Sand-casting, the most widely used metal-forming route by tonnage, is classified by mold material and bonding chemistry rather than by alloy, and the four practical families on a 2026 shop floor are green-sand, resin-bonded (no-bake and shell), lost-foam, and 3D-printed binder-jetted molds [S3]. The Aluminum Association AA 18 covers sand and permanent-mold aluminum castings as parallel specifications, and the FIFTEENTH EDITION reaffirms sand casting as a mainstream process for all types of aluminum applications [S4].
For a working engineer, the choice between these mold types is driven by dimensional tolerance, pattern cost, alloy reactivity, and lot size, not by tradition [S1]. A spec-first classification starts with the binder and the mold-making method, then works outward to pattern material, gating, and shakeout.
Green-Sand Molds: Clay-Bonded, High-Volume Ferrous Workhorse
Green-sand molds use a mixture of silica sand, bentonite clay (typically 6–10% by mass), water (3–5%), and coal/cereal additives, compacted around a pattern at ambient temperature and used in the wet state within hours of mixing [S3]. The term "green" refers to moisture content, not cure state, and the mold is not fired before pour. The Springer entry on sand-casting technique confirms the classic two-flask sequence: lower half filled and rammed, rolled 180°, upper half built on top, parting made, and both halves combined around a separately formed core [S3].
Green-sand is the highest-volume sand process because the sand mold is reusable after shakeout and reclamation, pattern costs are lowest (wood patterns are acceptable for runs as short as a few dozen parts), and cycle times for a manually rammed match-plate line are 8–15 minutes per mold. The trade-off is surface finish (Ra typically 6–25 µm), dimensional scatter (±0.8–1.5 mm on small castings), and the inability to hold under-cuts or fine internal passages without a separate core. AA 18 mechanical-property and dimensional references apply to the resulting castings regardless of whether they are poured in green-sand or a bonded system, and the standard treats both as equivalent aluminum-alloy routes [S4].
Resin-Bonded Sand Molds: No-Bake, Shell, and Warm-Box Sub-Families
Resin-bonded molds replace the clay-water bond with a chemically cured organic binder, which hardens the sand mass into a rigid shape that holds tighter tolerances, finer surface finish, and more complex geometry than green-sand. The Springer reference work distinguishes "molding sand" (used for the mold body) from "core sand" (used for internal cavities) — both are typically resin-bonded in a chemically hardened system [S3]. The three practical sub-families are no-bake (furan / phenolic-urethane / alkaline phenolic resin + acid or ester hardener, set at room temperature over minutes to hours), shell (Croning process, resin-coated sand dropped onto a heated pattern at 200–280 °C to build a 6–30 mm hardened shell), and warm-box / hot-box (core-grade resin cured at 150–250 °C in a box for high-throughput core production).
No-bake (also called chemical-sand or self-setting) is the dominant choice for medium-batch steel and iron castings above ~200 kg where green-sand dimensional control breaks down; surface finish drops to Ra 3–12 µm and tolerance to ±0.3–0.8 mm. Shell-mold casting is the precision end of the family — thin-walled shells give Ra 1–4 µm, ±0.2 mm on small parts, and are the standard route for small aluminum and copper-alloy castings with thin sections. The decision axis is: no-bake for large ferrous and steel castings and heavy cores; shell for thin-wall, high-finish, lower-melting-point alloys. AA 18 covers permanent-mold and sand castings in a single document because the buyer spec is the casting property, not the mold chemistry [S4].
Lost-Foam and V-Process: Polystyrene and Vacuum-Bonded Specialties
Lost-foam (expanded polystyrene pattern) molding unbonds the pattern from the sand entirely: the foam pattern is invested in unbonded or lightly bonded silica sand, coated with a permeable refractory wash, and vaporized by the incoming molten metal, which fills the void left behind. The process is a true near-net-shape route with no parting line and no draft requirement, and is widely used for aluminum cylinder heads, manifolds, and complex iron pump housings where internal coring would otherwise require assembly of multiple sand cores. [S3]
Vacuum (V-) process molds use thin thermoformed plastic sheets over a pattern, with dry unbonded sand held by external vacuum during pour; vacuum release after solidification collapses the sand for shakeout. Both lost-foam and V-process sit in the same "no-binder / low-binder" classification branch and share the same disadvantage: the unbonded or lightly bonded sand cannot support tall, thin vertical features the way a rigid resin-bonded mold can, so pour height and section thickness must be engineered conservatively. They also fall outside AA 18 in any property-specific clauses that assume a rammed-up bonded mold, so buyers spec the casting properties directly from AA 18 and call out the process separately on the drawing [S4].
3D-Printed Sand Molds: Binder-Jetting at the Sand/Casting Interface
Additive-manufactured sand molds, produced by binder jetting furan or phenolic resin onto silica sand layer-by-layer from a CAD file, are the fourth practical classification and the only one that can produce internal lattices, conformal cooling channels, and organic-shape cores that no traditional sand process can hold [S5]. The 2018 International Journal of Metalcasting paper documents that binder-jetted molds allow castings to be optimized for strength-versus-weight via periodic lattices inside the mold — geometry that is not possible with traditional sand-casting methods because the loose sand would collapse before metal arrives [S5].
The practical 2026 boundaries are: layer thickness 0.2–0.5 mm, print box size typically 1000 × 700 × 400 mm on industrial systems, lead time measured in hours from STL to finished mold, and per-mold cost dominated by sand + binder rather than labor. For complex iron and steel castings where conventional coring would require 6–10 assembled cores, a single printed mold can replace the assembly and eliminate parting-line flash. The trade-off is throughput: a binder-jetting system produces 10–40 L of mold volume per hour, which is two to three orders of magnitude slower than a green-sand molding line and a full order of magnitude slower than a no-bake line, so 3D-printed sand is a low-volume / high-complexity tool, not a high-volume production route. Mold material is also called out as one of the decisive factors for rapid sand-casting capability, alongside pattern and sand-mixing equipment [S1].
Selection Criteria: Tolerance, Alloy, Lot Size, and Geometry
For an engineer choosing between the four classifications, the spec gates are: (1) required surface finish and tolerance — green-sand ±0.8–1.5 mm and Ra 6–25 µm, no-bake ±0.3–0.8 mm and Ra 3–12 µm, shell ±0.2 mm and Ra 1–4 µm, 3D-printed comparable to no-bake on small volumes; (2) alloy reactivity and pour temperature — ferrous above ~1450 °C pushes toward silica-based green-sand and no-bake with chromite or olivine facing sands, while aluminum and copper fit any route; (3) lot size — green-sand and shell dominate above 5 000 pieces per year, 3D-printed sand dominates below 200; (4) geometric complexity — under-cuts, internal lattices, and conformal channels force 3D-printed sand or lost-foam, and reject the rest outright. The resin sand line equipment selection — mixer type, throughput, and reclamation — is tied to the no-bake and shell choice, and a sand mixer spec is what determines whether the resin system hits its bench-life and tensile window on the floor. [S2]
AA 18-2008 sets the casting-property envelope (tensile, elongation, hardness, dimensional classes) that any of the four sand routes must meet, so a spec-first approach is to define the casting requirement first and select the sand mold family second [S4]. The mold base and flask system are then chosen to match the chosen family's ramming or curing cycle, and the sand cooler is sized to keep returned sand within the moisture and temperature window the binder chemistry requires.
Who Each Classification Is For — and What It Is Not
Green-sand is the correct answer for high-volume ferrous castings (engine blocks, pump housings, valve bodies) and short-run non-critical aluminum where pattern cost dominates unit cost; it is the wrong answer for tight-tolerance machined-interface features or for any casting with internal passages that a core cannot assemble cleanly. Resin-bonded (no-bake and shell) is correct for steel and large iron castings requiring better-than-green-sand integrity, and for thin-wall aluminum and copper-alloy parts; it is the wrong answer for very short runs where pattern and core-box tooling cannot be amortized. Lost-foam and V-process are correct for complex single-piece castings that would otherwise require assembled cores (aluminum heads, certain iron pump housings) and where parting-line flash is unacceptable; they are the wrong answer for high-melting-point alloys in tall, thin sections where unbonded sand cannot support the hydrostatic head. 3D-printed sand is correct for prototype and low-volume production of complex castings — lattice-optimized brackets, hydraulic manifolds, and pattern-free short runs; it is the wrong answer for any series above a few hundred pieces per year where mold cost is dominated by machine time, and for castings with a 2D cross-section that any bonded route can produce at lower unit cost. The Sand-Casting, Technique and History of entry in the Springer encyclopedia is explicit that molding sand is customarily the term for the mold body and core sand for internal cavities, and both are typically resin-bonded in chemically hardened systems — which is why the equipment side of a foundry (mixer, cooler, reclamation) is dominated by the resin-sand process family [S3].
The next node to watch is whether binder-jetting machine throughput crosses the ~100 L/h mark in production-class systems — at that line, the unit-cost crossover point with no-bake moves from a few hundred to a few thousand castings per year and reshapes the spec gate for low-to-mid volume production. A second trackable signal is the Aluminum Association revision cycle for AA 18: a new edition adding explicit additive-manufactured-sand property data would let buyers spec 3D-printed sand castings directly against a property envelope rather than by exception.
For related coverage, see Linear Actuator Pros and Cons: A 2026 Spec-First Decision Map.