The use of hydrogen in gas turbines is promising to help with the energy transition since there are no carbon-based pollutants released during its combustion. However, the transition from fossil fuels to hydrogen in gas turbines involves significant challenges. Hydrogen flames exhibit higher flame speed, making burners more prone to flashback, which in turn can cause damage by propagating upstream into the premixing sections. To avoid this type of failure, it is necessary to understand the different mechanisms that can cause flame flashback in gas turbines and develop strategies to help reduce the flashback tendency of hydrogen flames. The FlameSheet™ burner, originally developed by Power Systems Manufacturing (PSM), is a two-stage burner that can be used with fuels having high content of hydrogen. The geometry of the main stage includes a 180-degree bend (U-bend). This feature, which abruptly changes the flow direction of the air-fuel mixture, creates a trapped vortex which helps in stabilizing the flame in the main stage. Such a design makes the burner more resistant to flame flashback.

In this thesis, a numerical study is done on an academic trapped vortex burner, inspired by the FlameSheet™ design using Ansys Fluent. Initially, a cold flow simulation of the flow inside the main stage of the burner was performed using RANS. This was followed by a parametric study (using RANS) of the cold flow patterns while varying several parameters like the shape of the inner liner tip, channel height ratio and the distance between the tip and outer wall of the U-bend. From these results, useful insights were gained by analyzing and comparing the velocity profiles in different parts of the U-bend. To gain more information from some of the relevant cases, large eddy simulations (LES) were used. With the LES results, two-point correlations and cumulative distribution functions of the velocity field were made to describe the cold flow behaviour in the main stage of the burner. These results confirmed that the tip shape and its distance from the U-bend are significant parameters since they affect the velocity profile in the boundary layers of the flow. This change in the velocity profile of its boundary layer can alter the flashback tendency of hydrogen flames in the burner. Along with these cold flow simulations, a reactive simulation was also performed with the original geometry of the trapped vortex burner to understand the flow features inside, in the presence of a flame. Super adiabatic flame temperature and non-uniform diffusion of hydrogen into the flame were observed at various points in the flame.

The focus of this thesis is the preliminary design optimization and performance analysis of non-cavitating centrifugal turbopumps using a 1D lumped-parameter method. Centrifugal pumps are critical components in many engineering applications due to their high efficiency in converting mechanical energy into fluid pressure. This study presents the development and validation of a computationally efficient model to design and predict the flow properties along different stations of a centrifugal pump. The methodology minimizes the reliance on expensive CFD simulations, making use of multiple design parameters to control the impeller, diffuser, volute, and exit cone configurations to ensure a robust and efficient performance.

A common challenge in turbopump design is the prediction and mitigation of cavitation, which occurs when the fluid static pressure falls below its vapor pressure. This leads to the formation and subsequent collapse of vapor bubbles, which can cause severe damage to pump components. Given the complex physics involved, this work makes use of reduced-order models to predict, and avoid cavitating regimes through design adjustments.

This thesis provides a comprehensive framework for optimizing the design of centrifugal turbopumps by integrating validated loss models and a reduced-order modeling approach. The proposed design methodology is applied to develop a centrifugal pump for the TU Delft ORCHID facility, demonstrating its feasibility and effectiveness. This research contributes a practical modeling tool for centrifugal turbopump optimization and performance analysis.

Diesel engines find application in a large number of sectors of modern industries, such as the automotive and maritime. The wide adoption of Diesel engines, along with the rise of environmental concerns created the needs for optimization of fuel efficiency and power output as well as minimization of exhaust gas emissions. These targets must be met while retaining the reliability of Diesel engines. One technique that is used to significantly enhance the reliability of Diesel engines is piston cooling. The working fluid is engine oil which is injected in the cylinder towards the lower side of the piston. The objective of the current study is to simulate the process of cooling with the aid of CFD. The numerical methodologies that are used in the current study are the Reynolds-Averaged Navier-Stokes (RANS) equations for the continuous phase, while for the disperse phase the modelling is conducted with the Lagrangian modelling. The numerical model which is built focuses on the effect that the mesh resolution has on the results of the simulation, especially on the disperse phase of the flow. It can be inferred that there is dependence between the quality of the results and the mesh resolution. It is shown that the grid has to be fine enough to produce quality results, but if there is excessive refinement, the quality of results drops because assumptions of the Lagrangian modelling are not satisfied. Following this phase of the study, a CFD model is built to simulate the phenomenon of droplet impingement on a solid wall. In this set of simulations, the goal is to compare the impingement model that is used in the software with published experimental results of droplet impingement. The results indicate agreement between the CFD model and the experimental process. In the final stage of the thesis, there is the numerical study of heat transfer involved in the process of spray impingement on a hot wall. A CFD model is built to compare the heat transfer which is predicted by the software model and results from relevant published experimental studies. It can be concluded that there is divergence between CFD and published experimental results, stemming from the modelling assumptions of the software. The results of the present thesis can be used in future studies of the company, either as a continuation of the spray cooling topic, or for novel studies such as gallery cooling.

This research contributes to addressing climate change challenges through the examination of hydrogen combustion. It investigates the flow dynamics within a simplified model of Ansaldo Energia's GT36 reheat combustor using Large Eddy Simulation (LES) at a high pressure of 20 bar, focusing on the autoignition flashback phenomena observed. More specifically, it explores predicting the apparition of this event through the use of a modularity-based clustering algorithm and its subsequent suppression by using water injection.

In the LES simulation, lean, premixed and high-pressure conditions are used, revealing an unsteady behaviour in the flame dynamics, i.e. an autoignition event in the mixing duct which repeats itself due to high-amplitude pressure waves and their subsequent convergence in the mixing duct. Using this simulation, time-series of multiple variables are acquired over the course of 8 flashback events. These variables are then pruned using a co-kurtosis PCA based dimensionality reduction technique. This method, by measuring the joint occurrence of outliers in the flow variables, identifies successfully which are the most important variables (temperature, density, pressure, velocity in the axial direction, mass fractions of HO2 and OH) that introduce a lasting change in the system and are potentially useful in finding a precursor to the flashback event.

The time-series of these variables are then introduced in the modularity-based clustering algorithm. This algorithm tessellates the phase space of the system and then transforms it into a graph, thus retaining the information about the dynamics of the system. Then, it clusters this graph using a metric called modularity. The method proves effective in finding a precursor for the autoignition event, resulting in an average prediction time for the 8 flashback events of 32.13 microseconds, which is over 50% of the time that the combustor is in the normal operating state. Furthermore, the algorithm performs very well in the number of false positives, and, due to a change made to the algorithm in this study, the number of true positives is increased to 100%. The algorithm is then put through several robustness tests, which include the use of different sampling locations, less features, unseen data and relying only on temperature and pressure information sampled at the combustor walls. Here, the algorithm retained its performance, with only a small decrease in the prediction time, demonstrating its potential towards its use in an online prediction scenario. Lastly, the level of fidelity of the LES simulation was increased by using the digital filter method to simulate turbulent fluctuations at the inlet and fully-developed velocity profile, where once again it was found that the algorithm retains a good prediction time.

For the second part of this research, a flashback event and its afferent prediction time were chosen to investigate the use of water injection and its potential at suppressing the flashback when the water is injected based on the prediction time. Here, following an empirical design approach, where the Sauter Mean Diameter (SMD), the mass flow, the diameter of the nozzle, the angles of the cone were varied, a preliminary design was sought. It was found that the SMD is highly influential towards the spread of the spray, with larger particles being preferred to their ability to retain their momentum and more quickly cover the mixing duct. In addition, large angles for the cone and a high injection velocity are again necessary for a good spread and a quick response time. An attempt was also made to place a set of sprays at the walls of the mixing duct, where it was found that the flashback event now takes place upstream of the spray due to aerodynamic blockage. The design process culminates in a setup where six spray are placed at the inlet of the mixing duct. In this setup, the spray is able to quickly cover the mixing duct and the flashback is suppressed, while also retaining an evaporation efficiency of 96.3%.

To meet the goals of the Paris agreement,sustainable aviation fuels, such as hydrogen, need be adopted on a large scale within the coming decades. Hydrogen-assisted combustion of kerosene is investigated as an intermediate step in the transition towards sustainable aviation. This study addresses the gaps in the current understanding of hydrogen blending into kerosene with respect to chemical kinetics and emission of NOx, CO and un burnt hydrocarbons.  A perfectly-stirred chemical reactor model is developed to study the fundamental effects of hydrogen blending into kerosene. Additionally, to focus on a more practical application the single-reactor model is extended to a chemical reactor network which represents a lab-scale lean premixed prevaporized combustor setup supporting multi-fuel combustion. Combustion characteristics and emission profiles are carefully studied for different hydrogen blending fractions using the developed models, as well as the reaction pathways in oxidation of hydrocarbons.  Hydrogen blending is observed to increase the reactivity of kerosene, mainly due to increased availability of OH and H radicals, as well as O to a lesser extent. For 20% of kerosene mass substituted with H2, ignition delay time at 1200 K and 1 bard ecreases by 55%. Peak laminar flame speed increases by up to a factor of 2.2,while peak adiabatic flame temperature rises by 74 K. In rich burning conditions, H2 additions remove hydrocarbon reaction loops, which causes a strong decrease in un burnt hydrocarbon emissions and potentially soot. CO emissions can be greatly reduced from H2 additions at very lean conditions as overall burning rate is increased, which is corresponding to a reduction in lean extinction limit. This limit is reduced in the chemical reactor network combustor model by ∆φ=0.09 already for substitution of kerosene mass by 20%.  For φ>0.6, dissociation of CO2 is enhanced from thermal effects and causes emission of CO to rise with hydrogen blending when normalized to carbon in fuel. This effect is reflected in the combustion efficiency, which is improved with H2 blending only for very lean conditions,φ<0.6. The reaction pathways of aromatics are modified with H2 addition,which results in a chemical inhibition of the mechanism responsible for thema jority of un burnt hydrocarbon emissions at very lean conditions. Furthermore,formation of NOx is increased via the thermal path since flame temperatures rise. The prompt NOx route becomes more efficient due to promotion of CH formation, while the N2O and NNH paths are enhanced from increased radical availability. Despite the increased emission of NOx at constant φ, blending H2under constant combustion power and combustor mass flow leads to decreasedflame temperatures and reduces total emissions of NOx, CO and unburnt hydrocarbons.

There is a need to examine the potential of alternative methods such as Variational Multiscale Method (VMM) for the ability to improve the efficiency of combustion simulations. However the computational expense of a LES run of a 3D turbulent non-premixed combustion problem is high due to the large variety of length and time scales in turbulent non-premixed combustion. Secondly, VMM is a sophisticated model and the models used to simulate turbulent non-premixed combustion are complex. And finally, the results of a full 3D turbulent non-premixed combustion simulation are difficult to interpret due to their complexity. Therefore the objectives as follows: Firstly to define a simplified model problem which has the important aspects of turbulent non-premixed combustion for comparing numerical methods. Secondly, to propose and to investigate the potential of a framework based on VMM formulation for non-premixed combustion computations. The results of the simplified model problem in a VMM framework are compared to the results when using a standard Smagorinsky model. Results show that in general the VMM model with the dynamic subscale assumption outperforms the Smagorinsky model when looking at the mean and RMS value of the velocity, mixture fraction and progress variable.

The energy transition from fossil fuels to renewable energy sources is crucial for climate resilience. In particular, hydrogen combustion in gas turbine combustors is expected to play an essential role in the energy transition. Hence, a fuel-flexible gas turbine combustor operating with a wide range of natural gas and hydrogen fuels is necessary. Low Swirl Burner(LSB), developed at Lawrence Berkeley National Laboratory (LBNL), is one of the promising solutions for fuel-flexible gas turbine combustors. However, hydrogen combustion in LSB at lean conditions is accompanied by several challenges, especially flame flashback. Flashback results in damage and shutdown of gas turbines. Recent experimental studies performed on LSB revealed the occurrence of flame flashback at a high volume percentage of hydrogen, but the exact cause of the flashback in LSB is yet to be discovered.

In this study, numerical modelling of low swirl premixed burner with turbulent flow and high hydrogen content is performed. RANS simulations are performed initially for the non-reacting flow and are validated with the experimental measurements. Flamelet combustion model with Zimont Turbulent Flame speed correlation is employed to model the reacting flow. The flame front position obtained from the Zimont correlation closely agrees with the experiments.
Further, Algebraic Flame Surface Wrinkling (AFSW) correlation is compared with the Zimont correlation for hydrogen-enriched flames. Based on the comparison, it is concluded that the AFSW correlation did not lead to a significant difference. However, both the correlations failed to model the flashback process. Therefore, the potential mechanism responsible for the flame flashback in LSB is examined. It has been found that none of the standard flashback mechanisms is accountable for the flame flashback in LSB. Furthermore, the effect of different flame shapes reported in the experiments is evaluated. It has been observed that ’M’ shape flame has more propensity to flashback than ’V’ shape flame. Then, a shear layer and turbulent wakes in the premix nozzle of the LSB were noticed. Hence, an increase in flame speed due to the local enrichment of thermo-diffusive unstable hydrogen flame and all the aforementioned effects might lead to a sudden flashback in LSB.

This thesis covers the experimental characterisation of a multi-nozzle research combustor capable of achieving flameless combustion (FC). The combustor is operated with methane as fuel and air as the oxidiser, under an overall lean equivalence ratio ranging from 0.6-0.9. The work focused on designing and applying different experimental techniques under varying operating conditions. Particle Image Velocimetry (PIV) is used to characterise the flow field for reacting and non-reacting conditions. Local gas temperature and gas composition measurements are performed to characterise the formation of NOx in the combustor. Emission data and local gas temperatures are further reported as a function of the equivalence ratio, thermal power input, and oxidiser's oxygen concentration. PIV revealed the flow field structures to be very similar between non-reacting and reacting flow conditions; the location of the recirculation zones remained unaffected. Significant differences exist in the axial flow velocities and turbulence levels in the central recirculation zone (CRZ). Radial temperature measurements near the reaction zone revealed the temperature profiles to flatten for a decrease in equivalence ratio, indicative of distributed combustion. Gas composition measurements are performed near the exhaust of the combustor. The measurements showed the combustor to perform very well from an emission standpoint. Measurements showed the NOx on the centreline to be ultra-low (<10ppm) for all cases investigated. The N2O and NNH mechanisms are identified as the major contributors to the production of NOx and not the thermal pathway, as the gas temperature is consistently below the thermal NOx threshold of 1800K. A reduction in the equivalence ratio and a reduction in the oxygen concentration both reduce NOx emissions. The measured levels of CO and CH4 were negligible near the exhaust for all operating conditions investigated, confirming that combustion is fully completed at this location. Measurements of CO and CH4 near the reaction zone indicate distributed combustion, where a reduction in the equivalence ratio below 0.7 resulted in elongation and widening of the reaction zone. Oxygen measurements near the exhaust and close to the reaction zone indicate the entrainment of cooling air into the combustion chamber, which would locally influence the equivalence ratio. It is recommended to redesign the exhaust duct connection to eliminate the entrainment of cooling air.

In reacting flows, detailed chemistry computations are usually avoided precomputing the thermochemical quantities as functions of a reduced set of variables such as the Flamelet Generated Manifold (FGM) approach[34]. Although it mitigates the calculations of detailed chemical mechanism, the memory requirement associated to store the lookup table and retrieve the information during numerical simulations is usually large (order of Gigabytes). Thereby, extending the FGM approach in order to include other conditions requires to add
other independent variables which will inevitable lead to increase the size of the lookup table. This will generate that Large Eddy Simulations (LES) cannot be performed such as is the case of the Diluted Air FGM (DA-FGM) approach developed by Xu Huang[9], limiting the simulations to Reynolds Average Navier-Stokes (RANS) approach. In this master thesis, the goal is to use Artificial Intelligence (AI)-Machine Learning (ML) techniques in order to reduce substantially the computational cost of storing lookup tables. In order to achieve this, the Artificial Neural Network (ANN) technique is used. First, a 4D FGM lookup table for hydrogen flames is simplified using the aforementioned AI technique. Then, this technique is used to replace a 6D lookup table generated using the DA-FGM approach. The accuracy and stability of the models provided by ANNs is measured by statistical indicators, providing high accurate and stable AI models. Finally, in the middle of this project, unexpected issues regarding the 4D lookup table were encountered, which lead to recreate the 4D lookup table. After studying carefully how the 4D lookup table was created, a new 4D lookup table is generated, providing excellent results and improving the AI models obtained.

During a severe accident in a water­-cooled nuclear power plant, large quantities of hydrogen can be generated in the reactor core. The hydrogen mixed with air presents a potential risk of combustion when the hydrogen concentration reaches flammability limits. If combustion occurs, pressure loads can damage safety systems and even compromise the structural integrity of the nuclear reactor walls. Thus, predicting the pressure loads is an important safety issue to ensure the reliability of critical structures in the event of a severe accident.

Computational Fluid Dynamics (CFD) codes can be used as numerical tools to assess the risks of hy­drogen combustion and predict detailed transients of the pressure loads. The scenarios simulated in nuclear safety engineering can be analyzed with low fidelity models based on reduction assumptions of the physical phenomena. Specifically, the flame gradient models are chosen to model the com­bustion phenomena based on a turbulent flame speed to which the combustion propagates. Different approaches to compute the turbulent flame speed are implemented and compared in this study, in­cluding the Turbulent Flame Closure (TFC) and the Extended Turbulent Flame Closure (ETFC). New models are proposed based on new experimental correlations specifically derived for lean mixtures of hydrogen. Other required libraries as flame radius calculation, radiative heat transfer properties determination, and axisymmetric Adaptive Mesh Refinement are implemented and evaluated. A study on the influence of the ignition source term is also carried out.

The implementation of the solver and the models is verified with the analytical solution of a planar one­-dimensional premixed flame moving into frozen turbulence. Moreover, three different experimen­tal cases, relevant to nuclear safety, are selected to perform the validation of the solver. The first one is a spherical combustion chamber which was experimentally tested with lean hydrogen mixtures and controlled measured homogeneous isotropic turbulence. The second one is the Thermal-hydraulics Hydrogen Aerosol and Iodine (THAI) facility, which is a larger vessel where effects as buoyancy be­ come more relevant. Finally, the third one is a flame acceleration enclosure called ENACCEF­2, where obstacles are placed along a cylindrical tube to promote flame acceleration and the transition from de­flagration to detonation.

Numerical results are presented in terms of the flame front location, the flame front velocity, and the pressure rise. The sensitivity of the results to mesh resolution, time discretization, and initial turbu­lence levels is assessed. The implemented combustion models are simulated with multiple turbulence models, analyzing the results and the influence on one each other. In general, the proposed mod­els represent an approach that provides good predictions in both flame development and pressure dynamics, being more robust than other models previously available

In the future, fossil fuels will be replaced by renewable sources. Hydrogen seems to be a promising energy carrier (fuel). A lot of research has been conducted on the combustion of hydrogen, but the effect of the high diffusivity of hydrogen compared to other species (differential diffusion) was often not included in simulations of turbulent flames because turbulent mixing is expected to suppress strong influence of differential diffusion. Nevertheless effects of differential diffusion have been reported in experiments. The main purpose of this thesis is to include the effect of differential diffusion in a CFD model of turbulent non-premixed hydrogen flames and to validate the models with available experimental data. The validation data set used in this study is a non-premixed turbulent jet flame of hydrogen diluted with nitrogen on 50/50 volume ratio. This flame is interesting because the role of differential diffusion in this flame has been a matter of discussion in the literature.

A review is given of previous work on turbulent non-premixed hydrogen flames and effects of differential diffusion. It is concluded that the Flamelet Generated Manifold (FGM) model and the Transported Probability Density Function (PDF) model are both promising turbulent combustion models for these flames. Next, a new comparative study is made of several turbulence and combustion models using Reynolds-averaged Navier Stokes simulations in Ansys Fluent, and focusing on FGM as turbulent combustion model. It is concluded that the standard k-e with a correction as suggested by Pope is the best choice for the turbulence model. The ANSYS Fluent implementation of FGM does not have the option to include differential diffusion. Therefore it is added in a separate way. To do this, flamelets are created with the help of CHEM1D, a tool developed by TU Eindhoven. These flamelets are combined to an FGM table also including the effect of turbulence via a PDF and this table is imported into Fluent with the help of a user-defined function overwriting the default Fluent FGM table.

In order to clearly see the effects of differential diffusion, simulations with and without differential diffusion are made. Good agreement is obtained between experimental results and the numerical simulation for the models without differential diffusion. This confirms that turbulence can suppress strong influence of differential diffusion. In experiments effects of differential diffusion have been observed at the base of the flame, close to the burner nozzle. The model simulations with differential diffusion included, provide a slightly more accurate prediction of the mean temperature close to the nozzle but still large discrepancies remain.

Although significant strides have been made with regards to increasing the fuel economy of commercial passenger aircraft, the reduction of the environmental footprint remains of the utmost importance. The flow development over an aero-engine spinner will affect the velocity distribution over the fan of the engine and this affects the estimation of the losses generated in the fan and consequently the performance of the engine itself. Previous experimental studies have shown the existence of spiral vortices formed due to boundary layer instabilities over a rotating cone under axial and non-axial inflow. To study the effects of hub-corner separation in detail, it is first important to check the efficacy of existing commercial numerical simulation tools in the prediction of this boundary layer transition development of flow over a rotating nose-cone.
The objective of the thesis was thus to simulate using URANS and LES, the formation of the counter-rotating vortices over a rotating cone. To this end. ANSYS DesignModeler and ICEM CFD were used to generate a meshed computational domain and ANSYS CFX was used as the solver.
With axis-symmetric cones considered as an idealised geometry for the spinner of an aircraft engine, a 15° half-angle cone with a diameter of 47 mm was chosen. The tip of this cone was blunted by a factor of 1/100^th of the cone-diameter. The domain diameter was set to 10 times the cone diameter to reduce the impact of the flow over the side walls. Two structured hexahedral meshes were generated.
The BSL EARSM was chosen for the URANS simulation and the WALE Model for the LES run. To check if the simulations were able to capture the vortices, the footprint left behind by them were visualised using the non-dimensionalised instantaneous wall shear values. 
The magnitude of the inlet flow velocity given was 2.46 m/s. A 2° incidence is given to the flow for the non-axial case. The effect of mesh refinement was studied for the axial case using both meshes and the finer mesh was used for the non-axial case. The axial case was studied using both the URANS and LES runs, while the non-axial case was studied using only the LES run.
The variation of local Reynolds number (defined using boundary layer edge velocity) with local rotation ratio (defined using local radius, boundary layer edge velocity, and rotation velocity of the cone) was also studied and compared with available experimental data. The wall parallel velocity profiles over the length of the cone were also studied. Using the contours of the wall parallel velocity, the momentum mixing in the boundary layer due to the vortices was visualised and compared with the results from the experiments.
To characterise the spatial development of the spiral vortices, a critical point was defined using the wall friction coefficient. This critical point was then compared to the one obtained through the experiments. The number of counter-rotating vortices formed over the length of the cone was also checked and the trend between the simulations and experiments were compared.