Abstract
A swirling pulverized coal flame is computationally investigated. A Eulerian–Lagrangian formulation is used to describe the two-phase flow. Turbulence is modelled within a RANS (Reynolds averaged numerical simulation) framework. Four turbulence viscosity- (TV) based models, namely the standard k-ε model, realizable k-ε model, renormalization group theory k-ε model, and the shear stress transport k-ω model are used. In addition, a Reynolds stress transport model (RSM) is employed. The models are assessed by comparing the predicted velocity fields with the measurements of other authors. In terms of overall average values, the agreement of the predictions to the measurements is observed to be within the range 20–40%. A better performance of the RSM compared to the TV models is observed, with a nearly twice as better overall agreement to the experiments, particularly for the swirl velocity. In the second part of the investigation, the resolution of the discrete particle phase in modelling the turbulent particle dispersion (TPD) and particle size distribution (SD) is investigated. Using the discrete random walk model for the TPD, it is shown that even five random walks are sufficient for an accuracy that is quite high, with a less than 1% mean deviation from the solution obtained by thirty random walks. The approximation of the measured SD is determined by a continuous Rosin–Rammler distribution function, and inaccuracies that can occur in its subsequent discretization are demonstrated and discussed. An investigation on the resolution of the SD by discrete particle size classes (SC) indicates that 12 SC are required for an accuracy with a less than 1% mean deviation from the solution with 18 SC. Although these numbers may not necessarily be claimed to be sufficiently universal, they may serve as guidance, at least for SD with similar characteristics.
Highlights
Combustion has been used as the main process for power and heat generation from fossil fuels for many decades [1]
Turbulence was modelled by different RANS models, including the turbulence viscosity- (TV)-based standard k-ε model [46] (S-KE), realizable k-ε model [47] (R-KE), RNG-KE, and stress transport (SST) models as well as the Reynolds stress model (RSM)
It was observed that the velocity fields predicted by the RSM are closer to the measured values compared to the TV models, especially for the swirl velocity, where the overall accuracy of the RSM accuracy is twice as good as that from the TV models
Summary
Received: 28 September 2021 Accepted: 28 October 2021 Published: 4 November 2021
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have