International Journal of Numerical Methods for Heat & Fluid Flow

Purpose – The purpose of this paper is to study the effect of a computational grid in computational fluid dynamics-based mathematical modeling, focusing on but not limiting the attention to industrial-scale boilers.

Design/methodology/approach – A full boiler model is used to show the difficulties related to judging iteration and discretization errors in boiler modeling. Then, a single jet is studied in detail to determine the proper degree of local grid refinement required in the vicinity of jets in the full boiler model. Both a nonreactive axisymmetric jet exhausting into a quiescent atmosphere and a reactive jet exhausting into a crossfiow are studied.

Findings – Over two million computational cells are required for the grid-independent solution for a single jet. Local grid refinement is shown to be a good option for improving the results consistently without an excessive increase in the number of computational cells. Using relatively coarse grids of tetrahedral cells with a finite-volume-based solver may cause serious errors in results, typically by overpredicting the jet spreading rate and underpredicting the mean axial centerline velocity. Relatively coarse grids of hexahedral cells are less prone to error in a case where a jet exhausts into a quiescent atmosphere. However, their performance deteriorates when a crossfiow is introduced. As assumed, the differences in the predicted reaction rate and species concentrations are significant in the reactive case. It is confirmed that the standard k-? model tends to overpredict the axisymmetric jet spreading rate. The estimated inlet turbulence intensity is not among the most critical factors in modeling. Estimations of the axisymmetric jet centerline velocity from the analytical correlation may not coincide with the modeling results.

Practical implications – The error caused by the computational grid may easily dominate the errors caused by simplifying models used in industrial-scale boiler modeling (turbulence, combustion, radiative heat transfer, etc.).

Originality/value – The present study deals with grid independency issues in industrial-scale boiler modeling in a systematic and profound manner.