In the development of large marine diesel engines, our company faced significant challenges with the production of L32/40 series ductile iron cylinder heads. These castings, made of QT400-15, have external dimensions of 543 × 540 × 706 mm and a gross weight of 600 kg. During the initial casting trials, the rejection rate was as high as 70% due to various sand casting defects including gas porosity in valve guide bores, shrinkage porosity on the top face, and water leakage under hydraulic pressure. This article presents a systematic investigation of these sand casting defects, the application of numerical simulation via MAGMA, and the subsequent process modifications that dramatically improved casting yield.
1. Original Process Description and Observed Defects
The original casting process employed alkaline phenolic resin self-hardening sand molds and triethylamine cold-box core-making. The melting was performed in a 20 t/h medium-frequency induction furnace, with a pouring temperature of 1365–1380 °C. To mitigate shrinkage, internal chills (45# steel rods) were placed at the valve guide positions, external chills were applied on the bottom face, and exothermic risers were designed on the top face. The gating system and chill layout are summarized in Table 1.
| Parameter | Specification |
|---|---|
| Mold material | Alkaline phenolic resin self-hardening sand |
| Core material | Triethylamine cold-box sand |
| Pouring temperature | 1365–1380 °C |
| Internal chills (valve guide) | 45# steel rods, diameter 20 mm, length 60 mm |
| External chills (bottom face) | Steel plates, 30 mm thick, area 200×150 mm |
| Riser type | Exothermic sleeves on top face |
| Number of risers | 3 (designated 1#, 2#, 3#) |
| Coating | Alcohol-based zirconia coating, single layer |
| Core drying | None |
| Casting yield (initial) | 30% |
Typical sand casting defects observed during the initial production are listed in Table 2, together with their locations and probable root causes based on visual and metallographic inspection.
| Defect type | Location | Frequency (%) | Observed features |
|---|---|---|---|
| Gas porosity (blowhole) | Valve guide bore, oil pool top face | 30 | Round or elongated cavities, shiny inner walls |
| Shrinkage porosity | Top face, under riser 2# and 3# | 25 | Irregular interconnected pores, dendritic pattern |
| Leakage (hydrostatic test) | Injector hole, valve guide area | 15 | Continuous water seepage at 0.8 MPa |
| Sand inclusion | Random | 10 | Embedded sand grains, rough surface |
| Other (slag, misrun) | Various | 20 | Slag entrapment, incomplete filling |
The overall rejection rate of 70% was dominated by gas-related sand casting defects (porosity) and shrinkage-related sand casting defects (shrinkage and leakage). The subsequent sections detail the analysis of these sand casting defects and the corrective measures implemented.
2. Root Cause Analysis of Sand Casting Defects
2.1 Gas Porosity in Valve Guide Bores and Oil Pool
Gas porosity originated from multiple sources: (1) moisture and binder decomposition in the sand cores; (2) rust on the internal steel chills; (3) insufficient venting pathways. The original process used alcohol-based coatings on mold and core surfaces, which generate volatile gases upon contact with molten iron. Additionally, the 45# steel chills, stored in ambient humidity, developed surface rust. During pouring, the rust reacted with carbon and oxygen in the melt, producing CO and H₂ bubbles. The oil pool core, located at the top of the casting away from the ingates, experienced cooler metal temperature and poor gas evacuation, leading to trapped gas pores. Table 3 quantifies the gas evolution sources.
| Source | Gas species | Estimated volume (cm³/kg metal) |
|---|---|---|
| Coating decomposition (alcohol-based) | CO, H₂, H₂O(g) | 50–80 |
| Rust on chills (Fe₂O₃·nH₂O) | H₂, H₂O(g) | 10–20 |
| Sand binder (resin) pyrolysis | CO, CH₄, H₂ | 30–50 |
| Moisture in core (residual) | H₂O(g) | 5–15 |
The total gas evolution exceeded the venting capacity, causing bubble entrapment. The gas pressure P inside a pore at solidification front can be approximated by:
$$ P_{\text{gas}} = P_{\text{atm}} + \frac{2\gamma}{r} + \rho g h $$
where γ is surface tension (≈1.5 N/m for ductile iron), r is pore radius, ρ is density, g gravity, and h depth. For small pores (r < 0.5 mm), the capillary term dominates, making gas pores stable against buoyancy.
2.2 Shrinkage Porosity and Water Leakage
Ductile iron (QT400-15) solidifies with a wide freezing range (≈40–50 °C), promoting a mushy solidification mode. The formation of shrinkage porosity is described by the Niyama criterion:
$$ N = \frac{G}{\sqrt{\dot{T}}} $$
where G is the thermal gradient and dot{T} is the cooling rate. When N < 1 (for ductile iron), microporosity is likely. The original process had N values below 1 in the regions under risers 2# and 3#. Table 4 lists the computed Niyama numbers from MAGMA simulation.
| Region | G (K/mm) | dT/dt (K/s) | Niyama (K^0.5·s^-0.5/mm) |
|---|---|---|---|
| Under riser 1# | 1.2 | 0.8 | 1.34 |
| Under riser 2# | 0.5 | 0.6 | 0.65 |
| Under riser 3# | 0.3 | 0.4 | 0.47 |
| Valve guide area | 0.8 | 1.0 | 0.80 |
The simulation also indicated isolated hot spots under risers 2# and 3#, where the thermal modulus of the riser neck was lower than that of the casting body. The thermal modulus M is defined as:
$$ M = \frac{V}{A} $$
For riser 3#, we measured:
$$ M_{\text{riser}} = 2.1 \, \text{cm}, \quad M_{\text{casting}} = 2.5 \, \text{cm} $$
Thus the riser solidified before the casting, failing to feed the shrinkage. This led to the sand casting defects of internal porosity and subsequent water leakage during hydrostatic testing.
3. Process Improvement Measures for Sand Casting Defects Elimination
3.1 Modification to Reduce Gas Porosity
To address gas-related sand casting defects, the following changes were implemented:
- Copper-plated chills: Replaced 45# steel rods with copper-plated steel chills to prevent rust formation. The copper layer (0.2 mm thick) effectively isolates the steel from moisture.
- Double coating and baking: All molds and cores received two coats of alcohol-based zirconia coating, followed by oven drying at 180–200 °C for 30 minutes. This reduced volatile content.
- Enhanced venting: Additional vent holes were drilled in the upper sand mold and the oil pool core. The core was redesigned with a central vent channel. Table 5 summarizes the venting modifications.
- Pouring within 1 hour after mold closure: To minimize re-absorption of atmospheric moisture.
| Component | Original vent area (mm²) | New vent area (mm²) | Increase (%) |
|---|---|---|---|
| Upper mold (top face) | 0 (none) | 4 holes × 10 mm diameter = 314 | ∞ |
| Oil pool core | 1 hole × 5 mm = 19.6 | 2 holes × 8 mm = 100.5 | 413 |
| Valve guide core (manual drill) | None | 3 holes × 6 mm = 84.8 | ∞ |
3.2 Optimization of Chills and Risers to Mitigate Shrinkage
Based on MAGMA simulation results, additional external chills were placed on the outer sand core (1-1# and 1-2#) and on the oil pool core (2-1# and 2-2#). The locations are schematically shown in the simulation section. The chills accelerate local solidification and promote graphitization expansion, which enhances self-feeding. The effect of chills on the thermal gradient can be expressed by the modified Fourier number:
$$ Fo = \frac{\alpha t}{L^2} $$
where α is thermal diffusivity, t time, and L characteristic length. With chills, the local L effectively decreases, increasing Fo and thus faster solidification. Table 6 lists the chill specifications.
| Chill ID | Location | Material | Dimensions (mm) | Thickness (mm) |
|---|---|---|---|---|
| 1-1# | Outer sand core, near riser 2# | Gray iron plate | 150 × 100 | 25 |
| 1-2# | Outer sand core, near riser 2# (opposite) | Gray iron plate | 150 × 100 | 25 |
| 2-1# | Oil pool core, under riser 3# | Cast steel | 120 × 80 | 20 |
| 2-2# | Oil pool core, adjacent to 2-1# | Cast steel | 120 × 80 | 20 |
4. MAGMA Simulation Results and Comparison
4.1 Hot Spot and Thermal Modulus Analysis
MAGMA simulations were run for both the original and improved processes. Figure 1 (not shown) compared the hot spot distributions. The key outputs are summarized in Table 7.
| Parameter | Original process | Improved process |
|---|---|---|
| Hot spot under riser 2# | Isolated hot spot in casting | Hot spot confined to riser and neck |
| Hot spot under riser 3# | Isolated hot spot in casting (severe) | Reduced intensity, still present |
| Thermal modulus of riser 2# neck (cm) | 2.0 | 2.3 |
| Thermal modulus of casting under riser 2# (cm) | 2.4 | 2.2 |
| Thermal modulus of riser 3# neck (cm) | 1.8 | 1.9 |
| Thermal modulus of casting under riser 3# (cm) | 2.5 | 2.4 |
| Maximum Niyama in region A (under riser 2#) | 0.65 | 1.1 |
| Maximum Niyama in region B (under riser 3#) | 0.47 | 0.55 |
The improved process eliminated the hot spot under riser 2# and increased the Niyama number above the critical threshold of 1.0. However, under riser 3#, the Niyama remained below 1.0 due to geometric constraints – the casting section was tall and narrow, and the riser neck could not be enlarged without affecting core assembly. Table 8 lists the simulated shrinkage porosity volumes.
| Region | Original | Improved |
|---|---|---|
| Under riser 1# | 0.2 | 0.1 |
| Under riser 2# | 3.8 | 0.3 |
| Under riser 3# | 5.1 | 2.8 |
| Valve guide area | 1.2 | 0.0 |
5. Production Validation and Final Results
A trial of 200 cylinder heads was produced using the improved process. The defect statistics are summarized in Table 9. The overall rejection rate dropped from 70% to 8.5%.
| Defect category | Original (100 pieces) | Improved (200 pieces) |
|---|---|---|
| Gas porosity (blowhole) | 30 | 0 |
| Shrinkage porosity (visual) | 25 | 8 |
| Hydrostatic leakage | 15 | 6 |
| Sand inclusion | 10 | 3 |
| Other | 20 | 0 |
| Total defective | 70 | 17 |
| Rejection rate (%) | 70 | 8.5 |
Sectioning of castings from the improved process confirmed that shrinkage porosity under riser 2# was completely eliminated. The residual leakage cases were all traced to the region under riser 3#, which matched the simulation prediction. Future work will focus on further optimizing the riser geometry or adding a local chill at that location.
6. Discussion: Quantitative Relationship Between Process Parameters and Sand Casting Defects
The success of the improvement can be rationalized by considering the feeding efficiency factor η defined as:
$$ \eta = \frac{V_{\text{riser}}}{V_{\text{shrink}}} $$
where Vriser is the liquid volume available from the riser and Vshrink is the volumetric shrinkage of the casting section. For ductile iron, the volumetric shrinkage during solidification is approximately 3–5% due to graphitization expansion. The riser must provide sufficient liquid to compensate for the net contraction. Additionally, the feeding distance Lf can be estimated from:
$$ L_f = \frac{k \cdot M_{\text{riser}}^2}{M_{\text{casting}}} $$
where k is a constant (≈ 0.5 for ductile iron). In the original design, Lf for riser 3# was only 150 mm, less than the actual feeding distance of 220 mm, leading to shrinkage sand casting defects. After adding chills, the effective Mcasting decreased locally, extending the feeding distance to 190 mm – still insufficient for full feeding but significantly reducing porosity.
The gas porosity elimination can be quantified by the gas volume fraction φg remaining in the casting:
$$ \phi_g = \frac{V_{\text{gas}} – V_{\text{vent}} – V_{\text{dissolved}}}{V_{\text{casting}}} $$
where Vgas is total gas evolved, Vvent is the volume evacuated through vents, and Vdissolved is the gas that remains in solution. With the improved venting area increased by 400% and baking reducing moisture content, φg dropped below the critical threshold of 0.5% for visible porosity.
7. Conclusion
This study systematically addressed the sand casting defects in large ductile iron cylinder heads through a combination of root cause analysis, simulation, and practical modifications. The key findings and outcomes are:
- Gas porosity was eliminated by using copper-plated chills, double coating with baking, and enhanced venting systems.
- Shrinkage porosity under riser 2# was completely resolved by adding external chills that promoted directional solidification and increased the Niyama number above 1.0.
- The residual shrinkage under riser 3# remains a challenge due to geometric constraints, but was reduced by 45% in volume.
- MAGMA simulation proved to be an indispensable tool for predicting sand casting defects and guiding process optimization.
- The overall rejection rate decreased from 70% to 8.5%, resulting in substantial cost savings and improved product reliability.
Future work will involve redesigning the riser 3# geometry or applying localized intensive cooling to fully eliminate the last remaining sand casting defects. The methodologies described here can be extended to similar large ductile iron castings prone to shrinkage and gas-related sand casting defects.
| Measure | Target defect | Effectiveness |
|---|---|---|
| Copper-plated chills | Gas porosity in valve guide | Eliminated |
| Double coating + baking | Gas porosity (general) | Eliminated |
| Increased venting | Gas porosity (top face) | Eliminated |
| Additional chills in outer sand core | Shrinkage under riser 2# | Eliminated |
| Additional chills in oil pool core | Shrinkage under riser 3# | Partial (45% reduction) |
| Pouring within 1 hour of mold closure | Re-absorption moisture defects | Eliminated |
$$\text{Total rejection rate reduction: } \Delta = 70\% – 8.5\% = 61.5\%$$
In conclusion, the systematic application of simulation-driven design and targeted process changes successfully mitigated the major sand casting defects in this challenging casting. The experience underscores the importance of addressing both gas evolution and solidification feeding in ductile iron production to achieve defect-free components.