Bachelor and Master Projects
For excellent students we also offer projects which should lead to journal publications.
The chair of Fluid Dynamics also supports students who want to arrange their projects in cooperation with industry - provided that there are no non-disclosure agreements.
To top students, we also offer external Master’s projects with friends and colleagues at other universities, for example in Trondheim (Norway), Newcastle (UK) or Berkeley (US). Some prior work as a HiWi with us is a prerequisite, so we can make sure to only send out excellent students.
On request outstanding students can also realize important projects as Research Assistants.
Please contact: project.cfd [at] uni-due.de
Master and Diploma Projects which should lead to publications
Understanding the wind field near the ground is crucial for various applications, including urban planning, the assessment of pollutant exposure risks, and the siting and optimization of wind turbines. Traditionally, such measurements rely on stationary installations—typically temporary towers constructed from metal scaffolding. However, advances in drone technology offer a more flexible and efficient alternative.
This project explores the use of drones equipped with appropriate sensors to measure near-surface wind fields. In theory, the data recorded by a drone's flight controller during steady hovering can be used to estimate wind speed, direction, and turbulence characteristics. In practice, data acquisition can be further simplified by using onboard logging tools such as the PhyBox app.
The aim of the project is to collect wind data using a drone at specific points, along predefined lines, or over larger areas (planes or volumes). Depending on the student's interest and progress, the project may be extended to include turbulence statistics, temperature field measurements, or even the detection of airborne pollutants.
While some familiarity with drone operation and programming is beneficial, the primary focus of the project will be on data processing and on developing models that relate sensor readings to flow phenomena. This project can be repeated with different students, particularly if they wish to explore new methodologies or propose improvements to the measurement approach.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
The "Lilium Jet" was meant to be a battery-electric vertical take-off and landing (VTOL) aircraft that was conceived to combine low-emission propulsion, ease of control during vertical operations, and efficient horizontal cruise flight. Initially presented as a breakthrough in sustainable aviation, the concept attracted substantial investment and led to the development of an uncrewed flying demonstrator. Despite this, the company behind the project ultimately filed for bankruptcy—raising questions about the technical and economic viability of the concept.
Critics have long argued that the performance targets set by the company were overly ambitious and could not be met with current or foreseeable battery and propulsion technologies.
The goal of this thesis is to conduct an independent feasibility study of the "Lilium Jet" concept. Using publicly available data, fundamental aerodynamic principles, and established methods in aircraft design, the project will develop a simplified performance model of such an aircraft. This model should estimate key quantities such as power consumption, flight endurance, and energy usage across different flight phases: vertical take-off, cruise, and vertical landing.
Where necessary, the model may be supported by basic CFD simulations to improve the fidelity of key aerodynamic parameters. Based on the performance model, both best-case and conservative worst-case scenarios should be evaluated. In addition, a sensitivity analysis will be conducted to investigate how key design and technology parameters—such as battery energy density, propulsion power density, fan diameter, aircraft mass, or safety margins—affect overall performance.
This study aims to provide a technically grounded assessment of whether such an aircraft concept could be realistically operated in meaningful use cases under current or near-future technological conditions.
The project will start with a (brief) planning phase to identify the steps and work-packages involved in the work.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
As global food demand continues to rise, there is a growing need for sustainable production technologies. One promising solution is the use of photobioreactors, which cultivate microalgae in closed, light-permeable systems. These reactors typically consist of transparent pipes—often made of PMMA—that allow sunlight to penetrate and support photosynthetic growth. Microalgae are highly sensitive organisms that require carefully controlled environmental conditions. In particular, conventional axial pumps used for fluid circulation often generate shear forces that can damage the cells, ultimately reducing biomass yield and quality.
This project aims to develop a low-shear axial pump specifically tailored for microalgae photobioreactors. Using computational fluid dynamics (CFD), the goal is to design a pump that minimizes shear stress while maintaining the necessary volumetric flow rate and pressure head. The design must also consider compatibility with lightweight materials and integration into the existing reactor system without compromising structural stability.
The project involves detailed flow analysis, optimization of pump geometry (including impeller, stator, and casing), and assessment of material constraints. A strong background in fluid dynamics, numerical simulation, and engineering design is essential. The thesis is offered in collaboration with Algoliner GmbH & Co. KG and supervised by Prof. Dr.-Ing. Andreas Kempf.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
In closed water systems, tank maintenance typically requires halting operation, draining fluids, and disassembling components. In industrial settings, this leads to production downtime, material losses, and associated ecosnomic costs. A promising alternative is the use of integrated high-pressure water jets that clean tank walls during ongoing operation. The effectiveness of such jets depends heavily on the nature and adhesion strength of the materials—referred to as substrates—attached to the inner surfaces. In photobioreactors, for example, microalgae can form deposits whose adhesion varies significantly depending on environmental and operational conditions.
This project aims to identify the jet parameters required to effectively remove such deposits without interrupting system operation. Key variables include jet pressure, nozzle diameter, penetration depth, and flow rate. In parallel, the project will experimentally characterize the adhesion properties of algae films using a Design of Experiments (DOE) approach, enabling a systematic investigation of how substrate variability affects cleaning performance.
The work requires a solid foundation in fluid mechanics and parametric design. The project is offered in collaboration with Algoliner GmbH & Co. KG and supervised by Prof. Dr.-Ing. Andreas Kempf.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
Flame kernels are the small, self-sustaining combustion zones that form immediately after spark ignition in internal combustion engines, marking the onset of flame propagation. When hydrogen is used as a fuel, the behavior of these kernels becomes particularly sensitive to instabilities due to hydrogen's high mass diffusivity. Thermo-diffusive instabilities arise when heat and mass diffuse at different rates, leading to uneven flame development. The geometry of the flame front is critical: convex regions can locally accelerate the flame due to hydrogen accumulation, while concave regions tend to decelerate as hydrogen is depleted. These local imbalances amplify perturbations, resulting in wrinkled or cellular flame structures that can influence engine performance, efficiency, and emissions.
This project aims to characterize hydrogen flame kernels in terms of their shape and propagation dynamics using high-fidelity computational fluid dynamics (CFD) simulations. The work will involve defining suitable initial and boundary conditions, running simulations within our in-house CFD framework, and developing post-processing tools for quantitative and visual analysis of flame evolution and structure.
The project requires a solid understanding of fluid dynamics and numerical methods, along with interest in combustion phenomena and simulation-based research.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
Modeling internal combustion engines presents significant geometric challenges due to the complexity of components such as intake and exhaust ducts, valves, the combustion chamber, spark plug, and injector. While these shapes are critical to engine performance, they are often difficult to describe using standard parametric surfaces. In many cases, 3D scanning is required to reconstruct the internal geometry, particularly for the ducts and combustion chamber. However, the resulting surface meshes frequently contain defects—such as overlapping triangles or non-physical intersections—that make modification and integration into simulation workflows difficult.
The goal of this project is to develop a fully parametric CAD model of the optical engine operated at the University of Duisburg-Essen. The model should accurately reproduce key physical and geometric properties of the real engine, including compression ratio, effective cross-sectional areas of the ducts, injector and spark plug placement, valve timing and lift profiles, duct orientations, and crevice volumes. This model will serve as a foundation for future simulation, visualization, and experimental comparison.
The project requires strong skills in CAD modeling, technical drawing, dimensional measurement, and engineering documentation. For more information, please contact Prof. Dr.-Ing. Andreas Kempf.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
Large-eddy simulations (LES) that include emission predictions rely on finite-rate chemistry, which requires the integration of stiff ordinary-differential equations for every grid point and time step. For realistic kinetic schemes, this procedure dominates total computing time and limits the practicality of LES in industrial design. A data-driven alternative is to replace the ODE solve with a neural-network surrogate. Such surrogates must reproduce reaction rates that span many orders of magnitude, preserve mass conservation, and maintain numerical stability when embedded in the strongly coupled solver; otherwise local errors can accumulate and terminate a simulation. While there are approaches to handle the first two challenges, keeping a simulation stable remains difficult.
The objective of this thesis is to identify a neural-network architecture that best mitigates the above criteria and to provide a neural network that can be successfully integrated into an existing LES framework. A representative database of thermochemical states and corresponding reaction rates will be provided, together with a ready-to-run LES infrastructure. Initial investigations will use a global mechanism involving five species to enable rapid prototyping; if successful, the study will be extended to a detailed scheme with more than twenty species, including short-lived intermediates and widely separated time scales.
The project requires proficiency in Python and, preferably, prior experience with PyTorch. Familiarity with combustion modelling or numerical methods is advantageous but not essential. Expected deliverables include a documented, pretrained network that demonstrably replaces finite-rate chemistry in LES without compromising accuracy or robustness, as well as a critical evaluation of its limitations and prospects for broader application.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
Neural networks can replace the stiff finite-rate chemistry solver in large-eddy simulations, but only when the training data covers the full thermochemical manifold that a turbulent flame actually explores. Databases taken from canonical one-dimensional laminar flames trace a narrow path through state space and miss the off-equilibrium mixtures generated by turbulence, heat loss, partial premixing, and emerging auto-ignition pockets. To construct a surrogate that remains reliable under such conditions, this thesis reverses the usual workflow.
First, a series of three-dimensional reactive LES cases will be carried out in which operating parameters are gradually shifted so that the combustor passes from a flame-propagation-dominated regime to one governed almost entirely by auto-ignition. The simulations will be mined for thermochemical snapshots, producing a high-fidelity catalogue of the states that truly occur in mixed-mode operation. An analysis routine will then assign to every stored cell a quantitative index that measures the local balance between propagation and auto-ignition, allowing the database to be filtered by stabilisation mechanism.
The next stage reconstructs these LES states with low-order reference problems: homogeneous reactors, freely propagating premixed flames, and counter-flow configurations. A similarity metric in an appropriately reduced state space must be developed that will identify gaps and drive targeted additions until coverage is statistically complete. The final, well-balanced dataset will serve to train a neural-network surrogate for species source terms; its accuracy and numerical robustness will be tested against the original finite-rate solution, and the expected computational speed-up for future deployment will be quantified.
The LES infrastructure, detailed reaction schemes, and preliminary post-processing scripts are available. The project is suited to a student with strong programming ability (Python or C++), familiarity with machine learning, preferably PyTorch, and an interest in combustion physics and numerical simulation. Deliverables include the LES database, the low-order reconstruction workflow, the trained surrogate model, and a critical assessment of its performance limits.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
In the numerical modeling of nanoparticle synthesis, the evolution of particle properties—typically represented as a particle size distribution (PSD)—is governed by the Population Balance Equation (PBE). One of the simplest approaches to solving this equation is the monodisperse method, which reduces the PSD to a single representative value by assuming a locally monodisperse distribution. This method has shown particular promise for simulating the formation of hetero-aggregates, i.e., particles composed of two or more different materials.
When coupled with Computational Fluid Dynamics (CFD), such models can be used to simulate complex three-dimensional flow and reaction fields in nanoparticle synthesis processes. However, in Large Eddy Simulations (LES), steep spatial gradients in the particle field often cannot be fully resolved on the computational grid. Previous studies at our chair have identified that standard filtering operations in LES can introduce significant errors into the prediction of particle distributions.
To address this, the Transported Filtered Density Function (FDF) approach offers a more accurate alternative. This method computes the evolution of a filtered probability density function in both physical and composition space, with all source terms appearing in closed form. In doing so, it inherently captures subgrid-scale effects and spatial fluctuations that influence particle dynamics.
The goal of this project is to integrate the FDF approach into an LES framework in order to derive and solve the transported FDF equation for particle number concentration in nanoparticle synthesis. In the first step, a general monodisperse model for hetero-aggregate formation will be implemented in our in-house CFD code PsiPhi. In the second step, this model will be extended using the LES-FDF method to simulate the spray flame synthesis of iron oxide (Fe₂O₃) nanoparticles in the well-characterized SpraySyn burner.
This project combines theoretical modeling, numerical method development, and practical simulation. It requires a solid foundation in fluid dynamics, numerical simulation, and statistical modeling of multiphase flows.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
To demonstrate the Chair of Fluid Dynamics' capabilities in tomographic reconstruction at scientific conferences and industrial exhibitions, this project aims to design and build a compact and portable measurement setup using commercially available hardware. The system will utilize a set of 6 to 8 Raspberry Pi cameras, controlled by 3 to 4 Raspberry Pi 5 boards, to capture optical data around a small cylindrical test region approximately 15 cm in diameter. This space will accommodate small-scale combustion systems such as candles or gas lighters.
For transportability and durability, the entire system is to be integrated into a rugged, protective equipment case similar to those used for professional camera gear. The project involves the mechanical design of the camera mounts and the test region, integration of the optical system, and routing of power and data lines. The camera mounts are to be manufactured via 3D printing to allow for iterative prototyping and adjustments throughout the development phase. Once constructed, the setup will be tested using known target geometries to validate image acquisition and alignment. Finally, a Python-based software tool will be developed for image acquisition and basic tomographic reconstruction of small objects.
The project may be extended in two directions. First, to enable the tomographic reconstruction of reactive flows, the system can be modified to safely accommodate small combustors and capture combustion-related data. Second, the PCIe interface of the Raspberry Pi 5 offers the possibility of attaching external GPU accelerators via adapters. This opens the potential for in situ hardware-accelerated postprocessing and, possibly, real-time low-resolution tomographic reconstruction.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
In both automotive applications and subsurface hydrogen storage facilities, hydrogen is stored at very high pressures—typically 350 to 700 bar for vehicles and 150 to 200 bar in caverns. A critical safety scenario arises in the event of a sudden release, such as a valve failure, where a highly underexpanded, supersonic hydrogen jet is discharged into the atmosphere. Under certain conditions, this jet can autoignite, posing serious hazards to surrounding infrastructure and personnel.
This Master's project investigates the feasibility of numerically predicting such autoignition events using PsiPhi, the in-house CFD code developed at the Chair of Fluid Dynamics. The project begins with a literature review to identify a suitable experimental case that demonstrates autoignition following hydrogen blow-out. This case will be used to define boundary conditions and geometry for a fully compressible, multispecies simulation in PsiPhi. The numerical setup will be used to study whether the onset of autoignition can be reproduced under realistic thermodynamic and flow conditions.
If autoignition can be reliably predicted, the model may be extended to include finite-rate chemistry to investigate whether the ignition event can transition into a sustained flame. Ideally, the simulation results will allow for comparison with experimental observations and enable further parametric studies to explore the sensitivity of ignition behavior to jet dynamics, turbulence intensity, and thermochemical properties.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
Wildfires are becoming an increasingly serious threat to ecosystems and human settlements, making accurate prediction and modeling tools essential for risk assessment and mitigation. This thesis focuses on validating the Large Eddy Simulation (LES) code PsiPhi, developed at the Chair of Fluid Dynamics, which combines LES with Lagrangian particle tracking and combustion modeling. The validation will be carried out using controlled wind tunnel experiments conducted by Finney et al., in which fire propagation through arrays of cardboard structures was studied under varying conditions of fuel density, wind speed, and geometric configuration.
The project involves reconstructing and simulating selected experiments to assess the accuracy of the model with respect to flame spread rates and temperature evolution. To achieve this, the student will identify key thermophysical and kinetic properties of cardboard, validate particle-scale heat transfer behavior using an existing solver, and implement a simplified geometric representation of the fuel array. A reference simulation will be performed using the Flamelet Generated Manifold (FGM) model, followed by comparison to experimental results. The robustness of the model will then be tested by applying it to additional experimental configurations.
Challenges include uncertainties in material parameters due to variability in cardboard composition and potential limitations in applying pyrolysis models developed for wood to cardboard-based fuels. For students interested in pursuing a Master's thesis, the project can be extended to include radiative heat transfer using the Discrete Ordinates Method (DOM), with separate validation of radiation-particle interactions and an evaluation of radiative effects in flame propagation. The work will require solid knowledge in fluid mechanics and heat transfer, interest in combustion processes, and confidence in numerical simulation and programming.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
The transition from deflagration to detonation remains a complex and not yet fully understood phenomenon. While detonation poses significant safety risks in industrial settings, its underlying mechanisms also offer opportunities for the development of advanced propulsion systems. Pulse detonation engines (PDEs) and rotating detonation engines (RDEs), for instance, promise less complexity and weight and higher efficiency at elevated Mach numbers compared to conventional turbojets or rocket engines. A better understanding of how detonations are initiated is therefore essential for both risk mitigation and technological innovation.
This project investigates the role of shock wave focusing in the initiation of gas-phase detonations. Using high-resolution numerical simulations, the project explores how different reflector geometries—such as three-wall, four-wall, and five-wall corners, as well as conical configurations—influence shock wave reflection patterns and focusing intensity. Of particular interest is how these geometric configurations affect local pressure and temperature fields, and thereby promote or suppress detonation onset in reactive gas mixtures.
The simulations will be conducted using the in-house CFD code PsiPhi, which enables fully compressible flow modeling with reactive species. The goal is to identify the conditions and geometries that lead to critical focusing effects, and to quantify their impact on detonation behavior through analysis of the resulting flow fields and reaction dynamics. The project requires an interest in high-speed flow phenomena, numerical methods, and reactive fluid dynamics.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
Modern gas turbine combustors are increasingly designed to operate under lean combustion conditions in order to reduce emissions. However, lean flames are particularly susceptible to thermoacoustic instabilities, which can compromise performance, reliability, and structural integrity. In industrial systems, these instabilities are typically mitigated using acoustic dampers, which consume a portion of the combustion air and consequently reduce the overall efficiency of the system.
To support the development of more efficient combustor designs, predictive tools have been developed to evaluate the thermoacoustic behavior of combustion systems at an early design stage. While high-frequency thermoacoustic modes can be resolved using Large Eddy Simulations (LES), these require high mesh resolutions due to the short acoustic wavelengths involved. To enable such predictions on coarser computational meshes, a correction method has recently been introduced to compensate for the acoustic damping effects caused by numerical filtering.
The aim of this thesis is to evaluate this correction approach by applying it to a second laboratory-scale test case—the DLR Hexagon combustor. Using OpenFOAM and in-house solver extensions, the student will carry out LES of cold and reactive flow conditions under both stable and unstable operating regimes. The tasks include generating a high-quality mesh using snappyHexMesh, validating the cold flow simulations, and conducting reactive simulations with different combustion models. The performance of the acoustic correction model will then be tested under representative multi-jet configurations.
The project requires experience with CFD, combustion modeling, and numerical simulation workflows. Simulations are implemented in OpenFOAM (C++), with postprocessing carried out in Python. A helpful introduction to thermoacoustic theory is provided in Chapter 1 of the thesis by Sharifi, available at: https://duepublico2.uni-due.de/receive/duepublico_mods_00074576.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
To meet increasingly strict emission regulations, modern gas turbine combustors are designed to operate under lean combustion conditions. One approach to achieving flame stabilization in such systems is the use of jet flame combustors, in which a high-speed jet of premixed air and fuel enters the combustion chamber and is ignited by recirculated hot gases. Depending on the specific operating conditions, this ignition may be delayed, resulting in a lifted flame—i.e., a flame where the main heat release zone is located downstream of the burner nozzle.
In reactive Large Eddy Simulations (LES), finite-rate chemistry is often used to model combustion by solving detailed reaction mechanisms based on the local gas composition in each computational cell. In simulations of recirculation-stabilized flames, accurate prediction of the mixing between the incoming jet and the surrounding hot gases is critical, as ignition delay times are highly sensitive to local mixture properties. In practice, insufficient resolution of the mixing process can lead to overpredicted ignition delays, and consequently, to artificially high flame lift-off heights.
The aim of this project is to investigate these effects through LES of a canonical slot burner configuration based on an existing reference case. Specifically, the study will examine the grid dependency of flame lift-off predictions, the influence of heat losses in the burnt gas region, and the computational cost of different approaches to solving the finite-rate chemistry—comparing classical solvers to modern, machine-learning-based alternatives such as neural networks.
Simulations will be performed using OpenFOAM (C++), with additional postprocessing carried out in Python. All computations will be run on the high-performance computing cluster amplitUDE. The results will contribute directly to the development and optimization of the in-house research code, and exceptional results may serve as the basis for publication in a peer-reviewed journal.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
Pore diffusion, the process by which molecules move through the interconnected voids or pores of a solid material, significantly influences reactive transport processes in porous media, impacting reaction kinetics, efficiency, and system performance in numerous applications, including catalysis, filtration, energy storage devices, drug delivery systems, and subsurface contaminant remediation. It also plays a critical role in industrial processes such as iron ore reduction, particularly in the context of hydrogen-based direct reduction technologies aimed at reducing CO₂ emissions. Despite its practical importance, a detailed understanding of how pore-scale diffusion interacts with reactive processes remains limited.
This project aims to theoretically and numerically investigate the role of pore diffusion in controlling reaction rates and efficiency in porous materials, using an in-house simulation code that has already been developed. Initially, the study will focus on data collection and an extensive literature review to identify established theories, models, and experimental insights relevant to pore diffusion and reaction coupling.
Based on the insights gained, numerical models will be developed and tested to simulate reactive transport within representative porous structures. These models will explicitly account for diffusion within pore-scale geometries and include relevant reactive mechanisms. Where appropriate, basic computational fluid dynamics (CFD) simulations may be employed to verify and refine pore-scale transport properties.
The validated models will then be utilized in parametric studies to systematically evaluate the sensitivity of reactive transport outcomes to pore diffusion parameters such as porosity, tortuosity, diffusion coefficients, and reaction kinetics. The results will enhance fundamental understanding and assist in optimizing porous media design for improved reaction efficiency and overall process performance.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
Thermophoresis is a phenomenon where small particles suspended in a gas phase migrate in response to temperature gradients, typically moving from hot to cooler regions. In the context of iron particle combustion this effect can have a significant influence on particle trajectories, local concentrations, and heat transfer, all of which may alter ignition delay and combustion characteristics.
The goal of this project is to investigate how thermophoresis affects the ignition and combustion behavior of micron-sized iron particles suspended in a reactive gaseous environment. The study will begin with a review of relevant thermophoretic theory and incorporate it into an existing particle combustion model.
The investigation will be carried out using carrier-phase direct numerical simulations (CP-DNS), with a focus on isolating and quantifying the impact of thermophoresis. A parametric analysis will be conducted to explore how variations in thermophoretic strength influence ignition delay, burn rate, and overall combustion behavior. The results aim to enhance understanding of iron particle combustion under realistic conditions and inform the design of particle-based energy systems.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
The laminar flame speed is an important quantity in the mathematical modeling of flame reaction kinetics and in turbulent combustion modeling. It is defined as the speed of a planar flame front propagating into an unburned, combustible mixture along a one-dimensional flow coordinate. This strong simplification of the flow situation allows to derive models which relate reaction kinetics, diffusion and heat conduction in ordinary differential equations.
Measuring laminar flame speeds using flow devices (burners) requires to balance the flow speed versus the burning velocity. This is easily achieved in a Bunsen burner where the flame angle is in direct relation to the flame and unburned gas velocity. Unfortunately, manifold effects cause deviations from the ideal one-dimensional flow situation: Heat losses through radiation, non-negligible flame thickness and heat losses to the burner pipe. Furthermore, small deviations in experimental conditions add a noise to the reproducibility of experiments. Accurate measurements of laminar flame speed which match modeling results require highly sophisticated, fragile setups.
The aim of this project will be to develop an automated calibration method, in order to account for the experimental non-idealities, using pattern matching and correlation strategies from the machine learning method portfolio. The work will be conducted with experimentalists which contribute the training data (images of the flames) and corresponding sets of process conditions.
The project requires fundamental knowledge in programming in Python and Large Language Model (LLM) prompt writing. Fundamental knowledge in combustion science is appreciated but not required.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
The thermodynamic properties like heat capacity, enthalpy of formation, and entropy of molecules are essential quantities for any type of modeling of thermal and chemical processes. These properties depend on the geometry of the molecule and the intramolecular force fields. Similar applies to computation of transport properties like gas viscosity or diffusion coefficients which depend on mass and geometry of the molecules, and the intermolecular force fields. The calculation of these quantities requires sophisticated quantum-mechanical simulations and high computational resources, and it fails for molecules and atoms with many electronic states (like e.g. metals).
The aim of this project is to develop and test a prediction model for thermodynamic properties based on machine learning techniques. The training set is provided by existing databases for the C-O-H-Si system.
The project requires fundamental knowledge in programming in Python and Large Language Model (LLM) prompt writing. Fundamental knowledge in combustion science is appreciated but not required.
If interested, please contact M.Sc. Leon Bernau (leon.bernau@uni-due.de)
If interested, please contact Prof. Dr. Andreas Kempf (andreas.kempf@uni-due.de)
If interested, please contact Prof. Dr.Andreas Kempf (andreas.kempf@uni-due.de)
If interested, please contact Prof. Dr.Andreas Kempf(andreas.kempf@uni-due.de)
If interested, please contact Prof. Dr.Andreas Kempf (andreas.kempf@uni-due.de)
The present project aims at the development of computer models and simulation tools for modeling the combustion of Ammonia. The project will involve the improvement of a numerical framework, the running of simulation, and the development of better reaction and transport models. Simulations will be conducted using the group’s in-house code PsiPhi, applying Servers and Clusters to provide the necessary computational power.
The project would necessitate a strong background in mathematics, numerics and fluid dynamics. If interested, please contact Prof. Dr.Andreas Kempf (andreas.kempf@uni-due.de) or Parsa Ghofrani (parsa.ghofrani@uni-due.de).
The present project aims to develop sub-models for modeling the formation of nanoparticles in the process of iron powder combustion, based on detailed large eddy simulations. The project will involve improving a numerical framework, running simulations, and developing better reaction and transport models.
The project would necessitate a strong background in mathematics, numerics and fluid dynamics. If interested, please contact Prof. Dr.Andreas Kempf (andreas.kempf@uni-due.de) or Parsa Ghofrani (parsa.ghofrani@uni-due.de).
In this project, the concept of beam bending that is similar to the schlieren or background oriented schlieren (BOS) methods is to be utilised for the design of an optical detection system. The changes in refractive index within a reacting gas (flame) will bend light rays. The deflection of the light rays can be measured either by imaging a background pattern behind the combusting flow, or by aligning a light emitter and detector on opposite sides of the volume of interest. In the first phase, a weak laser pointer (low enough energy to prevent damage to the sensor) and a surveillance camera will be positioned around a flame for measurements to test the concept. Additionally, the existing BOS method of the group will be utilised to measure deflections on a background pattern behind the flame using a second camera. The two methods should be tested in terms of applicability and sensitivity.
The candidate must have a good grasp of MATLAB and the capability to perform optical experiments in the lab. Additionally, C programming experience is considered advantageous. The results of this work can potentially be published in a peer-reviewed journal with international recognition.
For further information please contact(khadijeh.mohri@uni-due.de)
The mixing of hydrogen gas with air or other gases is of great importance for energy- and process-applications but also for studies on hydrogen safety. Research on laser-diagnostics has lead to elegant non-intrusive techniques for analyzing mixing, based on tracer molecules. These techniques have been established for gases with Schmidt- or Prandtl-numbers near unity, but the very high diffusivity of hydrogen may limit the suitability of such diagnostics. This Master’s or Bachelor’s project aims to conduct Direct Numerical Simulations (DNS) of hydrogen mixing in air, including an acetone tracer that would be applied to an experiment. The aim of the project is to establish the conditions, length and time-scales where an (organic) tracer is suitable for studying hydrogen mixing, and where it is not. Simulations will be conducted using the group’s in-house code PsiPhi, applying Servers and Clusters to provide the necessary computational power.
If interested, please contact Prof. Dr. Andreas Kempf (andreas.kempf@uni-due.de)
If interested, please contact Prof. Dr. Andreas Kempf (andreas.kempf@uni-due.de)
If interested, please contact Prof. Dr. Andreas Kempf (andreas.kempf@uni-due.de) or Prof. Wlokas (i.wlokas@uni-due.de).
If interested, please contact Prof. Dr. Andreas Kempf (andreas.kempf@uni-due.de).
If interested, please contact Prof. Dr. Andreas Kempf (andreas.kempf@uni-due.de)
This project can be considered „blue sky research“ and multiple students can work on it in parallel. A general interest in computer programming and mathematical modeling is expected, a reasonable background in the relevant subjects is required.
If interested, please contact Prof. Dr. Andreas Kempf (andreas.kempf@uni-due.de)
If interested, please contact Prof. Dr. Andreas Kempf (andreas.kempf@uni-due.de)
The project is challenging and requires strong skills in the field of fluid dynamics, thermo dynamics, reaction kinetics and numerics (FORTRAN, Python). It will be only attempted with very good candidates or long term HiWis.
If interested, please contact Prof. Dr. Andreas Kempf (andreas.kempf@uni-due.de)
If interested, please contact Prof. Dr. Andreas Kempf (andreas.kempf@uni-due.de)
Project works
161108: Optimised Numerical Schemes for the Large-Eddy Simulation of Turbulent Combustion
Large-Eddy Simulation is a modern CFD technique for accurately predicting turbulent reacting flows by affordable computer simulation. The method has evolved over the last 15 years and is finally becoming available in commercial software programs and being used by industry leaders.
However, the method requires numerical schemes that combine high numerical accuracy with low numerical oscillation. Our group has used a hybrid approach with good success, combining accurate central differencing schemes for momentum transport with non-dispersive TVD schemes for scalar transport. Where this hybrid approach combines good accuracy with low dispersion, it can lead to inconsistencies when applied with certain combustion models based on the "Flames Surface Density" approach.
The present project will apply different combinations of available numerical schemes to different test cases, to eventually assess the overall error resulting from the schemes. Based on the findings, further transport schemes (e.g. (W)ENO) shall be implemented and tested, aiming to improve overall accuracy, reliability and robustness of the simulations.
Students interested in this project will require a strong background in fluid-mechanics and should ideally have some background in numerical techniques, programming, combustion and turbulence modelling.
Please contact Prof. Andreas Kempf ( andreas.kempf [at] uni-due.de ) for further information.