Mesoscopic Numerical Methods for Reactive Flows: Lattice Boltzmann Method and Beyond

Reactive flows are ubiquitous in several energy systems: internal combustion engines, industrial burners, gas turbine combustors. Numerical modeling of reactive flows is a key tool for the development of such systems. However, computational combustion is a challenging task per se. It generally includes different coupled physical and chemical processes. A single model can come to deal with simultaneous processes: turbulent mixing, multi-phase fluid-dynamics, radiative heat transfer, and chemical kinetics. It is required not only of mathematically representing these processes and coupling them to each other, but also of being numerical efficient. In some applications, the numerical model needs to be able to deal with different length scales. For instance, a continuum approach to reactive flows in porous media burners is not adequate: processes occurring at the pore-scale are not taken into account properly. It is therefore fundamental to have numerical methods able to capture phenomena at the microscopic scales and incorporate the effects in the macroscopic scale. The lattice Boltzmann method (LBM), a relatively new numerical method in computational fluid-dynamics (CFD), summarizes the requirements of numerical efficiency and potential to relate micro-and macro-scale. However, despite these features and the recent developments, application of LBM to combustion problems is limited and hence further improvements are required. In this thesis, we explore the suitability of LBM for combustion problems and extend its capabilities. The first key-issue in modeling reactive flows is represented by the fact that the model has to be able to handle the significant density and temperature changes that are tipically encountered in combustion. A recently proposed LBM model for compressible thermal flows is extended to simulate reactive flows at the low Mach number regime. This thermal model is coupled with the mass conservation equations of the chemical species. Also in this case a model able to deal with compressibility effects is derived. To this purpose, we propose a new scheme for solving the reaction-diffusion equations of chemical species where compressibility is accounted for by simply modifying the equilibrium distribution function and the relaxation frequency of models already available in the literature. This extension enables one to apply LBM to a wide range of combustion phenomena, which were not properly adressed so far. The effectiveness of this approach is proved by simulating combustion of hydrogen/air mixtures in a mesoscale channel. Validation against reference numerical solution in the continuum limit are also presented. An adequate treatment of thermal radiation is important to develop a mathematical model of combustion systems. In fact, combustion incorporates also radiation process, which tends to plays a significant role if high temperatures (and solid opaque particles) are involved. In the thesis a LBM model for radiation is presented. The scheme is derived from the radiative transfer equation for a participating medium, assuming isotropic scattering and radiative equilibrium condition. The azimuthal angle is discretized according to the lattice velocities on the computational plane, whereas an additional component of the discrete velocity normal to the plane is introduced to discretize the polar angle. The radiative LBM is used to solve a two-dimensional square enclosure bechmark problem. Validation of the model is carried out by investigating the effects of the spatial and angular discretizations and extinction coefficient on the solution. To this purpose, LBM results are compared against reference solutions obtained by means of standard Finite Volume Method (FVM). Extensive error analysis and the order of convergence of the scheme are also reported in the thesis. In order to extend the capabilities of LBM and make it more efficient in the simulation of reactive flows, in this thesis a new formulation is presented, referred to as Link-wise Artificial Compressibility Method (LW-ACM). The Artificial Compressibility Method (ACM) is (link-wise) formulated by a finite set of discrete directions (links) on a regular Cartesian grid, in analogy with LBM. The main advantage is the possibility of exploiting well established technologies originally developed for LBM and classical computational fluid dynamics, with special emphasis on finite differences, at the cost of minor changes. For instance, wall boundaries not aligned with the background Cartesian mesh can be taken into account by tracing the intersections of each link with the wall (analogously to LBM technology). LW-ACM requires no high-order moments beyond hydrodynamics (often referred to as ghost moments) and no kinetic expansion. Like finite difference schemes, only standard Taylor expansion is needed for analyzing consistency. Preliminary efforts towards optimal implementations have shown that LW-ACM is capable of similar computational speed as optimized (BGK-) LBM. In addition, the memory demand is significantly smaller than (BGK-) LBM. Two- and three-dimensional benchmarks are investigated, and an extensive comparative study between solutions obtained through FVM. Numerical evidences suggest that LW-ACM represents an excellent alternative in terms of simplicity, stability and accuracy.