Cylinder heads close the combustion chamber and carry intake and exhaust ports, valve gear, injector or spark-plug bosses, and cooling passages. The combustion deck and exhaust ports operate under cyclic thermal and mechanical load, while the water jacket must remain open enough to remove heat. In lost foam castings, an expendable polymer pattern is coated with a refractory layer, embedded in dry unbonded sand, compacted by vibration, and poured under negative pressure. Liquid metal replaces the pattern; the polymer pyrolyzes, and the products escape through the coating into the sand and vacuum system. The pattern can define internal passages without separate sand cores, but coating integrity, sand compaction, pattern chemistry, and gas evacuation become central to quality.
By 2019, production of cylinder heads by lost foam casting in China was still limited. Quanchai Tianhe Casting was the only Chinese foundry then producing such heads at production scale, supported by imported pattern equipment and a Vulcan sand-compaction system. A production trial examined whether the process could be transferred to domestic equipment for a thin-walled, two-jacket head.
Cylinder Head Requirements and Failure Modes
The head was a four-valve single-cylinder design with two intake and two exhaust valves and two stacked water jackets. Its envelope was 295 mm × 230 mm × 145 mm. Intake and exhaust port walls were 5.5 mm thick, and the water-jacket baffle was 5 mm. These sections are thin enough that a carbonaceous deposit or inclusion can open a leak path. The upper jacket contained three small-clearance zones where sand is difficult to compact and penetration is likely.
The material was HT250 gray iron, a Chinese designation for gray iron with a nominal 250 MPa tensile strength. Tensile strength was specified as at least 240 MPa, and hardness as 190–240 HBW at the combustion-chamber face near the injector hole, with a maximum within-head hardness spread of 25 HBW. The main combustion-deck wall was 15 mm thick, and the section change around the injector hole produced poor heat dissipation. The compact envelope lowered distortion risk, but five failure modes had blocked earlier attempts:
- distortion after machining, changing port dimensions and engine performance;
- poor dimensional consistency, especially in glued intake and exhaust port patterns;
- carbon slag or inclusions in the ports, causing leakage;
- sand penetration in the water jacket, restricting water flow and affecting power, emissions, and reliability;
- unstable iron properties, reducing service life.

Melt Chemistry and Inoculation
Conventional HT250 practice produced test bars that passed, but the casting body reached only 224 MPa and 164 HBW, with A+C graphite and 70–75% pearlite. The sampling location—a 15 mm combustion deck with a marked section change around the injector hole—cooled slowly and developed a coarse microstructure and local graphite blocks. Inoculation added a second problem. Lost foam melt and pouring temperatures were 30–50 °C higher than in sand casting. The existing SiBa pre-inoculation level, developed for housing castings, was too high for the head: in thick sections it increased ferrite and promoted C-type graphite, lowering strength even though inoculation normally promotes graphitization.
Two melt variants were evaluated. Variant 1 added Cu; variant 2 added Cu and Sb. In both, pre-inoculation at tapping was reduced, while stream inoculation during pouring remained unchanged.
| Variant | Alloying addition | Tensile strength (MPa) | Mean hardness (HBW) | Graphite form / length | Pearlite (%) |
|---|---|---|---|---|---|
| 1 | Cu | 251 | 174 | A / 4 | 90 |
| 2 | Cu + Sb | 270 | 198 | A / 4 | 95–98 |
Variant 1 reached 251 MPa, but hardness was 174 HBW and pearlite 90%, so it failed the hardness specification. Variant 2 reached 270 MPa and 198 HBW, with A-type graphite of length class 4 and 95–98% pearlite. It met all targets. The comparison shows that alloying has to compensate for the head’s thermal mass, while inoculation cannot simply be carried over from thinner, simpler castings.
With variant 2, nine points on the combustion-chamber face were measured after 3 h and 36 h in the flask. The 3 h condition represented the fastest line takt. Mean hardness was 197–198 HBW in all four runs, and the largest within-part spread was 13 HBW, below the 25 HBW limit. Extended cooling did not produce a systematic hardness decrease; within the tested range, hardness did not depend strongly on shakeout delay.
| Cooling condition | Run | Hardness at points 1–9 (HBW) | Mean (HBW) |
|---|---|---|---|
| A: 3 h | 1 | 196, 198, 197, 191, 196, 200, 199, 197, 200 | 197 |
| A: 3 h | 2 | 198, 198, 199, 198, 200, 200, 197, 193, 201 | 198 |
| B: 36 h | 1 | 200, 194, 200, 197, 195, 192, 202, 201, 205 | 198 |
| B: 36 h | 2 | 201, 194, 198, 199, 191, 200, 201, 198, 201 | 198 |
Pattern Material and Gating Design
Pattern pyrolysis can leave carbonaceous residue. EPS and STMMA were compared as bead materials.
| Property | EPS | STMMA |
|---|---|---|
| Molecular formula | (C8H8)n | (C8H8)n·(C5H8O2)n |
| Carbon mass fraction (wt%) | 92 | 69.6–83.4 |
| Gas evolution (mL/g) | 521 | 672 |
| Gas evolution rate (mL/(g·s)) | 68.2 | 70.1 |
STMMA is a styrene–methyl methacrylate copolymer. Its lower carbon content and more gaseous decomposition reduce carbonaceous residue. Its gas evolution is higher, 672 mL/g versus 521 mL/g for EPS, so coating permeability and vacuum must handle a larger gas load. Within a bead grade, STMMA size varies by no more than 0.15 mm, while EPS varies by about 0.4 mm, which improves pattern fit and glue-line consistency. STMMA was selected.
The combustion-chamber face would normally be placed downward as the most important flat surface, but a downward face hindered sand compaction during vibration. The face was set vertically, and a side-bottom gating system was used with a 50 mm × 50 mm × 15 mm, 20 ppi ceramic foam filter. Neither machined nor non-machined surfaces showed carbon slag or inclusion defects.
Water-Jacket Penetration: Compaction and Coating
The first trial with two coating dips produced multiple penetration sites in the upper water jacket. Four mechanisms were identified: thin local coating or insufficient refractoriness; undried coating in blind corners; cracks in the coating; and insufficient sand compaction, allowing iron to push through the coating.
The compaction table clamped the flask at mid-height and carried four side-mounted vibration motors. Acceleration mapping showed local minima at points 3 and 12, so the line between them was the lowest-acceleration zone in the flask. The head was placed near the motor side, 200 mm from the motor-side flask wall. Compaction energy depends on motor eccentricity and frequency. Increasing eccentricity raised acceleration and amplitude; the usable frequency band first widened and then narrowed beyond an optimum. At fixed eccentricity, acceleration rose with frequency to a maximum and then fell. After adjusting eccentricity and frequency, one penetration site remained at a small through-hole. After coating, the hole measured only φ2–3 mm, so sand could not pass through it; the cavity acted as an upward blind hole with loose sand. A process change or a local product-structure change eliminated this site.
Coating is both a barrier and a vent. It must prevent liquid iron from entering the sand, while allowing pattern decomposition products to escape. To inspect the internal jacket, the pattern assembly was dipped once, dried, disassembled, and checked. Two defects appeared: exposed pattern areas, which remained thin after the second dip and led to penetration; and coating accumulation that cracked, allowing iron to form an iron-encased sand defect.
| Coating | Dipping / pretreatment | Exposed pattern points | Crack points |
|---|---|---|---|
| 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 exposed points and zero cracks; coating B gave fewer exposed points but persistent cracks. Inclined dipping with the process hole upward let the slurry level rise gradually and gas escape, reducing entrapped air. The pretreatment was a mixed liquid applied to the pattern and dried before coating; it improved wetting of the glue lines and the hydrophobic pattern surface. Coating A, inclined dipping, and pretreatment were selected.
The drying room was set at 45 °C and ran at 45–50 °C; after drying, relative humidity was ≤15%, and dryness was normally judged by time. Blind holes in the jacket have poor air exchange, and drying time increases disproportionately with coating thickness. Drainage and positioning were adjusted to reduce accumulation in blind holes and corners. For the remaining blind holes, patterns were sectioned after 12 h and the wet coating was tracked by color until no wet layer remained. With the dipping changes, this essentially eliminated water-jacket penetration.
Cleaning the Internal Water Jacket
Unlike sand casting, where coating is applied to molds and cores, lost foam coating envelops the pattern and remains tightly attached to the casting after pouring. The complex water jacket makes shot blasting difficult. Five cleaning routes were compared.
| Route | Operations | Result |
|---|---|---|
| 1 | Q384 hanger blast, 25 min, one reorientation | Clean opposite process holes; coating remained at corners; not acceptable |
| 2 | Shakeout 90 s + Q384 blast, 25 min, one reorientation | Some coating flaked off; residual layer remained; similar to route 1; not acceptable |
| 3 | Q384 blast, 20 min + dedicated nozzle sandblasting, 10 min | Water jacket and intake/exhaust ports clean; acceptable |
| 4 | Q384 blast, 20 min + high-pressure water rinse | Better than routes 1–2; powder residue in dead corners; not acceptable |
| 5 | Q384 blast, 20 min + ultrasonic cleaning | Heavy residue contaminated the bath; low efficiency; discontinued |
Route 3 met the cleanliness requirement. The dedicated nozzle was designed around the process-hole size and water-jacket shape, so the blast could reach surfaces that the Q384 hanger blast could not. The water jacket and intake/exhaust ports had no coating residue. High-pressure water rinsing left powdered coating in dead corners, and ultrasonic cleaning was abandoned because the large residue load contaminated the bath and made the operation inefficient.
Residual Stress
Residual stress was measured at four points around the injector hole with 8000-8-SM equipment and BE120-2AA(11)-P100 uniaxial strain gauges, gauge factor 2.08, using 40 mm × 40 mm × 30 mm reference blocks. Vacuum holding time and shakeout time were varied. No clear trend appeared; the maximum measured stress was 39 MPa, within the product requirement.
Production Implications
Under existing domestic equipment, the trial established the following:
- Melt control with Cu and Sb alloying, reduced pre-inoculation, and unchanged stream inoculation gave 270 MPa, 198 HBW, A-type graphite, and 95–98% pearlite.
- STMMA patterns, a vertical combustion deck, side-bottom gating, and a 20 ppi ceramic foam filter prevented carbon slag and inclusions on machined and non-machined surfaces.
- Placing the head 200 mm from the motor-side flask wall, inclined dipping, pattern pretreatment, and controlled drying essentially eliminated water-jacket penetration.
- A 20 min Q384 hanger blast followed by 10 min of dedicated nozzle sandblasting cleaned the water jacket and ports to the required cleanliness.
- Maximum residual stress was 39 MPa; dimensions, material properties, and residual stress met product requirements.
These results do not define a universal recipe. Alloy addition levels, inoculation, vibration settings, coating selection, and nozzle geometry depend on head design and on the sand and coating system. The important point is that defects in lost foam castings of thin-walled water jackets are coupled: pattern chemistry affects gas load, coating affects both penetration and venting, compaction affects coating survival, and cleaning must match the internal geometry. Treating these variables as one system made production feasible.
