Resumo:
This work investigates two-dimensional flows with subcritical Reynolds number around a circular
cylinder with rough surface and kept heated at a constant temperature; the analyses include
mechanisms of mixed convection heat transfer. The numerical approach discretizes the vorticity
field and the heat by using particles, which characterizes a purely lagrangian description.The
work contributes to the scientific literature in two ways: i) implementation of an accelerator
algorithm to incorporate LES-type turbulence modeling for the temperature field, and ii) proposition
of a surface roughness model for the thermal boundary layer. The second contribution
is made possible through LES-type turbulence modeling, allowing that roughness effects for
different average surface roughness heights interfere on behavior of pressure, drag and lift coefficients,
of the boundary layer separation angle, and heat transfer from body surface to fluid
domain. Those parameters are of great interest for engineering design involving various devices
subject to external flows, where there is heat exchange with the fluid domain. Simulations were
performed for a Reynolds number of Re = 100,000, Prandtl number of Pr = 0.71, and average
surface roughness heights of ε/D = 0.001 and 0.007. The results showed that mixed convection
heat transfer impacts the fluid dynamic loads in a complex manner, increasing the drag
coefficient and heat transfer via Nusselt number behavior. For the latter dimensionless number,
the value obtained for the simulation in the smoothed cylinder case showed some discrepancy
with the expected experimental date, indicating the need for a more detailed energy balance on
the thermal boundary layer. However, by implementing roughness effects, an increase in the
Nusselt number was observed in proportion to the experimental results, indicating that the new
two-dimensional roughness model is sensitive and provides physically consistent results. The
three characteristic kinds of particles interaction consume a high computational time, which
also required the use of parallel processing in OpenMP-Fortran.
