Conditional Source-term Estimation (CSE) methods are used to obtain chemical closure in turbulent combustion simulation. A Laminar Flamelet Decomposition (LFD) and then a Trajectory Generated Low-Dimensional Manifold (TGLDM) method are combined with CSE in Reynolds-Averaged Navier Stokes (RANS) simulation of non-premixed autoigniting jets. Despite the scatter observed in the experimental data, the predictions of ignition delay from both methods agree reasonably well with the measurements. The discrepancy between predictions of these two methods can be attributed to different ways of generating libraries that contain information of detailed chemical mechanism. The CSE-TGLDM method is recommended for its seemingly better performance and its ability to transition from autoignition to combustion. The effects of fuel composition and injection parameters on ignition delay are studied using the CSE-TGLDM method. The CSE-TGLDM method is then applied in Large Eddy Simulation of a non-premixed, piloted jet flame, Sandia Flame D. The adiabatic CSE-TGLDM method is extended to include radiation by introducing a variable enthalpy defect to parameterize TGLDM manifolds. The results are compared to the adiabatic computation and the experimental data. The prediction of NO formation is improved, though the predictions of temperature and major products show no significant difference from the adiabatic computation due to the weak radiation of the flame. The scalar fields are then extracted and used to predict the mean spectral radiation intensities of the flame. Finally, the application of CSE in turbulent premixed combustion is explored. A product-based progress variable is chosen for conditioning. Presumed Probability Density Function (PDF) models for the progress variable are studied. A modified version of a laminar flame-based PDF model is proposed, which best captures the distribution of the conditional variable among all PDFs under study. A priori tests are performed with the CSE and presumed PDF models. Reaction rates of turbulent premixed flames are closed and compared to the DNS data. The results are promising, suggesting that chemical closure can be achieved in premixed combustion using the CSE method.

Conditional Source-term Estimation (CSE) methods are used to obtain chemical closure in turbulent combustion simulation. A Laminar Flamelet Decomposition (LFD) and then a Trajectory Generated Low-Dimensional Manifold (TGLDM) method are combined with CSE in Reynolds-Averaged Navier Stokes (RANS) simulation of non-premixed autoigniting jets. Despite the scatter observed in the experimental data, the predictions of ignition delay from both methods agree reasonably well with the measurements. The discrepancy between predictions of these two methods can be attributed to different ways of generating libraries that contain information of detailed chemical mechanism. The CSE-TGLDM method is recommended for its seemingly better performance and its ability to transition from autoignition to combustion. The effects of fuel composition and injection parameters on ignition delay are studied using the CSE-TGLDM method. The CSE-TGLDM method is then applied in Large Eddy Simulation of a non-premixed, piloted jet flame, Sandia Flame D. The adiabatic CSE-TGLDM method is extended to include radiation by introducing a variable enthalpy defect to parameterize TGLDM manifolds. The results are compared to the adiabatic computation and the experimental data. The prediction of NO formation is improved, though the predictions of temperature and major products show no significant difference from the adiabatic computation due to the weak radiation of the flame. The scalar fields are then extracted and used to predict the mean spectral radiation intensities of the flame. Finally, the application of CSE in turbulent premixed combustion is explored. A product-based progress variable is chosen for conditioning. Presumed Probability Density Function (PDF) models for the progress variable are studied. A modified version of a laminar flame-based PDF model is proposed, which best captures the distribution of the conditional variable among all PDFs under study. A priori tests are performed with the CSE and presumed PDF models. Reaction rates of turbulent premixed flames are closed and compared to the DNS data. The results are promising, suggesting that chemical closure can be achieved in premixed combustion using the CSE method.

Conditional Source-term Estimation (CSE) methods are used to obtain chemical closure in turbulent combustion simulation. A Laminar Flamelet Decomposition (LFD) and then a Trajectory Generated Low-Dimensional Manifold (TGLDM) method are combined with CSE in Reynolds-Averaged Navier Stokes (RANS) simulation of non-premixed autoigniting jets. Despite the scatter observed in the experimental data, the predictions of ignition delay from both methods agree reasonably well with the measurements. The discrepancy between predictions of these two methods can be attributed to different ways of generating libraries that contain information of detailed chemical mechanism. The CSE-TGLDM method is recommended for its seemingly better performance and its ability to transition from autoignition to combustion. The effects of fuel composition and injection parameters on ignition delay are studied using the CSE-TGLDM method. The CSE-TGLDM method is then applied in Large Eddy Simulation of a non-premixed, piloted jet flame, Sandia Flame D. The adiabatic CSE-TGLDM method is extended to include radiation by introducing a variable enthalpy defect to parameterize TGLDM manifolds. The results are compared to the adiabatic computation and the experimental data. The prediction of NO formation is improved, though the predictions of temperature and major products show no significant difference from the adiabatic computation due to the weak radiation of the flame. The scalar fields are then extracted and used to predict the mean spectral radiation intensities of the flame. Finally, the application of CSE in turbulent premixed combustion is explored. A product-based progress variable is chosen for conditioning. Presumed Probability Density Function (PDF) models for the progress variable are studied. A modified version of a laminar flame-based PDF model is proposed, which best captures the distribution of the conditional variable among all PDFs under study. A priori tests are performed with the CSE and presumed PDF models. Reaction rates of turbulent premixed flames are closed and compared to the DNS data. The results are promising, suggesting that chemical closure can be achieved in premixed combustion using the CSE method. / Applied Science, Faculty of / Mechanical Engineering, Department of / Graduate

Combustion science must necessarily go through a deep process of innovation, as only improving the energy efficiency and the fuel flexibility it will be possible to mitigate the impact of the anthropogenic activities on the climate and the environment. Because of the strong relation that is observed in chemically reacting flows between the fluid-dynamic conditions and the chemical kinetics, the use of Computational Fluid Dynamics (CFD) simulations with detailed kinetic mechanisms represents the best tool to optimize and develop novel combustion systems. In fact, while the CFD provides for the possibility to retrieve information that cannot be extracted by using experimental means (such as the turbulence-chemistry interaction and the local straining rates) and it avoids the costs associated to the scale-up process from laboratory scale experiments, the use of detailed kinetic mechanisms offers the possibility to correctly describe process conditions which are relevant from an industrial point of view (i.e. in which the chemical and mixing time scales are comparable), as well as to predict the formation of complex chemical species, such as the pollutants. Nevertheless, the use of detailed kinetic mechanisms in numerical simulations adds a considerable number of differential equations to be solved (because of the large number of species which are taken into account), and therefore increases the computational complexity of the CFD model. Thus, Machine Learning (ML) algorithms and Reduced-Order Models (ROMs) can be effectively included in the numerical description of chemically reacting flows. In fact, they can be used either to reduce the computational cost associated to the large number of equations in CFD simulations carried out with detailed chemistry, or to leverage the detailed information which can be found in massive, high-fidelity, data obtained from Direct Numerical Simulations (DNS), for model development and validation. In this Thesis, unsupervised and supervised learning algorithms were employed to design a novel adaptive-chemistry approach: the Sample-Partitioning Adaptive Reduced Chemistry (SPARC). This framework can be used to reduce the computational effort required by detailed CFD simulations thanks to a kinetic reduction accomplished in light of the local conditions of the thermochemical field. Several machine-learning algorithms, such as the Principal Component Analysis (PCA), the Local Principal Component Analysis (LPCA), and Artificial Neural Networks (ANNs) were coupled with the Direct Relation Graph with Error Propagation (DRGEP), a graph-based tool for the automatic reduction of kinetic mechanisms. The aforementioned algorithms were compared to achieve the optimal formulation of the adaptive approach, such that the best performances, in terms of accuracy and computational speed-up with respect to the CFD simulation carried out with detailed kinetics, could be obtained. Finally, PCA-based algorithms were proposed and tested to perform feature extraction and local feature selection from high-fidelity data, which were obtained by means of a DNS of a n-heptane jet reacting in air. The PCA, as well as two formulations of LPCA, and the Procrustes analysis were employed and compared with the aim to extract the main features of the turbulent reacting jet in an unsupervised fashion (i.e. to perform data mining tasks), as well as to aid the formulation of local optimized ROMs. All the codes employed to perform the unsupervised and supervised machine learning tasks in the current work were also included in an open-source Python framework, called OpenMORe, designed to perform reduction, clustering and data analysis, and specifically conceived for reacting flows. In fact, although many open-source Python software are already available, they often cannot be adapted to the user’s specific needs, unlike OpenMORe. In addition, many features such as the PCA-based clustering algorithm, or the local feature selection via PCA, are not yet available on any commercial or open-source software, to the best of the author’s knowledge. / Doctorat en Sciences de l'ingénieur et technologie / This thesis is submitted to the Université Libre de Bruxelles (ULB) and to the Politecnico di Milano for the degree of philosophy doctor. This doctoral work has been performed at the Université Libre de Bruxelles, École polytechnique de Bruxelles, Aero-Thermo-Mechanics Laboratory, Bruxelles, Belgium with Professor Alessandro Parente and at the Politecnico di Milano, CRECK Modelling Lab, Department of Chemistry, Materials and Chemical Engineering, Milan, Italy with Professor Alberto Cuoci. / info:eu-repo/semantics/nonPublished

The mechanisms governing the formation and destruction of soot in turbulent combustion are intimately coupled to thermal radiation due to the strong dependence of sooting processes and radiative loss on temperature. Detailed computational fluid dynamics (CFD) predictions of the radiative and soot output from turbulent non-premixed flames are normally performed by parabolic algorithms. However, the modelling of combustion systems, such as furnaces and unwanted enclosure fires, often require a fully elliptic description of the flow field and its related physical phenomena. Thus, this thesis investigates the intimate coupling between radiative energy exchange and the mechanisms governing soot formation and destruction within a three-dimensional, general curvilinear CFD code. Thermal radiation is modelled by the discrete transfer radiation model (DTRM). Special emphasis is given to approximate solutions to the radiative transfer equation encompassing various models for the radiative properties of gases and soot. A new algorithm is presented, entitled the differential total absorptivity (DTA) solution, which, unlike alternative solutions, incorporates the source temperature dependence of absorption. Additionally, a weighted sum of gray gases (WSGG) solution is described which includes the treatment of gray boundaries. Whilst the DTA solution is particularly recommended for systems comprising large temperature differences, the WSGG solution is deemed most appropriate for numerical simulation of lower temperature diffusion flames, due to its significant time advantage. The coupling between radiative loss and soot concentration is investigated via a multiple laminar flamelet concept applied within the CFD simulation of confined turbulent diffusion flames burning methane in air at 1 and 3 atm. Flamelet families are employed relating individual sooting mechanisms to the level of radiative loss, which is evaluated by the DTRM formulated for emitting-absorbing mixtures of soot, C02 and H20. Combustion heat release is described by an eddy break-up concept linked to the k-c turbulence model, whilst temperature is evaluated from the solved enthalpy field. Detailed comparisons between prediction and experiment for the critical properties of mixture fraction, temperature and soot volume fraction demonstrate the effectiveness of this novel, coupled strategy within an elliptic flow field calculation.

LLean premixed combustion systems, attractive for low NOx performance, are inherently susceptible to thermo-acoustic instabilities - the interaction between unsteady heat release and excited acoustic wave effects. In the present work, a hybrid, coupled Large Eddy Simulation (LES) CFD approach is described, combining the computational efficiency of incompressible reacting LES with acoustic wave effects captured via an acoustic network model. A flamelet approach with an algebraic Flame Surface Density (FSD) combustion model was used. The ORACLES experiments - a perfectly premixed flame stabilised in a 3D sudden expansion - are used for validation. Simulations of the inert flow agree very well with experimental data, reproducing the measured amplitude and distribution of turbulent fluctuations as well as capturing the asymmetric mean flow. With reaction the measured data exhibit a plane wave acoustic mode at 50Hz. The influence of this plane wave must be incorporated into the LES calculation. Thus, a new approach to sensitise the incompressible LES CFD to acoustic waves is adopted. First an acoustic network model of the experimental geometry is analysed to predict the amplitude of the 50Hz mode just before the flame zone. This is then used to introduce a coherent plane wave at the LES inlet plane at the appropriate amplitude, unlike previous LES studies, which have adopted a "guess and adjust" approach. Incompressible LES predictions of this forced flow then show good agreement with measurements of mean and turbulent velocity, as well as for flame shape, with a considerable improvement relative to unforced simulations. To capitalise on the unsteady flame dynamics provided by LES, simulations with varying forcing amplitude were conducted and analysed. Amplitude dependent Flame Transfer Functions (FTFs) were extracted and fed into an acoustic network model. This allowed prediction of the stable/unstable nature of the flame at each forcing amplitude. An amplitude at which the flame changed from unstable to stable would be an indication that this coupled approach was capable of predicting a limit cycle behaviour. With the current simple FSD combustion model almost all cases studied showed a stable flame. Predictions showed considerable sensitivity to the value chosen for the combustion model parameter but specially to the acoustic geometric configuration and boundary conditions assumed showing evidence of limit cycle behaviour for some combinations. Nevertheless, further work is required to improve both combustion model and the accuracy of acoustic configuration and boundary condition specification.

Experiments and simulations were performed to examine the complex processes that occur in Rayleigh�Taylor driven mixing. A water channel facility was used to examine a buoyancy-driven Rayleigh�Taylor mixing layer. Measurements of �uctuating den- sity statistics and the molecular mixing parameter were made for Pr = 7 (hot/cold water) and Sc 103 (salt/fresh water) cases. For the hot/cold water case, a high- resolution thermocouple was used to measure instantaneous temperature values that were related to the density �eld via an equation of state. For the Sc 103 case, the degree of molecular mixing was measured by monitoring a di�usion-limited chemical reaction between the two �uid streams. The degree of molecular mixing was quanti- �ed by developing a new mathematical relationship between the amount of chemical product formed and the density variance 02. Comparisons between the Sc = 7 and Sc 103 cases are used to elucidate the dependence of on the Schmidt number. To further examine the turbulent mixing processes, a direct numerical simu- lation (DNS) model of the Sc = 7 water channel experiment was constructed to provide statistics that could not be experimentally measured. To determine the key physical mechanisms that in�uence the growth of turbulent Rayleigh�Taylor mixing layers, the budgets of the exact mean mass fraction em1, turbulent kinetic energy fE00, turbulent kinetic energy dissipation rate e 00, mass fraction variance gm002 1 , and mass fraction variance dissipation rate f 00 equations were examined. The budgets of the unclosed turbulent transport equations were used to quantitatively assess the relative magnitudes of di�erent production, dissipation, transport, and mixing processes. Finally, three-equation (fE00-e 00-gm002 1 ) and four-equation (fE00-e 00-gm002 1 -f 00) turbulent mixing models were developed and calibrated to predict the degree of molecular mix- ing within a Rayleigh�Taylor mixing layer. The DNS data sets were used to assess the validity of and calibrate the turbulent viscosity, gradient-di�usion, and scale- similarity closures a priori. The modeled transport equations were implemented in a one-dimensional numerical simulation code and were shown to accurately reproduce the experimental and DNS results a posteriori. The calibrated model parameters from the Sc = 7 case were used as the starting point for determining the appropri- ate model constants for the mass fraction variance gm002 1 transport equation for the Sc 103 case.

<p dir="ltr">Two-phase flows have many important industrial applications, including in nuclear reactor coolant systems. Due to the density difference between gas and liquid, buoyancy effects can have a significant impact on two-phase parameters, and thus the orientation of the flow is a major factor impacting system performance and safety. Despite this, most two-phase experiments and models have focused on vertical upwards flows. While recent work has greatly expanded the knowledge of horizontal and inclined flows, gaps still exist, especially regarding the relative motion between the two-phases. There are very few experiments in literature that measured both local liquid and gas velocity profiles, which are key to understanding and modeling the relative velocity. This work will seek to address this gap, first by performing experiments in various non-vertical orientations, specifically measuring the local gas and liquid velocities by means of a local four-sensor conductivity probe and Pitot-static probe, respectively. Twelve fully developed horizontal bubbly flow conditions are measured and presented in the current work, with the trends in relative velocity analyzed. It is found that the relative velocity is negative throughout the pipe cross-section when in a horizontal orientation. The relative velocity becomes more negative with increasing void fraction and remains negative as the void fraction approaches zero. With this newly established database, a model is then proposed to predict the relative velocity in horizontal bubbly flows, accounting for bubble wake interactions. The model is able to predict the void-weighted area-average velocity within 10%, and the local relative velocity with an average absolute percent difference of 15%, which is considered adequate given the experimental uncertainties. Area-averaged and drift-velocity correlations are also developed based on this model. There is no predictive explanation given in literature for the negative relative velocity measured in horizontal bubbly flows, which this model enables. To supplement this model, a new method for estimating the local distribution of the dispersed phase in bubbly two-phase flows is proposed. This is based on the geometric packing of the bubbles and reflects the random nature of two-phase flows and bubble interactions, and preliminary qualitative comparisons show promise compared to experimental data. Finally, multiphase computational fluid dynamics simulations are improved by implementing the new relative velocity model. The peak void fraction value is improved by on average 250%, with the area-averaged relative velocity improved by about 100%.</p>