3D Modelling and Simulation of Reactive Fluidized Beds for Conversion of Biomass with Discrete Element Method

The use of biomass as a CO2–neutral renewable energy source gains more importance due to the decreasing resources of fossil fuels and their impact on the global warming. The thermochemical conversion of biomass in fluidized beds offers an economic and sustainable contribution to the global energy supply. Although the fluidized bed has reached a commercial status since many decades ago, its hydrodynamic behaviour is not completely understood. The availability of detail experimental information from real facilities is extremely difficult because the lack of accessibility, the measurement costs and the associated inevitable reduction in production. The numerical simulation provides an effective complement to the costly measurements. This requires besides the calculation of a gas-solid flow, an accurate description of particle–particle/wall collisions. Furthermore, kinetic models for pyrolysis, homogenous reactions, heterogeneous reactions and the related heat and mass transfer processes should be considered. Basically, there are two different methods for the representation of the gas–solid flow, viz. Euler–Euler and Euler–Lagrange models. The solid phase is treated as a continuum in the Euler–Euler model, while each particle trajectory is determined in the Euler–Lagrange model. In the Euler–Euler approach, the single particle-particle or particle-wall collision can be considered using additional assumptions. In the Euler–Lagrange approach, the particle-particle/wall collisions can be stochastically modeled or deterministically detected. The aim of this study is to develop a 3D program for the numerical simulation of biomass conversion in fluidized beds. The particle–particle/wall and gas–solid interactions are modeled by tracking all individual particles. For this purpose, the deterministic Euler–Lagrange/discrete element method (DEM) is applied and further developed. The fluid–particle interaction is studied using a new procedure, known as the offset method. The proposed method is highly precise in determining the interaction values, thus improving the simulation accuracy up to an order of magnitude. In this work, an additional grid, so-called particle grid, in which the physical values of solid phase is computed, is introduced. The suggested procedure allows the variation of the fluid grid resolution independent of the particle size and consequently improves the calculation accuracy. The collision detection between particle–particle/wall is performed with the aid of the particle search grid method. The use of the particle search grid method enhances the efficiency of collision detection between collision partners. The improved Euler–Lagrange/DEM model is validated towards the measurements obtained from a cold quasi–2D fluidized bed. The results suggest that the extended Euler–Lagrange/DEM model can predict accurately the motion of particles and the gas bubble expansion in the bed. The received results from the DEM model are also compared with other numerical approaches, namely the Euler-Euler and stochastic Euler–Lagrange models. Compared to measurements, the results show that the Euler–Euler model underestimates the bubble sizes and the bed expansions, while the stochastic Euler–Lagrange model reaches faster the maximum bed expansions. The efficiency and accuracy of the Euler–Lagrange/DEM model is investigated in detail. Parameter studies are carried out, in which stiffness coefficient, fluid time step and processor number are varied for different particle numbers and diameters. The obtained results are compared with the measurements in order to derive the optimum parameters for Euler–Lagrange/DEM simulations. The results suggest that the application of higher stiffness coefficients (more than 10^5 N/m) improves the simulation accuracy slightly, however, the average computing time increases exponentially. For time intervals larger than five milliseconds, the results show that the average computation time is independent of applied fluid time step, while the simulation accuracy decreases extremely by increasing the size of fluid time step. The use of fluid time steps smaller than five milliseconds leads to negligible improvements in the simulation accuracy, but to exponential rise in the average computing time. The parallel calculation accelerates the Euler–Lagrange/DEM simulation if the critical number of domain decomposition is not reached. Exceeding this number, the performance is not anymore proportional to the number of processors and the computational time increases again. The critical number of domain decomposition depends on particle numbers. An increase in solid contents results in a shift of critical decomposition number to higher numbers of CPUs. The local concentrations of solid and gaseous species, the local gas and particle temperatures, the local heat release and heat transfer rates can also be calculated with the developed program. In combination with the simulation of the gas–solid flow, it is possible to model the biomass conversion in the fluidized bed. Three series of warm simulations in a quasi–2D fluidized bed model are performed, viz. combustion with fuel gas without and with inert sand particles as well as combustion with solid fuel (a mixture of inert sand and pine wood particles). The received results realise the coupling of the Euler–Lagrange/DEM model with chemical reaction mechanism. The extended Euler–Lagrange/DEM model under the consideration of thermochemical reaction model is able to simulate, by the same token, the conversion of other solid fuels such as coal in fluidized beds.