Lost foam casting (LFC) replaces a conventional mold cavity with a polymer pattern that is coated, buried in dry sand, compacted by vibration, and then vaporized as molten iron fills the pattern under vacuum. The approach can produce complex, thin-walled geometries without draft, but pattern decomposition and coating contact create a narrow process window. The coating must be strong enough to keep liquid iron out of the dry sand, yet permeable enough to let decomposition gases escape. For a cylinder head with internal water jackets, the consequences of losing either function are carbonaceous inclusions, slag defects, or sand burn-on in passages that are difficult to clean.

Component Requirements and Failure Modes
A four-valve single-cylinder head provides a representative case. The casting is HT250-grade grey iron with an envelope of 295 mm × 230 mm × 145 mm. The intake and exhaust port walls are 5.5 mm thick, and the water-jacket baffle is 5 mm. The head contains upper and lower water jackets, and the upper jacket has three narrow-gap regions that are difficult to compact and clean. The cast-on sample is taken at the injector bore in the combustion-chamber plane, a section about 15 mm thick where wall thickness changes sharply and heat extraction is poor. The specified properties are tensile strength ≥240 MPa, Brinell hardness 190–240 HBW, and a hardness difference no greater than 25 HBW within one casting.
LFC has seen limited use for cylinder heads because several failure modes interact:
- distortion after machining changes port geometry and engine performance;
- poor dimensional consistency in bonded pattern assemblies, especially at intake and exhaust ports;
- carbonaceous or slag inclusions in port walls cause leakage;
- sand burn-on inside water jackets restricts coolant flow and affects power, emissions, and reliability;
- unstable matrix and hardness reduce service life.
The production problem is therefore not simply obtaining a sound iron casting. It is controlling pattern chemistry, alloy solidification, coating integrity, sand compaction, and internal cleaning as one system.
Alloy Chemistry and Inoculation
Under a conventional HT250 melting and inoculation practice, separately cast test bars met the specification, but the cast-on sample reached only 224 MPa and 164 HBW, with A plus C graphite and 70–75% pearlite. The shortfall was structural. The sampling location is a heavy-to-light transition that cools slowly, producing a coarse matrix and local graphite blocks. LFC also requires melt and pouring temperatures 30–50 °C above conventional sand casting to supply the heat consumed by pattern vaporization. The existing practice used a relatively high Si–Ba pre-inoculation dose, which was appropriate for thinner housing castings. In the cylinder head, that dose increased ferrite in the thick section and promoted C-type graphite, reducing strength.
Two alloying variants were evaluated while tapping inoculation was reduced and stream inoculation at pouring was kept unchanged.
| Condition | Tensile strength (MPa) | Mean hardness (HBW) | Graphite | Pearlite (%) |
|---|---|---|---|---|
| Conventional HT250 practice | 224 | 164 | A + C | 70–75 |
| Scheme 1: increased Cu | 251 | 174 | A, length grade 4 | 90 |
| Scheme 2: increased Cu and Sb | 270 | 198 | A, length grade 4 | 95–98 |
Scheme 1 met the tensile requirement but remained below the hardness range. Scheme 2 met all material targets. The result shows that alloying, rather than a further increase in inoculation, was the effective lever for the thick sampling section. Cu and Sb raise pearlite stability and alloy hardenability; the Sb-bearing variant increased pearlite to 95–98% and brought hardness to 198 HBW.
With the Scheme 2 chemistry, nine fixed points on the combustion-chamber plane were measured after 3 h in the flask, corresponding to the fastest line takt, and after 36 h. Mean hardness was 197–198 HBW in both cases, and the measured within-head differences were 7–13 HBW, below the 25 HBW limit. Longer in-mold cooling did not produce softening. For this section size and alloy level, cooling time was not the controlling variable for hardness uniformity.
Pattern Material and Gating for Carbon and Slag Control
Pattern material affects both carbon pickup and pattern dimensional consistency. EPS and STMMA differ in carbon content, gas evolution, and bead-size distribution.
| Property | EPS | STMMA |
|---|---|---|
| Carbon mass fraction (%) | 92 | 69.6–83.4 |
| Gas evolution (mL/g) | 521 | 672 |
| Gas evolution rate (mL/(g·s)) | 68.2 | 70.1 |
| Bead-size spread within one grade (mm) | ≈0.4 | ≤0.15 |
STMMA was selected. Its lower carbon content and more gaseous decomposition reduce carbonaceous residue, and its tighter bead-size distribution improves pattern consistency. The higher gas evolution also places a greater demand on coating permeability and vacuum venting, linking pattern selection directly to coating design.
The gating layout used a vertical combustion-chamber plane, side-bottom gating, and a 50 mm × 50 mm × 15 mm, 20 ppi ceramic foam filter. The conventional face-down orientation was avoided because it hindered sand compaction around the combustion-chamber face. With this arrangement, no carbonaceous or slag defects were found on machined or non-machined surfaces.
Coating and Compaction Control
The first trial used the normal two-dip coating and produced multiple burn-on patches in the upper water jacket. Four mechanisms were identified: locally thin coating or insufficient refractoriness under pouring heat; undried coating in dead corners; cracks in the coating; and insufficient sand compaction that allowed iron to break through the coating. Coating quality and compaction therefore have to be controlled together.
The compaction table clamps the flask at the waist and carries four side-mounted vibration motors. Acceleration mapping showed local minima at measurement points 3 and 12, so the centerline between them was the least compacted region of the flask. The head was placed near the motor side, 200 mm from the motor-side flask wall, to improve local sand filling around the water jacket.
Sand compaction depends mainly on motor eccentricity and vibration frequency. Increasing eccentricity raised acceleration and amplitude, while the usable frequency band widened up to an optimum eccentricity and narrowed beyond it. At fixed eccentricity, acceleration rose with frequency to a maximum and then fell. A suitable eccentricity–frequency combination was selected empirically for the flask and pattern.
One upper-jacket location still showed burn-on after tuning. After coating, its through-hole was only φ2–3 mm, so dry sand could not pass through it. The hole behaved as an upward blind cavity, the sand remained loose, and iron penetrated the coating. The problem was eliminated by a process modification or a product-structure change that opened the sand path.
Coating quality was inspected by dipping the assembled pattern once, drying it, and opening the water jacket. Two defect types matter: bare pattern, where the coating remains thin even after a second dip, and coating accumulation that cracks. The latter can allow iron to encapsulate sand and scrap the casting. Coating A and Coating B were compared under different dip orientations and with or without pattern pretreatment.
| Coating and procedure | Bare spots | Cracks |
|---|---|---|
| A, vertical dip | 12 | 0 |
| A, inclined dip | 4 | 0 |
| A, inclined dip + pattern pretreatment | 0 | 0 |
| B, vertical dip | 2 | 3 |
| B, inclined dip | 0 | 4 |
| B, inclined dip + pattern pretreatment | 0 | 3 |
Coating A with inclined dipping and pattern pretreatment gave zero bare spots and zero cracks. During inclined dipping, the process hole was kept upward so the slurry level rose gradually and gas escaped instead of being trapped. Pretreatment with a prepared liquid improved wetting of the glue lines and the hydrophobic pattern surface before the main coating was applied.
The drying room was set at 45 °C and operated at 45–50 °C, with humidity controlled to ≤15%. Blind holes have poor air exchange, and thicker coating can multiply drying time. Drainage and placement were adjusted to reduce accumulation in blind holes and corners. Drying was verified by sectioning the blind regions starting after 12 h and checking the wet layer by colour until no wet coating remained. Together with the coating trials, this control essentially eliminated water-jacket burn-on.
Internal Cleaning by Shot Blasting
LFC cleaning differs from sand casting. The coating is applied to the pattern surface and remains tightly bonded to the casting after pouring, so the water jacket presents blind surfaces that a conventional hanger blast cannot reach uniformly. Five cleaning routes were compared.
| Method | Procedure | Result |
|---|---|---|
| 1 | Q384 hanger blast, 25 min, one change of hanging orientation | Areas facing the process holes were cleaned; corners retained coating; not acceptable |
| 2 | Vibration for 90 s, then Q384 blast for 25 min, one change of orientation | Some coating flakes detached, but a residual layer remained; similar to method 1 |
| 3 | Q384 blast for 20 min, then dedicated nozzle abrasive blasting for 10 min | Water jacket and intake/exhaust ports were clean; met the requirement |
| 4 | Q384 blast for 20 min, then high-pressure water washing | Better than methods 1 and 2, but powdery coating remained in dead corners; worse than method 3 |
| 5 | Q384 blast for 20 min, then ultrasonic cleaning | Heavy coating residue contaminated the bath and efficiency was low; discontinued |
Method 3 was adopted. The nozzle was designed around the process-hole size and water-jacket geometry so that abrasive could reach surfaces missed by the hanger blast. The final sequence was 20 min in the Q384 machine followed by 10 min of dedicated nozzle blasting.
Residual Stress and Production Readiness
Residual stress was measured with an 8000-8-SM strain-gauge system and BE120-2AA(11)-P100 uniaxial strain gauges with a sensitivity coefficient of 2.08. A 40 mm × 40 mm × 30 mm reference block was used, and four points around the injector bore were measured. Vacuum holding time and shakeout time were varied, but no clear trend appeared. The maximum measured residual stress was 39 MPa, within the product requirement.
The individual measures address different defects: alloying and inoculation control the matrix; STMMA and gating control carbonaceous inclusions; coating, drying, and vibration control burn-on; nozzle blasting controls internal cleanliness. These measures also interact. A more permeable coating can be weaker against metal penetration, and stronger compaction must not deform the pattern. The combined process produced cylinder heads that met the tensile strength, hardness, hardness-uniformity, cleanliness, dimensional, and residual-stress requirements. The result supports the production of LFC cylinder heads on the available domestic equipment, provided that pattern quality, coating integrity, sand compaction, and internal cleaning are controlled as a single process chain.
