I'm working on conjugate heat transfer simulations where the key output is the peak temperature of a solid component cooled by surrounding airflow/coolant. The challenge is that the predicted hotspot temperature is sensitive to how well the mesh resolves the solid-fluid interface and the near-wall region.
I wanted to share the approach I've settled on and hear how others here handle it:
- Monitor a single scalar — peak solid temperature — rather than judging convergence visually.
- Run progressively refined meshes (coarse → medium → fine) and plot peak temperature against cell count.
- Prioritize the boundary layer at the interface — adding inflation/prism layers and checking the y+ target for the turbulence model usually affects the result far more than refining the bulk mesh.
- Identify the plateau — the converged mesh is the coarsest one where further refinement shifts peak temperature by less than ~1–2%.
A couple of questions for the community:
- For thin fluid gaps (e.g. closely spaced components or channels), what inflation-layer setup has given you reliable interface temperatures without exploding cell count?
- For transient runs, how are you balancing timestep against mesh fidelity to keep solve times manageable?
Interested in how others approach the accuracy-vs-cost balance for thermal CHT cases.
