How you know the number is right
A solver that always returns an answer is easy. A solver you can stake a design review on is not. NovaThermal is built around one discipline: every result carries the evidence that it converged, conserved energy, and stayed inside its qualification envelope — or it fails closed and tells you why.
Converged, or it doesn't count
Every solve is driven to the platform's maximum-resolution mesh and the result is reported alongside its mesh-convergence delta. A gradient that still moves as the mesh refines is flagged, not published. The number you read is the number the physics has stopped changing.
Enthalpy, not a smeared heat capacity
Solidification and melting use a high-fidelity mushy-zone enthalpy method. Latent heat is carried as energy crossing the freezing range — conserved exactly — rather than by inflating specific heat over a guessed temperature band, which quietly loses or gains energy right at the interface you care about.
You can't spec a wall you can't build
Pressure-containing geometry is held to a 1.25 mm minimum wall, tied to A-Basis allowables at 700 psi and crewed pressure-systems practice (SSP-51721 lineage). The tool refuses below-qualification geometry so a promising result can't quietly assume hardware that would never fly.
Every result traces back
A result records the exact model version, solver build, mesh, boundary conditions, and convergence status that produced it. Re-open the run and re-derive it. That receipt is what a design review asks for — and what a spreadsheet can never give you.
The methodology isn't abstract. Watch a eutectic freezing front build its microstructure in directional lamellar growth (sweep the transverse and axial cross-sections in 3D), or replay a Bridgman solidification animated from genuine solver output.