3D Printing b4e334: How Korean EIA Agencies Build Wind Field Simulation Models

• 1
  Key takeaways
• 2
  Why Korean EIA Teams Print Wind Field Models
• 3
  Choosing Scale, Tolerance, and Surface Finish
• 4
  When Printing Is the Wrong Answer
• 5
  Material Choices and What They Cost You in the Data
• 6
  Checks That Keep the Model Review-Ready
• 7
  Step-by-Step: Building a Printable Wind Field Model
• 8
  Which Process for Which Part of the Model

Korean environmental impact assessment work runs in dense, hilly terrain. Coastal reclamation sites, apartment clusters on slopes, and industrial parks near mountains all create airflow that a pure CFD run struggles to validate. Agencies therefore build a physical model and put it in a wind tunnel or an open-jet rig, then compare measured velocity and turbulence against the simulation.

The model is not a display piece. It is a measurement reference. Every surface the anemometer passes over has to be dimensionally honest, or the comparison is worthless. That is the reason 3D printing b4e334 took over from hand-built foam and plywood terrain: a printed tile reproduces the same slope twice, and the tenth copy matches the first.

Printing also lets a team iterate. When a reviewer asks for a taller stack on one building or a revised bridge pier, you re-export the tile and reprint it overnight instead of rebuilding the whole board. For a project with three review rounds, that is the difference between one model and four.

The limit is not the printer. It is how you handle the joints, the surface texture, and the scale. Those three decide whether the results hold up in a review meeting.

Start from the tunnel working section, not from the building. If the test section is 1,000 mm wide and the site is 3 km across, you are locked near 1:3000. Most Korean district-scale EIA models land between 1:500 and 1:2000. Below 1:500 the model gets heavy, expensive, and hard to mount.

Tolerance matters most at mating features, not on the terrain skin. Terrain slope can drift 0.3 mm over a tile without changing the flow. A dowel hole that is 0.2 mm oversize will let tiles shift, and the seam becomes an artificial ridge that shows up in the velocity data. Hold mating faces and locating holes at ±0.005 mm where the geometry allows.

Surface finish sets the effective roughness. As-printed SLA at Ra 1.6–3.2 μm is acceptable for far-field terrain. Anything within about 20 mm of a probe or inside a street canyon should be smoothed to Ra 0.8–1.6 μm. Vapor smoothing, bead blasting, or a light sanding pass all work. Do not paint the surface; coating thickness varies and you lose the roughness reference.

One more rule: keep the vertical exaggeration at 1:1 unless the tunnel team explicitly asks otherwise. Exaggerated terrain looks convincing in photos and produces misleading separation behavior.

Printing wins on shape. It loses on fits, flatness over long spans, and anything that has to hold a seal or a bearing. If a part in your model has to bolt to a tunnel wall and stay air-tight, machine it. A printed flange will creep under bolt load and open a leak path that ruins the pressure reference.

Long base plates are the other trap. A 1,500 mm printed rail will bow, and the bow transfers straight into your terrain reference plane. We mill base rails and adapter plates at ±0.005 mm and use them as the datum for everything printed on top.

There is also a cost crossover. One-off complex terrain geometry is cheapest printed. A simple flat plate with holes is cheaper milled once you count print time, support removal, and flatness rework. For runs above roughly 20 identical small brackets, CNC usually lands close to printed cost with better repeatability.

If your model mixes both, that is normal. Terrain and facades print; rails, dowels, bushings, and tunnel interfaces get machined. We run both under one roof so the datums stay consistent.

• 1
  PrintTerrain, facades, roof plant, complex one-off shapes
• 2
  MachineBase plates, rails, dowel bushings, tunnel flanges, anything sealing
• 3
  Do not print then drillPrinted holes drift; bore them in a machined insert instead

Standard SLA resin is the default for terrain. It holds fine slope detail, sands easily, and stays dimensionally stable indoors. It is brittle, so protect tile edges during transport. Keep the model out of direct sunlight; UV exposure yellows the surface and slowly changes the skin.

For larger tiles where weight matters, SLS nylon gives a lighter part with good toughness. The trade-off is a rougher surface, roughly Ra 3.2–6.3 μm as built. If probes pass close to that surface, plan a bead blast or vapor smooth step. Nylon also absorbs moisture, so let tiles stabilize in the test room for 24 hours before the run.

FDM is fine for far-field terrain and for internal ribs that never see the flow. Do not use it in a street canyon at 1:500; the layer ridges are the same order as the building detail you are trying to measure.

MJF sits between SLS and SLA: good detail, reasonably smooth, and cheaper than SLA at volume. For a 12-tile district model, MJF is often the practical pick. Metals are rarely used for the model itself, but machined aluminum is the right choice for base plates and tunnel adapters.

Before the instrumented run, verify three things: seam step, tile flatness, and locating-hole fit. A dial indicator on each seam takes ten minutes and catches the errors that reviewers ask about first. Record the values; a documented check sheet is worth more than a verbal assurance.

Run a smoke or tuft pass at low speed. It shows seam-induced separation and any unintended blockage from support material left inside a canyon. Fix geometry issues before you spend tunnel time on instrumentation.

Keep the model and the CAD revision tied together. When one tile is reprinted, update the revision number on the base plate. Mixed revisions are a common source of confusion in multi-round EIA reviews.

We inspect 100% of parts before shipment, including mating faces and locating holes, and can supply dimensional reports on request. The tolerance we hold on machined interfaces is ±0.005 mm, with a 99.99% qualification rate across production.