I am structuring a conjugate heat-transfer model for a liquid-cooled cold plate made by laser powder bed fusion in copper alloy. The internal channels are as-built, so the measured inputs available for correlation are flow rate, pressure drop, inlet/outlet temperature, heat load, and the external contact-patch temperature.
converge the hydraulic model against measured pressure drop before tuning heat-transfer assumptions;
apply an effective channel roughness only within the measured range;
compare both a uniform heat-flux patch and a mapped power-density boundary;
correlate delta-T and thermal resistance at multiple flow points;
keep material conductivity temperature-dependent and run a mesh-independence check near the walls.
For SimScale CHT, would you calibrate wall roughness first and then freeze it for the thermal runs, or solve hydraulic and thermal calibration together? I want to avoid a model that matches one operating point but has no predictive value.
You have laid out an exceptionally robust validation sequence. LPBF copper cold plates are notoriously tricky to simulate accurately because the as-built internal channel stochasticity (powder agglomeration and layer stepping) heavily drives both pressure drop and convective heat transfer.
To answer your question directly: You should absolutely calibrate the hydraulic model (wall roughness) first, freeze it, and only then move on to the thermal calibration.
Here is why solving them together is a trap for this specific application:
1. Avoiding the “Curve-Fitting” Trap If you try to calibrate hydraulic and thermal parameters simultaneously, you create an ill-posed mathematical problem. You might accidentally offset a poor roughness assumption by artificially tweaking your thermal contact resistance or material conductivity. While this “curve-fitting” will perfectly match your single calibration point, it will destroy the model’s predictive value. As soon as you change the flow rate, the artificial balance will break, and your results will diverge from reality.
2. The Isothermal Baseline Your best bet is to run an isothermal, flow-only simulation first (Incompressible solver). Sweep across your multiple measured flow points and adjust the equivalent sand-grain roughness until your pressure drop (ΔP) curve matches the experimental data. LPBF as-built roughness is typically much higher than standard drawn pipes, so this step is critical.
3. The y+ vs. Roughness Caveat A quick warning when applying heavy roughness in SimScale: Pay close attention to your near-wall mesh (prism layers). If you are using a turbulence model like k−ω SST, the physical roughness height you input should generally not exceed the distance to the first cell center. If the roughness elements are “taller” than your first boundary layer cell, the wall functions can become unstable or produce unphysical heat transfer coefficients.
Once you have your hydraulic baseline locked in and your mesh validated against the roughness height, switch over to the Conjugate Heat Transfer solver, freeze that roughness value, and proceed with steps 3 through 5 of your sequence.
