Abstract: In this study, formation water sample №1082 from the Narimanov OGPD, together with crude oil samples from the Bulla-Deniz and Muradkhanli fields, were examined under laboratory conditions to evaluate the efficiency of chemical reagents. The Alkan-318 demulsifier, Marza-1 inhibitor, Difron-4201 depressor additive, and the combined ADM composition (Alkan-318 + Difron-4201 + Marza-1 in a 1:1:1 ratio) were tested for their effects on water separation, corrosion inhibition, sulfate-reducing bacteria activity, paraffin deposition, and pour point depression. Comparative experiments showed that the ADM composition demonstrated superior performance over individual reagents at equal concentrations. At an optimal dosage of 600 g/t, Alkan-318 and the ADM composition reduced residual water in Bulla-Deniz (75% water cut) and Muradkhanli (41% water cut) oils to 0.1% and 0.8%, respectively. For pour point depression, Difron-4201 (900 g/t) and ADM (600 g/t) achieved efficiencies of 169.2% and 176.9% in Bulla-Deniz oil, and 151.4% and 170.0% in Muradkhanli oil. Regarding deposit prevention, ADM reached 95.4% and 96.9% efficiency, significantly exceeding individual reagents. Corrosion tests revealed that Marza-1 and ADM provided up to 99.9% protection in aggressive H₂S and CO₂ environments, while ADM also exhibited a nearly complete bactericidal effect (99.8%) against sulfate-reducing bacteria, highlighting its multifunctional efficiency.

Abstract: Current paradigms in Self-Supervised Learning (SSL) achieve state-of-the-art results through complex, heuristic-driven pretext tasks such as contrastive learning or masked image modeling. This work proposes a departure from these heuristics by reframing SSL through the fundamental principle of Minimum Description Length (MDL). We introduce the MDL-Autoencoder (MDL-AE), a framework that learns visual representations by optimizing a VQ-VAE-based objective to find the most efficient, discrete compression of visual data. We conduct a rigorous series of experiments on CIFAR-10, demonstrating that this compression-driven objective successfully learns a rich vocabulary of local visual concepts. However, our investigation uncovers a critical and non-obvious architectural insight: despite learning a visibly superior and higher-fidelity vocabulary of visual concepts, a more powerful tokenizer fails to improve downstream performance, revealing that the nature of the learned representation dictates the optimal downstream architecture. We show that our MDL-AE learns a vocabulary of holistic object parts rather than generic, composable primitives. Consequently, we find that a sophisticated Vision Transformer (ViT) head, a state-of-the-art tool for understanding token relationships, consistently fails to outperform a simple linear probe on the flattened feature map. This architectural mismatch reveals that the most powerful downstream aggregator is not always the most effective. To validate this, we demonstrate that a dedicated self-supervised alignment task, based on Masked Autoencoding of the discrete tokens, resolves this mismatch and dramatically improves performance, bridging the gap between generative fidelity and discriminative utility. Our work provides a compelling end-to-end case study on the importance of co-designing objectives and their downstream architectures, showing that token-specific pre-training is crucial for unlocking the potential of powerful aggregators.

Abstract: Contactless monitoring of vital signs such as Heart Rate (HR) and Respiratory Rate (RR) has gained significant attention, with vibration-based sensors like geophones showing promise for accurate, non-invasive monitoring. However, most existing systems are developed with healthy subjects and may not generalize well to extreme physiological ranges, such as those observed in infants or patients with arrhythmia. Moreover, the underlying mechanisms of cardiorespiratory vibration dynamics remain insufficiently understood, limiting clinical adoption of these systems. To address these challenges, we present a programmable cardiorespiratory testbed capable of generating realistic HR and RR signals across a wide range (HR: 40–240 bpm, RR: 8–40 bpm). Our system uses a voice coil motor that acts as the vibration source, driven by a Raspberry Pi based control circuit. Unlike similar systems that use separate modules for heart and lung signals, our setup generates both signals using a single motor. The synthetic signals exhibit a strong correlation of 0.85 compared with data from 75 human subjects. We use this system to design signal processing based algorithms for vital signs monitoring and demonstrate their robustness for extreme physiological ranges. The proposed system enhances the understanding of cardiorespiratory vibration dynamics while significantly reducing the time and effort required to collect real-world data.

Abstract: This article presents PrevOccupAI-HAR, a new publicly available dataset designed to advance smartphone-based human activity recognition (HAR) in office environments. PrevOccupAI-HAR comprises two sub-datasets: (1) a model development dataset collected under controlled conditions, featuring 20 subjects performing nine sub-activities associated to three main activity classes (sitting, standing, and walking), and (2) a real-world dataset captured in an unconstrained office setting captured from 13 subjects carrying out their daily office work for six hours continuously. Three machine learning models, namely k-nearest neighbors (KNN), support vector machine (SVM), and random forest, were trained on the model development dataset to classify the three main classes independently of sub-activity variation. The models achieved accuracies of 90.94 %, 92.33 %, and 93.02 % for the KNN, SVM, and Random Forest, respectively, on the development dataset. When deployed on the real-world dataset, the models attained mean accuracies of 69.32 %, 79.43 %, and 77.81 %, reflecting performance degradations between 21.62 % and 12.90 %. Analysis of sequential predictions revealed frequent short-duration misclassifications, predominantly between sitting and standing, resulting in unstable model outputs. The findings highlight key challenges in transitioning HAR models from controlled to real-world contexts and point to future research directions involving temporal deep learning architectures or post-processing methods to enhance prediction consistency.

Abstract: The chili pepper (Capsicum annuum) is among the most widely consumed vegetables worldwide, valued for its sensory and nutritional properties. Still, it is highly vulnera-ble to deterioration due to its elevated moisture content. Effective preservation strate-gies, such as the addition of salt combined with drying, are therefore crucial to main-taining quality and extending shelf life. This study employed a modified Reaction En-gineering Approach (REA) to model the drying kinetics and temperature behavior of chili paste under continuous and intermittent conductive hydro-drying conditions. Thirty experiments were conducted, considering various salt concentrations (0, 7.5 y 15 g salt/100 g paste) , water temperatures in the hydro-dryer, and heating intermit-tency through on/off cycles. The modified REA model accurately predicted both mois-ture and temperature profiles, with determination coefficients of 0.9463 and 0.8820, respectively. In addition to direct validation with the complete dataset, cross-validation between cayenne and jalapeño varieties demonstrated the ability of the model to generalize across different formulations and structural characteristics. These results confirm the robustness of the proposed framework and its suitability as a predictive tool for heterogeneous food matrices. Overall, the model provides a reliable platform for analyzing, designing, optimizing, and controlling hydro-drying processes in semi-solid foods, supporting the development of more efficient and sustainable preservation strategies.

Abstract: Sustainability has become a strategic priority for Higher Education Institutions (HEIs), particularly in the context of the Sustainable Development Goals, where university research plays a key role in addressing environmental, social, economic, and institutional challenges. However, the evaluation of sustainability-oriented research models remains limited by fragmented indicators, descriptive approaches, and the absence of robust, data-driven assessment frameworks. This study proposes a comprehensive framework for assessing the sustainability orientation of university research models, integrating validated measurement instruments with advanced analytical and predictive techniques to support evidence-based decision-making in higher education governance. The framework is based on a multidimensional instrument comprising 26 indicators across environmental, social, economic, and institutional dimensions, developed through expert judgment using the Delphi method and statistically validated by Confirmatory Factor Analysis (CFA). The instrument was applied to 260 researchers from four public HEIs located in the Colombia–Ecuador border region, and perceived performance was contrasted with actual institutional indicators, revealing significant non-linear discrepancies. To address this complexity, an artificial neural network model was developed to estimate real sustainability performance based on survey data, achieving a predictive accuracy of 90.92%. Beyond institutional diagnosis, the proposed framework functions as a decision-support tool that enables HEIs to identify critical gaps, prioritize interventions, and guide continuous improvement strategies in research management. Due to its methodological rigor, scalability, and transferability, the framework can be adapted to different higher education contexts, contributing to the advancement of sustainability assessment methods and governance practices in universities.

Abstract: The Wallonia region of Belgium aims to transition to a modern hydrogen infrastructure. Since hydrogen is much lighter than natural gas, so it is important to understand its nature and behavior while transporting through pipelines. This research aims to observe the pressure loss in pipelines due to surface roughness with H2 and other singular losses to find a solution to minimize the amount of pressure loss that occurs during transportation. This study involves numerical methods and gas equation models to determine thse pres-sure loss. This analysis includes the properties of hydrogen gas, pipeline material used, friction factor, pipeline efficiency, and other relevant properties of hydrogen and the pipe-lines. To address this challenge, this study integrates numerical fluid dynamics methods with structural modelling of pipeline walls. It accounts for long-term friction effects, erosion over several years, radial pressure gradients (mixing pressure drop), acceleration effects, and gravity influences, considering the non-ideal behavior of gaseous hydrogen (GH2).

Abstract: Methane emissions from end-use installations in residential natural gas systems remain poorly quantified, despite their importance to both safety and climate policies worldwide. While distribution networks and appliances have received research attention, interior piping between the meter and appliances represents a critical knowledge gap. To address this gap, a systematic survey of 473 residential systems in Saarlouis, Germany was conducted using standardized pressure-decay tests (DVGW G 600). Measurements were performed during the installation of gas regulators necessitated by a grid pressure increase from 23 mbar to 55 mbar above ambient. This provided a unique opportunity to assess whole-system leakage under controlled conditions without installation modifications. Leak rates were standardized to reference pressure and converted to methane emissions using measured gas composition. A total of 411 (86.9%) installations showed no detectable leak rate (LDL: 0.2 l h-1). However, seven systems (1.5%) exceeded 1 l h-1, and one surpassed the unacceptable threshold of 5 l h-1. Mean emissions across all systems were 0.067 [0.041, 0.098] g h-1, with smaller installations showing higher volume-normalized rates. Critically, fewer than 1.48% of systems contributed more than 46% of total emissions, demonstrating a strongly skewed, heavy-tailed distribution. Scaled nationally using Monte Carlo methods accounting for sampling uncertainty and skewed distributions, residential interior piping contributes 12.30 [8.11, 18.55] Gg yr-1 to Germany's methane emissions. These results emphasize the need to include residential leak rates in emission inventories and highlight the efficiency potential of targeted mitigation strategies focused on high-emitting installations under evolving EU methane regulations.

Abstract: Injection molding is widely used for plastic parts, but its performance is limited by the cooling stage, which dominates cycle time and affects dimensional stability and energy consumption. Conformal cooling channels, which can be manufactured using additive technologies, improve thermal efficiency but introduce a high-dimensional design problem. This work proposes an integrated methodology for optimizing injection molds with conformal cooling channels that combines parametric CAD, simulation, nonlinear principal component analysis, artificial neural network, and multi-objective evolutionary optimization. The workflow is applied to a case study with five cooling layouts. An initial set of 36 metrics related to temperature gradients, warpage, shrinkage, and energy is reduced to a small number of latent objectives, simplifying the search space while preserving the main physical trends. Artificial neural networks surrogates accurately reproduce numerical results, enabling exploration of the design space at a fraction of the computational cost. The optimization yields diverse Pareto-optimal solutions that balance cycle time, dimensional stability, and energy consumption, assisting the design of more sustainable injection molds. Sensitivity analysis identifies mold temperature and channel position/diameter as key design levers. The proposed methodology reduces dependence on expensive simulations and is readily transferable to industrial mold design.

Abstract: Mechanical coffee dryers have been widely adopted to reduce weather dependence, improve yield, and stabilize product quality. However, their operation is still energy-intensive and often suboptimal in terms of controlling the temperature, airflow and moisture content of the grains. In parallel, digital twin (DT) technology has emerged to virtually replicate complex processes and enable model-based monitoring, optimization, and control. This article presents a systematic review based on PRISMA on mechanical coffee dryers and their modeling and control strategies and the current and emerging use of digital twins in drying processes, including agricultural and food products with technological analogies to coffee. The results show a large amount of research on mathematical modeling, energy evaluation, and quality evaluation of mechanical coffee drying. Rapidly growing but still predominantly conceptual literature on digital twins for food processing and drying. Finally, only a small convergence between the two fields, with no fully realized digital twin for mechanical coffee dryers having yet been reported. This review found key gaps in the detection, data infrastructure, and development of hybrid physical-informed AI models. Finally, lines of research are proposed for mechanical coffee dryers enabled with digital twins, aimed at energy efficiency, product traceability and quality assurance.

Abstract: The growing demand for high-protein dairy products, driven by the expanding markets for infant formula and nutritional supplements, has led to a higher incorporation of milk protein ingredients like milk protein concentrate (MPC) and whey protein isolate (WPI) in dairy formulations. However, the effects of these protein additives on the thermal stability and sensory attributes of dairy products remain insufficiently studied. This research examines the influence of thermal processing (80 °C for 30 min) and protein fortification (MPC, WPI, and their combination) on the denaturation of whey proteins, the formation of volatile compounds, and the sensory characteristics of milk. Specifically, whole milk was fortified with MPC, WPI, and their combination at concentrations of 4% MPC, 4% WPI, and 2% MPC + 2% WPI, respectively, to evaluate the impact of different protein fortifications on these properties. Our findings reveal that heat treatment significantly promoted the denaturation of β-lactoglobulin and α-lactalbumin, with protein fortification playing a role in modulating these changes. Notably, lactoferrin exhibited matrix-dependent antioxidant behavior, meaning its antioxidant activity varied based on the protein composition and structure of the milk matrix, influencing its stability and function under different fortification conditions. Volatile profiling indicated that MPC enhanced the formation of sulfur-containing compounds and aldehydes, whereas WPI favored ketones and Maillard-derived volatiles. Sensory analysis revealed that heated WPI fortified samples exhibited stronger cooked and dairy fat aromas, while unfortified milk retained milky and grassy notes. Correlation analysis highlighted the mechanistic links between protein denaturation and lipid-derived compounds. These results emphasize that protein type and composition play crucial roles in flavor development. The strategic blending of MPC and WPI offers a practical approach to balancing volatile profiles and mitigating off-flavors, providing insights for the formulation of thermally stable, protein-fortified dairy products with optimized sensory quality.

Abstract: Self-compacting concrete (SCC) offers significant advantages in construction due to its superior workability; however, optimizing SCC mixture design remains challenging because of complex nonlinear material interactions and increasing sustainability requirements. This study proposes an integrated, sustainability-oriented framework that combines machine learning (ML), SHapley Additive exPlanations (SHAP), and multi-objective optimization to improve SCC mixture design. A large and heterogeneous global dataset, compiled from 156 peer-reviewed studies and enhanced through a structured three-stage data augmentation strategy, was used to develop robust predictive models for key fresh-state properties. An optimized XGBoost model demonstrated high predictive performance, achieving coefficients of determination of R² = 0.835 for slump flow and R² = 0.828 for T50 time, with strong generalization to industrial SCC datasets. SHAP-based interpretability analysis identified the water-to-binder ratio and superplasticizer dosage as the dominant factors governing fresh-state behavior, providing physically meaningful insights into mixture performance. A cradle-to-gate life cycle assessment was integrated within a multi-objective genetic algorithm to simultaneously minimize embodied CO₂ emissions and material costs while satisfying workability constraints. The resulting Pareto-optimal mixtures achieved up to 3.9% reduction in embodied CO₂ emissions compared to conventional SCC designs without compromising performance. External validation using industrial data confirms the practical reliability and transferability of the proposed framework. Overall, this study presents an interpretable and scalable AI-driven approach for the sustainable optimization of SCC mixture design.

Abstract: Weed management is a critical component of soybean production. Efficient weed con-trol can improve both yield and crop quality. However, conventional spraying tech-niques exhibit low pesticide utilization and contribute to environmental pollution. To address these issues, this study proposes a deep learning–based precision target spray-ing method. A lightweight YOLOv5-MobileNetv3-SE model was developed by modi-fying the backbone feature extraction network and incorporating an attention mecha-nism. Field images of weeds were collected to construct a dataset, and the detection performance of the model was subsequently evaluated. A grid-based matching spray-ing algorithm was developed to synchronize target detection with spray actuation. The system time delay, including image processing delay, communication and control de-lay, and spray deposition delay, was analyzed and measured, and a time-delay com-pensation strategy was implemented to ensure accurate spraying. Experimental results demonstrated that the improved model achieved an mAP@0.5 of 86.9%, a model size of 7.5 MB, and a frame rate of 38.17 frames per second. Experimental results showed that weed detection accuracy exceeded 92.94%, and spraying accuracy exceeded 85.88% at forward speeds of 1–4 km/h. Compared with conventional continuous spraying, pesticide reduction rates reduced 79.0%, 72.5%, 55.8%, and 48.6% at weed coverage rates of 5%, 10%, 15%, and 20%, respectively. The proposed method provides a practi-cal approach for precise herbicide application, effectively reducing chemical usage and minimizing environmental impact.

Abstract: This study examines cyber vulnerabilities affecting critical infrastructure along NATO’s 2 eastern flank, with a focus on industrial control systems and operational technology. It 3 addresses how hybrid threats exploit legacy protocols and interoperability gaps across 4 mixed-generation IIoT environments, increasing the likelihood of disruptive events. We 5 propose an AI-enabled framework that links cyber resilience engineering to European 6 regulatory and operational requirements through two components: a Unified Compli- 7 ance Framework that maps legal obligations to implementable technical controls, and 8 an AI-enabled Cyber Resilience Index that consolidates detection, operational continuity, 9 governance, and supply-chain risk into a single scoring model. The methodology combines 10 regulatory-control mapping, OT-specific gap analysis, and engineering validation of real- 11 time constraints, supported by a digital-twin testing environment used to evaluate resilience 12 under representative adversarial scenarios. Results from the simulation-based evaluation 13 show consistent improvements in detection and response stability across tested scenar- 14 ios and provide an auditable evidence model for continuous assurance. The framework 15 supports risk-informed governance and investment decisions by translating compliance 16 objectives into measurable service-level targets and operational resilience indicators, while 17 promoting time-deterministic architectures, federated learning, and explainable AI for 18 accountable deployment in industrial settings.

Abstract: Injection well placement and rate allocation are among the most decisive factors in determining the efficiency and bankability of CCS projects. However, optimizing these parameters is notoriously complex: even a small number of injection wells leads to a vir-tually infinite set of injection scenarios, while traditional optimization techniques typically require thousands of high-fidelity reservoir simulations. For project developers, this computational burden can stall critical Final Investment Decisions (FID). The proposed approach here addresses this bottleneck by using a Design of Experiments (DoE) framework combined with nonlinear surrogate modeling, which efficiently maps the relationship between injection rates and storage performance, to identify near-optimal solutions with a minimal number of simulations. We show that our method achieves up to 97% of the initially targeted CO2 sequestration with as few as 15 simulations, demonstrating a step-change reduction in time and cost. From a business standpoint, CCS operators can de-risk projects earlier, accelerate FID timelines, and evaluate multiple site configurations in parallel while minimizing computational overhead. Rather than waiting weeks or months for exhaustive optimization, decision-makers can gain timely, reliable insights that directly support capacity commitments, regulatory submissions, and ultimately revenue realization.

Abstract: In electronic product manufacturing, quality control and production cost management pose challenges for enterprises. First, key factors for production decisions are identified: component inspection, assembly, inspection of semi-finished and finished products, and handling defective goods at different stages. Next, dynamic programming and genetic algorithms optimise sampling inspection. A production decision model covers multiple processes and component combinations. The relationship between the inspection and the final product quality is explored, showing that different decision paths affect the total cost and defect rate. Real-time monitoring and a dynamic Bayesian network guide production strategy adjustments to boost efficiency and reduce defects. This study proposes adaptable inspection and disassembly strategies that reduce costs and optimise resource use across production scenarios.

Abstract: Efficient placement of Virtual Machines (VMs) iscritical for optimizing resource utilization and ensuring servicereliability in cloud computing infrastructures. Existing validationmethods for VM placement algorithms, such as limited in-vivoexperiments and ad hoc simulators, often fail to reflect real-worldcomplexities and provide fair comparisons. This paper introducesVMPlaceS, a simulation framework built on SimGrid, designed toaddress these shortcomings by enabling the robust evaluation andcomparison of VM placement strategies. VMPlaceS facilitateslarge-scale scenario modeling with customizable parameters torepresent dynamic workloads and realistic platform conditions.By simulating centralized, hierarchical, and distributed algo-rithms, this study highlights the framework’s capability to assessscalability, reactivity, and SLA adherence in various deploymentscenarios. VMPlaceS emerges as a valuable tool for researchersand practitioners to explore innovative VM placement solutionsand advance the field of cloud computing resource management.

Abstract: This research implemented management indicators to increase produc-tivity at the ice plant, seeking to align with SDG 9 through the construc-tion of resilient infrastructure, promoting sustainability and innovation. The overall objective was to improve plant productivity through the ap-plication of management indicators. The research was applied, with a quantitative approach and a pre-experimental design, analyzing produc-tion records. Results showed a 52.49% increase in production and a 107.65% increase in gross profit, with 52.50% of production capacity be-ing utilized. These improvements were attributed to the implementation of a balanced scorecard, standardization, the Kanban board, enhanced maintenance management, and staff training. Additionally, raw material productivity increased by 7.66%, labor productivity by 130.02%, and total productivity by 53.37%. These improvements were demonstrated through inferential testing, yielding a p-value of 0.006 (p < 0.05), thus confirming the hypothesis. In conclusion, the establishment of management indica-tors enabled the implementation of significant improvement actions within the organization.

Abstract: This research aimed to optimize the continuous monitoring of residual chlorine in the Potable Water System (PWS) managed by the Water and Sanitation Management Board (JASS) of the Sinsicap district, through the implementation of a low-cost technology, thus contributing to Sustainable Development Goal 6 (SDG 6). The study was developed using an applied approach, with an experimental design and explanatory scope. A prototype was designed and validated, consisting of an I2C/UART chlorine sensor, a PCB board, an SD module, and an LCD screen, programmed to record automatic readings three times a day at four points in the distribution network. The data obtained were analyzed using SPSS software, applying one-sample t-tests and calibration correlations. The results showed a significant correlation (R² = 0.983; p < 0.004), ensuring compliance with the sanitary standard (0.5-1.5 mg/L). Furthermore, the system achieved 95% availability and cost savings of 90.83% compared to commercial equipment. It is concluded that the developed technology improves the efficiency, accuracy, and sustainability of chlorine monitoring, representing a viable and replicable alternative for the country’s Water and Sanitation Management Boards (JASS).

Abstract:

The increasing use of integrated renewable energy sources (RES) is undoubtedly reshaping the structure of power systems. In such conditions, achieving energy efficiency and sustainability requires the development and integration of digital solutions to manage energy flows and resources optimization. This paper aims to provide a comprehensive overview of the successful integration of artificial intelligence (AI) in the energy sector, particularly in relation to the increasing utilization of renewable energy. The paper presents trends and potential scenarios in the digitalization of energy, along with the associated challenges. It analyzes particular applications of AI tools in strategic areas of the energy sector. The article also attempts to summarize the current status, goals, key areas, and activities in the irreversible transformation of power structures into digital intelligent ones. Five key areas in the energy sector have been identified in which AI tools are applied.