Engineering Solutions to Prevent Surgical Complications in Horses: Optimizing Suture Patterns

13:30

Jellis Bollens, Lise Gheysen, Marlies Verkade, Janne Stael, Ann Martens, Patrick Segers

Complications related to surgical incisions are common in horses following abdominal surgery, which typically involves an incision through the abdominal wall. In the classical approach, this incision is made along the linea alba, a fibrous tissue running longitudinally (cranial-caudal direction) along the ventral abdomen. After surgery, the abdominal wall is closed with sutures, where different suture patterns and parameters can be selected. These variations have shown to impact the rate of incisional complications. Therefore, this study aims to assess how different suture patterns and parameters influence the stresses in the tissue. The mechanical properties of the equine linea alba were first characterized using tensile tests. Tissue samples from 13 horses were loaded uniaxially in either the longitudinal or the transverse direction. The resulting stress-strain curves were used to fit the parameters of the anisotropic hyperelastic Gasser-Ogden-Holzapfel (GOH) material model. This calibrated model was then applied to develop a finite element model (FEM) of the sutured linea alba, emulating an interrupted suture pattern. The study varied the bite size (distance between the incision and the suture entry point) and the step size (distance between adjacent stitches) to analyse the effects on tissue stress. Additionally, a continuous suture pattern was modelled to compare its impact on stress against the interrupted pattern. The tensile tests revealed stiffer behaviour of the linea alba in the longitudinal direction compared to the transverse. Moreover, the GOH model fitting showed that the fibre orientation was generally aligned with the longitudinal axis, but that a large fibre dispersion was present. FEM results from the interrupted suture model demonstrated stress concentrations at the suture entry points, with maximum principal stresses increasing as both bite size and step size increased independently. Finally, the continuous suture pattern was found to induce higher tissue stress compared to the interrupted suture patterns.

Next-Generation EV Powertrains: Design and Challenges of High Power Density Multifunctional On-Board Chargers with Integrated Converters and Reduced Passive Components

13:35

Homayoun Soltani Gohari, Peter Sergeant, Hendrik Vansompel

Electrification is a transformative advancement in transportation, driving the shift toward sustainable and energy-efficient solutions. As electric vehicles (EVs) become more widespread, further innovations are required to enhance their performance and ensure their growth. A critical element in this progression is the powertrain, which combines various mechanical and electrical components vital to an EV’s operation. Among these, the on-board charger plays a pivotal role, as it directly impacts the vehicle’s performance during both charging and driving. One of the most promising strategies to enhance powertrain efficiency is the integration of motor stator windings as Power Factor Correction (PFC) filters, which significantly increases the system’s power density. However, this approach presents a range of technical challenges that must be addressed. Key issues include ensuring zero-torque production during the charging process to keep the vehicle stationary, managing the mutual inductance between different motor coils, and reconfiguring the motor windings to function optimally in the charging mode. Additionally, cooling system requirements must be carefully considered to prevent overheating, and the controller design must be adapted to ensure seamless operation under these complex conditions. This project focuses on designing a high power density, multifunctional on-board charger that supports both single and three-phase EV charging. The charger will be capable of integrating renewable energy sources, which aligns with broader sustainability goals. By addressing the challenges associated with using motor coils as PFC filters, this innovative solution aims to enhance the overall efficiency and performance of EVs. The outcome will not only contribute to more efficient EV powertrains but also support the ongoing transition to cleaner energy in the transportation sector.

Anti-Brownian Electrokinetic Trapping of Fluorescence-free Nanoparticles in Water

13:40

Farshad Rezakhanloo, Yera Ussembayev, Kristiaan Neyts, and Filip Strubbe

Anti-Brownian electrokinetic (ABEL) trapping has proven effective for confining individual nanoparticles in solution by counteracting Brownian motion with electric forces. However, its application has been largely limited to fluorescently labeled nanoparticles, restricting its broader utility. In this study, we introduce a novel axial ABEL trapping method that relies on the light-scattering properties of label-free nanoparticles. Using an evanescent field created by total internal reflection (a), we monitor the scattered light intensity (b) to generate a real-time feedback voltage (c) which is applied to electrodes to create an electric field, precisely controlling the nanoparticle’s position relative to a glass surface. This innovative approach enables high-resolution trapping and observation of a nanoparticle’s behavior in response to applied voltages (d) at kilohertz frequencies without the need for fluorescence labeling. Our method opens new possibilities for studying a diverse range of nanoparticles, paving the way for advanced research in many fields such as cell membrane biology, surface chemistry, and single-molecule biophysics. This label-free trapping technique offers immense potential for industrial applications, including developing novel biosensors, drug delivery systems, and diagnostics, making it highly relevant to academic research and commercial innovation.

Preoperative Endovascular Embolization Planning for Cerebral Arteriovenous Malformations: Patient-Specific Compartmentalisation Identification

13:45

Jessie Duquesne, Saar Vermijs, Mohammadjavad Sedghizadeh, Elisabeth Dhondt, Luc Defreyne, Uri Singfer, Vincent Keereman, Peter Vanlangenhove, Danilo Babin, Charlotte Debbaut

Cerebral arteriovenous malformations (cAVMs) are complex vascular anomalies in the brain that pose significant risks of haemorrhage and neurological dysfunction. In young adults (under 40 years), cAVMs are the most common cause of brain haemorrhage in absence of head trauma. Endovascular embolization treatments aim to occlude the cAVM vasculature with an embolic agent injected via a catheter. To avoid incomplete embolization or recanalization, it is essential to gain insight into the patient-specific cAVM compartmentalisation, determining which feeding/draining vessel corresponds to what part of the cAVM. Although previous studies demonstrated that compartmentalisation may be based on either anatomical [1] or hemodynamic features [2], no optimized and validated compartmentalisation methods are available. Therefore, this feasibility study proposes an anatomically inspired three-dimensional cAVM compartmentalization model. Building upon our previous work (kidney perfusion zones [3]), the patient-specific 3D model starts from segmentations in Mimics (Materialise, Belgium) after which centrelines were calculated using VMTK (vmtk.org) to label arterial feeders and draining veins. Seed points were placed at the final artery/vein centerline point before entering/leaving the cAVM, respectively. The seeds initiated a Python-based region-growing algorithm, executed in a voxelised bounding box surrounding the cAVM and its feeding/draining vessels, resulting in cAVM-specific feeding and/or draining compartments. The anatomically-based compartmentalisation model resulted in the amount of feeding compartments equalling the number of feeding arteries. The algorithm detected compartments of varying volumes, each with its own drainers when seeded from the arterial side. Additionally, it identified a fistula (i.e. high-flow and wide artery-vein passageway), which is important to account for regarding intervention safety. Future development will focus on integration of hemodynamic features including but not limited to information derived from intensity analysis of digital subtraction angiography images. [1] Pertuiset et al., Neurol Res., 1982. [2] ApSimon et al., Acta Radiol, 1986 [3] De Backer, Vermijs et al., European Urology, 2023.

Structural Integrity of a Power Cable

13:50

Esteban Cadavid Gil, Kris Hectors, Wim De Waele

Floating offshore wind turbines are a clean and sustainable technology for renewable energy generation. The electricity generated by the turbines is transmitted through dynamic power cables to the shore. The cables face challenging environmental and loading conditions due to the motions of the floating platform and the sea currents. These harsh loading scenarios pose a risk of fatigue damage in the cables, potentially diminishing their operational lifespan. This study presents a numerical framework to accurately compute the fatigue driving stresses in dynamic power cable components by integrating global and local analysis methodologies. The coupling of these methodologies will facilitate a comprehensive assessment of the cable’s response to varying loading conditions. In the global analysis, the overall mechanical behaviour of the power cable in response to the currents and floating platform motions is modelled. From the global deformation modes (bending, tension and torsion), section loads can then be extracted to drive a local analysis. In the local analysis, sub-models of the power cable are used to calculate the time series of stresses acting in individual components. These can subsequently be used to estimate the fatigue life of each component. Key aspects of the local analysis include a well-considered trade-off between the level of modelling detail and the computational cost. Critical parameters that need to be accounted for are the helical shape of the components, inelastic material behaviour, contact conditions, and large deformations. This research will contribute to a deeper understanding of the mechanical performance of dynamic power cables, vital for ensuring their reliable operation in the offshore floating wind industry.

Interior Insulation: A Risky Business?

13:55

Kaat Janssens, Valentina Marincioni, Nathan Van Den Bossche

Buildings represent 40% of the EU’s total energy use, making their renovation a top priority for reducing energy consumption and greenhouse gas emissions. The EU aims to upgrade its entire building stock to achieve an Energy Performance Certificate (EPC) label A. In Flanders, this goal necessitates renovating 95,000 homes annually—an equivalent of 260 homes each day. Facade insulation is essential for minimizing energy losses, but it presents unique challenges, especially as climate conditions change. Heat, Air, and Moisture (HAM) modelling tools can assess the hygrothermal behaviour of various building components under diverse climate scenarios. By simulating heat transfer, airflow, and moisture movement, HAM tools enable detailed performance analysis of wall assemblies over time, offering insights into building resilience over time. Despite their value, HAM tools have limited industry adoption due to their complexity and the advanced knowledge required to interpret detailed datasets. Their computational demands also pose challenges for everyday use. Addressing this gap, this research has developed an online web tool called HAMalyser. This tool uses thousands of simulations to generate accessible, customized predictions and straightforward advice on building performance. By simplifying HAM data interpretation, HAMalyser aims to support the construction industry in making informed decisions on energy-efficient building designs that align with the EU’s sustainability goals.

Electromagnetic entomology: Exposure of the Aedes aegypti mosquito

14:00

Eline De Borre, Arno Thielens

Radio-frequency electromagnetic fields originating from our telecommunication systems expose everything in our surroundings, including insects. The effects of electromagnetic fields on insects are not well known, while many insect are often a crucial part of our ecosystems. Some insects are vectors for pathogens, like the Aedes aegypti mosquito, spreading Dengue, Yellow fever and Zika and the control of this insect is important in many countries. The interaction of electromagnetic fields with insects is different than for humans and other invertebrates, due to their tissue and their size. For electromagnetic fields with wavelengths comparable to the insect size, resonance effects could cause a higher power absorption, possibly leading to dielectric heating. In this study, we use the Aedes aegypti mosquito, as a model organism to investigate the effect of radio-frequency electromagnetic fields on the development of insects. Early life stages of Ae. aegypti were exposed during experiments and wing length and development time were analyzed. Before the exposure experiments were executed, numerical simulations revealed the dose received by the insect during the experiments, taking into account the exposure setup and fact that both the larvae and pupae are aquatic. Also the position of the insect and frequency dependency were investigated. The simulations used new 3D models of a Aedes aegypti larva and pupa generated from micro CT-scans.

Electrochemical CO2 Reduction to CO for In Situ Resource Utilisation on Mars

14:05

Paulina Govea-Alvarez; Jason Song; Zhiyuan Chen; Yi Ouyang; Kevin M. Van Geem

The development of sustainable technologies for in-situ resource utilisation (ISRU) on Mars is critical for future human missions. This study investigates the electrochemical reduction of carbon dioxide (CO2) to carbon monoxide (CO) in laboratory conditions, focusing on the performance of silver-based gas diffusion electrodes (Ag GDE) in various catholytes. We evaluate key metrics such as pH and different salt concentrations on the catholytes to evaluate which current density obtains the highest faradaic efficiency (FE%) towards CO; this means the optimum condition at which CO is preferred over the hydrogen evolution reaction. This is the first step towards developing an ISRU process for Mars utilisation since CO2 is the primary component of the Mars atmosphere, and converting it to CO can benefit other chemical processes that generate fuel. The results indicate that selected sulphate catholytes, especially potassium, significantly increase CO yield while minimising hydrogen formation at specific lower current densities. Moreover, a more acidic media results in more hydrogen generation instead of CO, while a lower but still acidic media (such as pH 3) promotes the reduction of CO2 into CO using an Ag GDE. This research encompasses the use of salts as electrolytes and acidic media to enhance the reduction of gaseous CO2 towards CO by using a silver GDE in a microfluidic cell.

Pseudonymity using Solid and the Decentralised Identity Foundation

14:10

Gertjan De Mulder, Ben De Meester

Nowadays, a user’s digital identity is primarily controlled by centralised platforms (such as Google, Facebook, etc.), which allow them to track and profile a user’s online presence, posing a threat to privacy. These rising privacy concerns instigated a fundamental shift towards decentralised ecosystems, such as Solid and the Decentralised Identity Foundation (DIF), which place privacy and control at the core. However, Solid still needs to address some challenges when considering pseudonymity as a vital privacy-enhancing technology (PET) that allows users to perform accountable actions without exposing their true identity. Solid only vaguely covers pseudonymity in current specifications or use cases despite becoming an increasingly prominent Web-based and decentralised data-sharing ecosystem. Meanwhile, DIF is gaining momentum in establishing Self-Sovereign Identity, i.e., user-controlled identity. Although Solid and DIF share similar objectives, each ecosystem emphasises distinct aspects of decentralisation. This divergence in focus presents a compelling rationale for examining their complementarities. In this work, we researched the technical requirements and design choices needed to achieve pseudonymity. Moreover, we developed two solutions to pseudonymity: one solely based on Solid, the other applying a complementary approach combining Solid and DIF. Furthermore, we performed a privacy threat analysis (PTA) to identify gaps and vulnerabilities when implementing pseudonymity solely based on Solid and to determine to what extent these are solved and mitigated by complementing Solid with DIF. Our PTA shows the complementary approach reduces linkability, detectability, and data disclosure threats. This further strengthens the incentive towards complementary approaches between decentralised ecosystems such as Solid and DIF, thereby providing users with more control over their digital existence (both identity and data) on the Web.

Extending path memory to improve kinetics calculations in rare molecular processes

14:15

Elias Wils, Wouter Vervust, Titus Van Erp, An Ghysels

Molecular dynamics (MD) simulations serve as a computational microscope to study the intricate kinetics of crucial biological processes, such as the binding of a cancer drug molecule or permeation of a molecule through cell membranes. One of the main challenges in this field is the significant disparity in timescales: while these processes occur rapidly, simulating them can require years of CPU time. To solve this, we use Transition Interface Sampling (TIS), an enhanced sampling algorithm that bridges the gap between these widely separated timescales. This is done by quantifying the progress of a reaction using a single variable and placing checkpoints ("interfaces") along the way, after which we cleverly sample MD trajectories ("paths") at strategic points. However, TIS can struggle with simulating biological processes with complex energy landscapes, often leading to paths getting stuck in metastable states, resulting in lengthy simulation times. PPTIS (Partial Path TIS) removes this problem by cutting paths short, which unfortunately also truncates the history of particle trajectories, potentially leading to a significant loss in accuracy. To tackle these shortcomings we introduce StapleTIS, an innovative approach that preserves essential trajectory memory while keeping paths manageable in length. The paths are terminated by turns (occurring when encountering an energetic barrier for example) on both sides, mimicking the shape of a staple. This ensures paths "remember" a longer segment of their trajectory, enabling them to navigate a potential energy landscape more efficiently and giving us more accurate information about the kinetics of the system. We demonstrate the effectiveness of StapleTIS using 1D and 2D toy systems, achieving performance comparable to traditional TIS. Our next steps involve applying this method to real-world biological systems to obtain valuable information about their kinetics, ultimately enriching our knowledge about various biological mechanisms.

Network-Centered Resource Management for HPC Networks

Dante Van Poucke, Wouter Tavernier, Didier Colle

The significance of High-Performance Computing (HPC) cannot be overstated in today's research and industry landscape. Supercomputers play a crucial role in solving the most challenging problems in various domains, such as climate modelling, decoding the genome of cancerous cells, and training language models like ChatGPT. To preserve the development of innovative ideas and technologies, supercomputers must continue to grow in size. Consequently, the significance of the interconnection network that connects the computing resources becomes increasingly critical. However, the communication demands of common HPC workloads often cause bottlenecks in the supercomputer’s network, limiting the utilization to just 3% of the theoretical peak performance. This underutilization does not only waste money and energy but also limits innovation. Addressing this inefficiency presents an opportunity to enhance system efficiency through network-oriented resource management. The goal of our research is to improve the effective computing power of supercomputers by developing new resource management strategies. We advocate that flexibility and direct control over resources are the fundamental principles for improving efficiency. State-of-the-art approaches concentrate solely on managing the compute resources and only consider static application and network behaviour. We plan to improve resource managers by introducing methods for real-time monitoring and modelling large-scale networks. Utilizing these methods, we aim to develop scalable strategies for managing both compute and network resources. Furthermore, we will explore the potential of elastic job scheduling. Through these efforts, we aim to overcome the network bottlenecks and prepare HPC systems for the future.

Hydrogen production from decomposition of ethanol using a non-self-sustained plasma discharge at atmospheric pressure: Hydrogen selectivity and energy efficiency evaluation

Aftab Javaheri, Victor Lievens, Mikhail Gromov, Anton Nikiforov, Rino Morent, Nathalie De Geyter

Today, hydrogen stands out as one of the most strategically important substitutes for carbon-based fuels, offering numerous advantages. These include a reduction in reliance on finite reserves of fossil fuels, mitigation of the greenhouse effect by minimizing CO2 emissions, and the release of more energy during combustion compared to traditional fuels. However, over 95% of the current hydrogen production stems from fossil fuels, predominantly through steam methane reforming (SMR), an energy-intensive process that generates substantial quantities of CO2. Hence, there's a growing interest in exploring alternative techniques, such as leveraging plasma sources. Non-thermal plasmas (NTPs) have emerged as a promising avenue for hydrogen production through the decomposition of hydrocarbons, alcohols, and various polymers, given their CO2 neutrality and compatibility with renewable energy sources. NTPs can efficiently dissociate almost any H-rich compounds into molecular hydrogen and other valuable products such as C2H6 and CO. Among H-rich feedstocks, ethanol has captured researchers' attention due to its derivation from renewable bio-ethanol obtained from biomass. However, enhancing the conversion efficiency and selectivity of this process remains of importance. Within the realm of NTPs, non-self-sustained plasma discharges, operating at atmospheric pressure, are known to prevent instabilities and create large-volume plasmas crucial for chemical reactions. Nonetheless, they represent a relatively underexplored yet promising frontier for hydrogen generation from ethanol. In the current study, we employ a plasma discharge operated in a non-self-sustained regime for ethanol decomposition into molecular hydrogen. Ethanol in the gas phase is introduced into the discharge and the resulting decomposition products are analyzed using gas chromatography. We investigate the impact of different gas flow rates on the gas vortex inside the reactor. Moreover, we explore ethanol decomposition under various atmospheres such as argon and nitrogen. Our findings reveal complete ethanol decomposition, with hydrogen being the predominant product detected by gas chromatography, alongside other compounds such as CO, CO2, carbon black, CH4, and C2H6. Furthermore, we determine key parameters like energy yield and selectivity across a wide range of experimental conditions.

Mechanical characterisation and fibre morphology analysis in ABS and short carbon fibre composites: influence of different polymer processing techniques

Ilke Pelgrims, Ellen Fernandez, Hannelore Ohnmacht, Mariya Edeleva, Ludwig Cardon

Polymer composites have attracted considerable interest due to their potential to enhance mechanical properties of polymers for advanced applications. Yet, challenges exist, including complex mechanical behaviour and poor affinity between reinforcement material and polymer matrix. This study investigates the influence of different processing methods on the mechanical properties of acrylonitrile butadiene styrene (ABS) reinforced with 15 wt% short carbon fibres (sCF), with a main focus on the fibre morphology. By analysing injection moulded (IM), fused filament fabricated (FFF) and compression moulded (CM) parts, this research aims to elucidate the relationship between the processing method, fibre orientation and mechanical properties. Improved affinity is achieved through nitric acid surface treatment. Scanning electron and optical microscopy are used to analyse the morphology, revealing insights into void presence and fibre roughness, length, distribution, and orientation. Additionally, mechanical properties are assessed through impact, tensile and flexural tests. Significant differences in morphology among ABS/sCF composites produced by different methods are revealed. Microscopy images show that in IM parts, fibres are highly aligned in the shear layer but exhibit a more varying orientation in the core, with minimal voids. FFF parts exhibit excellent sCF alignment but show voids between different layers, while CM parts display varying fibre orientations and small air inclusions. Test bars produced via IM demonstrate superior stiffness, tensile strength, and tensile strain at break, followed by FFF (which shows a very high impact strength), while CM parts exhibit the least favourable properties, partly due to their isotropic nature and random sCF orientation. These outcomes can be immediately linked to the resulting morphologies. The findings of this work highlight the critical role of processing methods and fibre morphology in determining the composite performance.

Investigating the Influence of Rheological Parameters on Mechanical Performance of 3D-printed Lattice Structures Using Doehlert Design

Laia Farràs-Tasias, Max Vermeerbergen, Francisco A. Gilabert, Ludwig Cardon, Flavio H. Marchesini

Lattice structures are composed of repeating unit cells arranged in a three-dimensional pattern. usually exhibit higher stiffness and strength compared to conventional designs, making them ideal for lightweight, high-performance applications in the fields of aerospace, automotive, or medicine. Due to their complex geometry, they are often impossible to manufacture with traditional methods, positioning Additive Manufacturing (AM) as the preferred solution. While these structures are typically manufactured using rigid materials, flexible materials are gaining interest due to their enhanced damping properties under compression.When manufactured through processes such as Fused Deposition Modelling(FDM), the rheological behaviour of the material during printing plays a critical role in determining the mechanical properties of the final structure. Challenges such as poor adhesion, can induce anisotropy and affect the failure mechanisms of these structures. During extrusion, variations in shear rate and temperature significantly impact the material’s performance, making control over parameters like viscosity and shear rates essential.This study aims to optimize the rheological properties of printable materials to enhance the mechanical performance of lattice structures using a Doehlert design of experiments strategy.A challenging flexible lightweight lattice structure, made from Thermoplastic Polyurethane (TPU), is chosen to be mechanically improved for a morphing wing application. Flow curves of the molten material are obtained across different temperatures, and the time-temperature superposition principle is applied to obtain the exact rheological conditions used in the study. By integrating a Response Surface Methodology (RSM) with a Doehlert design, the extrusion parameters –specifically viscosity and shear rates–are fine-tuned to achieve optimal mechanical properties. This approach demonstrates that high-performance lattice structures can be efficiently produced with minimal experimentation, overcoming traditional trial-and-error methods commonly used in the field.

Improving the treatment of neurological disorders by understanding the mechanisms of ultrasound neuromodulation

Joaquín Gázquez, Wout Joseph, Emmeric Tanghe, Thomas Tarnaud

Ultrasound neuromodulation (UNMOD) is the alteration of nerve activity by delivering mechanical pressure waves to a targeted area of the brain to treat neurological disorders. Low-intensity focused UNMOD is a rapidly emerging technology due to being a remotely non-invasive technique and having a millimeter-scale precision. Despite numerous experimental settings demonstrating that ultrasonic waves can modulate the neural activity in different brain areas, the interaction between the ultrasonic source and the neuronal tissue is still not fully understood. By understanding the underlying mechanism, ultrasound neuromodulation therapy can be improved by optimizing the stimulation protocol in silico. (3R principles: Replacement). This study focuses on the intramembrane cavitation mechanism, which occurs when ultrasonic pressure waves cause gas cavities to form between the phospholipid membrane leaflets. These gas cavities cause capacitive displacement currents, leading to membrane charge accumulation. In this study, we present work in progress towards simulation predictions obtained by morphologically realistic cortical cell models with the intramembrane cavitation (IC) mechanism. This allows us to determine if this mechanism can predict experimentally observed cell-type specificity and sensitivity to the ultrasonic protocol. Examining the activation site, the importance of charge overtones, and comparison of the optimized model with experimental results will help determine how various ultrasonic parameters affect cortical neuromodulation

The Virtual Mechanical System control law: the key to manipulating vibrational energy flows

Sarah Geyskens, Jasper Juchem, Kevin Dekemele, Mia Loccufier

As industries strive for sustainability, the use of lightweight flexible materials and complex designs to save energy also introduces non-linearities and vibrations. Furthermore, striving for fewer components translates to the need for underactuated control, which means that there are degrees of freedom on which the controller cannot exert a force. To avoid invoking damage or sacrificing accuracy and speed, a revolutionary innovation of vibration control techniques is needed. For these reasons, a new active control law is developed: The Virtual Mechanical System (VMS) control law. Combining insights from the fields of nonlinear dynamics and control, energy can be efficiently transferred and vibrations in the host system are mitigated. The host to be controlled can be modelled as a branched or unbranched chain of mass-spring-damper elements with nonlinearities, allowing for applications in machines and factories (machining tools, robots…). The way energy transfer is realised is by exploiting non-linear phenomena. More specifically, Nonlinear Normal Modes (NNMs) exhibit energy redistribution properties, altering both frequency and spatial energy distribution. This transfer of vibration energy is aimed towards a virtual system or a passive vibration absorber. Combining active and passive vibration control strategies, called hybrid control, is particularly interesting for decreasing the energy needs of actuators. This project aims to create a novel vibration control framework for underactuated, nonlinear systems, and so, paves the way for more efficient and sustainable industrial practices.

Structural integrity of the WAAM-substrate interface

Robin Motte, Wim De Waele

Structural integrity of the WAAM-substrate interfaceRobin Motte, Kris Hectors, Anil Sudhakar, Wim De WaeleWire + Arc Additive Manufacturing (WAAM) is a form of directed energy deposition for additive manufacturing of metals. An electric arc is used to melt metal feedstock wire, allowing layer-by-layer material deposition. In addition to newly manufacturing workpieces, WAAM could also be used to remanufacture damaged or worn components(such as forging or casting dies or machine parts in steel production plants), contributing to a more circular economy.Dueto high cyclical heat input during WAAM, a heterogeneous microstructure arises. In particular, a strong microstructural gradient may exist at the interface between the substrate and the first deposited layer. Therefore, material properties may also vary within this region. Furthermore, the welding wire’s alloy may differ from that of the substrate,also contributing to the gradient in the interface region.This poster presents ongoing research to numerically model the effect of mechanical loads at the interface of steel components remanufactured by WAAM, for both static and cyclic (fatigue) loading conditions. Local material properties are characterised by hardness mapping, after which correlations are used to construct the plastic stress-strain curve based on a hardness value. The interface region is then modelled based on element-wise material assignment.However, the limited accuracy of this approach is revealed after comparison with experimental data from tensile tests instrumented with Digital Image Correlation. Future work will investigate the use of Profilometry-based Indentation Plastometry, a novel technique to directly characterise local stress-strain properties, to replace the hardness map as input for this model. Additionally, fatigue loading is simulated using an eXtended Finite Element Modelling (X-FEM) framework, again based on a hardness map. Modelling of Compact Tension specimens with notches at different locations relative to the interface indicates a fatigue crack path deviating towards the material with lower Young’s modulus. In the future, this model will be experimentally validated.

Wear Quantification of Wind Turbine Blades by Drone-Captured High-Resolution Images

Jonathan Sterckx, Michiel Vlaminck, Hiep Luong

Erosion of wind turbine blades, caused by the continuous impact of airborne particles such as water and dust, is a significant challenge for maintaining aerodynamic efficiency and maximizing energy output. Over time, this wear not only reduces blade lifespan but also leads to energy losses. To address this issue, we present a novel approach that combines drone-based high-resolution image capture with advanced 3D modeling techniques to predict and quantify erosion on turbine blade leading edges with a high accuracy. Our workflow involves capturing sharp, low-noise images despite the challenges posed by wind and blade motion. We employ techniques such as LiDAR-based localization, high dynamic range (HDR) imaging, and structure-from-motion (SfM) reconstruction to create precise 3D models of blade surfaces. Once we have an accurate 3D reconstruction, we perform defect detection and quantification to measure critical erosion metrics such as volume, surface area, and depth. Additionally, we address the computational challenge of processing 3D point clouds efficiently by leveraging depth maps, which represent the 3D model by a set of 2D images. This allows the fusion of visual and geometric information for improved defect detection. Compared to purely image-based detection, this fusion helps resolve visual ambiguities and allows the computation of physically meaningful defect metrics. Our method has demonstrated significant improvements in 3D reconstruction accuracy, with a reduction in reprojection error by more than 38% compared to state-of-the-art methods. The result is a denser, more reliable model of the turbine blades, which enables a more effective analysis of the surface defects. In turn, this will translate to more effective predictive and preventive maintenance strategies. Future work will focus on integrating joint deblurring and 3D reconstruction techniques, as well as incorporating prior knowledge to further refine defect quantification and prediction.

career path recommendation using large language models

Iman Johary, Alexandru Mara, Tijl De Bie

In the contemporary job market, efficient and precise resume analysis is essential for high-level tasks such as career path recommendations and job matching. This work introduces a novel approach to streamlining the process by harnessing the power of large language models. The primary objective is to develop a comprehensive pipeline for information extraction, data mapping, and subsequent utilization in downstream recommendation and matching tasks. Resumes are often parsed from different formats, resulting in unstructured and noisy text that poses a significant challenge for automated analysis. We aim to enhance the accuracy and efficacy of information extraction by leveraging recent developments in natural language processing, particularly large language models. The proposed pipeline begins with an information extraction module that efficiently retrieves key data points, including skills, qualifications, and experiences. Subsequently, a data mapping component maps different sections, such as job titles, places, and degrees, to a smaller taxonomy. The extracted information is then employed in downstream tasks, such as career path recommendations and job matching. For information extraction, we propose different methods, such as prompt engineering and LoRa. For data mapping, we experimented with fine-tuning and heuristic methods to increase accuracy. We evaluate the proposed pipeline using a rich dataset provided by VDAB, encompassing resumes in Dutch, French, and English. We further enhanced this dataset by annotating it to facilitate resume analysis and information extraction.

ABCFair: an Adaptable Benchmark approach for Comparing Fairness methods

MaryBeth Defrance, Maarten Buyl, Tijl De Bie

Numerous methods have been implemented that pursue fairness in AI with respect to sensitive features by mitigating biases in machine learning. Yet, the problem settings that each method tackles vary significantly, including the stage of intervention, the composition of sensitive features, the fairness notion, and the distribution of the output. Even in binary classification, these subtle differences make it highly complicated to benchmark fairness methods, as their performance can strongly depend on exactly how the bias mitigation problem was originally framed. Hence, we introduce ABCFair, a benchmark approach which allows adapting to the desiderata of the real-world problem setting, enabling proper comparability between methods for any use case. We apply ABCFair to a range of pre-, in-, and postprocessing methods on both large-scale, traditional datasets and on a dual label (biased and unbiased) dataset to sidestep the fairness-accuracy trade-off.

Automatic kinetic model generation: A novel modeling approach for liquid-phase processes

Robin Vercauteren, Kevin De Ras, Gust Popelier, Lander Nelis, Joris W. Thybaut, Kevin M. Van Geem

The liquid-phase oxidation of cyclohexane is an important industrial process for the production of cyclohexanol and cyclohexanone, key intermediates in the production of nylon. Despite its industrial significance, its complex free radical mechanism is not yet fully understood, posing challenges for process optimization and reactor design towards a more sustainable future. Notable advances have been made in the past decade regarding the automatic construction of kinetic models for various gas-phase processes. ALKIMO, which stands for “Automatic Liquid-phase Kinetic Modeler”, aims to fill the present gap for liquid-phase processes. This enables the automatic transformation of a kinetic model developed for the gas phase to the liquid phase, by incorporating the impact of a liquid. ALKIMO was applied to the liquid-phase oxidation of cyclohexane as a proof of concept, and the developed model will be validated against experiments performed in a tubular reactor operating in the slug flow regime. Such a regime consists of alternating cyclohexane liquid slugs and oxygen gas bubbles, resulting in diffusional limitations between the phases, and necessitating the development of an accurate slug flow reactor model. The developed slug flow reactor model uses an overall volumetric mass transfer coefficient to describe these diffusional limitations. Simulated conversions form a more gradual profile compared to a plug flow reactor model, where perfect mixing between the phases is assumed at the inlet. Subsequently, the kinetic model was extended to include formation of 6 hydroxyhexanoic acid, a significant by-product observed experimentally. The extended kinetic model, in combination with the developed slug flow reactor model, will be validated through newly acquired experimental data in the future. The model correctly predicts the expected products to be the major products in the reactor outlet, namely cyclohexane, cyclohexanol, cyclohexyl hydroperoxide, and 6-hydroxyhexanoic acid.

SemantiPy: A tool to semantically describe Python code to provide understanding and ensure provenance

Pol Nachtergaele

This research explores the transformative potential of the semantic web, emphasizing how it can help individuals regain control over their data by making the processes behind data handling more transparent. Unlike traditional data structures, the semantic web connects data meaningfully, representing real-world entities like people and places. This research introduces SemantiPy, a tool developed to automatically describe code semantically with the Resource Description Framework (RDF), enabling users to visualize how their data is processed. By leveraging ontologies such as the Function Ontology and the Provenance Ontology, SemantiPy elucidates the flow of data through cleaning pipelines. This enhances transparency regarding how data was manipulated. However, challenges remain, including enhancing accessibility for non-programmers and improving the tool's accuracy. Future enhancements aim to integrate SemantiPy with broader semantic web applications, fostering knowledge sharing and collaboration among developers while promoting transparency in machine learning processes.

Spark-Ignited Mixing-Controlled Combustion

Victor Sileghem - Sebastian Verhelst

This poster presents a new research trajectory, aiming at the efficient use of renewable fuels, such as hydrogen, methane and methanol, in heavy-duty combustion engines. These fuels have proven their advantages, both in terms of efficiency and emissions, for light-duty engines, but their use in heavy-duty engines requires novel combustion strategies. One example is Spark-Ignited Mixing-Controlled Combustion (SI-MCC), in which a fuel jet or spray is directly ignited by a spark, to establish Diesel-like combustion. Literature provides a limited proof-of-concept for SI-MCC using hydrogen and methane, but the ignitability of other fuels remains uncertain, especially for liquids. This research trajectory studies SI-MCC ignitability with hydrogen, methane and methanol and aims to understand the key phenomena influencing ignition. Experiments in an optically accessible combustion chamber will investigate the ignition success for different ignition timings and locations of the spark over the jet/spray, and provide fundamental data. The local conditions at the spark are assessed via simulations, while high-speed videos provide detailed insight into the ignition process. The conclusions will be presented as a conceptual model that shows how different fuels ignite, by indicating the phenomena of interest and describing the influence of fuel properties. These fundamental experiments are also used to design engine-experiments, which will assess the actual potential of SI-MCC as an engine concept.

Flow Boiling in Confined Spaces under Spatially and Temporally Varying Heat Flux Conditions

Jana Rogiers, Hendrik Vansompel, Michel De Paepe

Flow boiling heat transfer has great potential for heating and cooling purposes in various applications due to its high heat transfer rate and uniform surface temperature. However, challenges such as dry-out, flow instabilities, and limited understanding of the underlying physical phenomena remain. Most research has focused on uniform and constant heat flux conditions, but many real-world applications are characterized by temporally and/or spatially varying heat flux conditions. This study presents a new experimental setup designed to evaluate the impact of these non-constant and non-uniform heat flux conditions on the heat transfer to low global warming potential refrigerants in different flow regimes. The setup's calibration and validation are detailed using R1233zd(E). Experimentally obtained Nusselt numbers for single phase liquid convective heat transfer were within a 10% error margin of the Gnielinski correlation and the intended flow regimes can successfully be reached. Further validation is required to investigate the effects of parameters like flow rate, pressure, and temperature before conducting two-phase flow tests, as well as time-varying and non-uniform heat flux tests. The final goal is to develop heat transfer correlations that capture the effect of the non-constant and non-uniform heat flux conditions. Those correlations could then be used in the development of cooling systems.

Optimizing Write Performance in Decentralized Personal Data Ecosystems

Jitse De Smet, Ruben Taelman, Ruben Verborgh

In the past 2 decades, the once decentralized web has become increasingly centralized. This centralization is bad for the web and its users, there is a need to decentralize the web again. A decentralized web, allows individuals to regain control over their personal data, and allows industry to increase data quality while decreasing data management costs. Decentralization initiatives such as Solid strive to provide a personal data store to each user. These stores differ both in data stored by the user, and interface exposed. Heterogeneity is key for decentralization, but makes interacting with data stores complex for data consumers. There is a need to abstract these complexities, shielding consumers from the heterogeneity inherent to decentralization. Previous and ongoing research focusing on abstracting read queries proves to be successful. No research has been done to abstract updates to these data stores, but updates are essential to data management. This PhD project focuses on abstracting update queries over a permissioned decentralized environment, optimizing write performance. To do so, research on the requirements of interfaces, interface trade-offs, transactions, and query processing techniques is needed. Future research will be able to refine the algorithms proposed and create alternatives. In research about read-queries, adaptive query planning seems promising. Future research might try to incorporate that idea in update-queries.

Biomimicry of Reptile Eggshells Based on Keratin Nanofibrous Membranes

Yana Maudens, Gerben Debruyn, Eva Loccufier, Cinzia Tonetti, Claudia Vineis, Alessio Varesano, Liliana D’Alba, Matthew D. Shawkey, Lode Daelemans, Karen De Clerck

Eggs are multifunctional structures that enabled vertebrates to colonize the land millions of years ago. Life on Earth imposes a series of environmental challenges on eggs, often presenting conflicting or contradicting demands. For example, eggs must be crack-resistant yet allow breakage from the inside, be impermeable to bacteria yet breathable and in some cases allow water absorption. Eggshells also offer protection from harmful radiation while allowing some light transmission. As a result, they have evolved into multifunctional systems with extraordinary properties, including unique combinations of high flexibility and strength, strong water absorption and antimicrobial filtration. Biomimicking of these natural structures could lead to the development of new materials with advanced properties for filtration/separation technology, sensors and biomedical applications. Through chemical and physical analysis of sixty-two reptile eggshells (Debruyn et al., 2024), the key structural components - namely a proteinaceous layer (primarily keratin) and an inorganic calcium carbonate component - were identified and used to develop biomimicry models. The non-woven layer of proteinaceous fibers closely resembles randomly oriented nanofibrous membranes produced via solvent electrospinning. Hence, keratin, extracted through a sulphitolysis process, was electrospun from formic acid. A subsequent heat treatment ensured crosslinking of residual amine and carboxyl groups in the keratin chains, imparting hydrophobicity and improved water stability to the nanofibrous membranes. To further approximate the eggshell’s inner proteinaceous layer, the keratin nanofibers were embedded in an egg white matrix using a dip-coating procedure. The mineral component, typically found in Testudines and Crocodylia eggshells, was mimicked by depositing calcium carbonate particles onto the keratin-based membranes. This biomimetic approach paves the way for the production of multifunctional, biocompatible keratin-based membranes tailored to the specific needs of various end-applications such as filtration membranes, wound dressings, smart textiles, etc. Debruyn, G., Geltmeyer, J., Schoolaert, E., Nicolaï, M. P. J., Xie, W., Wynant, M., Shawkey, M. D., De Clerck, K., & D’Alba, L. (2024). Hydric Environment and Chemical Composition Shape Non-avian Reptile Eggshell Absorption. Integrative and Comparative Biology, 64(1), 107–119. https://doi.org/10.1093/ICB/ICAE040

Compact and Impact-resilient Robot Actuators via Lightweight Overload Clutch with Wedged Rollers

Frederik Ostyn, Guillaume Crevecoeur

Robots avoid collisions as these can lead to hardware damage. To deploy robots in less structured environments, robust hardware is required. Traditional approaches that improve impact-resilience result in bulky actuators, increased backlash or decreased torque density. A lightweight and compact clutch principle based on wedged rollers is presented as well as an experimental proof-of-concept. A use case involving the design of clutched planetary gearbox actuators tailored to humanoid robot legs shows a reduction in axial length of 25% up to 40% depending on the maximum expected overload.

Moisture-resilient Bio-based Walls: Innovation or Illusion?

Ruben Van den Bossche, Nathan Van Den Bossche, Marijke Steeman

Bio-based materials show a promising potential for the CO2 transition in the current construction industry from a linear to a circular economy. They are renewable, locally available, fast-growing, valorise rest streams of well-established industries and sequester CO2 resulting in low embodied energy. Despite an increasing use as sustainable building methods, their exact hygrothermal response in vapour-open walls is still under research. Degradation, upscaling, material variability, legislation, lack of standards, poor perception and absence in confidence lead too often to applications primarily limited to low-rise buildings and committed (self-)builders. The purpose of this research is to integrate bio-based materials in futureproof vapour-open wall construction systems. There will be focused on different building methods, e.g. monolithic walls, masonry walls with interior retrofit and classic wood framing. Thus, the research is applicable to both new and renovated buildings. The research will be done by a laboratory characterisation and in situ monitoring campaign representative for the Flemish building context, complemented by simulations. The hygrothermal analyses will be carried out for existing and new robust wall systems suitable for wider application and upscaling. Heat, Air and Moisture (HAM) simulations will be carried out to study the hygrothermal behaviour of the wall systems, investigate variants and define key parameters. The experimental work will function both as input and validation for the simulations, because bio-based materials are scarcely represented in simulation databases. The most promising wall systems will then be built and investigated by lab testing. The research methodology aims to be reproducible for future-emerging bio-based materials and lower the threshold for bio-based building systems.

SELF-INDUCED AND MUTUAL-INDUCED EMISSIONS OF GRID TIED CONVERTERS

Abdellatif M. Aboutaleb, Marwa S. Osheba, Jan Desmet, Jos Knockaert

The increased use of grid-tied converters (GTCs) leads to power conversion harmonics in low voltage networks in supraharmonics range. The effect of using grid-tied converters with switching rates that are non-integer multiples of the grid frequency is studied. It is observed that both high frequency (HF) and low frequency (LF) emissions increase more than what is expected and occur at many interharmonic frequencies, due to a combination of feedback mechanisms. When a GTC is switched with a non-integer multiple of the grid frequency, extra harmonic bands appear around the main LF and HF harmonics emitted from the GTC. The frequencies of these additional harmonics depend on the deviation of the HF emissions from the nearest integer multiple of the grid frequency. Two definitions are presented, the self-induced emissions which represent the extra high frequency emissions generated by the device itself, and the mutual-induced emissions which define the extra emissions of a device around its switching frequency and its multiples as a reflected effect of the high frequency emissions of a neighboring device. Simulations using MATLAB software are performed to clarify the results. An experimental setup is used to verify the outcomes with two case studies.

Studying the interference consequences for switching AC/DC converters subjected to a grid polluted with Supraharmonic distortion

Marwa S. Osheba, Abdellatif M. Aboutaleb, Jan Desmet, Jos Knockaert

AC/DC converters, controlled by pulse width modulation (PWM) and used as power factor correction (PFC), is considered one of the main contributors to emissions in the range 2 kHz–150 kHz, recently known as the supraharmonic (SH) range. This study looks at the impact of SH grid distortion on the LF (<2 kHz) and HF (>2 kHz) emission of an AC/DC converter. The PFC boost converter is used as a particular case for validation of the results. It is observed that the AC/DC converters emit additional LF interharmonics and subharmonics when the grid voltage contains interharmonic components in the SH range. A mathematical analysis is provided to study and assess the interference between the SH in the background distortion and the AC/DC converters. Experimental studies are then performed for a PFC boost setup based on dSPACE MicroLabBox for the purposes of validating the mathematical analysis.

A trade-off in polymer waveguides for interfacing PICs

Toon De Baere

In the datacom and telecom markets, there has been a recent increase in demand for data throughput and processing. The challenge to meet this demand by increasing compute efficiency, can only be tackled with the support of more advanced and faster packaging methods. An optical redistribution layer (ORDL), that routes optical signals between fibers, from fiber to PIC and from PIC to PIC, is needed to provide the processing units with the required data-flow. This solution eliminates the need for pluggable modules that are less flexible and allows the opto-electronic conversion to happen in the same package as the processing unit. The polymer waveguide material platform offers promising characteristics as a foundational element for building an ORDL. In this poster, we present two methods for interfacing a photonic integrated circuit (PIC) with a polymer waveguide: adiabatic coupling and edge coupling, both of which use an inverse taper to expand the mode in the PIC. We compare two commercially available polymer materials for both methods, focusing on dimensions, and coupling efficiency. The modal overlap at different cross-sections is analysed, and time and frequency domain simulations are used to validate the coupling performance. A key material property is the refractive index, as a higher index contrast between the core and cladding of the polymer waveguide results in smaller modes. This can help reduce the taper length and minimize modal mismatch at the interface. However, a trade-off occurs, as it is decreases the alignment tolerance and makes it harder to couple the smaller modes to an optical fiber.

A resource-efficient variational hybrid classical-quantum algorithm for genome sequence reconstruction

Hongfeng Zhang, Aritra Sakar, Koen De Bosschere, Koen Bertels

Reference-free DNA sequence construction is a well-known NP-hard problem that presents challenges in computational biology. Traditional methods often struggle with the complexity and computational cost on reconstructing entire genomes from smaller DNA or RNA fragments. Quantum computing offers a promising alternative for addressing this issue, leveraging quantum mechanics to outperform classical methods. This research introduces a novel quantum algorithm designed for reference-free DNA sequence construction. Through utilizing fewer qubits and quantum gates than existing methods, our algorithm enhances computational efficiency, making it feasible to conduct longer DNA sequence reconstructions on currently available quantum computers with limited resources. This algorithm is implemented on a gate-based quantum simulator, demonstrating its practical applicability. Despite the limitation of existing quantum computing technology, this work provides a compelling proof that quantum approaches can be effectively used for genome assembly tasks. The results indicate that our algorithm not only has fewer quantum resources but holds the promising of accelerating the reconstruction process. Our findings contribute valuable insights to the potential of quantum computing in solving complex biological problems and pave the way for future advancements in this field. Quantum approaches in biological research could evolve our understanding of genetic information, offering new avenues for innovation in medicine and biotechnology.

3D reconstruction of welded joints from an unordered point cloud

Jiacheng Qi, Kris Hectors, Wim De Waele

Various fatigue assessment methods have been proposed to achieve more accurate fatigue life estimations over the past decades. However, most of these methods are proposed and developed based on idealized geometries of welded joints. Even slight variations in geometry details will significantly influence the precision of fatigue life estimations. Both global and local geometries should be taken into consideration to address this challenge. In this poster, we present a framework for 3D reconstruction of welded joints from an unordered point clouds. A high-resolution laser scanner was used to obtain the point cloud from a cruciform welded joint with capturing the geometry details. The point cloud was then sliced along the welding direction to obtain a series of sub-volume point clouds. In each sub-volume, the point cloud was ordered by using KD-Tree and then fitted with a quadratic B-Spline curve. 2D cross section models along the welding direction were reconstructed as an intermediate step. Based on the reconstructed 2D models, a 3D model was finally reconstructed using iso-surface extraction method. Finite element simulations were conducted on the reconstructed 3D model. The proposed framework supports stress analysis and fatigue assessment based on the as-built geometries of welded joints. Future work will focus on evaluations of different fatigue assessment methods with the aid of this framework.

Remote Detection of Cement in Heritage Mortars Using Multispectral Imaging

Philippe Soubrier, dr. Michiel Vlaminck, prof. Hiep Luong, prof. Nathan Van Den Bossche

This study proposes an experiment to detect and quantify the ratio of lime to cement in mortar compositions on heritage facades using multispectral imaging. Cement-based mortars, often incompatible with historic masonry, pose risks such as reduced breathability, thermal rigidity, and loss of heritage integrity. Identifying the precise amount of cement in mortars remotely could enhance non-invasive inspections, especially in culturally significant buildings. The proposed experiment will create a database of mortars with lime-cement ratios in 10% intervals, using visually distinct materials: white and grey cement and white and beige natural hydraulic lime. Spectrometers will capture full reflectance spectra, while multispectral cameras simulate realistic, field-relevant detection conditions. This remote sensing approach could support drone-based assessments of building facades, revealing material composition in otherwise hard-to-access areas. By providing restauration architects with more accurate data on building materials, this study aims to support restoration efforts, allowing more sensitive and informed restoration of cultural heritage structures.

Novel Computational Memristive Bio-detection for CRP Inflammatory Biomarker and PSA Detection

Manel Bouzouita, Ioulia Tzouvadaki, Fakhreddine Zayer, Sandro Carrara, Hamdi Belgacem

During the last years, memristive computational modelling has been gaining more fame for outlining memristive dynamics and behaviour. Moreover, their capability to link biological processes to their electrical properties explains the emergent need for memristive biosensing simulation models. In the last decades, several research studies have been reported, aiming at a better understanding of the memristive bio- sensing mechanisms through equivalent circuits, as well as multiphysics dynamics allowing for advanced pre-fabrication structural and performance optimization. In our research work, we develop a new methodology for modelling memristive biosensors within COMSOL Multiphysics and we demonstrate our model for C-reactive Protein (CRP), an inflammatory biomarker, and Prostate Specific Antigen (PSA), one of the main biomarkers for Prostate Cancer, as cases of study. A series of simulations are conducted to explore the model’s multiphysic dynamics. The variation of different inputs such the electro-geometric material properties, the inlet concentration, and the association-dissociation coefficients, is employed to examine their impact on the memristive bio-sensors output resistive state (RS) and antibody-antigen complex binding. Equally important, this study frameworks the feasibility of simulated data’s extraction and mapping for artificial intelligence algorithm deployment. Specifically, the binding concentration of the complex CRP-anti CRP is proved in this work to effectively train a support vector machine classifier (SVMC) for data categorization, achieving a notable accuracy rate of 97%. Concluding, empowering computational memristive biosensing models with artificial intelligence (AI) algorithms, improves various sides of the detection method including, data quality, noise-to-signal ratio, output data reading and labelling, which guarantees an advanced, cost-effective bio-signal recognition with reduced noise, improved specificity and clear read-out. Concisely, this approach introduces a proof-of-concept for machine learning integration with simulated memristive bio-sensing paving the way for validating our model with existing experimental paradigms for label-free prostate cancer and inflammatory biomarkers continuous diagnosis.

Blade leading edge erosion damage due to droplet impact

Alireza Shadmani; Dieter Fauconnier; Wim De Waele

Wind turbines play an increasingly vital role in the global renewable energy landscape, offering a sustainable alternative to fossil fuel-based power generation. Offshore wind energy holds great promise for producing substantial electrical output due to the larger swept area resulting from longer blades. Currently, the largest wind turbines have diameters of up to 236 m, with blade tip speeds surpassing 100 m/s. These high velocities cause interactions between blade tips and raindrops or other airborne particles, leading to gradual erosion, wear, and degradation over time. The leading edge of the wind turbine is the primary area impacted by raindrops, causing wear at both surface and sub-surface levels. Initially, damage is minimal, which is known as the incubation period. Following this phase, fatigue damage accelerates during the mass-loss-rate increasing period, necessitating the investigation of fatigue damage through the coating layers. To address this, a continuum fatigue damage model is proposed for lifetime prediction of coated substrates during the incubation period. This model offers several advantages, including the ability to accurately represent the progressive accumulation of damage over time and to capture complex interactions between material properties and loading conditions. By providing a continuous representation of fatigue damage, the model allows for better prediction of material behavior, especially during the early stages of degradation. Therefore, a finite element model was developed to analyze the coating fatigue lifetime under multi-point pressure impact with respect to three raindrop sizes with 50 m/s velocity during the incubation period. This model helps predict erosion rates and identify vulnerable regions on the blade surface that are most susceptible to accelerated wear.

Advanced Lifetime Prediction for Wind Turbine Gearboxes: Integrating Non-linear Damage Models with Industry Standards

Arsalan Muhammad Ahmad, Prof. dr. Kris hectors, Prof. dr. Wim De Waele

This research presents a Python-based tool designed to improve fatigue life predictions for wind turbine drivetrain components by integrating industry-standard methods with a non-linear fatigue damage accumulation model. The tool applies the L10 method (IEC 61400, ISO 281) for bearings and ISO 6336 for gears, accounting for variable load, speed, and torque spectra. Enhanced by the Aeran damage model, the tool provides a more accurate reflection of damage interactions across stress levels, overcomming the limitations of traditional Miner’s rule.

Data-driven lifetime assessment of a wing sail system

Mekete Mebratu, Prof. dr. ing. Kris Hectors, Prof. dr. ir. Wim De Waele

Wind Assisted Propulsion Systems (WAPS) are crucial in the maritime industry's decarbonization efforts. As part of the Horizon Europe Orcelle Wind project, a new wind-powered roll-on/roll-off (ro-ro) cargo vessel is being developed with commissioning targeted for late 2026 or early 2027. This vessel incorporates a Wing System (WS) technology, utilizing aerodynamic thrust forces for propulsion. Early identification of mechanical malfunctions and component degradation is essential to ensure the WS operates safely and effectively. UGent's primary contribution to this initiative involves the development of numerical tools to support data-driven structural integrity assessments for the WS, focusing on fatigue degradation and lifetime analysis. A probabilistic fatigue reliability framework has been developed to evaluate the structural reliability of the WS under uncertain conditions. This approach accounts for load, material, and geometric parameter variability, leveraging a non-intrusive Stochastic Finite Element Model (SFEM) for structural response assessment. Uncertainties in wind loads, material properties, and geometric factors are captured through Sobol sequence-based low-discrepancy sampling, modeling nine random variables to cover a broad probability space. This study presents only the wind-induced fatigue stresses on the critical components within the wing system. The stochastic characteristics of these stresses are derived from Finite Element (FE) simulation results, based on representative combinations of uncertain parameters. Future work will incorporate additional assessments of inertial loads across varying wind angles, and surrogate models will be developed to predict structural responses using sensor data. Stress-life methodologies will subsequently be applied to estimate the fatigue life of individual components, with the overall fatigue reliability of the wing system being assessed through Bayesian Networks.

Refrigerant flow and heat transfer prediction for propane in heat pump applications

Marina Brancaccio, Steven Lecompte, Bernd Ameel (Daikin Europe), Michel De Paepe

As global warming intensifies, the heating and cooling sector is turning to heat pumps as sustainable alternatives. The phase-out of high global warming potential (GWP) refrigerants is driving the transition towards natural refrigerants, such as propane, which have significantly lower GWP. However, the high flammability of propane introduces new safety challenges, requiring system modifications to comply with stricter safety regulations. One approach to address these concerns is reducing the refrigerant charge while developing more energy and material efficient components. In this context, multiport tube heat exchangers are replacing conventional round tube designs as evaporators. These heat exchangers work with lower refrigerant charge, but they are prone to flow maldistribution, negatively impacting the system performance. To mitigate this, standard distributors alone are insufficient, but are installed in combination with headers. The design of the header is critical in improving the system efficiency, making it essential to investigate the flow behavior within the header. Furthermore, relevant experimental data are scarce in the current literature. This research aims to bridge this gap by developing experimental correlations for heat transfer and pressure drop during flow boiling of propane in multiport tube heat exchangers. Additionally, the distribution of the two-phase refrigerant within both the header and the heat exchanger will be investigated to better understand maldistribution effects. The findings will be used to calibrate and improve models for designing headers and heat exchangers, enhancing the reliability of simulations tool and reducing need for trial and error in design optimization. Specifically, this study focuses on operating conditions with mass fluxes between 5 and 50 kgm2/s, heat fluxes ranging from 0.3 and 1.3 kW/m2 and saturation temperatures between -20 and 5 °C – conditions currently unexplored in the literature.

Melting regime maps to characterize heat transfer during solid-liquid phase change

Maité Goderis, Wim Beyne, Michel De Paepe

Latent thermal energy storage (LTES) systems with phase change materials (PCMs) are a promising technology to store intermittent, renewable energy. However, the development of accurate and efficient design methods still remains a challenge, due to the transient behavior of these systems. This study investigates the melting process in rectangular enclosures heated from one side wall at constant temperature. To study the melting behavior, the melting regime theory of Jany and Bejan (1988) is applied, which states that the melting process can be divided into distinct melting regimes, each with their own characteristic heat transfer behavior. First, heat transfer occurs mainly through conduction. Once enough liquid PCM is formed, natural convection will come into play and become the dominant mode of heat transfer. Previous studies often overlook the distinction between the different regimes, potentially compromising the accuracy of their results. This project aims to address this limitation by separately characterizing the heat transfer in each melting regime and identifying the conditions under which the transitions occur. The combination of the correlations for heat transfer and the transitions is referred to as ‘melting regime map’. The poster presents an overview of our current findings based on literature data, and the remaining research gaps.

The Nonlinear Vibration Absorber for Mitigating Torsional Vibrations

Harikrishnan Venugopal, Kevin Dekemele, Mia Loccufier

From causing mild inconveniences to catastrophic failures, vibration affects all structural and mechanical components. The situation becomes especially critical when the structure is subjected to an exciting force that induces resonance. Industry trends towards lightweight construction compounds this issue by lowering the resonance frequency . The existing solution of attaching auxiliary linear vibration absorbers, is not so robust to changes in the structure, nor is it effective for mitigating other potential resonances. The Nonlinear Vibration Absorber (NVA) solves these issues by introducing nonlinear elements to the vibration absorber, offering a robust and broadly tuneable vibration control solution. Here, a piecewise-stiffening NVA is used to mitigate the torsional vibrations of a benchmark system. The piecewise-linear stiffness of the NVA mimics a cubic nonlinear stiffness, by the different angular engagement of flexural rods. The dynamics of the NVA-host system is studied analytically, by assuming the response to be of a single frequency; referred to as the first-order Complexification-Averaging scheme. The frequency response thus obtained has undesirable Isolated Resonance Curves and regions of quasi-periodic response, but shows 80%-87% reduction in the resonant response.

Fatigue analysis of floating offshore wind substructures

Ir. Victor Rappe, Prof. dr. ir. Wim De Waele, Prof. dr. Muk Chen Ong, Prof. dr. ing. Kris Hectors

As offshore wind energy advances, floating offshore wind turbines (FOWTs) provide a crucial solution for harnessing wind power in deep waters where fixed-bottom wind turbines are not viable. Traditional FOWT analyses model structural behaviour as a rigid substructure connected to a flexible or rigid tower, allowing easy extraction of tower base loads. Consequently, most fatigue research for FOWT substructures focuses on the tower base, neglecting welded joints, which are known to be fatigue-critical locations. Hence, accurate assessment of the loads, and particularly the local stresses at these joints, is vital. This poster introduces a multidimensional time-domain strategy to estimate the fatigue life of a FOWT substructure, highlighting a novel load mapping method that translates global dynamic simulation loads into a detailed finite element model (FEM). The initial step applies hydrostatic, inertial, mooring, tower base, and Morison drag loads to the FEM. Hydrodynamic forces like diffraction, radiation, and Froude-Krylov loads, however, are commonly oversimplified as concentrated forces at the substructure’s centre of gravity. For better precision, these forces should be represented as distributed pressures across the wetted surface of the substructure. To achieve this, a boundary element method (BEM) solver calculates these distributed pressures in the frequency domain, which are then converted to time-domain data using dynamic simulation inputs, i.e. wave elevation for diffraction and Froude-Krylov loads, and platform motion for radiation loads. This high-fidelity load modelling on the FEM enables the determination of accurate stresses, and, in extension, fatigue life predictions for the FOWT substructure. These predictions are crucial for optimising the structural integrity and operational lifespan of offshore wind energy systems.

A 74 Gbps SiGe BiCMOS Electro-Absorption Modulator Driver for Space Applications

Kieran De Bruyn, Arijit Karmakar, Warre Geeroms, Michael Vanhoecke, Laurens Bogaert, Günther Roelkens, Johan Bauwelinck

This poster presents a radiation-hardened-by-process 74 Gbps electro-absorption modulator driver designed in a 130 nm SiGe BiCMOS technology for application in optical intra-satellite links. The driver is designed to enable optical communications to replace the RF waveguides that are currently used to connect the antennas of a satellite to its interior core. By substituting these waveguides with optical fiber, the size and weight of the connecting networks is significantly decreased, all while allowing for a more flexible and scalable network. However, the outer space environment differs greatly from the terrestrial environment these optical communications devices are typically designed for, mainly because of the presence of both ionizing radiation and high-energy particles. To combat this, several design considerations are taken, and experiments are performed to assess the vulnerability of the driver against radiation exposure in the space environment. The tested samples show little change when irradiated with X-rays up to a total accumulated dose of 1.2 Mrad(Si). During the heavy-ion test, no latchup was observed and transients introduced bit errors at a rate of only 6.4 x10^-13. Both of these experiments demonstrate the radiation hardness of the tested driver, and show that it is indeed suitable to successfully replace the heavy metal waveguides in satellites.

Converting laser scans of tubular joints to finite element models

Ir. Jelle Plets, Prof. dr. ing. Kris Hectors, Prof. dr. ir. Wim De Waele

Jacket type substructures for offshore wind turbines gained prominence in the 2010s, especially in wind farms where monopiles became less feasible due to increasing water depths. As these turbines near the end of their design life, extending their operational lifetime or repowering offers significant economic and ecological advantages compared to decommissioning. However, these strategies both require an accurate assessment of the remaining fatigue life of the substructure. Previous studies showed that the actual weld geometry strongly affects local stress concentrations, influencing the fatigue life of offshore substructures. Current fatigue assessments in the design phase assume perfectly cylindrical tubular members with simplified welds, while also omitting surface degradation effects. This research aims to improve the fatigue assessment of offshore jackets by incorporating the as-built joint geometry, including ovality, out-of-roundness, surface irregularities, and the real weld geometry, to better estimate stress concentrations and fatigue damage accumulation. Four full-scale tubular joints were scanned. The joints originate from a decommissioned gas platform jacket that was installed in the North Sea in the 1980s and was decommissioned in 2023. A handheld 3D laser scanner was used to capture the real geometry of the joints. Since the scanning principle is based on the reflection of light, only the outer surface can be scanned. In this work, a framework was developed to automatically extend the tubular members and reconstruct the inner parts that could not be scanned.

Hollow square core fibre sensor: simultaneous measurement of pressure, temperature, and curvature

João P. Santos*, Diana Pereira, Jörg Bierlich, Micael Nascimento, Marta S. Ferreira, Jeroen Missinne

A multiparameter fibre sensor for pressure, temperature, and curvature detection is proposed. The sensor is constituted by a single section of a hollow square core fibre (HSCF) spliced between a single mode fibre (SMF) and a long section of a silica capillary tube (SCT). The splicing of different fibres creates parallel interfaces that can give rise to interferometers that propagate in the fibre sections. In a reflection scheme, several Fabry-Perot (FP) cavities are enhanced in different areas of the HSCF. In a single 439 µm long sensing head, three FP cavities are excited. Using the Fourier band-pass filter method, each cavity is individually monitored towards the measurand variation. The sensor revealed a maximum pressure sensitivity of (3.23 ± 0.04) nm/MPa for FP1 within a pressure variation of 0.4 MPa. As for the temperature response, FP3 attained the highest sensitivity of (9.6 ± 0.3) pm/°C up to 110°C. On the other hand, the sensor revealed a maximum curvature sensitivity of (22 ± 1) pm/m-1 for FP3, in a range of 9 m-1. The distinct responses of the FP cavities allow for a hybrid application for simultaneous measurement of pressure, temperature, and curvature. Additionally, a sinusoidal behaviour is found in the curvature response when bending is applied for different relative positions of the fibre. The proposed sensor is robust with simple fabrication and small dimensions, demonstrating promising potential for applications where the simultaneous measurement of several physical parameters is required.

Optimization method for drivetrain settings in a calorimeter

Adham M. K. Albanna, Daan P. K. Truijen, Korneel De Viaene, Kurt Stockman, Jeroen D. M. De Kooning

The optimization of both machine and process efficiency has surged in importance, as it is a key part of the energy transition. This poster addresses the need for enhanced energy efficiency in electromechanical systems by identifying an optimal switching frequency that minimizes total power losses while adhering to the precision constraints of calorimetry. The research aims to determine the optimal switching frequency for a drive system that minimizes power losses in a permanent magnet synchronous motor (PMSM) and drive combination. Measurements are conducted using a precision calorimeter, providing highly accurate power loss data. Obtaining precise measurements using a calorimeter is time-intensive, as each reading requires the system to reach a steady state. To speed up the process, an initial quick sweep of switching frequencies is performed, spanning from low to high frequencies and back, repeated six times to assess power loss variance. The collected data are then used for polynomial fitting, creating a model of power loss as a function of switching frequency. This model supports a specially developed gradient descent algorithm, which tests a reduced number of points and iteratively adjusts the switching frequency by tuning specific parameters to converge on the frequency that minimizes power loss. The experimental results demonstrate the effectiveness of the optimized switching frequency selection. The developed algorithm identifies a switching frequency of 4211 Hz as the optimal frequency, with a minimized total power loss of 1132 W across five tested points. Additionally, experimental tests were conducted with eight points to confirm that the minimum power loss frequency was identified. These tests show that the switching frequency yielding the lowest observed power loss is 3341 Hz, resulting in a total power loss of 1035 W.

Investigating the stability of plant-based emulsions: Towards high quality 3D-printed meat analogues

Elise Caron, Alexandra Alicke, Davy Van de Walle, Koen Dewettinck, Flavio Marchesini

Over the last decade, research on 3D food printing (3DFP) as novel processing technology for meat analogues has emerged. This combines bioscience and chemical engineering to understand every aspect of the process. Protein-rich and lipid-rich inks are applied, where the animal fat may be mimicked by an emulsion, i.e. a mixture of two immiscible liquids. Despite the plant-based trend, stabilizing multiphasic food products with plant-based proteins as stabilizers instead of conventional animal proteins still faces some challenges. The main reason is that plant proteins perform differently at the interface as they exhibit lower aqueous solubility and thus form weaker viscoelastic films. Therefore, the formulation and long-term stability of plant-based emulsions need further investigation. Preferably, one can find a way to predict the long-term emulsion stability. This work aims at (i) understanding the stabilization of oil-in-water (O/W) emulsions with plant-based proteins, (ii) extracting interfacial and bulk rheological data to enhance the rheological comprehension of O/W emulsions, and (iii) relating the stability to the used methods. O/W emulsions with oil/water content of 80/20 were formulated. The oil phase consists of soybean oil, whereas the water phase concerns a soy protein isolate (SPI) dispersion. Two different treatments were performed on the SPI dispersions. Whereas interfacial tension, interfacial rheology, and bulk rheology were investigated. Only the interfacial rheological results could explain the difference in emulsion stability after 30 days. In summary, this work provides a relation between the interfacial properties and long-term stability of plant-based emulsions. The reported findings acknowledge the importance of rheology and can pave the way to include novel lipid-rich inks in 3D-printed plant-based meat.

Investigating the effect of template head models on Event-Related Potential source localization

Emma Depuydt, Yana Criel, Miet De Letter, Pieter van Mierlo

Introduction: Mapping brain activity from EEG signals helps researchers study how different brain regions respond during specific tasks. This mapping accuracy depends on the head model used to guide where the signals are coming from within the brain. Two popular types of models—Boundary Element Models (BEM) and Finite Element Models (FEM)—differ in complexity, with FEM providing more anatomical detail. Additionally, head models can either be individualized using each person’s MRI or standardized using a template model. This study explores how these modeling choices affect the accuracy of EEG brain mapping. Methods: BEM and FEM head models were compared using both personalized and template-based setups. The models were first tested using simulated EEG data, after which they were applied to real EEG data collected during a face recognition task. Each model’s performance was analyzed by measuring how accurately each model localized the sources of brain activity, looking at factors like precision and error rates. Results: FEM models, especially those based on individual MRIs, were most accurate in identifying brain areas responsible for EEG signals. Template models showed reduced accuracy, with BEM in particular producing broader and sometimes incorrect mappings. While template models are useful without individual MRIs, they carry a higher risk of errors in pinpointing exact brain regions. Discussion: These results suggest that while template models provide a practical alternative when individual MRIs are not available, subject-specific FEM models deliver the most reliable results for studies requiring precise localization. BEM models, though easier to compute, showed limitations in more complex brain regions. Conclusion: For EEG brain mapping, personalized FEM models offer the highest accuracy, making them ideal for studies requiring detailed brain region identification. Template-based BEM models, while convenient, should be used cautiously due to their lower precision, with implications for research in cognitive and clinical neuroscience.

Performance of analytical models predicting upstream-to-downstream water depth ratios in asymmetric open-channel confluences with width discordance

Saeed Hashemikia, Greet Deruyter, Tom De Mulder

Most analytical models to predict the upstream-to-downstream water depth ratios for asymmetric open-channel confluences only apply when the downstream-to-upstream width ratio is one and assume equal water depth ratios for the lateral branch and the main channel. Both limitations are overcome by the model of Shabayek or the Luo variant thereof. The performance of both versions is assessed against a recently published, large dataset produced by a 3D numerical model for geometries with width discordance. The performance of both model versions is quite similar, with an average (resp. maximum) deviation of the water depth ratios of ±1% (resp. ±3%) against the numerical data.

Development of a general supercritical heat transfer correlation based on measurements on low GWP refrigerants

Jera Van Nieuwenhuyse, Stijn Van Isterdael, Jana Rogiers, Steven Lecompte, Michel De Paepe

In order to design the heat exchangers of supercritical systems (e.g. vapor generators in transcritical ORCs), it is important to accurately predict the heat transfer. For this purpose, heat transfer correlations are developed. They should be specific enough to handle the exact geometry and fluid properties. Yet, even correlations that have been developed for horizontal flow of specific supercritical refrigerants, lack accuracy when applied to other refrigerants. This is an issue because at the moment most correlations are developed for refrigerants with a high Global Warming Potential (GWP). In addition, directives when to apply which correlation are also missing. In this work, the methodology of creating a generally applicable heat transfer correlation for supercritical refrigerants flowing horizontally under heating conditions is elaborated. First, the in-house built experimental test rig to perform heat transfer measurements of supercritical refrigerants is presented. Then, based on data available in literature, existing correlations are evaluated. Especially at the top of the tube, the existing correlations from literature have a low prediction capability. Next, the prediction capability of several correction factors commonly applied to heat transfer correlations for supercritical flows is investigated. Finally, a new, more generally applicable, heat transfer correlation based on this analysis is proposed. Both for the top and the bottom of the tube, incorporating a density-based correction factor to take into account the radial property variations improves the prediction capability. The prediction capability at the top can be further enhanced by incorporating an additional correction term accounting for buoyancy effects. For the bottom of the tube, 91% of the data can be predicted within a relative error below 30%. For the top, this is 90%.

Seizure Propagation Maps in Patients with Refractory Epilepsy

Zoé Bolluijt, Emma Depuydt, Serge Vulliémoz, Pieter van Mierlo

Epilepsy impacts approximately 50 million people worldwide, yet pharmacological treatment fails to achieve seizure freedom in one-third of patients, resulting in drug-resistant epilepsy. The most effective treatment to render refractory epilepsy patients seizure-free is surgical resection of the seizure onset zone. Nevertheless, only around 55% of patients become seizure-free after epilepsy surgery. Eligibility for surgery is assessed through various tests, including long-term electroencephalography (EEG), to localize the seizure onset zone and track the spread of seizure activity. Using Dynamical Imaging of Coherent Sources (DICS), a technique that traces the origin of oscillatory brain activity based on the seizure EEG data, the seizure propagation over time can be mapped in the brain. These propagation maps provide insights into patient-specific seizure dynamics, highlighting how seizure spread may vary across subjects, seizures, and frequency bands. In this study, we developed the pipeline to map the seizure propagation and tested it using multiple seizures. In the future, we want to validate these propagation maps based on intracranial records and seizure semiology. Furthermore, we want to identify biomarkers for predicting post-surgical seizure freedom based on these propagation patterns.

POTR: Post-Training 3DGS Compression

Bert Ramlot, Peter Lambert, Glenn Van Wallendael​

Creating 3D scenes and viewing them in real time are essential technologies in computer graphics and virtual reality. Recently, 3D Gaussian Splatting (3DGS) has emerged as the most promising contender to Neural Radiance Fields (NeRF) in 3D scene reconstruction and real-time novel view synthesis. 3DGS outperforms NeRF in training and inference speed, but falls short in storage requirements, with typical unbounded scenes requiring 250 MB to 1.5 GB of space. To remedy this downside, we present POTR, a post-training codec for 3DGS scenes that uses two novel compression techniques to achieve compression ratios around 100x while minorly affecting visual acuity. First, a modified 3DGS rasterizer accurately and efficiently calculates a splat's contribution and the change in PSNR upon its removal. Both metrics are subsequently used to remove over 80% of splats, while introducing only minor visual artifacts and without altering any other attributes. Splat removal also significantly boosts inference speeds, with scenes more than doubling in framerate. Second, a novel spherical harmonics energy compaction method is employed to lower the AC lighting coefficients' entropy substantially. Using a heavily modified version of ridge regression, up to 95% of AC lighting coefficients are set to zero while simultaneously reducing their L2 norm. Additionally, this energy compaction method can be used for a wide range of post-processing spherical harmonics operations, such as increasing or decreasing the degree of spherical harmonics after training or removing hallucinated colors.

Models and methodology for the design and control of residential district heating systems with thermal energy storage

Alixe Degelin, Ilya T'Jollyn, Michel De Paepe

District heating networks for building applications will need integrated thermal energy storage (TES) systems to be able to include more renewable energy into their operation. Thermal energy storage models currently require too much calculation time, to use them for district heating network topology optimization. They show the same deficiencies when control during operation is considered. This research aims at developing fast and reliable models for both sensible and latent thermal energy storage systems and integrating them into district heating models for topology optimization and development of control strategies.

Synthesis of Polystyrene Nanoparticles via a Gas Aggregation Cluster Source

Jasna-Tinea Jelinek, Maryam Nilkar, Zdeněk Krtouš, Ondřej Kylián, Rino Morent, Nathalie De Geyter

In this study, we present for the first time, a solvent-free synthesis of polystyrene (PS) nanoparticles via radiofrequency (RF) magnetron-based gas aggregation cluster source (GAS). A PS target, 81 mm in diameter and 4 mm thickness, is bombarded by high-energy plasma species, mainly ions generated by the magnetron, leading to the ejection of atoms, molecules, or molecular fragments from the target. The ejected species then travel and condense onto the surrounding surfaces. The PNPs were synthesized under constant pressure of 164 Pa and a constant flow of 40 sccm of argon for 30 minutes. The effect of power was investigated on the morphology and chemical characteristics of the synthesized PNPs. The PNPs synthesized at lower power (40 W) exhibited spherical morphology with a diameter of approximately 100 nm, while higher powers (60 and 80 W) led to a cauliflower-like morphology marginally larger than the 40 W particles. For further comprehensive analysis of the formed PNPs advanced techniques such as X-ray photoelectron spectroscopy (XPS) and Fourier-transform infrared (FT-IR) were employed to provide insight into the elemental composition and surface functional groups. Due to particles appearing transparent in the scanning-electron microscopy (SEM) images, ultraviolet-visible spectroscopy was utilized to elucidate optical properties and the potential application of the synthesized PNPs in optoelectronic devices.

Tribological investigation of potential green alternatives to PFAS-Teflon (PTFE)

Á. Kalácska, R. Vergieu, P. De Baets

PFAS (per- and polyfluoroalkyl substances) are a group of man-made chemicals that have been widely used in various industrial and consumer products since the 1940s. These compounds serve various purposes, however, PFAS face criticism due to health and environmental concerns. The problem with PFAS lies in their persistence, bioaccumulation, potential health effects, and widespread environmental contamination, highlighting the need for comprehensive strategies to mitigate their impact on human health and the environment. Polytetrafluoroethylene (PTFE) is a PFAS thermoplastic fluoropolymer, widely used in tribological applications (e.g. bearings, coatings) in the industry and common household products as well, due to its favourable tribological properties (low coefficient of friction, chemically inert, self-lubricating nature and high thermal stability). When seeking substitutes, it is essential to explore the relative wear resistance compared to PTFE and meet specific conditions encountered in its application (varying temperatures, chemical resistance, impact resistance, machinability/manufacturability, and acceptable friction behaviour). Currently, no direct substitute exists within the range of technical materials. The main objective is to gather tribological information (friction and wear performance) about potential PTFE replacement eco-polymers to be used in sliding bearings. The work aims to investigate the abrasive wear behaviour, including wear micro-mechanisms, surface damage, wear debris generation and micro-geometry changes along with the friction in contact. The pin-on-disc laboratory tests representative of the sliding bearing application will be conducted under various pressure-velocity conditions, providing a basis for comparison. By incorporating a better understanding of the wear/lifetime and surface degradation processes of green alternatives, the application of PTFE and its composites can be drastically reduced. The information to be obtained is crucial for sustainable development and could aid material developers and product manufacturers in further developing their products to comply with the more strict rules about environmental and human health protection.

Leveraging human strategy in Bayesian optimization for industrial commissioning

András Retzler, Tom Levebvre, Guillaume Crevecoeur

Commissioning industrial processes is often a tedious and difficult task, because certain settings, such as parameters of a controller, need to be adjusted repeatedly until a certain goal is reached. It is also difficult to pass on the information on how to tune a system from experienced operators to junior colleagues, because such knowledge is often tacit and hard to formalize. We approach the tuning problem with Bayesian optimization (BO). We discuss different ways of combining the best of both worlds, i.e., the insights of the expert with the goal-driven exploration of BO. With the goal of using recorded tuning sessions of experts in BO, we dive into two main directions. First, we explore the effect of context changes on a Human First-Computer Last approach, where the expert does the initial rough tuning, and BO does the fine-tuning afterward. Second, we lay out initial steps toward a subgoal-based BO method, based on the observation that humans often break up complex tasks into multiple subgoals to achieve in a given sequence. We evaluate these methods on two simulation setups featuring a cascade controller of a pusher-slider robotic system, and a projectile trajectory planning problem.

The design and evaluation of a personal data store application

Tim Theys

Although personal data stores (PDS) offer users greater transparency and control over personal data, adoption remains limited. This study explores user perceptions of PDS to uncover barriers to adoption, evaluating four use cases: HR: PDS helps career platforms match job seekers with positions by sharing qualifications and interests, allowing companies to contact suitable candidates. Media: PDS enables sharing of media preferences across platforms, enhancing personalized recommendations. Finance: PDS consolidates financial data, offering insights into income, expenses, and assets. Health: PDS stores medical, fitness and wearable data which can be shared with a doctor for more personalized health monitoring. A survey (n=2500) in which participant rated the four use cases based on a textual description revealed a significantly higher behavioral intention for the Health use case, with intent decreasing across use cases by participant age. Following the survey, a mock-up prototype was designed for each use case and reviewed by experts (n=9), leading a second iteration of the prototype. This refined prototype is now undergoing usability testing with potential end users (n=15) to assess its effectiveness and user perceptions.

Expanding Photonic Integrated Circuits (PIC) into the Ultraviolet-C Region: Design and Fabrication of Low Loss SiOx Waveguide

Chenming Su, Chupao Lin, Roel Baets, Nicolas Le Thomas

Photonic integrated circuits (PICs) have enabled a large number of applications by using light from the visible to mid-infrared spectrum, but not yet in the ultraviolet-C (UVC, λ = 200 – 280 nm) spectral region. Considering that most of the biomolecules have strong absorption in the UVC, PICs operating in this wavelength range are envisioned to spark new biosensing and biomedical applications. However, the development of PICs in the UVC region is hindered by significant scattering and absorption losses, and limited choices of waveguide materials. Addressing these hurdles, we propose a suspended waveguide design with air-cladding on a silicon substrate. The core of the waveguide is made of thermal silicon oxide. Simulations and optimizations were conducted to identify a single-mode regime at λ = 266 nm. The waveguides were fabricated with a two-step electron beam lithography patterning process. As a result, we achieved propagation losses of 5 dB/cm for single-mode waveguides and 2.4 dB/cm for multi-mode waveguides at λ = 266 nm. Furthermore, we delved into the limiting factors of propagation losses in the UVC region. This work demonstrates on-chip low-loss waveguides in the UVC range for the first time and paves the way for on-chip UVC resonance Raman spectroscopy.

Radio-frequency electromagnetic exposure of Aedes aegypti mosquito in early life stages

Eline De Borre, Charles De Massia, Matthieu N. Boone, Pie Müller, Arno Thielens

Radio-frequency electromagnetic fields are emitted by our telecommunication systems into the environment, where they expose living organisms, including insects. The effects of electromagnetic fields on insects are not well known, while many insect are often a crucial part of our ecosystems. Some insects are vectors for pathogens, like the Aedes aegypti mosquito, spreading Dengue, Yellow fever and Zika and the control of this insect is important in many countries. The interaction of electromagnetic fields with insects is different than for humans and other invertebrates, due to their tissue and their size. For electromagnetic fields with wavelengths comparable to the insect size, resonance effects could cause a higher power absorption, possibly leading to dielectric heating. In this study, we use the Aedes aegypti mosquito, as a model organism to investigate the effect of radio-frequency electromagnetic fields on the development of insects. Early life stages of Ae. aegypti were exposed during experiments and wing length and development time were analyzed. Before the exposure experiments were executed, numerical simulations revealed the dose received by the insect during the experiments, considering the exposure setup and the fact that both the larvae and pupae are aquatic. Also the position of the insect and frequency dependency of the dose were investigated. The simulations used new 3D models of a Aedes aegypti larva and pupa generated from micro CT-scans.

Plasma-assisted fabrication of antimicrobial coatings on titanium implants

Asif Ali, Maryam Nilkar, Louis Boissau, Anton Nikiforov, Rino Morent, Kim Verbeken, and Nathalie De Geyter

Titanium (Ti) and its alloys are the most commonly used materials for the production of metallic medical implants due to their promising characteristics, including bioinertness, corrosion resistance, and biocompatibility. Nevertheless, these materials exhibit inadequate osseointegration capabilities, leading to implant failure. Additionally, Ti lacks inherent antibacterial properties, and bacterial-induced inflammation is another major contributor to implant loosening. To address these challenges, we successfully deposited zinc (Zn) doped TiO2 coatings onto commercially pure Ti (grade 2) using the unipolar pulsed plasma electrolytic oxidation (PEO) method under diverse experimental conditions i.e. Zn concentration in the range of 10 – 100 mM and treatment time of 3 and 5 min. Subsequently, the coatings were characterized microscopically, spectroscopically, and mechanically. Spectroscopic and microscopic analyses revealed that the deposition parameters have a significant influence on the coating morphology and degree of crystallinity. Scanning electron microscopy (SEM) illustrated that the increased Zn concentration likely promotes the formation of a more textured morphology and molten-like deposition with increased pore density and roughness, while increased coating time leads to larger and interconnected pores with microcracks at higher concentration. X-ray diffraction (XRD) spectra confirmed that elevating the Zn concentration and extending the treatment time yielded dense, and crystalline TiO2 coatings. Furthermore, mechanical analysis revealed a substantial enhancement in surface hardness (approximately 100% compared to the bare substrate) for the coatings deposited at a higher concentration (50 mM) and subjected to a longer treatment time (5 min). Consequently, plasma-assisted fabrication of doped bioactive coatings emerges as a promising approach for enhancing implant performance.

From Text to Treatment Effects: A Meta-Learning Approach to Handling Text-Based Confounding

Henri Arno, Paloma Rabaey, Thomas Demeester

One of the central goals of causal machine learning is the accurate estimation of heterogeneous treatment effects from observational data. In recent years, meta-learning has emerged as a flexible, model-agnostic paradigm for estimating conditional average treatment effects (CATE) using any supervised model. This work examines the performance of meta-learners when the confounding variables are expressed in text. Through synthetic data experiments, we show that learners using pre-trained text representations of confounders, in addition to tabular background variables, achieve improved CATE estimates compared to those relying solely on the tabular variables, particularly when sufficient data is available. However, due to the entangled nature of the text embeddings, these models do not fully match the performance of meta-learners with perfect confounder knowledge. These findings highlight both the potential and the limitations of pre-trained text representations for causal inference and open up interesting avenues for future research.

Holistic assessment of optimal (collective) renovation scenarios in existing neighbourhoods

Hannelore Scheipers, Eline Himpe, Arnold Janssens

The renovation rate of the building stock must increase to reach the 2050 climate goals. One of the methods to reach this goal, is to shift the focus to district renovations. Therefore the Horizon Europe Mission Project Neutralpath has been initiated. This project investigates how PCEDs (Positive and Clean Energy Districts) can be achieved in 5 European cities, in an inclusive manner. PCEDs are districts that produce more energy than they consume, and only use ‘clean’, i.e. renewable energy. The design of these PCEDs is done through participative and human-centre principles. This research is part of the Neutralpath project and investigates the most optimal renovation scenarios in existing districts in a holistic manner, with the Muide Meulestede district in Ghent as case study. Energy use, the environmental impact over the full life cycle and costs will be considered. To make sure the results are accurate, dynamic simulations will be performed and the energy performance gap, future climate data, a dynamic electricity mix and current costs from the building sector will be used. District-specific parameters, such as the maximum available renovation budget of the inhabitants, are also included. At first, different building envelope renovation scenarios will be evaluated. As the renovation budgets might be limited, deep renovations will not always be feasible. Therefore, alternative renovation solutions, with a focus on sufficiency, will be explored. For these different renovation scenarios, collective and individual heating/cooling and domestic hot water solutions are investigated and compared to each other. Thus, the outcome of this research will be optimal renovation scenarios, taking into account the specific boundary conditions of the district. There will also be a trade-off between collective and individual heating and cooling solutions, indicating which is the best solution. The results of this study will be scalable to other similar districts.

Electrochemical CO2 Reduction to CO for In-Situ Resource Utilisation on Mars

Paulina Govea-Alvarez; Jason Song; Zhiyuan Chen; Kevin M. Van Geem; Yi Ouyang

The development of sustainable technologies for in-situ resource utilisation (ISRU) on Mars is critical for future human missions. This study investigates the electrochemical reduction of carbon dioxide (CO2) to carbon monoxide (CO) in laboratory conditions, focusing on the performance of silver-based gas diffusion electrodes (Ag GDE) in various catholytes. We evaluate key metrics such as pH and different salt concentrations on the catholytes to evaluate which current density obtains the highest faradaic efficiency (FE%) towards CO; this means the optimum condition at which CO is preferred over the hydrogen evolution reaction. This is the first step towards developing an ISRU process for Mars utilisation since CO2 is the primary component of the Mars atmosphere, and converting it to CO can benefit other chemical processes that generate fuel. The results indicate that selected sulphate catholytes, especially potassium, significantly increase CO yield while minimising hydrogen formation at specific lower current densities. Moreover, a more acidic media results in more hydrogen generation instead of CO, while a lower but still acidic media (such as pH 3) promotes the reduction of CO2 into CO using an Ag GDE. This research encompasses the use of salts as electrolytes and acidic media to enhance the reduction of gaseous CO2 towards CO by using a silver GDE in a microfluidic cell electrolyser.

Deep Reinforcement Learning For Inventory Control

Naim Al Khoury, Dieter Claeys, Zaharah Bukhsh

This research investigates the potential of Deep Reinforcement Learning (DRL) to optimize spare parts inventory control by incorporating real-time degradation data. While DRL has shown promise in various inventory control applications, its application to spare parts control leveraging degradation data remains relatively unexplored. We propose a framework to develop intelligent spare parts ordering policies for systems with multiple machines and stochastic lead times. Three state-of-the-art DRL algorithms – DQN, PPO, and DCL – are employed to learn optimal ordering strategies directly from the system dynamics and degradation data. To evaluate the performance of the DRL-based policies, we compare them against the Proactive Base Stock Policy (ProBSP), a high performing spare parts policy that leverages degradation data. Our numerical experiments analyze the impact of various problem parameters, such as degradation rates and lead time variability, on the performance of the DRL and ProBSP policies. The results demonstrate that DCL consistently outperforms DQN and PPO, and in certain scenarios, surpasses ProBSP. This research contributes to the advancement of data-driven decision-making in spare parts inventory control and provides valuable insights for practitioners in industries with critical equipment.

One-Shot Team Recognition and 3D Pose Estimation of Cyclists for Augmented Reality Visualization

Winter Clinckemaillie, Jelle Vanhaeverbeke, Steven Verstockt, Maarten Slembrouck

Applying advanced computer vision and machine learning tech- nologies transforms how we experience sports events. This research focuses on enhancing the viewing experience of cycling races by identifying and classifying teams from helicopter footage, address- ing challenges posed by fast movements and often similar team uniforms. State-of-the-art object detection and one-shot team recog- nition via siamese neural networks are implemented to provide effi- cient team classification with minimal labeling. A range of advanced computer vision models has been tested for their effectiveness in accurately recognizing teams, with the siamese neural networks achieving a classification accuracy of 95%. Furthermore, tempo- ral tracking and post-processing techniques have been applied to strengthen classification performance. These methods improve the quality of metadata during broadcasts by adding detailed team clas- sification, team visibility, and positioning, thereby facilitating a more informative viewing experience. The developed software also facilitates post-race analyses with visualizations that offer insights into team performances. The research further explores using aug- mented reality (AR) and 3D pose estimation to enhance the visual presentation of live broadcasts. This involves integrating real-time data such as the riders’ names or speeds, enriching the broadcast’s informational value. The combination of augmented reality and advanced computer vision opens up new possibilities for enhancing live sports broadcasts.

Is exposure to 4G and 5G signals from your mobile phone higher in Ghent compared to Deinze?

Bram Stroobandt, Han Van Bladel, Adriana Fernandes Veludo, Kenneth Deprez, Sam Aerts, Leen Verloock, György Thuróczy, Piotr Politanski, Kinga, Polanska, Gabriella Tognola, Marta Parazzini, Joe Wiart, Mònica Guxens, Martin Röösli, and Wout Joseph

Microenvironmental studies aim to characterize typical radio-frequency electromagnetic field (RF-EMF) exposure in various microenvironments, such as schools or business areas. They enable a personal assessment of RF-EMF exposure. With the deployment of 5G technology, mobile phone usage has become a significant factor in personal RF-EMF exposure due to the dependency on individual mobile phone usage. This study investigates typical transmit power levels in Ghent and Deinze, Belgium, within the broader context of a large-scale European assessment of RF-EMF exposure from mobile phone usage. Using the network monitoring application QualiPoc, measurements were conducted along predefined routes in various microenvironments across different European countries, while performing a scenario of maximum uplink usage, i.e., repeatedly uploading a large file. Average transmit powers of 4G and 5G were higher in Deinze compared to Ghent by 0.5 dB in median value and 5.6 dB in minimal value, aligning with broader European trends where rural areas exhibit 0.6 to 2.1 dB higher transmit powers than urban city centres. This suggests that the density of the network infrastructure, which is generally higher in urban than rural areas, is an important factor influencing exposure levels. This study provides essential measurement data for policymakers and epidemiologists to enhance the accuracy of personal RF-EMF exposure assessment. Future work will evaluate the influence of the 5G evolution on exposure across Europe.

Bridging Real and Virtual Worlds: A study on 21st century skills in VR

Dennis Osei Tutu

This research introduces a novel framework to evaluate communication and collaboration skills within virtual reality (VR)—skills recognized as essential 21st-century competencies across disciplines. Traditional assessment methods often rely on subjective observations and lack the precision to capture non-verbal nuances effectively. In contrast, VR offers a controlled, immersive environment, enabling a structured and consistent approach to study these complex behaviours. Our framework leverages VR’s advanced tracking capabilities—such as eye-tracking, facial expression analysis, and gesture recognition—to capture real-time interactions. We designed a collaborative task where two individuals work together to complete a 3D puzzle as quickly as possible. This setup generates sensor-based data, capturing markers of eye contact, joint attention, one-way gaze, and specific gestures (e.g., pointing, thumbs up/down, and stop/wait). These markers allow us to observe the intricate dynamics of communication and collaboration in a measurable way. However, interpreting these markers is challenging. Our initial findings reveal no straightforward correlation between task completion speed and these interaction markers, highlighting the complexity of creating meaning from raw data alone. Without established baselines or guidelines to define “positive” versus “non-positive” behaviours, it becomes difficult to assess skill proficiency accurately. Current literature lacks detailed standards for these competencies within VR, underscoring the need to develop a dedicated rubric to guide further analysis and establish meaningful benchmarks. This study embraces the subjectivity inherent in communication skills, using it to inform our objective analysis. Our findings underscore VR’s potential as a powerful tool for training and assessing interpersonal skills, with promising applications in educational and professional development contexts.

Evaluating Energy Calculation Methods for Energy-conscious Occupants in Three Belgian Heritage Townhouses

Luca Maton, Klaas De Jonge, Eline Himpe, Arnold Janssens

Achieving the EU's carbon-neutral target by 2050 requires a significant reduction in energy use within the building sector, which accounts for 40% of the total energy use. European historical cities, particularly heritage townhouses, pose a unique challenge due to their architectural and cultural value. The EU-Horizon project HeriTACE is addressing this challenge through a holistic renovation approach. This study investigates the impact of energy-conscious occupant behaviour on primary energy use for space heating in heritage townhouses in Ghent, Belgium. Three realities are compared: energy performance calculations, dynamic simulations, and actual energy use measurements. The energy performance calculations, based on the Belgian EPB software, tend to overestimate energy use by assuming a continuous heating system and a constant indoor temperature of 18°C. Dynamic simulations using Modelica incorporate actual occupancy behaviours and flexible heating patterns, leading to significantly lower energy use estimates. This actual heating patterns are derived from indoor temperature measurements in the case study buildings. Real energy use data, derived from energy bills, reveals that actual consumption can be up to seven times lower than EPB calculations. The findings suggest that dynamic simulations, which account for realistic occupant behaviours and lower comfort expectations, are more accurate in predicting energy use and are essential for developing effective renovation strategies for heritage townhouses.

Spatial distribution of wind driving rain and drying on façades and associated hygrothermal response: An external coupled CFD and HAM model.

Vanderschelden Bruno, Kubilay Aytac, Cnudde Veerle, De Kock Tim, Van Den Bossche Nathan.

Heat, Air, and Moisture (HAM) transfer models have proven their value in the renovation and restoration sector. They allow for analyzing the causes of moisture accumulation and related issues, as well as studying the impact of renovation measures. Wind-driven rain (WDR) is one of the most important moisture sources, while evaporation serves as the main drying mechanism; together, they determine the hygrothermal performance and durability of building components. In traditional HAM modeling, WDR and the convective heat transfer coefficient (CHTC) are often simplified by applying generic, uniform values across the façade. However, this approximation fails to account for the spatial and temporal variations in WDR and neglects the significant variations in CHTC across the windward façade due to surrounding wind flow conditions. These oversimplifications can limit the accuracy of hygrothermal performance predictions, particularly when assessing moisture-related risks for heritage structures or during building renovation planning. In this paper, the spatial distribution of the catch ratio and CHTC are implemented through an externally coupled computational fluid dynamics (CFD) model with a hygrothermal response model (HAM). Steady Reynolds-averaged Navier-Stokes (RANS) equations and Eulerian multiphase CFD simulations are performed for CHTC and WDR, including the turbulent dispersion of raindrops. The coupling model is tested on a cubic low-rise building, with an analysis of the effects on surface degradation phenomena, including frost damage, salt crystallization, and algae growth. Results indicate that conventional approaches tend to underestimate critical rain loads, while the drying capacity is overestimated. Furthermore, the specific combination of rain loads and drying patterns results in unexpected performance patterns.

Mapping wear mechanisms and wear failure modes in metal contacts

J.C. Poletto, P.D. Neis, N.F. Ferreira, D. Fauconnier, P. De Baets.

A worn component can exhibit different types of failure on its surface. Pitting, indentation, grooving, and adhesive failure are typical examples of wear failure modes often encountered in metal contacts. These wear failure modes are the result of one or more wear mechanisms acting throughout the lifetime of a mechanical component. Literature often classifies wear mechanisms between abrasion, erosion, surface fatigue, tribochemical and adhesion. Nonetheless, linking the type of failure mode to its underlying cause (i.e. wear mechanism) is challenging. Inconsistencies and ambiguities in nomenclature arises due to differences in historical development or industrial application, increasing the complexity of this relationship. In this work, a diagram was developed to relate wear mechanisms of abrasion, erosion, surface fatigue, tribochemical and adhesion to the wear failure modes of grooving, indentation, pitting, and adhesive failure. Moreover, worn topographies from experimental applications are presented to illustrate the distinguished appearance of each of these wear failure modes. The proposed diagram shades light on the existing relationship between wear mechanisms and wear failure modes. This contributes to minimize discrepancies in terminology across various industries, supporting the development of consistent wear assessment.

Interconnection Optimization: A Treasure Trove for the EDA of Digital Electronic Circuits

Xiaoke Wang

As Moore's Law approaches physical limits, optimizing interconnection complexity emerges as a key opportunity in VLSI. Our findings reveal that by measuring and controlling Rent's exponent—a key indicator of interconnection complexity—in FPGA netlists, metrics like runtime, wirelength, and critical path length exhibit a double-exponential relationship with complexity. This underscores the potential of interconnection as a resource for optimization. Based on these insights, strategies such as sparse packing and duplication can simplify the interconnection structure with acceptable circuit size sacrifice, ultimately enhancing EDA physical design efficiency.

Large Language Models Reflect the Ideology of their Creators

Maarten Buyl, Alexander Rogiers, Sander Noels, Iris Dominguez-Catena, Edith Heiter, Raphael Romero, Iman Johary, Alexandru-Cristian Mara, Jefrey Lijffijt, Tijl De Bie

Large language models (LLMs) are trained on vast amounts of data to generate natural language,1 enabling them to perform tasks like text sum- marization2 and question answering3. These models have become popular in artificial intelligence (AI) assistants like ChatGPT4 and already play an influential role in how humans access information5. However, the behavior of LLMs varies depending on their design, training, and use.6 In this paper, we uncover notable diversity in the ideological stance exhibited across different LLMs and languages in which they are accessed. We do this by prompting a diverse panel of popular LLMs to describe a large number of prominent and controversial personalities from recent world history, both in English and in Chinese. By identifying and ana- lyzing moral assessments reflected in the generated descriptions, we find consistent normative differences between how the same LLM responds in Chinese compared to English. Similarly, we identify normative disagree- ments between Western and non-Western LLMs about prominent actors in geopolitical conflicts. Furthermore, popularly hypothesized disparities7 in political goals among Western models are reflected in significant normative differences related to inclusion, social inequality, and political scandals. Our results show that the ideological stance of an LLM often reflects the worldview of its creators. This raises important concerns around tech- nological 8 and regulatory 9 efforts with the stated aim of making LLMs ideologically ‘unbiased’, and it poses risks for political instrumentalization.

Renal Spatial Profile of IVIM Parameters: Humans and Dogs

Luis Sanmiguel, Amber Hillaert, Pieter de Visschere, and Pim Pullens

Intrarenal diffusion dynamics were studied by comparing the spatial profile of Intravoxel Incoherent Motion (IVIM) parameters in humans and dogs. Despite previous renal IVIM studies in humans, limited data exists on canine kidneys, and cross-species studies, to evaluate the canine kidney model, are sparse. To address these gaps, this study evaluates the spatial heterogeneity of renal IVIM parameters, including diffusion (D), pseudo-diffusion (D*), and perfusion fraction (fp), across different kidney regions. Nine healthy human participants and eight beagles were scanned using similar diffusion-weighted imaging (DWI) protocols on a 3T MRI. Kidneys were divided into nine anatomical regions, categorized by cortical, outer, and inner medullary layers from the cranial to caudal poles. Using Bayesian fitting, IVIM parameters were extracted, and differences were analyzed across regions. Results showed significant spatial variation in IVIM parameters for both species. In humans, D values were higher in cortical regions compared to the medulla, while D* and fp showed elevated values in inner medullary regions. Conversely, dogs reported consistently higher D values across all regions, with significant differences in D* and fp profiles compared to humans. Canine cortical regions displayed higher perfusion fractions than medullary regions, unlike the trends observed in human kidneys. This cross-species analysis provides insights into renal microcirculation and diffusion, highlighting distinct spatial profiles in human and canine kidneys. The findings may advance the understanding of renal physiology and pathophysiology, supporting the relevance of canine models in translational kidney research. Further studies are required to validate these observations and improve imaging protocols for enhanced reproducibility and accuracy in renal IVIM assessments.

Explainable and cohesive clustering of embeddings with InfoClus Dashboard

Fuyin Lai, Edith Heiter, Jefrey Lijffijt

Developing an understanding of the top-level structure of high-dimensional data can be facilitated by visualizing that data using a dimensionality reduction method. However, low-dimensional embeddings are often not so straightforward to interpret. To alleviate this problem, we introduce a method for clustering with explanations that partitions the data on the embedding into groups that each have a sparse explanation. Our objective function uses information theory to measure how much we learn from the explanations of the data partitioning, and also how complex the explanations are. This combination yields a tunable optimization objective. To optimize this objective, we employ hierarchical clustering to narrow the search space, and show the objective can then be effectively optimized using greedy search, resulting in a scalable algorithm. We conduct a qualitative and quantitative analysis of InfoClus on three data sets, and find the method efficiently yields useful insights. We contrast the results on the cytometry data with its associated published manual analysis. This comparison highlights that a thresholding procedure used in the manual analysis may lead to false negatives, unlike in the bottom-up analysis using t-SNE and InfoClus. Empirical scalability results for InfoClus show close to linear scaling in number of data points and dimensions. We conclude that InfoClus can automatically create good starting points for the analysis of dimensionality-reduction-based scatter plots.

Pattern formation in planar electrophoretic cells

Martijn van den Broek, Filip Beunis, Mohammadreza Bahrami, Yuanze Chen

In experiments on planar electrophoretic cells with nonpolar solvent and surfactant we observe an intriguing pattern formation of suspended pigment particles. The observed dependency of the patterns on concentrations, cell dimensions and applied voltage profile is so far not understood. A possible explanation is that the phenomenon is an expression of spinodal decomposition or turing pattern. We compare our observations with earlier reports of pattern formation in different environments. The results are also part of a recently started project that aims to better understand electrokinetics in complex geometries and nonpolar environments.

Deep Learning-Based Microcatheter Localization in the Brain for Pretreatment Planning of Arteriovenous Malformation Embolization

Mohammadjavad Sedghizadeh, Jessie Duquesne, Elisabeth Dhondt, Luc Defreyne, Uri Singfer, Vincent Keereman, Peter Vanlangenhove, Charlotte Debbaut, Danilo Babin

Endovascular catheter use for medical interventions has increased due to shorter post-procedural recovery times and a relatively low complication rate. Endovascular interventions are one of the preferred treatments for arteriovenous malformations (AVMs). An AVM is an abnormal vascular structure that irregularly connects arteries and veins, occurring most commonly in the brain. However, navigating a catheter close to the nidus of an AVM is often challenging because of limited visualization, complex vessel structure, and limited manoeuvrability of the guidewire[1]. Locating the catheter within 3D vascular maps is still difficult in current AVM procedures[2]. Therefore, this research focused on providing a detection of the microcatheter inside the brain using 2D X-ray angiography (XRA) images obtained during interventions. The 2D catheter localization can then be mapped into the 3D segmented brain vessels, helping clinicians better navigate the catheter to the target. A deep learning-based approach is proposed to localize the microcatheter. Small patches within vessel structures were extracted from the XRA images, as the vessels represent the region of interest (ROI) for microcatheter presence. The ROI is selected using Digital Subtraction Angiography (DSA) images, showing contrast-enhanced blood vessel structures during the procedure. The patches were then classified using a convolutional neural network to determine the presence or absence of the microcatheter, based on visual features extracted from XRA images. For now, the model was trained on XRA data from 17 patients and evaluated on a different test set of two patients. A post-processing algorithm was implemented to reduce false positives based on the connectivity of positive patches, ensuring that isolated detections are removed. Our proposed approach demonstrates potential in detecting the microcatheter, even with limited data. Further optimization and validation within brain vascular maps is needed for precise 3D catheter localization. [1]Schimmel, K. et al.Int J Mol Sci.2021 [2]Liu, C. et al.Front Neurol.2023

On the Physical Mechanisms in Ultrasound Neuromodulation

Tom Plovie, Thomas Tarnaud, Emmeric Tanghe

Ultrasound Neuromodulation is a technique that can potentially be used for non-invasive neuromodulation. A transducer is placed on the scalp and insonicates the brain. By focusing the ultrasound beam, a subcortical region in the brain can be targeted with high resolution in the order of centimeters in axial direction and millimeters in the lateral direction. However, the mechanism of interaction between the ultrasound wave and the neuronal tissues is still unclear. Multiple mechanisms have been proposed (e.g.: intramembrane cavitation, mechanosensitive ion channels, changes in membrane capacitance). It is likely that they all contribute but it is unclear to what extent. These mechanisms are playing a role on the neuron level and are therefore called biophysical mechanisms. This study focuses on the physical mechanisms of the wave inside the brain and not on the beforementioned biophysical mechanisms. To this end, a realistic head model is used. The medium properties are retrieved from a CT scan. This is used to run a pressure simulation. From the pressure field, the acoustic radiation force (ARF) is calculated. The ARF is then used as an input to simulate the displacement of the brain tissue. Prior to the simulation of the head model, the grid sensitivity needed for both simulations is investigated in order to have reliable results. It is found that the locations of maximal pressure and maximal ARF are the same. This is also the case for the displacement immediately after insonication. However, as time progresses, the displacement field changes and so do the point of maximal displacement and the targeting resolution.

Hybrid Unicast-Broadcast Video Delivery for Scalable Low-Latency Live Streaming

Casper Haems, Jeroen van der Hooft, Hannes Mareen, Peter Steenkiste, Glenn Van Wallendael, Tim Wauters, Filip De Turck

The demand for high-quality, low-latency video streaming is placing strain on conventional internet infrastructures. This paper proposes a hybrid unicast-broadcast video delivery framework designed to address this challenge by integrating advanced 5G broadcast technologies with traditional unicast methods. By offloading popular content to a broadcast network, the approach aims to alleviate congestion and enhance overall streaming efficiency. To ensure reliable video segment delivery over the broadcast network, regardless of the physical layer, we incorporate packet recovery (PR) and Forward Error Correction (FEC) mechanisms. Additionally, Temporal Layer Injection (TLI) is employed to further improve video quality while maintaining reduced bandwidth requirements compared to traditional unicast-only approaches. This innovative framework leverages 5G terrestrial broadcasting within over-the-top (OTT) streaming environments, enabling seamless delivery of adaptive video content with sub-1-second live latency. Comprehensive experimentation and evaluation through large-scale emulation demonstrate the efficacy of this hybrid approach in meeting the evolving demands of modern multimedia delivery systems. Notably, when broadcasting the top 3 most commonly watched video streams, 63% of viewers no longer need to request video segments via unicast, as they are efficiently delivered over broadcast channels. This hybrid model offers significant scalability, cost reduction for an internet service provider (ISP), and efficiently delivers content directly to user devices without additional intermediaries, improving viewer experience through low-latency, high-quality streaming.

FlexIntensity: Zeeland electrical grid model

Louis De Backere, Collin Bohncke, Lieven Vandevelde

To reach the climate goals set out by the European Union, a real energy transition is necessary; from fossil based to renewable energy. The province of Zeeland, the Netherlands, will be heavily affected by this energy transition. It hosts one of the largest petrochemical clusters of North-West Europe and the largest hydrogen cluster of the Benelux. Zeeland is therefore heavily dependent on fossil resources. In the coming years, up to 9 GW of large-scale low emission energy production could be connected to the Zeeland grid. This gives the possibility for industry to electrify its energy and hydrogen demand, but raises two grid challenges: intermittency of power and grid congestion. Hereto three solutions are proposed, each with their own advantages and disadvantages; large-scale electrolysis installations, industrial flexibility and grid infrastructure investments. A grid solution will consist of a combination of these three aforementioned solutions. To determine adequacy of a proposed future grid, a grid model is developed in the open-source Python library PandaPower. This grid model simulates a proposed grid on hourly basis over four typical weather years. The grid model uses open-source topology and component data. The time series data for residential and industrial loads are based on synthetic load profiles from literature. The resulting power flows allow to determine the adequacy of the solution. In a next step, the grid model will be used to determine an optimal future grid and give policy recommendations.

Extending path memory to improve kinetics calculations in rare molecular processes

Elias Wils, Wouter Vervust, Titus Van Erp, An Ghysels

Molecular dynamics (MD) simulations serve as a computational microscope to study the intricate kinetics of crucial biological processes, such as the binding of a cancer drug molecule or permeation of a molecule through cell membranes. One of the main challenges in this field is the significant disparity in timescales: while these processes occur rapidly, simulating them can require years of CPU time. To solve this, we use Transition Interface Sampling (TIS), an enhanced sampling algorithm that bridges the gap between these widely separated timescales. This is done by quantifying the progress of a reaction using a single variable and placing checkpoints ("interfaces") along the way, after which we cleverly sample MD trajectories ("paths") at strategic points. However, TIS can struggle with simulating biological processes with complex energy landscapes, often leading to paths getting stuck in metastable states, resulting in lengthy simulation times. PPTIS (Partial Path TIS) removes this problem by cutting paths short, which unfortunately also truncates the history of particle trajectories, potentially leading to a significant loss in accuracy. To tackle these shortcomings we introduce StapleTIS, an innovative approach that preserves essential trajectory memory while keeping paths manageable in length. The paths are terminated by turns (occurring when encountering an energetic barrier for example) on both sides, mimicking the shape of a staple. This ensures paths "remember" a longer segment of their trajectory, enabling them to navigate a potential energy landscape more efficiently and giving us more accurate information about the kinetics of the system. We demonstrate the effectiveness of StapleTIS using 1D and 2D toy systems, achieving performance comparable to traditional TIS. Our next steps involve applying this method to real-world biological systems to obtain valuable information about their kinetics, ultimately enriching our knowledge about various biological mechanisms.

Heart-Carotid Pulse-Wave Velocity via Laser-Doppler Vibrometry as a Biomarker for Arterial Stiffening: a Feasibility Study

Simeon Beeckman, Smriti Badhwar, Rosa Maria Bruno, Nilesh Madhu, Patrick Segers

Background: Pulse-wave velocity (PWV) is the most commonly used biomarker of arterial stiffness and has been proven useful in monitoring arterial stiffness and predicting cardiovascular risk. We previously introduced a Laser doppler vibrometry (LDV) prototype which can measure carotid-femoral PWV (cfPWV). We assessed the feasibility of using the same device to infer heart-carotid PWV (hcPWV). The advantage of hcPWV over cfPWV is that the ascending aorta, which is the most distensible segment of the aorta contributing most to total arterial compliance, is included in the PWV pathway. Methods: Signals were simultaneously acquired from a location on the chest (near either the base or apex of the heart) and the (left/right) carotid artery in 100 subjects, age 19 to 81 (44M, 38F). Fiducial points on the heart waveforms are associated with opening (S1) and closure (S2) of the aortic valve, which can be combined with, respectively, the foot and dicrotic notch of the carotid waveform to retrieve heart-carotid pulse transit times (hcPTT) for hcPWV. Considering two distinct heart-signal measurement sites (base and apex), four hcPTTs estimations are evaluated. Correlations of these hcPTTs with cfPTT, age and blood pressure were evaluated. Results: hcPTTs could be measured in 94% of the subjects. hcPTT derived from S2 from a heart-measurement site located at the base of the heart with the carotid dicrotic notch provided the strongest correlations with age (r=-0.377, P<0.001) and cfPTT (r=0.475, P<0.001). Conclusions: We conclude that LDV may provide a corollary biomarker of arterial stiffness, encompassing the ascending aorta.

Beam-Enhanced Time Shaping

Mohammadreza Heydarian, Didier Colle, Wouter Tavernier

Full automation in Industry 4.0 demands pervasive and reliable connectivity. Due to the high mobility of industrial robots, wired connections are often impractical, making wireless connections essential. Ensuring reliable time-sensitive critical traffic in these scenarios is challenging due to the unpredictable nature of wireless environments. Our approach leverages the significant potential of spatial reuse through beamforming to guarantee TSN traffic while optimizing best-effort traffic. This strategy enables a path toward full automation with a balanced and cost-effective network solution.

Demo: Selection tool for renewable energy integration on a livestock farm

Manon Everaert, Jarissa Maselyne, Steven Lecompte

Globally, only 4 % of energy used on livestock farms comes from renewable sources. Yet, it is estimated that around 7.6 % of a livestock farm’s total emissions are from on-farm energy use (e.g. heating and various electricity use). Although there is potential to shift to renewable energy sources, farmers have identified several barriers. These include: unclear regulations, lack of knowledge about existing systems, insufficient understanding of the required capacities, and limited investment capital. To address these challenges, an interactive selection tool was developed. This tool allows users to explore different scenarios — different combinations and variations of simulated systems — based on investment cost, total cost of ownership, and energy related greenhouse gas emissions. Users can choose from three different use cases: a pig farm in Belgium, a dairy farm in Germany, or a poultry farm in Greece. The tool can display results for up to 100,000 scenarios per case. A Pareto front is used to identify optimal scenarios.

Demo: From CT Scans to 3D Models: A Comprehensive Approach for Insect-Based Research

Felipe Oliveira Ribas, Pieterjan De Boose, Maria Bouga, Jürg Fröhlich, Fani Hatjina, Anke Huss, Menelaos Stavrinides, Zoi Thanou, Antonios Tsagkarakis, Andri Varnava, Marco Zahner, Arno Thielens

Insects are vital to ecosystem health but may face increased risks due to RF-EMF (Radio-Frequency Electromagnetic Field) exposure from advancing telecommunication networks, especially with the shift from 4G to 5G. Understanding these effects requires detailed biological models for accurate simulations. This project presents a comprehensive workflow for converting insect CT scans into anatomically accurate 3D models, designed to handle the specific challenges of insect data, such as missing body parts and non-ideal positioning. Using free software tools, particularly Blender with custom Python scripts, the workflow enables the transformation of raw CT data into refined, watertight 3D meshes. The primary goal is to establish a reusable and adaptable pipeline that meets diverse research needs, including assessing environmental impacts of RF-EMF exposure on insects. To demonstrate the workflow’s application, a 26GHz electromagnetic simulation was conducted in Sim4Life, showcasing RF-EMF interaction with insect anatomy. Following simulation, the models were imported into Blender for enhanced manipulation and analysis. This workflow, along with the CT scan data, is accessible on Sketchfab and published in open repositories, adhering to FAIR (Findable, Accessible, Interoperable, Reusable) principles to broaden accessibility and support reproducibility. By providing an open, accurate approach to biological modeling, this project facilitates advances in environmental impact research and aids in understanding the potential effects of technology on critical insect populations.

Unveiling Sex Dimorphism in Healthy Cardiac Anatomy: A data-driven morphological analysis on the UK Biobank Population

Beatrice Moscoloni, Cameron Beeche, Julio Chirinos, Patrick Segers, Mathias Peirlinck

Despite growing awareness of sex differences in cardiovascular disease, many questions about sex dimorphism in cardiovascular physiology remain. Recent studies have highlighted both sex- and age-related differences in cardiac morphology that are not fully explained by simplistic scaling models. This underscores the need for a more detailed, sex-specific analysis of cardiac anatomy that considers factors like age, blood pressure, and body size. Statistical shape modeling, coupled with data-driven techniques, provides an opportunity to analyze these complex shape variations and how they interact with such factors. We performed statistical shape analysis on 234 biventricular anatomies from a healthy subset of the UK Biobank CMR dataset. We segmented ventricular volumes, and generated an anatomical atlas from 3D meshes. We identified the main modes of anatomical variation and conducted MANOVA tests to evaluate the effects of sex, age, blood pressure, and body size (BMI, BSA, height, weight) on cardiac morphology. We used multivariate linear regressions to adjust shape coefficients for confounders and leveraged logistic regression models to assess how well cardiac morphology, before and after adjustment, could discriminate males from females. The logistic regression coefficients were then analyzed to identify key shape patterns between the sexes. Our results show significant effects of sex, age, blood pressure, and body size on cardiac morphology. Importantly, sex dimorphism persisted in cardiac size and obliqueness after adjustment for confounders, underscoring the need for sex-specific diagnostic criteria in cardiovascular care.

On the use of a plasma jet for atomic oxygen production: Source characterization and application in cultural heritage

Michalis Poupouzas, A. Nikiforov, R. Morent, A. Sobota

Non-thermal radio frequency (RF) driven Atmospheric Pressure Plasma Jets (APPJs) are efficient sources of reactive oxygen and nitrogen species (RONS). APPJs can be operated in ambient air, so that generated RONS can be effectively delivered onto target surfaces without need of complex vacuum system. Recently our team started investigating the applications of plasma technology in cultural heritage for the cleaning of different samples/artifacts. In this work an RF atmospheric pressure plasma jet is used to induce atomic oxygen chemistry in order to treat various targets. The beam is produced by a coaxial electrode configuration. An RF discharge is established between a single axial pin electrode and the grounded stainless steel nozzle. The plasma operates in a mixture of He and O2. By applying the gas flow the reactive species generated inside the nozzle are transported in the ambient air forming an effluent rich of atomic oxygen. As shielding gas, different noble gases such as Ar and He are used. The purpose of the shielding gas is the enhancement of O transport downstream to the sample and protection of the plasma chemistry from entrance and mixing with ambient air. Within the context of this study, a parametric investigation of emission spectra is conducted by varying the ratio of the flows of main gas and shielding gas. Moreover, the thermal effect of shielding gas is investigated by means of a precision temperature probe placed directly into the effluent. The effect of power in gas temperature is measured with a commercial RF power meter.