National Technical University of Athens

The rapid evolution of Artificial Intelligence (AI) and Machine Learning (ML) has significantly heightened computational demands, particularly for inference-serving workloads. While traditional cloud-based deployments offer scalability, they face challenges such as network congestion, high energy consumption, and privacy concerns. In contrast, edge computing provides low-latency and sustainable alternatives but is constrained by limited computational resources. In this work, we introduce SynergAI , a novel framework designed for performance- and architecture-aware inference serving across heterogeneous edge-to-cloud infrastructures. Built upon a comprehensive performance characterization of modern inference engines, SynergAI integrates a combination of offline and online decision-making policies to deliver intelligent, lightweight, and architecture-aware scheduling. By dynamically allocating workloads across diverse hardware architectures, it effectively minimizes Quality of Service (QoS) violations. We implement SynergAI within a Kubernetes-based ecosystem and evaluate its efficiency. Our results demonstrate that architecture-driven inference serving enables optimized and architecture-aware deployments on emerging hardware platforms, achieving an average reduction of 2.4× in QoS violations compared to a State-of-the-Art (SotA) solution.

• An active learning Kriging methodology for Bayesian updating is presented. • The BUS method recasts the updating problem into an equivalent reliability problem, solved with a Kriging metamodel. • A multi-fidelity approach is employed in the active learning Kriging model, where models of different accuracy are adopted. • Multi-fidelity Bayesian optimization is used for calculating the constant c of the BUS method. • An improved version of the Quantified Active learning Subset Simulation method is proposed to define the posterior samples. The paper presents a multi-fidelity framework for efficient Bayesian updating using active learning Kriging models. The updating problem is solved as an equivalent probabilistic problem using the Bayesian updating with Structural reliability (BUS) method. The proposed approach leverages an ensemble surrogate modeling strategy that combines two Kriging models of multiple fidelities, thereby enhancing predictive capacity while reducing computational cost by exploiting information from models of different fidelity levels. A key contribution is the integration of multi-fidelity Bayesian optimization in order to determine the optimal BUS constant, ensuring accurate transformation of the posterior into a reliability problem. Furthermore, an improved version of the Quantified Active Learning Subset Simulation (qAK-SuS) method, previously proposed by the authors, is proposed in order to efficiently estimate small failure probabilities and obtain the posterior distribution. The proposed framework, called qAK-BUS, offers a robust solution for Bayesian inference in engineering systems by smartly balancing model fidelity, prediction variance, and learning efficiency. Three numerical examples and an engineering application that involves the updating of a reinforced concrete bridge are examined for demonstrating and validating the proposed framework.

This study provides a detailed parametric computational fluid dynamics (CFD) assessment of the thermal performance of floating photovoltaic (FPV) systems under a range of environmental conditions. A three-dimensional multiphase model was developed using the Volume of Fluid (VOF) method in ANSYS Fluent 2024 R1. The panel vertical height above the water surface, wind velocity, water temperature, air temperature and solar irradiance were parameterized to evaluate their impacts on performance. Each panel thermal influence on downstream neighboring panels was evaluated with consideration of cumulative panel heating interference along the linear FPV array. The key findings were that in terms of increased panel thermal losses, being installed close to the water surface can reduce convective cooling from air to the FPV panels. An installation height of 2 meters effectively balanced heat loss potential with thermal efficiencies. Power outputs using uniform temperature estimates across each FPV panel can lead to under predictions as much as 0.81% at low heights while higher heights were over-predicted by as much as 0.41%. Under heat wave conditions the optimal height is calculated at 0.8 meters. The model results were verified analytically and validated with empirical models. Collectively, these results advance accurate thermal modeling and reduce uncertainty in the performance of FPV installations, particularly in locations subject to highly variable wind conditions. • 3D VOF CFD model captured the thermal behavior of floating PVs. • Optimal panel height of 2.0 m minimized the PV operating temperature. • Results agreed well with analytical and empirical models. • Power output error: +0.81% at low and −0.41% at high panel heights. • Height of 0.8 m maximizes efficiency under heat wave conditions.

ABSTRACT Background Intervertebral disc degeneration (IDD) involves complex inflammatory and oxidative stress pathways within nucleus pulposus (NP) cells, leading to catabolic remodeling and loss of disc function. Traditional single‐analyte assays capture only a small part of these coordinated networks. Leveraging the multiplex capability of Luminex, this study uses targeted immunoassays to simultaneously quantify 73 extracellular proteins and 21 intracellular signaling molecules in primary human NP cells stimulated with IL‐1β, Pam2CSK4, or hydrogen peroxide, enabling comparison of shared and stimulus‐specific responses. Methods Human NP cells isolated from patients undergoing spine surgery were plated in monolayer or encapsulated in alginate and stimulated for 1 week with either: 100 pg/mL or 1 ng/mL Interleukin (IL)‐1β, or the Toll‐Like Receptor (TLR) 2/6 agonist Pam2CSK4 at 100 ng/mL, or 50 μM Hydrogen Peroxide (H 2 O 2 ). The role of the stimuli was investigated using mRNA expression analysis and Luminex multiplex immunoassays. Results TLR2/6 agonist and IL‐1β stimulation significantly up‐regulated IL‐6, IL‐8, IL‐1β, and MMP‐3 mRNA expression, while H 2 O 2 failed to induce any changes. The secretome showed that IL‐8, IL‐6, MMP‐7, Intercellular Adhesion Molecule 1 (ICAM‐1), and Growth‐Regulated Oncogene‐α (GRO‐α) were the most responsive proteins across treatments and culture conditions. Intracellularly, phosphorylation of Jun Proto‐Oncogene AP‐1 Transcription Factor Subunit (cJUN), Nuclear Factor Kappa B (NF‐kB), and Inhibitor of Nuclear Factor Kappa B Alpha (IKBA) was observed following IL‐1β and Pam2CSK4 stimulation. Conclusions Across inflammatory and innate immune activation and under both monolayer and 3D culture, a concise five‐marker panel, namely IL‐6, IL‐8, GRO‐α (CXCL‐1), ICAM‐1, and MMP‐7 captured NP cell responses with high consistency, while phosphoproteomics demonstrated convergence on NF‐κB/AP‐1 signaling. These markers are well established within disc‐degeneration pathways (inflammation, innate immunity, and ECM remodeling); thus, they offer practical, reusable readouts for comparative pharmacology, assay qualification, and early biomarker development in translational spine research.

A novel approach for detecting preliminary signals designating upcoming entrance of a loaded system to the critical stage of impending fracture is assessed. The approach is based on the analysis of a time series of the cumulative number of acoustic events, the amplitude of which exceeds the respective average value of all the events recorded during loading. Using the "sliding window" technique, the average slope of the evolution of this time series is quantified, either against conventional or natural time (the latter provides a more detailed view of the stage before macroscopic fracture, during which the "information" gathered is very densely packed in a short interval). For the needs of this study, data from a previously published experimental protocol are exploited. The protocol comprised notched, beam-shaped specimens, made of either plain or fiber-reinforced concrete, under three-point bending. It is concluded that the slope of the evolution of the above time series systematically attains a value equal to unity slightly before the applied load attains its peak value. The results of the present analysis are in qualitative agreement with the respective ones based on either the instantaneous frequency of generation of acoustic events or the Euclidean distance between the sources of acoustic signals.

Steels remain among the most widely used structural and engineering materials in modern infrastructure, energy systems, and industrial facilities. Their long-term reliability depends critically on the early detection of corrosion damage and microstructural degradation. This review surveys recent advances in two complementary families of non-destructive evaluation (NDE) methods: magnetic techniques, including magnetic Barkhausen noise (MBN), magnetic flux leakage (MFL), eddy current testing (ECT), and magnetic hysteresis analysis; and electrochemical methods including electrochemical impedance spectroscopy (EIS), linear polarization resistance (LPR), scanning vibrating electrode technique (SVET), and electrochemical noise (EN). Recent progress in sensor miniaturization, signal processing algorithms, and multi-technique integration is reviewed. Particular attention is given to the sensitivity of these methods to microstructural changes reported in the literature, including carbide dissolution, phase transformations, temper embrittlement, and sensitization in stainless steels, as well as to the conditions under which such sensitivity has been demonstrated. The potential synergy between magnetic and electrochemical monitoring is discussed as a possible pathway toward more robust, condition-based maintenance frameworks. Challenges related to field deployment, environmental interference, calibration, and data interpretation are identified, and future directions—including machine learning-assisted analysis and multi-physics sensor arrays—are outlined.

The authors wish to make the following correction to this paper [...]

Landslides in the Western Corinth Rift reflect a mix of long-term “set-up” conditions—such as terrain, rock type, and fault-related structure—and short-term triggers linked to transient deformation and changing rainfall patterns. To represent these interacting processes in a clear and interpretable way, we propose a two-phase, multi-scale landslide susceptibility workflow based on explainable XGBoost. At Phase 1 (watershed scale) we develop a baseline susceptibility model using a standardized set of conditioning factors. These include (i) terrain and geomorphometric variables (elevation, slope, aspect, profile curvature, plan curvature and topographic wetness index (TWI) and (ii) lithological and structural controls (lithology and hydrolithology classes, distance from river network and fault-influence proxies such as distance to faults). The model is trained using historical landslide inventories, whereas interpretability was built in through explainable AI tools, such as SHAP, allowing us to quantify both global and site-specific contributions of conditioning factors, including key interactions. The result is a set of susceptibility maps paired with readable diagnostics that explain why certain areas are critical. At Phase 2 (local refinement and activity confirmation) focuses on the Krini–Gkrekas–Pititsa sector, where observations are denser and more reliable. Here, we evaluate whether susceptibility hotspots from Phase 1 align with evidence of ongoing or emerging instability. We add dynamic indicators and independent validation using: European Ground Motion Service InSAR ground motion, SBAS historical InSAR data; GNSS trend metrics and antecedent precipitation indices from station data. The goal is not just to refine local interpretation, but to test whether predicted patterns make physical sense, by checking consistency between (a) areas predisposed by lithology and structure and (b) present-day deformation signals and rainfall forcing. The workflow aims to produce decision-ready, interpretable outputs at two complementary scales: (1) watershed-scale susceptibility that highlights where failures are more likely based on relatively stable controls, and (2) a localized assessment that strengthens confidence where susceptibility coincides with measured deformation and hydrometeorological conditions. This improves trust and usability of AI-assisted landslide hazard assessment in tectonically active landscapes.KeywordsLandslide susceptibility; XGBoost; explainable AI; SHAP; multi-scale modeling; watershed analysis; lithology; active faults; EGMS; InSAR ground motion; GNSS; antecedent precipitation index; Western Corinth Rift.