Introduction and engineering context. I approached the marine cylinder liner as a large, variable-wall-thickness gray iron component that must be produced reliably by sand casting. The liner is a core part of a high-power marine diesel engine. It experiences combustion pressure, cooling-water thermal cycling, intake and drainage pipe thermal stress, and high-speed friction against the piston. Because these conditions directly affect engine reliability and service life, the casting process must deliver dense, uniform, and defect-free material. I selected HT250 gray cast iron because it offers a favorable combination of wear resistance, thermal conductivity, castability, and cost. However, a conventional sand casting route can still generate shrinkage cavities, shrinkage porosity, inclusions, cracks, and gas defects. These defects reduce performance and lifetime, especially in a large liner with a large bore and variable wall thickness. For this reason, I treated the sand casting process design as a coupled thermal, filling, feeding, and chilling problem rather than a simple mold layout exercise.
I designed the process around a top-poured rain-type gating system, an annular riser, external steel chills, internal graphite chills, and a self-hardening furan resin sand mold. I then used ProCAST numerical simulation to examine mold filling, temperature fields, solidification fields, and defect distribution. The simulation results guided the industrial trial. The measured temperature curve at the large-end center agreed closely with the predicted curve, and the finished casting met the required production quality. This experience confirmed that sand casting can be optimized for large marine cylinder liners when the gating, risering, chilling, and simulation parameters are treated as one integrated system.
Material specification and thermal basis. I used HT250 gray cast iron for the liner. The specified composition range used in the sand casting design is summarized in Table 1. I used the material database to compute density, enthalpy, thermal conductivity, and related thermophysical properties as functions of temperature. Those properties were then assigned to the casting and gating system. The computed liquidus temperature was 1228 °C, and the solidus temperature was 1117 °C. The eutectic arrest appeared near 1150 °C. These temperatures governed the feeding distance, the chilling strategy, and the final solidification location.
| Element | Specified value or limit (mass fraction, %) |
|---|---|
| C | 3.1 |
| Si | 1.0 |
| Mn | 0.5 |
| P | 0.25 |
| S | ≤ 0.06 |
| Cu | 1.4 |
| V | ≤ 0.15 |
| B | 0.02 |
I treated the gray iron as a solidifying alloy with a mushy zone, not as a pure metal with a sharp freezing front. This distinction is important in sand casting because the release of latent heat over a temperature interval changes the local cooling rate and the effectiveness of chills and risers. The energy transport during solidification can be written in the form shown in Equation (1), where the last term represents the latent heat contribution associated with the solid fraction rate.
$$\rho c_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + \rho L \frac{\partial f_s}{\partial t}$$
Equivalently, I used the enthalpy formulation in Equation (2), which is convenient for numerical simulation of sand casting because it avoids explicitly tracking the latent heat source term at every time step.
$$\rho \frac{\partial H}{\partial t} = \nabla \cdot (k \nabla T)$$
Here, \(H\) is the specific enthalpy, \(k\) is thermal conductivity, \(\rho\) is density, \(c_p\) is specific heat, \(L\) is latent heat, and \(f_s\) is solid fraction. The enthalpy is defined by Equation (3).
$$H = \int c_p \, dT + L f_s$$
Cast geometry and wall-thickness classification. The liner was a large cylindrical casting with a net mass of approximately 16,100 kg, a total height of 4,140 mm, and an inner diameter of 890 mm. The wall thickness changed significantly along the height. I divided the casting into three zones for process reasoning: the top large-end zone, the middle zone, and the bottom small-end zone. The top large end behaved as a thick-walled cylinder with a wall thickness of about 230 mm. The middle zone had a wall thickness of about 180 mm. The bottom small end had a wall thickness of about 110 mm. Because these zones contained different metal masses and lost heat at different rates, the sand casting design had to impose a controlled solidification sequence from the bottom and thin sections toward the top and thick sections, with final feeding into the riser.
| Feature | Value |
|---|---|
| Material | HT250 gray cast iron |
| Casting net mass | approximately 16,100 kg |
| Total height | 4,140 mm |
| Inner diameter | 890 mm |
| Top large-end wall thickness | approximately 230 mm |
| Middle wall thickness | approximately 180 mm |
| Bottom small-end wall thickness | approximately 110 mm |
| Overall shape | large cylindrical liner with variable wall thickness |
| Main casting risk | shrinkage porosity, shrinkage cavities, hot spots, thermal stress |
Sand casting process route and gating design. I selected sand casting because it offers low cost, high production flexibility, and broad applicability for medium and large thick-walled castings. In this sand casting route, the mold was made from self-hardening furan resin sand, and a cast iron core bone with a diameter of 700 mm supported the sand core. I chose a top-poured rain-type gating system. A top gating system supports bottom-to-top solidification and provides good feeding capability. The rain-type ingates divide the liquid metal into multiple streams, which reduces erosion of the sand mold during filling and distributes heat more evenly over the top surface. I used two sprue channels with a diameter of 90 mm. The sprues connected to a runner, and the runner delivered metal through 32 ingates with a diameter of 22 mm. The ingates were distributed uniformly around a circle with a diameter of 978 mm.
| Parameter | Value |
|---|---|
| Gating type | top-poured rain-type gating system |
| Sprue count | 2 |
| Sprue diameter | 90 mm |
| Ingate count | 32 |
| Ingate diameter | 22 mm |
| Ingate distribution circle diameter | 978 mm |
| Mold material | self-hardening furan resin sand |
| Core support | cast iron core bone, diameter 700 mm |
| Pouring orientation | vertical, top pouring |
For a top-poured sand casting, the pouring time must be consistent with the pouring weight and the desired filling rate. I estimated the nominal pouring time from the total pouring mass and the pouring rate using Equation (4).
$$t_p \approx \frac{W_p}{\dot{m}}$$
With a pouring mass of approximately 22,150 kg and a pouring rate of 175 kg/s, the nominal time is about 126.6 s. The process design used a nominal filling time of about 120 s, which is close to this estimate and was acceptable for the simulation and trial. The filling behavior in sand casting is also affected by the gating geometry, the permeability of the sand, and the local flow resistance. I therefore did not treat the pouring time as a single isolated number; I evaluated it together with the simulated filling pattern and the observed trial casting.
Riser design and feeding strategy. I placed an annular riser at the top of the large end. The riser had an outer diameter of 1,494 mm and a height of 703 mm. Its mass was approximately 5,750 kg. The riser served three functions in this sand casting: it stored liquid metal for feeding, it collected inclusions and slag, and it guided the final solidification to a location outside the finished liner. The casting mass was 16,100 kg, and the total pouring mass was 22,150 kg. The casting yield was calculated using Equation (5).
$$Y = \frac{W_c}{W_p} \times 100\%$$
$$Y = \frac{16{,}100}{22{,}150} \times 100\% \approx 72.69\%$$
In sand casting of a thick-walled cylinder, the riser must remain liquid longer than the casting section it feeds. A simple modulus criterion is often used as a first approximation, as shown in Equation (6), where \(M\) is the modulus, \(V\) is volume, and \(A\) is cooling surface area.
$$M = \frac{V}{A}$$
The riser modulus should exceed the modulus of the thickest casting section by a suitable factor. I used this principle together with the simulation results to confirm that the annular riser remained the last region to solidify. Chvorinov’s rule, given by Equation (7), also helped me reason about local solidification time, where \(B\) is a mold constant and \(n\) is an exponent usually close to 2 for sand casting.
$$t_s = B \left(\frac{V}{A}\right)^n$$
In this sand casting design, the large-end wall was the dominant hot spot. The annular riser was therefore placed directly above this hot spot. The intended solidification sequence was thin bottom section first, middle section next, large-end section afterward, and riser last. That sequence was later confirmed by the solidification field.
| Parameter | Value |
|---|---|
| Riser type | annular riser |
| Riser location | top large end |
| Riser outer diameter | 1,494 mm |
| Riser height | 703 mm |
| Riser mass | approximately 5,750 kg |
| Casting net mass | approximately 16,100 kg |
| Total pouring mass | approximately 22,150 kg |
| Casting yield | approximately 72.69% |
Chill design for directional solidification. The variable wall thickness made natural cooling alone insufficient. I used external steel chills and internal graphite chills to strengthen directional solidification and reduce shrinkage defects. Graphite chills have high thermal conductivity, strong chilling capacity, and good resistance to fusion welding, making them suitable as internal chills for heavy hot spots and for protecting the inner wall quality. Steel chills are strong and economical, making them suitable as external chills that also support the sand mold structure and can be reused. I placed five rows of steel external chills on the outside of the casting. Each row contained 24 blocks, and each block had dimensions of 40 mm × 40 mm × 200 mm. On the inside, I used a dedicated core chill arrangement with graphite chills.
| Chill type | Material | Quantity or arrangement | Main purpose |
|---|---|---|---|
| External chill | steel | 5 rows, 24 blocks per row | increase heat extraction and support sand mold |
| External chill size | steel | 40 mm × 40 mm × 200 mm | local chilling at outer surface |
| Internal chill | graphite | dedicated core chill | reduce internal hot spots and protect inner wall |
| Core support | cast iron core bone | diameter 700 mm | support the sand core during pouring |
The chills changed the local heat flux at the casting-mold interface. The interfacial heat transfer was modeled using Equation (8), where \(h\) is the interfacial heat transfer coefficient, \(T_c\) is the local casting temperature, and \(T_m\) is the local mold or chill temperature.
$$q = h (T_c – T_m)$$
I assigned a casting-to-sand heat transfer coefficient of 500 W/(m²·K) and a casting-to-chill heat transfer coefficient of 500 W/(m²·K). The external surface of the mold lost heat to air with an effective coefficient of 10 W/(m²·K). These values were selected from the simulation database, practical foundry experience, and the need to reproduce the observed cooling behavior. In sand casting, the interfacial heat transfer coefficient is not a universal constant; it depends on coating, gap formation, surface roughness, and local solidification shrinkage. I therefore treated it as a calibration parameter and checked the final simulation against the measured temperature curve.
Numerical simulation setup. I built the three-dimensional model of the liner and the gating system, converted it to a format readable by the simulation software, and then performed surface meshing and volume meshing. I checked overlapping faces, intersecting faces, and mesh quality. Because the casting was large, I avoided an unnecessarily fine global mesh. However, the gating system and chills contained small features, so I used local mesh refinement. The runner and ingates were meshed at 15 mm, the casting body at 25 mm, and the sand mold and cast iron core bone at 50 mm. The final mesh contained approximately 340,000 surface elements and 3.7 million volume elements.
| Region | Mesh size | Reason |
|---|---|---|
| Runner and ingates | 15 mm | preserve small gating features and flow paths |
| Casting body | 25 mm | balance thermal accuracy and computational cost |
| Sand mold and core bone | 50 mm | reduce total element count |
| Surface mesh | approximately 340,000 elements | sufficient geometric representation |
| Volume mesh | approximately 3.7 million elements | capture solidification and filling behavior |
The casting and gating system were assigned the custom HT250 composition. The mold was assigned resin-bonded sand. The external steel chills were assigned an iron material, and the internal graphite chills were assigned a graphite chill material. The initial temperature of the mold and chills was 30 °C. The pouring temperature was 1,340 °C. I used a gravity casting setup with a pouring rate of 175 kg/s and a nominal pouring time of 120 s. These parameters are summarized in Table 7.
| Parameter | Value |
|---|---|
| Casting material | HT250 gray cast iron |
| Mold material | resin-bonded sand |
| External chill material | iron or steel chill |
| Internal chill material | graphite chill |
| Pouring temperature | 1,340 °C |
| Pouring rate | 175 kg/s |
| Nominal pouring time | 120 s |
| Initial mold temperature | 30 °C |
| Initial chill temperature | 30 °C |
| Gravity direction | vertical, downward |
| Liquidus temperature | 1,228 °C |
| Solidus temperature | 1,117 °C |
| Eutectic arrest | approximately 1,150 °C |
The filling and solidification simulation was based on the continuity equation, the momentum equation, and the energy equation. Equation (9) gives the continuity relation for the liquid and mushy metal.
$$\frac{\partial \rho}{\partial t} + \nabla \cdot (\rho \mathbf{v}) = 0$$
Equation (10) gives the momentum balance, including gravity and a source term for mushy-zone resistance.
$$\frac{\partial (\rho \mathbf{v})}{\partial t} + \nabla \cdot (\rho \mathbf{v} \otimes \mathbf{v}) = -\nabla p + \nabla \cdot (\mu \nabla \mathbf{v}) + \rho \mathbf{g} + \mathbf{S}_m$$
In the mushy zone, I used a Darcy-type resistance to represent flow through the solid-liquid network. Equation (11) shows the permeability relation, where \(K\) is permeability, \(K_0\) is a reference permeability, and \(f_l\) is liquid fraction.
$$K = K_0 \frac{f_l^3}{(1 – f_l)^2}$$
Equation (12) gives the associated mushy-zone velocity relation.
$$\mathbf{v} = -\frac{K}{\mu} (\nabla p – \rho \mathbf{g})$$
For defect prediction, I used the thermal gradient and cooling rate. The Niyama criterion is commonly written as Equation (13), where \(G\) is the magnitude of the temperature gradient and \(\dot{T}\) is the cooling rate.
$$N_y = \frac{G}{\sqrt{|\dot{T}|}}$$
Regions with a low Niyama value are more likely to form shrinkage porosity. I used this criterion together with the solid fraction field and the location of the last liquid regions. In sand casting, the criterion must be interpreted with care because graphite expansion in gray iron can partially compensate for solidification shrinkage. Nevertheless, the combination of thermal gradient, cooling rate, and final liquid location gave a useful prediction of where defects would concentrate.
| Interface | Heat transfer coefficient | Comment |
|---|---|---|
| Casting to sand mold | 500 W/(m²·K) | typical sand casting interface value |
| Casting to chill | 500 W/(m²·K) | calibrated against trial cooling behavior |
| Mold external surface to air | 10 W/(m²·K) | natural air cooling |
Temperature-field analysis. The temperature-field results showed that the metal filled the mold completely. In the early stage of solidification, the external and internal chills extracted heat rapidly, so the inner and outer walls of the large end and middle section cooled faster than the central region. The large-end thick wall remained hotter than the bottom thin wall. This produced a temperature gradient along the height and through the wall thickness. In the later stage, heat accumulated in the riser, creating a clear hot spot. That hot spot was intentional: it kept the riser liquid and allowed shrinkage porosity and shrinkage cavities to concentrate there rather than in the functional liner body.
I extracted temperature-time curves at the center of the large end, the middle section, and the bottom small end. Each curve showed three stages: fast cooling, eutectic transformation, and slow cooling. The fast cooling stage occurred from 0 to about 2 h in all three locations, with the temperature dropping from about 1,340 °C to about 1,150 °C. The cooling rate was high because the initial temperature difference between the liquid metal and the mold was large. After the eutectic stage began, the temperature remained near 1,150 °C for a period that depended on local wall thickness. The large end had the longest eutectic plateau, from about 2 h to 7 h. The middle section had a plateau of about 3 h. The bottom small end had a plateau of about 1 h. After eutectic solidification, the casting entered the slow cooling stage. The large end, middle section, and bottom section began slow cooling at about 7 h, 6 h, and 3 h, respectively.
| Stage | Time range | Temperature range | Physical meaning |
|---|---|---|---|
| Fast cooling | 0 h to approximately 2 h | approximately 1,340 °C to 1,150 °C | large temperature difference between metal and mold |
| Eutectic transformation | approximately 2 h to 7 h depending on section | approximately 1,150 °C | latent heat release and mushy-zone solidification |
| Slow cooling | after approximately 3 h to 7 h depending on section | below approximately 1,150 °C | reduced thermal gradient and lower heat extraction |
| Location | Fast cooling duration | Eutectic plateau duration | Slow cooling start |
|---|---|---|---|
| Top large end | 0 h to 2 h | 2 h to 7 h | approximately 7 h |
| Middle section | 0 h to 2 h | approximately 3 h | approximately 6 h |
| Bottom small end | 0 h to 2 h | approximately 1 h | approximately 3 h |
Solidification-field analysis. The solidification field confirmed the intended sequence. In the axial direction, the bottom thin wall solidified first. The large-end thick wall solidified later. This happened because the thin wall had a larger surface-area-to-volume ratio and therefore lost heat faster. The regions in contact with chills also solidified quickly because of the high local heat extraction. In the radial direction, solidification proceeded from the inner and outer surfaces toward the center. This behavior is consistent with layered mold filling and with the external and internal chilling arrangement. The last region to solidify was the riser. The simulation also indicated that geometric transitions where the radial dimension changed could experience thermal stress concentration. I therefore paid attention to those transitions during process review and trial inspection.
| Direction | First to solidify | Later to solidify | Last region |
|---|---|---|---|
| Axial | bottom small end | middle section and top large end | riser |
| Radial | inner and outer surfaces | central wall region | riser and central hot spot |
| Chill-affected regions | areas near external and internal chills | areas away from chills | riser |
Shrinkage defect prediction. The predicted shrinkage porosity and shrinkage cavities were concentrated in the riser. In the top view, the defects formed a ring-like pattern in the center of the wall thickness. This distribution indicated that the riser performed effective feeding. The defects were not scattered through the functional body of the liner. The simulation therefore supported the selected sand casting design: top rain gating, annular riser, external steel chills, internal graphite chills, and resin-bonded sand mold. The defect prediction is summarized in Table 12.
| Defect type | Predicted location | Interpretation |
|---|---|---|
| Shrinkage porosity | mainly in the riser | riser remained liquid and fed the casting |
| Shrinkage cavity | riser, ring-like central region | final solidification occurred in the riser |
| Inclusions | not dominant in the casting body | top gating and riser helped collect impurities |
| Gas defects | not dominant | sand permeability and filling design were adequate |
| Cracks | not dominant | directional solidification reduced thermal stress concentration |
Industrial trial and production validation. I conducted an industrial trial based on the simulation results. The mold filling was complete, and the casting surface was smooth. The internal structure was uniform, and no obvious shrinkage porosity or shrinkage cavity was found in the functional body of the liner. These results matched the simulation prediction. The sand casting process was therefore considered capable of producing the required marine cylinder liner. I also used a thermocouple to measure the temperature at the center of the large end during production. I compared the measured curve with the simulated curve. The liquidus temperature was 1,228 °C, and the solidus temperature was 1,117 °C. The two curves showed good agreement throughout the cooling process.
| Item | Trial observation | Comparison with simulation |
|---|---|---|
| Mold filling | complete | consistent |
| Surface quality | smooth | consistent |
| Internal structure | uniform | consistent |
| Shrinkage defects | no obvious defects in the casting body | consistent with riser concentration |
| Large-end cooling curve | measured by thermocouple | well fitted to simulated curve |
To evaluate the fit between the measured and simulated temperature curves at the large end, I used the root mean square error and the mean absolute error. Equation (14) gives the root mean square error.
$$RMSE = \sqrt{\frac{1}{N} \sum_{i=1}^{N} \left(T_{sim,i} – T_{exp,i}\right)^2}$$
Equation (15) gives the mean absolute error.
$$MAE = \frac{1}{N} \sum_{i=1}^{N} \left|T_{sim,i} – T_{exp,i}\right|$$
The root mean square error was 8.62 °C, and the mean absolute error was 7.07 °C. These values are small relative to the temperature range of the process and indicate that the sand casting simulation was reliable for this large gray iron liner. The measured curve and the simulated curve both showed the same fast cooling stage, eutectic plateau, and slow cooling stage. The close agreement increased my confidence that the thermal boundary conditions, material properties, and interfacial heat transfer coefficients were reasonably calibrated.
| Metric | Value | Interpretation |
|---|---|---|
| Root mean square error | 8.62 °C | good overall fit |
| Mean absolute error | 7.07 °C | small average deviation |
| Liquidus temperature | 1,228 °C | material thermal reference |
| Solidus temperature | 1,117 °C | material thermal reference |
| Cooling behavior | three-stage response | predicted and measured curves matched |
Discussion of the sand casting design logic. The success of this work depended on treating sand casting as a thermal management problem. A large marine cylinder liner cannot be produced well by simply pouring metal into a sand mold and hoping for the best. The mold must fill without excessive turbulence, the riser must remain liquid until the casting has solidified, and the chills must impose a favorable thermal gradient. In this design, the top rain gating system reduced sand erosion and distributed the incoming metal around the circumference. The annular riser created a final hot spot at the top. The external steel chills and internal graphite chills increased heat extraction at selected surfaces. The resin-bonded sand mold provided the necessary strength, collapsibility, and permeability for a large sand casting.
The variable wall thickness was the central difficulty. The bottom small end had a wall thickness of about 110 mm, while the top large end had a wall thickness of about 230 mm. If the casting were allowed to cool naturally, the thin bottom would solidify first and the thick top would remain liquid for a long time. That alone is not sufficient, because the thermal center of the thick wall can still form isolated liquid pockets. The chills and riser together prevented those pockets from appearing in the functional body. The solidification simulation showed that the last liquid was in the riser. The defect prediction showed that shrinkage porosity and shrinkage cavities were concentrated in the riser. The industrial trial confirmed that the functional body was sound. This chain of evidence supports the use of sand casting with simulation-based process optimization for large gray iron cylinder liners.
I also note that gray iron is not a simple alloy. The carbon and silicon contents influence graphite formation, eutectic temperature, and expansion behavior. The copper content can affect matrix structure and hardness. The phosphorus and sulfur levels must be controlled to avoid undesirable phases and defects. In sand casting, the cooling rate interacts with composition. A slow-cooling thick section may develop a different graphite morphology and matrix than a fast-cooling thin section. For this reason, the thermal design and the metallurgical specification must be considered together. The temperature-time curves obtained in this study provide a useful basis for understanding how the large end, middle section, and small end cool at different rates.
Key process equations and their practical meaning. The heat conduction equation, the enthalpy equation, the continuity equation, the momentum equation, the Darcy relation, the Chvorinov rule, the modulus criterion, the Niyama criterion, and the error metrics formed the quantitative backbone of the sand casting design. Table 15 summarizes these equations and their role in the project.
| Equation or criterion | Expression | Role in this work |
|---|---|---|
| Heat conduction with latent heat | \(\rho c_p \frac{\partial T}{\partial t} = \nabla \cdot (k \nabla T) + \rho L \frac{\partial f_s}{\partial t}\) | describe solidification heat transport |
| Enthalpy formulation | \(\rho \frac{\partial H}{\partial t} = \nabla \cdot (k \nabla T)\) | numerical simulation of phase change |
| Continuity | \(\frac{\partial \rho}{\partial t} + \nabla \cdot (\rho \mathbf{v}) = 0\) | mass conservation during filling |
| Momentum | \(\frac{\partial (\rho \mathbf{v})}{\partial t} + \nabla \cdot (\rho \mathbf{v} \otimes \mathbf{v}) = -\nabla p + \nabla \cdot (\mu \nabla \mathbf{v}) + \rho \mathbf{g} + \mathbf{S}_m\) | fluid flow and gravity effects |
| Darcy mushy-zone flow | \(\mathbf{v} = -\frac{K}{\mu} (\nabla p – \rho \mathbf{g})\) | flow resistance in the mushy zone |
| Permeability | \(K = K_0 \frac{f_l^3}{(1 – f_l)^2}\) | represent liquid flow through solid network |
| Interfacial heat flux | \(q = h (T_c – T_m)\) | casting-mold and casting-chill heat exchange |
| Modulus | \(M = \frac{V}{A}\) | riser and feeding design |
| Chvorinov rule | \(t_s = B \left(\frac{V}{A}\right)^n\) | estimate local solidification time |
| Niyama criterion | \(N_y = \frac{G}{\sqrt{|\dot{T}|}}\) | predict shrinkage porosity risk |
| Pouring time | \(t_p \approx \frac{W_p}{\dot{m}}\) | connect pouring mass and filling rate |
| Casting yield | \(Y = \frac{W_c}{W_p} \times 100\%\) | evaluate riser and gating efficiency |
| Root mean square error | \(RMSE = \sqrt{\frac{1}{N} \sum_{i=1}^{N} (T_{sim,i} – T_{exp,i})^2}\) | validate simulated temperature curve |
| Mean absolute error | \(MAE = \frac{1}{N} \sum_{i=1}^{N} |T_{sim,i} – T_{exp,i}|\) | validate average prediction error |
Production implications and quality control. In production, I would control the sand casting process using several checkpoints. The first is the sand mold quality. The self-hardening furan resin sand must have sufficient strength, permeability, and collapsibility. The core assembly must be stable because the casting is tall and heavy. The second is the gating and pouring practice. The rain-type ingates must be clean and properly aligned. The pouring temperature and pouring rate must be kept within the validated range. The third is the chill condition. Steel chills must be clean, dry, and correctly positioned. Graphite chills must be inspected for cracks or contamination. The fourth is the riser condition. The annular riser must be large enough to feed the thick section and must not freeze before the casting body. The fifth is the thermal record. A thermocouple at the large-end center provides a useful process signature. If the measured curve deviates significantly from the validated curve, I would investigate the mold, melt, pouring, or chill conditions.
| Control point | Target | Reason |
|---|---|---|
| Mold sand | resin-bonded sand with adequate strength and permeability | withstand filling and allow gas escape |
| Core assembly | stable, centered, and supported by core bone | maintain wall thickness and dimensional accuracy |
| Gating system | clean sprues, runners, and 32 ingates | avoid turbulence and sand erosion |
| Pouring temperature | approximately 1,340 °C | ensure fill and feeding without overheating |
| Pouring rate | approximately 175 kg/s, nominal 120 s fill | match simulation and avoid cold shut |
| External chills | 5 rows, 24 steel blocks per row | promote directional solidification |
| Internal chills | dedicated graphite core chills | reduce internal hot spots |
| Riser | annular, 1,494 mm outer diameter, 703 mm height | feed the large-end hot spot |
| Thermal record | large-end center thermocouple | verify simulation and process stability |
Comparison with purely empirical sand casting development. A purely trial-and-error sand casting approach would require repeated large melts, repeated mold making, repeated machining, and destructive sectioning. For a 16,100 kg liner, that path is expensive and slow. Numerical simulation reduces the number of trials by predicting filling, temperature, solidification, and defects before the first industrial pour. In this work, simulation showed that the riser would be the last region to freeze and that defects would concentrate there. The industrial trial confirmed this prediction. The measured temperature error was small. I therefore regard simulation as a practical extension of sand casting process knowledge, not as a replacement for foundry experience. The best result comes from combining simulation with material knowledge, mold design, melting practice, and shop-floor observation.
Role of sand casting in large marine component manufacturing. Sand casting remains attractive for large marine components because it can produce complex internal cavities, thick sections, and large overall dimensions at reasonable cost. The sand mold can accommodate chills, cores, risers, and gating systems that would be difficult or impossible in permanent mold casting. For a cylinder liner, the sand casting route also allows the foundry to adjust the mold and core materials for the required surface finish and dimensional tolerance. The main disadvantage is that sand casting introduces process variability: sand compaction, binder content, moisture, coating, chill contact, and pouring practice can all affect the result. Simulation helps reduce this variability by identifying the critical thermal windows. In this project, the critical window was the large-end hot spot and its feeding path. Once that window was controlled, the rest of the sand casting process became more robust.
Thermal gradient and feeding distance. The feeding distance in gray iron depends on the thermal gradient, the solidification interval, and the graphite expansion behavior. A higher thermal gradient helps feed liquid metal toward the last solidifying region. The chills increase the thermal gradient between the chilled surface and the central hot spot. The riser then provides the liquid reservoir. In this design, the external steel chills and internal graphite chills created a dual cooling effect. The outer surface and inner surface cooled toward the center, while the riser remained hot at the top. This is a three-dimensional feeding pattern. The simulation showed that the center of the wall was the last region in the radial direction, but the riser was the last region overall. This means the feeding path was directed upward and inward toward the riser, which is exactly what I wanted. Equation (16) expresses the local thermal gradient magnitude.
$$G = |\nabla T|$$
Equation (17) expresses the local cooling rate.
$$\dot{T} = \frac{\partial T}{\partial t}$$
Equation (18) expresses the Fourier number, which is useful for reasoning about transient heat conduction in the sand casting mold.
$$Fo = \frac{\alpha t}{L^2}$$
Here, \(\alpha\) is thermal diffusivity and \(L\) is a characteristic length. A larger Fourier number indicates that heat has diffused more deeply into the mold during the time of interest. For a large sand casting, the mold does not reach a uniform temperature during the pour. The thermal field is transient throughout much of the solidification process. This is one reason why chills and risers must be designed together rather than independently.
Filling behavior and gating hydraulics. The top-poured rain gating system divided the metal into 32 streams. This division reduced the momentum of any single stream and limited direct impingement on the sand mold. It also increased the area over which the hot metal entered the mold, which helped distribute heat at the top. The two 90 mm sprues supplied the runner. The 32 ingates of 22 mm diameter controlled the final distribution. In sand casting, the gating system must avoid excessive velocity because high velocity can erode the mold, entrain air, and create turbulence that leads to oxidation and gas defects. It must also avoid too low a velocity because slow filling can cause cold shuts or misruns. The selected pouring rate of 175 kg/s and nominal time of 120 s were a compromise between these risks. The simulation showed complete filling, and the trial casting had a smooth surface, which supported the gating design.
Thermal stress and cracking risk. Large castings with variable wall thickness are also at risk of thermal stress and cracking. The thin bottom section cools and contracts before the thick top section. If the mold or core restricts contraction, tensile stresses can develop. The sand casting route helps because the resin-bonded sand mold and core can collapse after the metal solidifies. The graphite and steel chills must not lock the casting excessively. The riser also helps by keeping the top region hot and reducing the temperature difference between adjacent sections. The simulation did not predict dominant crack formation, and the trial casting showed no obvious cracks. This does not eliminate the need for careful shakeout and heat treatment practice, but it confirms that the thermal design was reasonable.
Graphite expansion and gray iron feeding. Gray cast iron can exhibit graphite expansion during eutectic solidification. This expansion can partly compensate for liquid and solidification shrinkage. However, the compensation is not uniform. It depends on carbon equivalent, cooling rate, inoculation, and mold stiffness. In a large sand casting, the mold can expand or yield slightly under internal pressure. The riser and chills must still be designed to handle the net shrinkage. I did not rely on graphite expansion as a substitute for feeding. Instead, I used it as a secondary factor that can reduce the required riser size, while the primary feeding design remained based on modulus, Chvorinov’s rule, and simulation. The predicted defects concentrated in the riser, which means the net feeding demand was satisfied outside the functional body.
Simulation calibration and uncertainty. The accuracy of sand casting simulation depends on many inputs: thermophysical properties, interfacial heat transfer, mold properties, chill properties, filling rate, and initial temperatures. I treated the simulation as a calibrated model rather than an exact prediction. The most uncertain parameter was the interfacial heat transfer coefficient. A value of 500 W/(m²·K) was used for both casting-to-sand and casting-to-chill interfaces. In reality, these coefficients vary with temperature, gap formation, coating, and contact pressure. The fact that the measured and simulated curves agreed within an RMSE of 8.62 °C suggests that the selected values were reasonable for this sand casting process. If the mold coating, chill condition, or sand binder were changed, I would recalibrate the model. This is an important point for production: simulation is most valuable when it is anchored to measured plant data.
Scale-up and repeatability. The liner is a large casting, and the process must be repeatable. A single successful trial is not enough for full production. I would use the validated simulation and the trial thermal record to establish a process window. The pouring temperature could vary within a small range around 1,340 °C. The pouring rate could vary around 175 kg/s. The mold and chill initial temperatures should remain near 30 °C. The sand properties should be controlled within a narrow range. The large-end thermocouple record can serve as a fingerprint for each pour. If the fingerprint is consistent, the solidification sequence is likely consistent. If it shifts, the foundry can investigate before committing more castings. This is a practical way to apply simulation and sand casting knowledge together.
| Variable | Recommended window | Monitoring method |
|---|---|---|
| Pouring temperature | near 1,340 °C | immersion thermocouple or optical pyrometer |
| Pouring rate | near 175 kg/s | pouring mass and time record |
| Nominal fill time | near 120 s | pour recorder |
| Mold initial temperature | near 30 °C | mold surface thermocouple |
| Chill initial temperature | near 30 °C | chill inspection and temperature check |
| Sand strength and permeability | foundry specification | sand testing |
| Large-end center cooling | match validated curve | embedded thermocouple |
Summary of results. The sand casting process produced a complete fill and a sound marine cylinder liner. The solidification sequence was directional: bottom small end first, large end later, and riser last. The temperature-time curves showed a fast cooling stage, a eutectic plateau, and a slow cooling stage. The large end had the longest eutectic plateau because it had the greatest wall thickness. The shrinkage porosity and shrinkage cavities were concentrated in the riser. The external steel chills and internal graphite chills helped control the thermal field. The industrial trial matched the simulation, and the measured large-end temperature curve agreed with the predicted curve within 8.62 °C RMSE and 7.07 °C MAE. The finished liner had a smooth surface, uniform internal structure, and no obvious casting defects in the functional body.
Conclusions. I designed and validated a sand casting process for a large HT250 gray iron marine cylinder liner with a large bore and variable wall thickness. The process combined a top-poured rain-type gating system, an annular riser, external steel chills, internal graphite chills, and a self-hardening furan resin sand mold. The gating system used two 90 mm sprues and 32 ingates of 22 mm diameter arranged on a 978 mm circle. The annular riser had a 1,494 mm outer diameter and a 703 mm height. The casting mass was approximately 16,100 kg, the riser mass was approximately 5,750 kg, the pouring mass was approximately 22,150 kg, and the casting yield was approximately 72.69%. The pouring temperature was 1,340 °C, the pouring rate was approximately 175 kg/s, and the nominal fill time was approximately 120 s.
The simulation showed complete filling, a controlled solidification sequence, and defect concentration in the riser. The bottom small end solidified first, the middle section followed, the large end solidified later, and the riser remained last. The temperature-time curves showed a fast cooling stage from 0 to about 2 h, an eutectic plateau near 1,150 °C, and a slow cooling stage after 3 h to 7 h depending on wall thickness. The large end had the longest eutectic plateau from about 2 h to 7 h. The defect prediction showed that shrinkage porosity and shrinkage cavities were mainly in the riser, which confirmed effective feeding. The industrial trial produced a sound casting with a smooth surface and uniform internal structure. The measured large-end temperature curve matched the simulated curve with an RMSE of 8.62 °C and an MAE of 7.07 °C. These results confirm that simulation-based sand casting design is practical, reliable, and beneficial for large marine cylinder liners. The same approach can be applied to other large gray iron castings with variable wall thickness, where the main challenge is to combine filling control, directional solidification, riser feeding, and chill design into a single robust sand casting process.