Hex meshing automation for complex CAD geometry

Hex meshing is widely understood to produce more accurate FEA results than tetrahedral meshing for structural problems — fewer elements, better stiffness matrix conditioning, lower solve time. Every engineer knows this. But the complaint 'hex meshing is its biggest challenge' and 'geometry editing for complex brick hex meshing is bit complicated' has been consistent across every preprocessor for twenty years. The reason it hasn't been solved inside the big tools is structural: hex meshing requires decomposing arbitrary geometry into mappable regions, which is a hard computational geometry problem that doesn't parallelize neatly and produces failures that are difficult to surface to the user in a meaningful way. Vendors ship tet meshing because it always produces a result, even if the result is suboptimal. Hex meshing failures just leave engineers stuck.

The specific gap is in the decomposition step. Engineers working on cast or forged metal parts — engine brackets, suspension components, pressure vessel flanges — need to define sweep paths and blocking topology manually in current tools, which requires expert knowledge of the geometry and hours of cleanup work. 'Manual meshing is very difficult for complex modelling like for example in composite analysis' and 'geometry creation is not easy in case errors or mismatches occur due to the internal mesher' describe the same root failure: the link between CAD topology and hex block topology is invisible and manual.

What's buildable now that wasn't viable five years ago is using mesh-to-geometry feature recognition to suggest valid decomposition strategies automatically — showing the engineer three or four blocking options ranked by expected element quality, letting them pick one, then generating the hex mesh. The engineer isn't removed from the loop; they're given a starting point instead of a blank canvas. This is meaningfully different from what auto-hex meshers currently do, which is either produce an all-hex mesh with poor quality in concave regions or fail silently.

The business case is straightforward: a structural analyst who spends two days creating a good hex mesh for a complex casting, and then has to redo it every time the CAD changes, is the direct buyer. The cost per meshing iteration is measurable in engineering hours. This recurs every design cycle.

What to build

Build a geometry decomposition tool that reads STEP or Parasolid files of solid bodies, identifies valid sweep directions and blocking topologies using feature recognition, presents ranked hex meshing strategies with preview quality metrics, and exports the confirmed mesh to Nastran BDF or Abaqus INP format.

Where to start

Start with axisymmetric and near-axisymmetric geometries — pump housings, motor end caps, bearing races — where sweep decomposition is well-defined and automatable, and where the analyst community is concentrated enough to find early adopters through NAFEMS forums and SAE technical sessions.

The hard part

The hardest technical trade-off is that fully automatic hex decomposition still fails on genuinely complex geometry — you will need to decide early whether to constrain the tool to a geometry class (axisymmetric parts, thin-walled brackets) where automation succeeds reliably, or ship an interactive tool that just reduces manual effort, which is a harder value proposition to communicate.

How it makes money

Per-seat annual license at $3,000–5,000/year, targeting small simulation consultancies and Tier 1 supplier teams that run more than ten structural programs simultaneously and can't justify a full preprocessor seat for every analyst.