[umop-3pisdn]

I'm running a steady-state simulation of an ice hockey rink using a segregated flow solver and the intensity-viscosity ratio turbulence model w/ default params. The thermal loads present are spectators (boundary condition modeled as heat flux and species flux on the bleacher surfaces), hockey players (modeled as thermal and species volumetric generation in a region spanning 1-1.5m above the ice), and lighting/LCD screens (boundary condition modeled as heat flux on surfaces). All other surfaces (walls, roof, ice) are defined as isothermal at known temperatures. Air supply diffusers are modeled using a velocity inlet boundary condition. The exhaust air grilles are modeled using a split flow outlet boundary condition. I've initialized the simulation with a temperature gradient based on measurements from similar hockey rinks.

No matter what i do, i observe EXTREMELY high velocities near the thermal flux BC surfaces (bleachers, light fixtures, LCD screens). The velocity field is initially at rest, and the boundary-normal velocity inlet BCs representing the HVAC diffusers range in magnitude from 1-5 m/s.... but i'm seeing velocities from 80-150 m/s near the flux BC surfaces. I've let the simulation iterate through 500+ iterations and the velocity profiles appear to be settling at these high rates. the residuals for Tke, Tdr, and momentum are rather flat (declining.... but VERY slowly), but still on the order of +1 - +2.

I've tried this with all the flux BCs converted to adiabatic walls. I STILL see these hurricane-level velocities near these surfaces. I am boggled as to how/why these extraordinarily high velocities are generated with no apparent energy sources to cause them.

i've tried different supply/exhaust air boundary approaches, including mass flow inlet + pressure outlets (resulted in recirculation between outlets), and mass flow inlet + mass flow outlet + 1 pressure outlet (converged fine, but the velocity profile was physically unrealistic, with very low velocities exiting the diffusers).

I am at my wits' end trying to figure this out. Any ideas?

[ping]

test it without ideal gas and gravity and see what happens, assuming you want convective flow modelled, to see if the model has more basic issues

try constant density with gravity on and the boussinesq approximation method instead of ideal gas since it is quite accurate for small density variations but remember to set the thermal expansion coefficient correctly for air

the coupled solver is also worth using with convective flow and also be aware that such flow is inherently unstable and might need the unsteady solver