[edwardsh]

Hello!

I'm somewhat new to turbulence modelling and just need some hints about which model I should be using. I've been finding it a bit hard to sort through all the jargon on my own.

I'm modelling a parallel flow field from a fuel cell in steady state. The channel cross section is quite small (0.9 mm x 1 mm) with a larger header cross section (1.8 mm x 1 mm). The inlet mass flow rate is 4.325e-5 kg/s of air, with a 70 kPa back pressure. I've upped modified the density slightly based on a temperature of 70C and humidity of 100%, but the model is just simulating fluid flow and nothing else.

Based on my hand calculation for Reynold's number, it should be firmly in the laminar region (Re_D = 1511). However, what I've found during my mesh indepdence study, is that convergence stalls at high mesh densities (~3 million elements total). From a closer look, this appears to be due to flow separation as the fluid turns the corner between the header and the parallel channels.

My understanding is that flow separation is inherently unsteady, so I need a RANS model to be able to capture this behaviour. My question is WHICH ONE? The flow through all of the parallel channels quickly becomes laminar again, with some having barely any flow through them at all. Is there a turbulence model that is good at both these turbulent regions and also the super low Re flow at other parts of the model?

As a side note, do I need to worry about Y+? I've just been using a uniform cartesian mesh through the entire model up until now, with the goal of finding the mesh converged value for the global pressure drop, then optimizing the mesh from there by adding inflation layers and increasing the size in less sensitive areas.

Sorry for the novel and thanks for any advice!