Molten Metal Casting (3D)

A Blender-modelled foundry: tap the furnace, pour the mould and watch the cut face freeze with real heat conduction, Chvorinov's rule and shrinkage porosity.

Switch to 2D

Presets

Playback

Casting shape

Metal and mould

View

Camera

Layers

Quality

Model check

The science

What is modelled

A sand-casting mould is cut open and the cut face shows the live temperature field. The mould is a 0.8 m x 0.5 m section (per unit depth, i.e. a long bar) with a pouring cup, a sprue, a plate casting and a riser. Metal runs in at the Torricelli rate, then heat flows out through the mould while the metal freezes. Everything is computed on a 160 x 100 grid of 5 mm cells with the enthalpy method.

Equations

  • Filling: head H from the cup to the plate, flow per unit depth Q = Cd w √(2 g H) through the choke w = min(sprue, plate thickness) with Cd = 0.8; fill time tf = Acavity / Q. The cells are poured in breadth-first order from the cup (down the sprue, along the plate, up the riser) at the pouring temperature.
  • Heat conduction: ∂H/∂t = ∇·(k ∇T) where H is the volumetric enthalpy. For the metal T = T0 + H/(ρ c) below the melting point, T = Tm while ρ L of latent heat is absorbed (liquid fraction fl = (H - Hs)/(ρ L)) and T = Tm + (H - Hs - ρ L)/(ρ c) above it. Face conductivities use the harmonic mean; across the metal-mould interface 1/keff = 1/(2km) + 1/(2ks) + 1/(hi Δx). Explicit time stepping at 90 % of the stability limit.
  • Chvorinov's rule: ts = B (V/A)2 with B = (π/4) [ρ L' / (Tm - T0)]2 / (k ρ c)mould and L' = L + c ΔTsuperheat. The modulus M = V/A of each part (casting, riser, sprue) is measured from the grid as area divided by the length of its surface that touches the mould.
  • Riser rule: the riser must freeze after the casting, triser > 1.25 tcasting, so that liquid can still feed the shrinkage. Any liquid pocket that loses its connection to the open riser or cup surface is marked as shrinkage porosity, with area β times the pocket area (β = volumetric solidification shrinkage of the metal).
  • Exact reference: for a solid layer growing against a semi-infinite mould with perfect contact the shell is s = 2 λ √(αs t) where λ solves (Tm - T0) c/L = λ √π eλ² [ ks√αm / (km√αs) + erf λ ] (Neumann).

Checks run in the page

  • The solver against the exact Neumann two-region solution (bronze in sand).
  • Freezing time of two slab thicknesses: the exponent must be 2, and the result is compared with the closed-form B M2.
  • Energy accounting on the real mould layout with an insulated top.
  • The pouring-basin drain ODE integrated with RK4 against its closed form.

Honest simplifications

Each metal freezes at a single temperature (the pure-metal limit; real alloys freeze over a range and form a mushy zone), with equal solid and liquid heat capacity and no density change when it freezes, so the shrinkage porosity is an estimate from the isolated-pocket area, not a pressure-driven feeding model. A riser that is attached to a casting also loses heat through its neck once the casting beneath it is solid, so its measured freezing time can fall below the stand-alone Chvorinov prediction; the prediction is what the foundry rule uses. Fluid flow, convection in the melt, mould-gas escape, radiation and the cooling of the metal during the short fill are not modelled (the poured cells simply enter at the pouring temperature). The mould is a thermally homogeneous solid with constant properties; the contact coefficient is a single number for the whole interface. Material properties are typical handbook values. The mould section is two-dimensional (a long bar), so the modulus of the casting is area over perimeter, not volume over area of a finite block. Real castings in aluminium, bronze or iron need specialised software and foundry experience; this is a teaching model. The foundry hall, ladle and furnace are modelled in Blender for illustration and the animation is scripted.

What to try

  • Make the riser narrower than 40 mm: it freezes before the plate, the green ring jumps to the casting and a dark shrinkage pocket appears.
  • Halve the plate thickness: the freezing time drops by about four (the square in Chvorinov's rule).
  • Switch the mould from green sand to a steel die: the shell grows many times faster and the effusivity readout jumps by an order of magnitude.
  • Pick an air-gap contact: the first seconds are slower and the casting freezes later than the formula predicts.
  • Pour hotter: superheat adds to the effective latent heat and delays freezing.
Loading model...
Loading Blender model...
Temperature at four probe points
Solid shell thickness against the exact Neumann solution
Freezing time measured / Chvorinov prediction
Filling and solid fraction
Made with 🟠 Blender · Three.js