sensing where observations may be disturbed by clouds, have irregular revisits and different sensors. Current transformer-based approaches have enhanced the modeling of distant time relationships, but they are still constrained by specific task designs and poor adaptability to various time-series patterns. To solve these problems, we present TimeHome, a universal sparse transformer basic model for handling heterogeneous time series. TimeHome incorporates a Heterogeneous Mixture-of-Experts (H-MoE) component, where different expert types are chosen dynamically based on a low-rank temperature-controlled gating mechanism to fit various sequence features. Moreover, a hybrid local–global attention mechanism is designed to consider both short-term variations and long-distance correlations, while specific heads are used for unified prediction, missing value estimation and abnormal event detection. TimeHome is pretrained on TS-200B, a huge database of time series including various temporal patterns from different domains. Comprehensive tests on several benchmark datasets and remote sensing extended evaluations show that TimeHome performs well in long-term prediction, missing value replacement and abnormal event detection. The model also exhibits good zero-shot adaptation ability and fast inference speed by adjusting experts dynamically. The source code and pre-training data will be released publicly. Full article

Current thermal validation in food engineering commonly depends on rigid, temperature-only endpoints that generate thermal stress and the associated energy cost in the case of high-caliber matrices. To overcome this, this study provides a proof of concept for post-heating cooling inertia as the main kinetic driver of cumulative integrated lethality in mass-transfer-limited geometries (p value framework), followed by separate enzymatic and not microbial (or, e.g., oxidative) degradation following product heat transfer. Time-series and multi-spatial thermodynamic profiling was performed for the commercial validation runs of three pasteurization processes of large-diameter (90 mm) finely minced cooked chicken emulsions packed in polyamide casing. When active heating was stopped at a lower core limit of 71 °C, the thermal cooling inertia led total integrated lethality to increase by more than 216%, leading to an ultimate core p value of 76.87–89.56 min. It systematically exceeded the mandatory food safety limit (p ≥ 40 min) in all spatial coordinates in the forced convection chamber, confirming absolute thermal homogeneity. In a subsequent cold-chain stability challenge at around 70 days (0–4 °C), foodborne pathogens were not detected in the product, and saprophytic microflora was limited below critical spoilage bounds (total viable counts (TVC) and dominant lactic acid bacteria (LAB) progressively increased to 3.7 and 3.5 log10 CFU/g on day 70). The autoxidation pathways for primary and secondary lipids remained low (peroxide at 0.00 mmol/kg and malondialdehyde at less than 0.15 mg MDA/kg). On the other hand, a linear progression of free fatty acids (R² = 0.9673) determined that the single enzymatic lipid hydrolysis is responsible for the final biochemical limiting effect of the sensory lifespan. Shelf life evaluation via rigorous quantitative descriptive analysis showed that all organoleptic attributes remained well above market acceptance criteria until day 70. At the same time, this study supports a scalable, prediction-based thermodynamic paradigm that ensures biological safety and fundamentally optimizes industrial energy use. Full article

Artificial intelligence is increasingly used to improve industrial quality control, but its practical value depends on whether models remain accurate under different operating conditions and fault classes. This study evaluates an artificial-intelligence-based workflow for gear quality control using vibration signals measured on a real two-stage reduction gearbox. Two orthogonal vibration channels were analyzed for six health states, three shaft speeds, and two load levels. Because the time-series data were only partly stationary, the dataset was divided chronologically into training and test segments. A 54-feature representation was built from rolling-window statistics and operating variables, and six classifiers were compared: Light Gradient Boosting Machine (LGBM), Extreme Gradient Boosting (XGBM), random forest, decision tree, multilayer perceptron (MLP), and logistic regression. LGBM achieved the best overall accuracy (0.9728) while maintaining substantially lower training time than several competing nonlinear models. Class-wise precision, recall, and F1-score ranged from 0.95 to 1.00, and the nominal response time for most operating-condition transitions was approximately 0.0998 s. The results show that vibration-based machine learning can support robust, near-real-time fault identification in mechanical manufacturing environments. The study also highlights the importance of chronological validation, feature engineering over multiple time windows, and the trade-off between predictive performance and deployment efficiency. Because the validation dataset originates from one gearbox platform, the results should be interpreted as promising internal evidence rather than as proof of universal industrial robustness. Full article

Designing modern disassembly systems requires the integration of industrial automation equipment. Due to the support of various communication protocols, PLCs not only perform control tasks but also act as intelligent data centers in distributed production systems. PLC solutions increasingly combine traditional approaches to automation with modern digital technologies, enabling predictive maintenance, real-time data analysis, as well as remote process management and integration with digital twin simulation. The takt time of manual disassembly may vary due to human and technical factors; therefore, its estimation is a problem in many processes including, for example, Bluetooth speakers. This article discusses the issue of PLC-based control systems for a sorting process of dismantled parts, and the methodology of a digital twin framework in FlexSim software. A prototype of a sorting line based on a conveyor belt with an S7-1200 series PLC controller and a full digital twin development cycle were presented. The explicit assessment of takt-related temporal fidelity in PLC-mediated event streams remains less developed. Therefore, this article addresses this gap by using a Digital-Twin-in-the-Loop (DTiL) configuration as a digital twin validation setup in which a source process model generates PLC-mediated events and a separate resulting digital twin model is evaluated against this source. The article focuses on temporal fidelity, PLC-mediated event transfer, and takt-time estimation. Thus, the gathered empirical time data were then fed into the digital twin model and analyzed to obtain information about the time delay of the PLC signals. This article separates the general digital twin architecture from one specific validation scenario implemented in a digital twin in-the-loop configuration with FlexSim, Siemens TIA Portal, PLCSim Advanced, and a local network communication chain. Delay analysis is based on photocell event timestamps and inter-event time differences, which reduce the effect of initial clock mismatch. The results indicate that, under the tested local-network DTiL configuration, absolute event delays are visible, while inter-event timing and aggregated takt statistics remain highly consistent between the source and resulting models. These findings support the preliminary feasibility of PLC-mediated takt-oriented monitoring for long manual operations. Nevertheless, broader validation under different controller configurations, communication conditions, and operating scenarios is required before generalizing the proposed approach. Full article

The palladium market represents a critical role in supporting key industrial sectors and facilitating the energy transition, as it is widely used in the automotive industry, electronics, chemical manufacturing, and hydrogen energy. These sectors influence a steady demand amid tightening environmental regulations and the development of green technologies. The aim of this study is to assess the structural transformation of the global palladium market through 2030 and to project Russian palladium production for 2026–2028 amid the energy transition by applying economic-mathematical methods, including linear regression, the Grey forecasting model, exponential smoothing, and Autoregressive Integrated Moving Average (ARIMA) time-series modeling. Particular attention is paid to the analysis of factors influencing the dynamics of the global palladium market, including the electrification of transportation, the substitution of palladium with alternative materials, and changes in global supply chains. The simulation results showed that the exponential smoothing model possesses the highest predictive accuracy, enabling it to estimate future palladium production volumes. The market is undergoing a structural transformation: declining demand from the traditional automotive sector is partially offset by the development of new applications in hydrogen energy, electronics, and advanced materials, suggesting that technological improvements can compensate for the loss of conventional demand segments. The key findings are (1) exponential smoothing (R² = 0.9812) outperforms linear regression, Grey model, and ARIMA; (2) Russian palladium production is projected at 74–130 tonnes (2026), 63–141 tonnes (2027), and 54–150 tonnes (2028); and (3) the decline in automotive demand is partially offset by new applications. Our findings confirm the need for Russian producers to adapt their strategies to the structural transformation of global demand, deepen domestic processing, and develop new high-tech applications for palladium to maintain their competitive positions amid the energy transition. Full article

This study examines the impact of various pretreatments on the composting of white grape skin (Vitis vinifera cv. Graševina) over a 30-day aerobic composting period. The pretreatments included the extraction of bioactive compounds from the grape skin and grinding. The research used an integrated approach combining physicochemical and microbiological analyses with advanced multivariate statistics and Artificial Neural Network (ANN) modeling. The combination of grinding and bioactive compound extraction was associated with improved composting performance under the investigated experimental conditions. All treatments exhibited reduced phytotoxicity, with germination indices exceeding commonly reported thresholds for phytotoxicity assessment. Spearman’s correlation and Principal Component Analysis (PCA) identified composting time and organic matter transformation as the primary drivers of variability, with the first three components explaining over 72% of the total variance. Furthermore, time-series Multi-Layer Perceptron (MLP) models successfully predicted the evolution of key maturity indicators, such as conductivity, organic matter content, and germination index, with high accuracy using composting time as the sole input. This integrated framework suggests that optimized conditions and predictive modeling can effectively transform winery by-products into stabilized compost, supporting sustainable waste management and circular economy practices in the wine industry. Full article

Predictive maintenance (PdM) has proven to be a critical strategy for minimizing downtime and optimizing operational efficiency in industrial systems. In this work the authors propose a data-driven PdM framework for circulating water pumps (CWPs), combining statistical analysis with machine learning. This begins with the analysis of a multivariate time-series dataset consisting of 51 sensors monitoring motor electrical parameters, pump hydraulics, vibration, temperatures, broken states, recovery states and finally normal states. With the help of a statistical analysis, utilizing boxplots and both Pearson and Spearman correlation indices, the aim is to identify the key degradation parameters. Notably, motor phase current (Pearson r = −0.872), pump vibration (r = −0.809), and discharge pressure (r = −0.732) showed the strongest negative correlation with failure states, revealing a complete system shutdown pattern during failure conditions. Applying these results to the development of a Random Forest classification model was the next step. Due to class imbalance and label interpretation challenges, i.e., normal state, broken state and recovery state, the reformulation of the multi-class problem into a binary task focusing on “at-risk” states was executed. Therefore, this led to the critical finding of the reversal of conventional state labels, where RECOVERING states were found to correspond to inactive post-shutdown conditions requiring maintenance intervention, whereas BROKEN states exhibited characteristics more consistent with partial degradation. Lastly, the statistically selected feature model achieved a Recall of 0.76, a Precision of 0.70 and an F1-score of 0.73. Furthermore, a physics-based grouped subsystem representation combining vibration, electrical, hydraulic, thermal and rotational measurements substantially improved classification performance, achieving a Precision = 0.997, a Recall = 0.992 and an F1-score = 0.995. These results demonstrate the effectiveness of combining statistical analysis, engineering knowledge and machine learning for the predictive maintenance of industrial pumping systems. Full article

Failure prediction of industrial machinery remains a concern for numerous researchers and practitioners. In this research, the improvement of failure predictions was highlighted by including contextual factors due to the fact that machine operation is a socio-technical process influenced not only by historical data but also by machine operating conditions, maintenance policies, and environmental conditions, etc. The paper proposes a contextual framework for the prediction of failure, which involves the application of various quantitative and qualitative methods, such as statistical analysis, time-series, machine learning algorithms, interviews, and survey questionnaires. The results of the application of the model indicate that it is possible to obtain quantitative estimates of the time between failures (TBF) and the downtime (DT) by applying the aggregated expert assessments of contextual factors. The key findings of the research indicate that the development of TBF and DT prediction models, in a contextual sense, represents an important segment for the further planning of maintenance of machine systems. Full article

This paper introduces an IIoT-based hybrid predictive maintenance system for industrial steam boilers, which responds to the increased demands for making intelligent and accurate decisions by leveraging data-driven analytics in complex industrial environments. The proposed approach presents comparative hybrid predictive monitoring frameworks based on Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models integrated with Statistical Process Control (SPC) and Cumulative Sum (CUSUM) monitoring techniques for industrial boiler monitoring; it allows accurate system behavior prediction coupled with enhanced anomaly detection across interconnected subsystems. To ensure practicability, the framework is implemented in an integrated operation technology and information technology (OT–IT) architecture with one year of real operation data from an industrial steam boiler in an oil refinery. A two-phase validation strategy is employed to overcome the gap between offline model development and application. During the initial phase, predictive models are developed and tested based on multivariate time-series data to model both the time dependence of the processes and the mechanical variables. The second phase involves the online deployment of the predictive monitoring framework through a Hardware-in-the-Loop (HiL) implementation with Programmable Logic Controller (PLC)-based and Open Platform Communications Unified Architecture (OPC UA) communication to enhance realistic system validation under emulated boiler process conditions without disrupting live plant operations. The experimental results indicate that the GRU model outperforms the LSTM, achieving good R² (0.8956) and mean absolute percentage error (MAPE, 0.6345%), demonstrating strong predictive accuracy across key operational variables. In addition, SPC is used to set up adaptive operational thresholds based on normal industrial process behavior, and then CUSUM is applied to the prediction residuals to improve the detection of the gradual degradation of the system. Real-time validation ensures system stability, low latency, and bidirectional data transfer between the OT and IT layers, enabling continuous monitoring and real-time decision-making. The proposed solution provides a practical and scalable predictive maintenance framework in an industrial context, particularly in oil and gas operations, that helps to transition to Industry 4.0 and intelligent asset management. Full article

This paper proposes an intelligent time-series early warning framework based on a production LSTM–Transformer network for petrochemical refining processes. A cascaded encoder–decoder architecture is designed, where the LSTM extracts local temporal patterns and medium-term memory from noisy industrial data, while the Transformer models global dependencies and cross-unit interactions via multi-head self-attention. An adaptive feature fusion layer bridges the representational gap between the two networks. A multi-stage preprocessing pipeline tailored for refining MES data handles missing values, outliers, and mixed operating conditions. Using 120 variables from five units of a fluid catalytic cracking unit, the framework predicts the regenerator bed temperature up to 8 h (48 steps) ahead. Comparative experiments show that the production LSTM–Transformer achieves a mean MAE of 0.088, a mean RMSE of 0.113, and the lowest median MAPE of 19.91% among all models, outperforming standalone LSTM (MAE 0.095, MAPE 20.85%) and Transformer (MAE 0.088, MAPE 20.49%). Robustness analysis confirms stable performance under strong noise (down to 5 dB) and missing rates up to 50%, with a median MAE of 0.1027 across tags. This work provides an effective, end-to-end predictive early warning solution that balances accuracy, production importance coverage, and industrial robustness, offering a generalizable data-driven paradigm for process industries. Full article

Porosity-related quantities from industrial X-ray CT depend on segmentation and inference choices. When inference defaults are omitted from the report, void or phase fractions can shift by amounts comparable to slice-to-slice variability. The contribution is metrological rather than architectural: we document a reproducible nnU-Net 2D workflow on Dataset601 CTVoid from semantic labels to slice-wise void fraction, optional two-dimensional connected-component pore summaries, isotropic three-dimensional stacking at 0.058 mm spacing, and spatial axis diagnostics, with region of interest and voxel spacing stated explicitly. The main results pair a weak export policy, defined as a single forward pass per slice without multi-scale fusion or test-time augmentation, with a strong policy that enables multi-scale fusion and flip-based augmentation on the same slice exports and identical weights, on one hundred consecutive slices from one cementitious industrial stack of 1028 × 1028 pixels. In parallel we report trainer validation on eight named Dataset601 validation cases and mirroring-based test-time augmentation off versus on re-inference on those same cases; case identifiers and the cross-validation split appear in the main text. These quantities answer different questions and must not be substituted for one another or for independent full-stack ground truth. Porosity-related scalars from industrial X-ray CT depend on how segmentation and inference are configured; when defaults are omitted, void fractions can shift by amounts comparable to slice-to-slice variability. For fixed nnU-Net weights on one cementitious industrial slice stack (1028 × 1028 pixels), we benchmark weak inference (single forward pass, no multi-scale fusion or test-time augmentation) against a strong export policy (multi-scale fusion and flip-based augmentation) on 100 paired slices, and report parallel trainer validation and TTA-off versus TTA-on re-inference on eight Dataset601 hold-out cases. For the industrial dataset, mean void-class IoU between modes is 0.716 (SD 0.043), while strong inference is ~2.6× slower and predicts lower mean void area (2.37% vs. 3.04%). The full weak export gives a 3D void ratio of 2.44% and integrated void volume of 5175 mm³. On validation patches, mean void Dice/IoU against the reference are 0.835/0.728, while weak–strong void IoU reaches 0.924 under the nnU-Net-native TTA contrast—quantities that must not be interchanged across domains or definitions. The present benchmark does not include a systematic polymer dosage series, and the study does not equate semantic void with open porosity but provides a reproducible disclosure template relevant to porous and polymer-modified cementitious CT reporting. Full article

The issue of water resources in a semi-arid country such as Morocco has been present for many years and is becoming increasingly critical. The droughts experienced over recent decades have demonstrated the country’s extreme vulnerability to any water deficit. In this context, the Berrechid plain represents a relevant case study illustrating both the practical and theoretical challenges of groundwater governance. The aquifer is heavily exploited to satisfy agricultural, industrial, and domestic needs. This study develops a hybrid “grey-box” modeling approach for predicting groundwater depth (GWD) fluctuations under climate change (CC). Unlike conventional black-box machine learning models, our framework combines a deterministic physical engine with a stochastic machine learning corrector. The physical component simulates aquifer mass balance using the Hargreaves method for evapotranspiration, linear drainage, climate memory via exponential decay, and an anthropogenic trend parameter (xi). The machine learning component—XGBoost with quantile regression—is trained exclusively on physical model residuals and predicts the 5th, 50th, and 95th percentiles, providing explicit 90% confidence intervals. Hydrological states (dry, normal, wet) are identified via K-means clustering for context-aware correction. The model is calibrated using historical data (1972–2019) and validated using blocked time-series cross-validation. Climate projections under the RCP 4.5 and RCP 8.5 scenarios were used to forecast GWD up to 2100. At piezometer 3933/20, the best performance was achieved, with an RMSE of 0.347 m and a KGE of 0.742 during the validation period. The proposed approach is suitable for seasonal GWD forecasting and offers practical value for water managers and decision-makers in the Berrechid region. Full article

With the adoption of more ambitious emission reduction strategies in the shipping industry by the International Maritime Organization and the resulting stricter greenhouse gas emission reduction requirements, it is particularly important for all stakeholders in the global maritime shipping industry to assess the energy efficiency of shipping vessels. Forming predictive capabilities for ship fuel consumption and Carbon Intensity Indicator (CII) annual ratings, for example, are two important works. This article adopted 14 different algorithms in three categories of data-driven approaches, i.e., statistics, machine learning and deep learning, including polynomial regression, ridge regression, adaptive boosting, categorical boosting, elastic net, etc., and built the ship fuel consumption prediction model using ship noon report as the data source. The prediction accuracy and computational efficiency of model training were compared based on metrics of coefficient of determination, mean absolute percentage error and floating-point operations per amount of training data. Cross-validations were performed for all 14 algorithms to analyze their sensitivities to their respective tuned parameters. Comparisons indicated that algorithms of the statistics approach were sensitive to the quality of the data source, compared with the machine learning and the deep learning approaches. The accuracy of the elastic net algorithm was sensitive to the tuned parameters. Two algorithms, light gradient boosting machine and random forest, were selected based on their performances of prediction accuracy and computational efficiency of model training. Then, the selected algorithms were separately combined with long short-term memory as the time-series prediction algorithm to form their respective coupled framework. Both of the coupled frameworks achieved successful prediction of the CII annual discriminant and rating of the studied ships. The prediction accuracy was validated to be sufficient. Full article