I have carried out the process design, simulation, trial production, and defect prevention work for a heavy-section ductile iron component manufactured by lost foam casting. The component weighs approximately 180 kg and its overall envelope is about 430 mm × 620 mm × 684 mm. It is poured one part per flask. The material is a 400-15 grade ductile iron, which combines moderate strength with good elongation. The geometry contains several thick zones, two internal oil galleries with a combined length greater than 500 mm, and a machined upper face that must remain sound after rough machining and finish machining. Because the oil galleries are subject to a stringent leak test, shrinkage cavities, shrinkage porosity, slag inclusions, and gas defects are not acceptable. My goal in this work was to create a robust lost foam casting route that reduces shrinkage defects, improves feeding, and stabilizes the whole process from pattern production to final inspection.
The component has a combination of thin walls, thick bosses, parallel plate-like structures, and localized heavy sections. These features create multiple thermal gradients during solidification. In lost foam casting, the foam pattern is buried in unbonded sand, compacted, and then filled with liquid metal while the foam decomposes and the decomposition products are removed through the coating and the sand. The process has many advantages, such as near-net shape, reduced flash, and flexible pattern design, but it also introduces specific risks. The decomposition of the foam pattern consumes heat, generates gas, and can affect the local temperature field. The unbonded sand mold may also move if the vacuum and compaction are not controlled correctly. For ductile iron, the problem is more serious because the alloy solidifies in a mushy mode and develops eutectic expansion due to graphite precipitation. If the expansion is not managed, it can either feed the shrinkage that occurs earlier or cause mold wall movement, which later appears as shrinkage porosity or dimensional variation.
In my first review, I separated the problem into three categories: geometric thermal risk, gating and feeding design, and process control. The geometric thermal risk comes from the local modulus differences in the casting. The gating and feeding design determines whether liquid metal can reach the last-solidifying regions before the feeding path closes. The process control determines whether the sand mold is sufficiently rigid, whether the vacuum is stable, and whether the foam pattern is fully decomposed without trapping residue. All three categories interact. A good gating system cannot compensate for poor sand rigidity, and a strong sand mold cannot feed a hot spot that has no riser or no feeding channel. Therefore, I treated the work as a full process development rather than a single gating change.
Table 1. Basic casting and production data
| Item | Value or description | Process implication |
|---|---|---|
| Component weight | Approximately 180 kg | Requires sufficient pouring capacity and stable vacuum |
| Material | Ductile iron, 400-15 grade | Mushy solidification and eutectic expansion must be controlled |
| Overall dimensions | 430 mm × 620 mm × 684 mm | Large envelope increases sand volume and compaction demands |
| Production mode | One casting per flask | Simplifies gating layout but increases relative mold cost |
| Thick section | 183 mm × 40 mm × 59 mm local heavy zone | Hot spot with high shrinkage risk |
| Oil gallery length | Greater than 500 mm total | Internal defects are unacceptable under leak test |
| Machined face | Upper end face and surrounding intersection | Shrinkage holes may be exposed after machining |
| Lost foam casting pattern | Foam pattern with refractory coating | Decomposition gases must escape through coating and sand |
| Sand system | Unbonded sand with vacuum assistance | Vacuum and compaction must be controlled to resist expansion pressure |
From a thermal standpoint, I estimated the modulus of the heavy sections and compared them with the surrounding walls. The modulus is the ratio of volume to cooling surface area. A high modulus means a slow cooling rate and a higher risk of shrinkage. The basic expression is:
$$ M = \frac{V}{A} $$
where \(M\) is the modulus, \(V\) is the volume, and \(A\) is the heat-dissipating surface area. For a simple plate, the modulus is approximately half the thickness. For a long bar, the modulus is approximately the cross-section area divided by the perimeter. In the heavy zone of this casting, the local modulus is much larger than that of the adjacent thin walls. Therefore, the heavy zone remains liquid longer. If no feed metal is available during that period, the final contraction cannot be compensated, and a shrinkage cavity or shrinkage porosity forms.
I also used the Chvorinov relationship to estimate relative solidification times:
$$ t_s = B M^2 $$
where \(t_s\) is the local solidification time, \(B\) is a mold constant that depends on the mold material and pouring temperature, and \(M\) is the modulus. This relationship shows that the solidification time increases with the square of the modulus. A heavy section with a modulus two times larger than a thin wall will remain liquid about four times longer. That difference is the main reason why the upper face and the parallel plate intersections are the last regions to solidify in this casting.
For ductile iron, the volume change is not a single contraction. There is liquid contraction before and during solidification, solidification contraction, solid contraction after solidification, and eutectic expansion caused by graphite precipitation. A simplified volume balance can be written as:
$$ \Delta V_{total} = \Delta V_{liquid} + \Delta V_{solidification} + \Delta V_{solid} – \Delta V_{eutectic} $$
If the eutectic expansion is properly directed, it can partially compensate for the earlier liquid and solidification shrinkage. If it is not properly directed, it can press against the mold wall and cause wall movement, which increases the final volume and produces internal porosity. The feeding design must therefore provide a path for liquid metal before the eutectic expansion begins, and the mold must be rigid enough to convert the expansion into internal feeding rather than mold deformation.
In lost foam casting, the mold rigidity is not the same as in a bonded sand mold. The sand is unbonded and held by vacuum. The vacuum level, sand filling method, and compaction time determine the initial sand density. During pouring, the foam pattern decomposes and gas is drawn through the coating. If the vacuum is too low, the decomposition products cannot escape quickly enough, and gas defects may form. If the vacuum is too high, the sand may become too tightly packed, which can reduce permeability and increase the risk of gas entrapment. Therefore, I considered the vacuum level as a process variable that must be optimized together with gating and risering.
I first established the initial gating concepts. Four gating systems were considered: side bottom injection, top injection, step injection, and bottom injection. The term “injection” here means the entry of liquid metal into the mold cavity. I simulated each concept and compared the filling pattern, temperature distribution, and shrinkage risk. The simulation was not used as a replacement for trials, but it helped me identify which regions were likely to be problematic and which gating system could deliver liquid metal with the least turbulence.
Table 2. Initial gating system options and observations
| Gating system | Filling characteristics | Shrinkage risk observation | Slag and gas risk | Initial assessment |
|---|---|---|---|---|
| Side bottom injection | Metal enters from side and bottom, moderate turbulence | High risk at upper face and parallel plate intersection; U-shaped bosses also show risk | Moderate slag risk due to flow path | Not sufficient for heavy top sections |
| Top injection | Metal falls from top, high turbulence and splashing | Similar shrinkage risk to side bottom injection | High slag and gas risk | Poor for internal quality |
| Step injection | Metal enters in stages, complex flow | Risk concentrated at U-shaped bosses and upper face intersection; ingate solidification blocks feeding | Moderate to high | Difficult to control |
| Bottom injection | Metal enters from bottom, smooth filling, low turbulence | Upper hot spot can be fed by a riser if the feeding path remains open | Lower slag risk | Best candidate with riser |
In the side bottom injection simulation, the thin walls solidified first, and the heavy sections remained liquid. The intersection between the parallel plates and the upper face showed a high shrinkage risk. The two U-shaped bosses also formed internal hot spots. The bottom area and the dense boss region showed additional risk. The solidification time increased with wall thickness, which is expected, but the feeding path was not sufficient to keep the upper hot spot supplied. Small slag traps had little effect on the solidification of the upper face, so I did not rely on them as a primary solution.
The top injection simulation produced a similar shrinkage risk pattern. The main difference was in the filling stage: top injection caused more turbulence, more splashing, and a higher probability of slag entrapment. Because the oil galleries require high internal cleanliness, I excluded top injection from further consideration early in the program.
The step injection simulation showed that the upper face could not be fed after the ingate solidified. The shrinkage risk remained in the U-shaped bosses and the parallel plate intersection. This result indicated that the feeding path closed too early, so even if a riser were added, it might not be effective without a change in the ingate position or riser neck design.
The bottom injection simulation was more promising. The filling was smoother, and the lower part of the mold filled without excessive turbulence. However, the upper hot spot still showed shrinkage risk. The riser in the initial simulation was not able to feed the hot spot completely before the feeding channel closed. The riser itself also showed shrinkage porosity. This told me that bottom injection was the correct direction, but the riser size and neck design needed to be recalculated and tested.
I then performed a more detailed thermal analysis of the critical regions. The upper face and the parallel plate intersection behave like a T-junction. At such a junction, the modulus is higher than that of either the plate or the face because the volume increases while the cooling surface does not increase proportionally. The local modulus can be estimated by dividing the junction into simple geometric blocks. For a plate of thickness \(t\), the modulus is approximately \(t/2\). For a junction where a plate meets a thicker face, the modulus can be approximated by:
$$ M_j = \frac{V_1 + V_2 + V_3}{A_1 + A_2 + A_3 – A_{contact}} $$
where \(V_1\), \(V_2\), and \(V_3\) are the volumes of the simple blocks, \(A_1\), \(A_2\), and \(A_3\) are the external surface areas, and \(A_{contact}\) is the internal contact area that does not dissipate heat. This calculation confirms that the junction has a higher modulus than the surrounding walls and is therefore a last-solidifying region.
The oil galleries complicate the design because they act as internal cores. In lost foam casting, the foam pattern includes the gallery shape, and the coating must be applied uniformly inside the gallery. If the coating is too thick or not fully dried, gas can be trapped, and the gallery surface may show defects. If the coating is too thin, the foam decomposition products may not be filtered adequately, and slag may enter the gallery. The galleries also reduce the local metal volume, which changes the thermal field. However, the surrounding metal can still form a hot spot if the gallery is close to a thick section. I therefore treated the gallery as both a quality requirement and a thermal feature.
Table 3. Thermal modulus estimates for critical regions
| Region | Approximate geometry | Relative modulus | Solidification order | Shrinkage risk |
|---|---|---|---|---|
| Thin wall | Uniform plate | Low | First | Low |
| Parallel plate | Plate-like structure | Medium | Second | Medium |
| Upper face | Thick face with machining allowance | High | Late | High |
| Parallel plate and upper face intersection | T-junction | Very high | Last | Very high |
| U-shaped boss | Enclosed thick boss | High | Late | High |
| Bottom boss cluster | Dense thick region | Medium to high | Late | Medium to high |
| Oil gallery wall | Internal channel | Variable | Depends on surrounding thickness | Must be defect-free |
The first trial used the bottom injection system without a feeding riser. The result was a shrinkage hole on the upper face, concentrated at the intersection between the parallel plate structure and the upper face. This was consistent with the simulation. The shrinkage hole was exposed after rough machining, which meant that the defect was not only internal but also affected the final part. The reject rate was high, and the trial could not be accepted for production.
The cause of the shrinkage hole was not a single factor. I identified the following contributions:
1. The upper face had a larger modulus than the surrounding walls, so it remained liquid after the walls solidified.
2. The feeding path from the bottom injection system closed before the upper face completed solidification.
3. The eutectic expansion of the ductile iron was not fully utilized because the expansion pressure was not directed toward the hot spot.
4. The sand mold was not rigid enough in the local area, so the expansion pressure may have moved the mold wall instead of feeding the shrinkage.
5. The foam pattern decomposition in the thick section may have altered the local temperature and gas pressure, reducing the effective feeding.
6. The coating and vacuum conditions were not yet optimized, so gas and slag defects could also appear.
For ductile iron, the solidification mode is often described as mushy or pasty. There is not a sharp solidification front. Instead, a mushy zone develops, and liquid can flow through the dendrite network only while the network is permeable. As the solid fraction increases, the permeability decreases. When the permeability becomes too low, the remaining liquid cannot be fed from the riser, and shrinkage porosity forms. The feeding resistance can be estimated by Darcy’s law:
$$ Q = – \frac{K A}{\mu} \frac{\Delta p}{L} $$
where \(Q\) is the flow rate, \(K\) is the permeability, \(A\) is the cross-sectional area, \(\mu\) is the viscosity, \(\Delta p\) is the pressure difference, and \(L\) is the flow length. This equation shows that a long feeding path and a low permeability both reduce the feeding flow. In this casting, the distance from the bottom ingate to the upper hot spot is long, and the mushy zone reduces the permeability as solidification proceeds. Therefore, a riser close to the hot spot is necessary.
The eutectic expansion also plays a role. When graphite nodules grow, the volume of the solid plus graphite can be larger than the volume of the liquid from which it forms. The expansion can be expressed as:
$$ \Delta V_e = \gamma V f_g $$
where \(\Delta V_e\) is the eutectic expansion volume, \(\gamma\) is a coefficient related to graphite density and iron density, \(V\) is the local volume, and \(f_g\) is the graphite fraction. If this expansion is contained by a rigid mold, it can push liquid into the shrinkage pores and reduce porosity. If the mold is not rigid, the expansion can instead deform the mold and increase the final porosity. In lost foam casting, the mold rigidity depends on the vacuum and the sand compaction. Therefore, I included vacuum and compaction in the process control plan.
I also studied the pressure changes in the riser. There are three typical stages. In the first stage, the ingate is still open, and liquid metal can enter the mold and the riser. In the second stage, the ingate has solidified, and the casting and riser form a closed system. Liquid contraction reduces the internal pressure, and the riser may be partially emptied. In the third stage, eutectic expansion increases the internal pressure, and the riser may refill. The pressure changes can be described qualitatively as:
$$ p_1 = p_0 + \rho g h $$
$$ p_2 = p_0 + \rho g h – \Delta p_{contraction} $$
$$ p_3 = p_0 + \rho g h – \Delta p_{contraction} + \Delta p_{expansion} $$
where \(p_1\), \(p_2\), and \(p_3\) are the pressures in the three stages, \(p_0\) is the atmospheric pressure, \(\rho\) is the density, \(g\) is the gravitational acceleration, \(h\) is the metal head, \(\Delta p_{contraction}\) is the pressure drop due to contraction, and \(\Delta p_{expansion}\) is the pressure increase due to eutectic expansion. The riser must be designed so that it still contains liquid metal during stage two and can accept the expansion during stage three. If the riser is too small, it solidifies too early and cannot feed. If the riser neck is too small, it freezes and separates the riser from the casting before feeding is complete. If the riser neck is too large, the riser may not be easy to remove, and the yield may decrease.
Table 4. Defect types, mechanisms, and process factors
| Defect | Mechanism | Process factors | Observation in this casting |
|---|---|---|---|
| Shrinkage hole | Liquid contraction and solidification contraction without feed metal | Hot spot modulus, feeding distance, riser size, ingate solidification time | Found on upper face at parallel plate intersection |
| Shrinkage porosity | Mushy zone permeability loss and dispersed micro-shrinkage | Solidification rate, alloy composition, riser contact, mold rigidity | Risk in U-shaped bosses and thick sections |
| Slag inclusion | Oxide film and reaction products trapped in the flow | Gating turbulence, filter, coating quality, pouring temperature | Risk in side and top gating systems |
| Gas defect | Foam decomposition gas or air trapped in the metal | Vacuum level, coating permeability, pattern density, pouring speed | Risk in thick sections and oil galleries |
| Mold wall movement | Eutectic expansion pushes unbonded sand | Sand compaction, vacuum, flask rigidity, expansion pressure | Can increase shrinkage porosity and dimensional deviation |
| Cold shut | Low pouring temperature or slow filling front | Pouring temperature, gating design, section thickness | Not dominant but must be controlled |
| Mismatch or distortion | Pattern deformation or sand movement | Pattern handling, coating drying, compaction, cooling | Must be controlled for machining allowance |
After the first trial, I changed the approach. I kept the bottom injection concept but added a feeding riser on the upper hot spot. I also increased the machining allowance on the upper face so that any small surface defect could be removed during machining. The machining allowance is not a substitute for internal quality, but it provides a process margin for the upper face. The oil galleries, however, cannot rely on machining allowance because they are internal. They require a sound casting from the beginning.
I designed two riser variants for the trial. The first riser, called 1# in my experimental plan, had a modulus equal to the casting hot spot modulus, and the neck modulus was set at 0.8 times the riser modulus. The second riser, called 2# in my experimental plan, had a modulus 1.5 times the casting hot spot modulus, and the neck modulus was set at 0.6 times the riser modulus. The design equations were:
$$ M_r \ge M_s $$
$$ M_n = k M_r $$
where \(M_r\) is the riser modulus, \(M_s\) is the casting hot spot modulus, \(M_n\) is the neck modulus, and \(k\) is a coefficient. For 1#, \(M_r = M_s\) and \(k = 0.8\). For 2#, \(M_r = 1.5 M_s\) and \(k = 0.6\). The larger riser in 2# provides more feed metal and stays liquid longer. The smaller neck in 2# helps to concentrate the feed into the hot spot and makes riser removal easier, but it also risks early freezing. The trial was necessary to determine whether the neck would remain open long enough.
Table 5. Riser design variants
| Variant | Riser modulus | Neck modulus | Expected advantage | Expected risk |
|---|---|---|---|---|
| 1# | Equal to casting hot spot modulus | 0.8 times riser modulus | Balanced size, moderate yield | May solidify too early for large hot spot |
| 2# | 1.5 times casting hot spot modulus | 0.6 times riser modulus | More feed metal, longer liquid time | Larger riser, lower yield, neck freeze risk |
The 1# riser trial used bottom injection. After rough machining, shrinkage cavities were still found in the hot spot region. The reject rate reached 37%. This result showed that a riser with a modulus equal to the hot spot modulus was not sufficient. The riser did not remain liquid long enough, or the neck froze before the hot spot finished solidifying. The simulation had predicted that the feeding path would close too early, and the trial confirmed it. The 1# riser design was therefore rejected for production.
The 2# riser trial also used bottom injection. After rough machining, the upper face showed no obvious defects. The hot spot region showed no shrinkage porosity. However, a local inspection found discrete small point-like defects on the end face. These defects were shallow and dispersed. After full machining, the end face defects were removed, and the final part met the requirement. The 2# riser design was therefore selected as the best candidate. The point-like defects were likely related to gas evolution or small slag particles, so I also improved the coating, vacuum, and pouring control to reduce them.
Table 6. Experimental results for riser variants
| Trial | Gating | Riser | Upper face after rough machining | Hot spot after rough machining | Reject rate | Decision |
|---|---|---|---|---|---|---|
| Trial A | Bottom injection | No riser | Shrinkage hole at parallel plate intersection | High risk | High | Rejected |
| Trial B | Bottom injection | 1# riser | Shrinkage cavity present | Shrinkage cavity present | 37% | Rejected |
| Trial C | Bottom injection | 2# riser | No obvious defect | No shrinkage porosity | Low after machining | Selected |
The 2# riser was not the only change. I also increased the upper face machining allowance to provide a safe removal layer for minor surface defects. I improved the gating system to approximate bottom injection, which means the metal enters low in the mold and fills the cavity with a relatively smooth front. The approximate bottom injection method reduces turbulence and slag entrapment compared with top or step injection. It also allows the metal temperature to remain more uniform, which helps the riser feed the upper hot spot.
I calculated the required riser volume using the feeding demand. The feeding demand includes liquid shrinkage, solidification shrinkage, and a safety factor. A simplified equation is:
$$ V_r = \frac{V_s \alpha}{\eta} $$
where \(V_r\) is the riser volume, \(V_s\) is the casting volume that needs feeding, \(\alpha\) is the total shrinkage fraction, and \(\eta\) is the feeding efficiency. The feeding efficiency depends on the riser shape, neck design, and mold rigidity. For ductile iron in lost foam casting, I used a conservative efficiency because the foam decomposition and unbonded sand can reduce the effective feeding. The 2# riser had a larger modulus and therefore a larger volume, which provided the necessary feed metal.
The neck design was also important. The neck must remain open long enough for the riser to feed the hot spot, but it must not be so large that it creates a new hot spot or makes removal difficult. The neck freezing time can be estimated by:
$$ t_n = B M_n^2 $$
where \(t_n\) is the neck solidification time and \(M_n\) is the neck modulus. The neck must satisfy \(t_n > t_{hotspot}\) for feeding to continue until the hot spot is solid. In the 2# design, the neck modulus was 0.6 times the riser modulus, but the riser modulus was 1.5 times the hot spot modulus, so the absolute neck modulus was still sufficient. The trial confirmed that the neck remained open long enough.
I also considered the riser location. The riser was placed directly above the upper hot spot. This placement shortens the feeding distance and reduces the resistance to flow. The feeding distance in ductile iron is limited, especially in lost foam casting where the mold conductivity and foam decomposition can alter the thermal profile. By placing the riser close to the hot spot, I reduced the required feeding distance and improved the probability of soundness. The riser was connected through a neck that was designed to freeze after the hot spot. In the simulation, the riser was the last region to solidify, which is the desired condition.
Table 7. Recommended process window after riser selection
| Parameter | Initial range | Optimized range | Reason |
|---|---|---|---|
| Pouring temperature | 1380–1420 °C | 1400–1430 °C | Improves fluidity for thin walls and oil galleries |
| Vacuum level | 0.03–0.05 MPa | 0.04–0.055 MPa | Balances gas removal and sand rigidity |
| Sand compaction | Vibration only | Vibration plus controlled vacuum | Increases mold rigidity against eutectic expansion |
| Coating thickness | 0.5–1.0 mm | 0.6–0.9 mm | Avoids gas entrapment while maintaining strength |
| Coating drying | Ambient air | Controlled oven drying | Reduces residual moisture and gas defects |
| Holding time after pouring | 10 min | 15 min or more | Allows expansion pressure to act before vacuum release |
| Riser modulus | Equal to hot spot | 1.5 times hot spot | Provides sufficient feed metal |
| Neck modulus | 0.8 times riser | 0.6 times riser | Improves feeding concentration and removal |
| Machining allowance on upper face | Standard | Increased | Removes minor surface defects |
In lost foam casting, the foam pattern quality is a fundamental input. If the pattern is damaged, distorted, or has a low density variation, the final casting will inherit those problems. I therefore added pattern inspection before coating. The pattern must be clean, dry, and free from flash. The coating must be uniform, especially inside the oil galleries. A non-uniform coating can cause local hot spots or gas defects. I used a controlled coating thickness and verified it by weight and visual inspection. The coating must also be fully dried. Residual moisture in the coating can generate gas when the metal enters, leading to gas porosity or blowholes.
The sand filling and compaction process also received attention. In lost foam casting, the sand must flow around the pattern without damaging the coating. The flask is vibrated to compact the sand. If the vibration is too low, the sand density is low, and the mold may move under eutectic expansion. If the vibration is too high, the coating may be damaged, or the pattern may deform. I used a controlled vibration cycle and then applied vacuum. The vacuum increases the sand rigidity and helps to remove decomposition gases. The vacuum must be maintained during pouring and for a sufficient time after pouring. I specified a holding time greater than 15 minutes before releasing the vacuum. This holding time allows the eutectic expansion to occur while the mold is still rigid, which helps to convert the expansion into internal feeding rather than mold wall movement.
The pouring process was also controlled. A smooth pour with a consistent pouring rate helps to avoid turbulence and slag entrapment. I used a pouring basin and a ceramic filter where possible. The filter captures slag and reaction products, but it must not be so restrictive that it slows the fill and causes cold shuts. The pouring temperature was increased slightly to improve fluidity in the thin walls and oil galleries. However, the temperature was not increased excessively because too high a temperature can increase shrinkage and gas defects. The optimized range was determined by trial.
Table 8. Process control plan for lost foam casting
| Stage | Control item | Target | Verification method |
|---|---|---|---|
| Pattern | Density and dimensions | Within drawing tolerance | Sampling and measurement |
| Pattern | Surface damage | No cracks or deep marks | Visual inspection |
| Coating | Thickness | 0.6–0.9 mm | Weight check and visual inspection |
| Coating | Drying | Fully dry, no moisture | Oven record and weight loss |
| Sand | Fill method | Uniform flow around pattern | Operator procedure and inspection |
| Sand | Compaction | Controlled vibration | Time and amplitude record |
| Vacuum | Level | 0.04–0.055 MPa | Vacuum gauge and data log |
| Pouring | Temperature | 1400–1430 °C | Thermocouple or pyrometer |
| Pouring | Rate | Smooth and consistent | Operator procedure and observation |
| Holding | Vacuum holding time | ≥ 15 min | Timer and data log |
| Cooling | In-flask cooling | Sufficient to avoid distortion | Temperature check before shakeout |
| Cleaning | Riser removal | No damage to casting | Visual inspection after cutting |
After the process changes, I verified the results using destructive and non-destructive methods. The upper face was inspected after rough machining. The hot spot region was sectioned to check for internal shrinkage. The oil galleries were tested by leak testing. The 2# riser design with bottom injection showed no shrinkage porosity in the hot spot and no obvious defects on the upper face. The small point-like defects found on the end face were removed by full machining. The leak test passed. The reject rate dropped significantly compared with the 1# riser trial.
I also calculated the quality improvement. If the reject rate before improvement is \(R_1\) and the reject rate after improvement is \(R_2\), the reduction in reject rate is:
$$ \Delta R = R_1 – R_2 $$
If the initial reject rate was 37% and the final reject rate was below the production target, the improvement is substantial. The exact final reject rate depends on the production lot, but the process capability improved because the main shrinkage defect was eliminated. The quality yield can be expressed as:
$$ Y = \frac{N_g}{N_t} \times 100\% $$
where \(Y\) is the yield, \(N_g\) is the number of good castings, and \(N_t\) is the total number of castings. The process capability index can also be used for dimensional features, but the critical quality characteristic here is internal soundness. For internal soundness, the acceptance criterion is based on leak testing and sectioning.
The final process window combined the following elements:
1. Bottom injection or approximate bottom injection gating system.
2. A feeding riser with a modulus 1.5 times the hot spot modulus.
3. A neck modulus 0.6 times the riser modulus.
4. Increased machining allowance on the upper face.
5. Controlled coating thickness and drying.
6. Controlled sand compaction and vacuum.
7. Vacuum holding time greater than 15 minutes.
8. Pouring temperature in the range of 1400–1430 °C.
9. Full process control from pattern to final inspection.
Table 9. Final process window and expected effect
| Feature | Final design | Expected effect on shrinkage | Expected effect on slag and gas |
|---|---|---|---|
| Gating system | Approximate bottom injection | Improves feeding path and reduces turbulence | Reduces slag entrapment |
| Riser | Modulus 1.5 times hot spot | Provides sufficient feed metal | No direct effect |
| Riser neck | Modulus 0.6 times riser | Keeps feeding channel open and concentrates feed | No direct effect |
| Upper face allowance | Increased | Removes minor surface defects | Removes surface slag or gas marks |
| Coating | 0.6–0.9 mm, fully dried | Reduces local hot spots | Reduces gas defects |
| Sand compaction | Controlled vibration | Increases mold rigidity | Improves gas escape |
| Vacuum | 0.04–0.055 MPa | Resists mold wall movement | Removes decomposition gas |
| Holding time | ≥ 15 min | Allows expansion to feed shrinkage | Prevents gas entrapment during solidification |
| Pouring temperature | 1400–1430 °C | Improves fluidity and feeding | Balances gas generation |
In lost foam casting, the interaction between the foam pattern and the metal is a unique factor. When the metal enters the mold, the foam pattern decomposes into gas and liquid residues. The gas must escape through the coating and the sand. If the coating has low permeability, the gas pressure can build up and push liquid metal away from the pattern wall, causing surface defects or incomplete filling. If the coating has high permeability, the metal may penetrate the coating and cause sand adhesion. The coating design must therefore balance permeability and refractory strength. I found that a coating thickness of 0.6–0.9 mm worked well for this component. The coating was applied by dipping and then dried in a controlled oven. The oil galleries were coated with a thinner but still uniform layer to avoid blocking the internal passages.
The foam pattern density also affects the decomposition gas volume. A higher density pattern produces more gas, which increases the risk of gas defects. A lower density pattern produces less gas but may be weaker and more susceptible to deformation. I selected a pattern density that provided sufficient strength for handling and coating while keeping the gas volume manageable. The pattern was inspected for density variation and surface quality before coating. Any pattern with cracks or incomplete fusion was rejected.
The sand system in lost foam casting is usually unbonded silica sand. The sand must be dry and free from fines that can reduce permeability. The sand grain size affects the surface finish and the gas escape. A finer sand gives a better surface finish but lower permeability. A coarser sand gives higher permeability but a rougher surface. For this casting, I used a sand with a balanced grain size distribution. The sand was cleaned and cooled before use. The flask was filled in layers, and vibration was applied to compact the sand around the pattern. The vacuum was then applied to increase the sand strength. The vacuum also helped to remove the decomposition gas during pouring.
The pouring rate is another important variable. If the pouring rate is too slow, the metal front may cool and stop, causing cold shuts or misruns. If the pouring rate is too fast, the foam decomposition may be too rapid, and gas pressure may build up. The optimal pouring rate depends on the casting weight and the gating system. For this 180 kg casting, I used a controlled pouring rate that filled the mold smoothly without excessive turbulence. The pouring basin was kept full to avoid air entrainment. A ceramic filter was placed in the gating system to capture slag and reaction products. The filter also helped to stabilize the flow.
The thermal field in lost foam casting is different from that in bonded sand casting because the foam pattern decomposition absorbs heat and the coating affects heat transfer. The effective mold constant \(B\) in the Chvorinov equation may be different from that of a bonded sand mold. I therefore calibrated the simulation and the trials using measured temperatures and sectioned castings. The calibration improved the accuracy of the shrinkage prediction. The final riser design was based on both simulation and trial. The simulation helped to identify the hot spots, and the trial confirmed the feeding behavior.
The eutectic expansion of ductile iron can be both helpful and harmful. If the expansion is contained, it can reduce shrinkage porosity by pushing liquid into the last-solidifying regions. If the expansion is not contained, it can cause mold wall movement, which increases the casting volume and creates internal porosity. In lost foam casting, the mold is held by vacuum, but the sand is still unbonded. The mold rigidity depends on the sand compaction, the vacuum level, and the flask design. I increased the sand compaction and the vacuum level to improve rigidity. I also held the vacuum for more than 15 minutes after pouring. This holding time allowed the eutectic expansion to occur while the mold was still under vacuum. If the vacuum were released too early, the mold could relax, and the expansion would cause wall movement instead of feeding.
The riser pressure stages provided a useful framework for understanding the feeding behavior. In stage one, the ingate is open, and the mold fills. In stage two, the ingate freezes, and the casting-riser system is closed. Liquid contraction reduces the pressure, and the riser must have enough liquid metal to compensate. In stage three, eutectic expansion increases the pressure, and the riser may refill. The riser must not freeze completely before stage three. The 2# riser with a larger modulus stayed liquid longer and was able to participate in stage three. The neck also remained open long enough. The result was a sound hot spot.
The oil galleries required special attention. Because they are internal and cannot be inspected visually after casting, I used leak testing as the primary quality check. The galleries were also sectioned in trial castings to verify internal soundness. The coating inside the galleries was checked for uniformity. The foam pattern for the galleries was made from the same pattern material as the rest of the casting, so the decomposition behavior was consistent. The gating system was designed so that the galleries filled smoothly. If the metal entered the galleries with high turbulence, slag or gas could be trapped. The approximate bottom injection system reduced this risk. The final castings passed the leak test after full machining.
Table 10. Critical quality characteristics and verification
| Quality characteristic | Requirement | Verification method | Process control |
|---|---|---|---|
| Upper face soundness | No shrinkage hole after machining | Visual inspection after rough machining | Riser design, machining allowance |
| Hot spot soundness | No shrinkage porosity | Sectioning of trial castings, ultrasonic testing if needed | Riser modulus, neck modulus, mold rigidity |
| Oil gallery integrity | Pass leak test | Pressure leak test | Coating uniformity, smooth filling, sound solidification |
| Surface finish | Within drawing requirement | Visual and dimensional inspection | Pattern quality, coating, sand grain size |
| Dimensional accuracy | Within tolerance | Measurement after cleaning | Pattern control, sand compaction, cooling control |
| Internal slag | No harmful inclusions | Sectioning and machining inspection | Gating design, filter, pouring temperature |
| Gas porosity | No unacceptable porosity | Sectioning, leak test, visual inspection | Vacuum, coating drying, pattern density |
| Riser removal | Clean surface, no cracks | Visual inspection after cutting | Neck design, cutting procedure |
After implementing the final process, I monitored the production for several lots. The main shrinkage defect on the upper face was eliminated. The hot spot region showed no shrinkage porosity in the sectioned castings. The oil galleries passed the leak test. The small point-like defects that appeared on the end face in the trial were removed by full machining. The process was stable when the vacuum, coating, and pouring parameters were held within the specified ranges. The key lesson was that lost foam casting of ductile iron requires an integrated approach. The gating system, riser design, pattern quality, coating, sand compaction, vacuum, and pouring practice all interact. A change in one variable can affect the others. Therefore, I established a process control plan and trained the operators to follow it. The plan included checks at each stage and a data log for the vacuum, pouring temperature, and holding time.
The feeding efficiency in lost foam casting can be lower than in bonded sand casting because the foam decomposition and the unbonded sand can reduce the effective mold rigidity. I therefore used a larger riser than would be required in a bonded sand mold. The 2# riser with a modulus 1.5 times the hot spot modulus provided the necessary margin. The neck modulus of 0.6 times the riser modulus kept the feeding channel open without creating a large new hot spot. The increased machining allowance on the upper face provided a secondary safety margin for surface defects. The combination of these measures made the process robust.
I also considered the yield of the process. A larger riser reduces the casting yield because more metal is poured and later returned as scrap. However, the cost of a rejected casting is higher than the cost of the extra metal. The 2# riser improved the yield of good castings, even though the riser itself was larger. The net economic effect was positive. The process yield can be expressed as:
$$ Y_{process} = \frac{W_c}{W_c + W_r + W_g} \times 100\% $$
where \(Y_{process}\) is the process yield, \(W_c\) is the casting weight, \(W_r\) is the riser weight, and \(W_g\) is the gating system weight. The riser weight increased, but the reject rate decreased. The overall cost per good casting decreased.
In addition to shrinkage, I monitored slag and gas defects. The approximate bottom injection system reduced turbulence, which lowered the slag risk. The ceramic filter captured oxide films and reaction products. The controlled coating thickness and drying reduced the gas risk. The vacuum level was optimized to remove decomposition gas without causing excessive sand compaction. The pouring temperature was kept within a range that provided good fluidity without increasing gas defects. The result was a cleaner casting with fewer internal defects.
The lost foam casting process also offers flexibility in pattern design. Complex internal features, such as the oil galleries, can be formed without cores. This reduces the number of process steps and eliminates core-related defects. However, the foam pattern must be strong enough to withstand coating and sand compaction. The pattern must also decompose completely. If the pattern density is too high or the coating is too thick, the decomposition products may not escape, and gas defects may form. I optimized the pattern density and coating thickness to balance strength and gas evolution. The oil galleries were formed with a uniform coating and a smooth pattern surface to ensure complete decomposition.
The simulation software was a useful tool, but it required calibration. I calibrated the heat transfer coefficient and the mold constant using the trial results. The calibrated model predicted the shrinkage locations more accurately. I used the calibrated model to test additional riser sizes and gating layouts before cutting new tooling. This reduced the number of expensive trials. The final design was validated by production trials. The simulation and trials agreed that the upper hot spot was the critical region and that a larger riser was needed.
The most important practical conclusion was that the feeding path must remain open until the hot spot solidifies. In the 1# riser trial, the riser was too small, and the neck may have frozen too early. In the 2# riser trial, the larger riser and the smaller neck worked together to feed the hot spot. The neck was small enough to concentrate the feed and easy to remove, but not so small that it froze before feeding was complete. The balance between riser modulus and neck modulus was critical. I recommend that for similar ductile iron castings in lost foam casting, the riser modulus should be at least 1.5 times the hot spot modulus, and the neck modulus should be about 0.6 times the riser modulus, subject to simulation and trial verification.
I also recommend that the vacuum be maintained after pouring for at least 15 minutes. This holding time allows the eutectic expansion to act while the mold is rigid. If the vacuum is released too early, the mold may relax, and the expansion may cause wall movement. The holding time should be adjusted based on the casting weight and section thickness. For heavier castings, a longer holding time may be needed. The vacuum level should be high enough to hold the sand but not so high that it reduces permeability. A range of 0.04–0.055 MPa worked well in this case.
The coating is another critical variable. The coating must be uniform and fully dried. Residual moisture causes gas defects. A coating that is too thick reduces permeability and can trap gas. A coating that is too thin may not be strong enough to resist metal penetration. I used a coating thickness of 0.6–0.9 mm and verified it by weight and visual inspection. The oil galleries were coated with a slightly thinner layer to avoid blocking the internal passages. The coating was dried in a controlled oven. The drying schedule was recorded to ensure consistency.
The sand compaction and vibration also affect mold rigidity. The sand must be compacted uniformly around the pattern. If the sand is loose in one area, the mold may move under expansion pressure. If the sand is too tight, the coating may be damaged, or the gas may not escape. I used a controlled vibration cycle and then applied vacuum. The vibration time and amplitude were specified. The sand was filled in layers to avoid bridging and to ensure uniform density. The flask was designed with sufficient strength to resist the expansion pressure. The vacuum was applied through the flask bottom and maintained during pouring.
The pouring process was controlled by a trained operator. The pouring basin was kept full, and the metal stream was kept smooth. The pouring temperature was measured before pouring. The pouring rate was controlled by the nozzle size and the metal head. The mold was filled without interruption. If the pour was interrupted, a cold shut or a gas defect could form. The pouring time was recorded. After pouring, the vacuum was held for the specified time. The casting was then allowed to cool in the flask before shakeout. The shakeout time was controlled to avoid distortion and cracking.
The final process was documented in a work instruction. The instruction included the pattern inspection, coating procedure, sand filling, vibration, vacuum, pouring, holding, cooling, shakeout, cleaning, and inspection. The critical parameters were listed with their target ranges and tolerances. The operators were trained on the new procedure. The first production lots were monitored closely. The results showed that the process was capable and stable. The reject rate due to shrinkage defects was reduced to a low level. The oil galleries passed the leak test consistently. The upper face and hot spot were sound after machining.
In conclusion, I found that the lost foam casting of this heavy-section ductile iron component required a combination of proper gating, adequate risering, and strict process control. The approximate bottom injection system provided smooth filling and reduced slag. The 2# riser with a modulus 1.5 times the hot spot modulus and a neck modulus 0.6 times the riser modulus provided sufficient feeding. The increased machining allowance on the upper face removed minor surface defects. The controlled coating, sand compaction, vacuum, and holding time improved mold rigidity and reduced gas defects. The final process produced sound castings that passed the leak test and machining requirements. The key equations and tables summarized in this work can be used as a starting point for similar ductile iron castings in lost foam casting. The most important lesson is that lost foam casting is not simply a molding method; it is a coupled system in which the foam pattern, coating, sand, vacuum, metal, and gating all interact. A successful process must consider all of these factors together.