I treated the large marine cast steel stem post as a high-risk sand casting project from the first review meeting. The component is long, curved, thin in many regions, and locally massive, so every sand casting decision has to balance feeding, distortion, dimensional control, and inspectability. My objective was not merely to produce a sound casting once, but to establish a repeatable sand casting route that could survive full non-destructive examination, mechanical property verification, and final dimensional acceptance. I therefore combined foundry experience, thermal analysis, sand casting simulation, and cross-software validation to identify why an initial sand casting route produced shrinkage porosity and how the process should be redesigned.
1. First-Person View of the Component and Its Sand Casting Challenge
I began by defining the geometry because sand casting performance is controlled by geometry, modulus, and feeding distance. The stem post is more than 3500 mm in length, with a curved length close to 4000 mm. Its maximum width is 2570 mm, while the minimum width is only 290 mm. The maximum wall thickness reaches 300 mm, and the minimum wall thickness is 50 mm. This is a severe variable cross-section casting. In sand casting terms, the thin sections cool quickly, the thick sections remain liquid longer, and the transition zones become natural hot spots where feeding channels can collapse before the liquid metal has finished solidifying.
| Parameter | Value | Sand casting consequence |
|---|---|---|
| Overall length | More than 3500 mm | Long feeding path and high distortion risk |
| Curved length | Approximately 4000 mm | Difficult pattern layout and core location |
| Maximum width | 2570 mm | Large sand mold area and possible mold wall movement |
| Minimum width | 290 mm | Thin, fast-cooling zones |
| Maximum wall thickness | 300 mm | Hot spots and feeding demand |
| Minimum wall thickness | 50 mm | Risk of cold shuts and incomplete fill |
| Mass ratio target | 0 to +3% | Tight sand casting dimensional control |
| Dimensional tolerance | DCTG 10 and DCTG 12 | Strict length and width acceptance |
| Out-of-plane deviation | Not more than 3 mm | Requires camber and fixture control |
I expected the sand casting route to be difficult because the stem post tends to straighten during cooling and heat treatment. The internal cavity and the external curved shell also create competing thermal gradients. When I evaluate sand casting routes for such a part, I do not only ask whether the mold can be filled. I ask whether every local hot spot can be fed, whether the sand cores can resist deformation, whether the chill system can create directional solidification, and whether the cleaning and riser removal operations will introduce new defects.
2. Material Requirements and Their Effect on Sand Casting
The material is a cast steel with a specified minimum yield strength of 370 MPa. The chemical composition window is narrow enough that I treated melting and pouring as part of the sand casting quality system rather than as separate operations. Phosphorus and sulfur are each required to be below 0.01% in product analysis, so I had to control raw material cleanliness, deoxidation, and slag conditions carefully. In sand casting, residual elements and gas content influence feeding behavior, hot tearing tendency, and impact toughness.
| Element | Required range or maximum, mass fraction | Sand casting process implication |
|---|---|---|
| C | 0.10 to 0.13% | Balances strength, hardness, and shrinkage |
| Si | 0.10 to 0.60% | Deoxidation and fluidity control |
| Mn | 0.40 to 0.80% | Desulfurization and toughness control |
| Ni | 1.30 to 1.70% | Toughness and hardenability |
| Cr | Not more than 0.040% | Restricted for the specified chemistry |
| V | Not more than 0.20% | Trace strengthening control |
| Cu | 0.50 to 0.80% | Strength contribution and corrosion consideration |
| P | Less than 0.01% | Reduces hot tearing and embrittlement risk |
| S | Less than 0.01% | Controls sulfide inclusions and toughness |
I used a carbon equivalent expression to screen the melt before pouring. Although carbon equivalent is more commonly used for cast irons, the expression still helps me compare the combined effect of alloying elements on solidification and shrinkage tendency:
$$CE = C + \frac{Mn}{6} + \frac{Cr + Mo + V}{5} + \frac{Ni + Cu}{15}$$
The mechanical property requirements are also demanding for sand casting. The test blocks must represent the production casting, be poured from the same heat, and undergo the same heat treatment. I required samples from near-surface, one-quarter wall thickness, and one-half wall thickness positions so that I could verify through-thickness consistency. This matters because sand casting solidification gradients can produce local differences in microstructure, especially in heavy sections and near feeding zones.
| Property | Requirement | Verification concern |
|---|---|---|
| Tensile strength, Rm | At least 490 MPa | Confirms alloy and heat treatment |
| Yield strength, ReL | At least 370 MPa | Primary design requirement |
| Elongation, A | At least 20% | Ductility and cleanliness |
| Reduction of area, Z | At least 40% | Internal soundness and inclusion control |
| Impact energy at minus 40 degrees C, KV2 | At least 27 J | Toughness and heat treatment control |
I required 100% magnetic particle testing on surfaces and ultrasonic testing on all external surfaces with a sensitivity equivalent to a 3 mm flat-bottom hole. For sand casting, this means the process must deliver more than a visually acceptable skin. It must deliver internal soundness. Any shrinkage porosity, slag inclusion, sand inclusion, or gas porosity that exceeds acceptance limits becomes a rejection risk. In my view, the non-destructive testing requirements transformed the sand casting design from a conventional feeding exercise into a fully coupled thermal, metallurgical, and dimensional problem.
| Inspection method | Coverage | Acceptance intent |
|---|---|---|
| Magnetic particle testing | 100% of accessible surfaces | Detect surface and near-surface discontinuities |
| Ultrasonic testing | All external surfaces | Detect internal shrinkage, cracks, and inclusions |
| Dimensional inspection | Full casting and interfaces | Confirm line type, length, width, and distortion |
| Mechanical testing | Representative test blocks | Confirm strength, ductility, and toughness |
3. Castability Analysis Before the First Sand Casting Trial
I analyzed castability using the modulus method. The modulus is the volume-to-surface-area ratio:
$$M = \frac{V}{A}$$
For a simple plate of thickness t, the modulus is approximately half the thickness when both faces cool:
$$M \approx \frac{t}{2}$$
For a long cylinder of diameter D, the modulus is approximately one-quarter of the diameter:
$$M \approx \frac{D}{4}$$
Chvorinov’s rule relates solidification time to modulus:
$$t_s = k M^n$$
where k is a mold and metal constant and n is usually close to 2. I used this relationship qualitatively because the stem post is not a simple shape. The thick sections have a high modulus and require feeding. The thin sections have a low modulus and freeze quickly. If the feeding path from a riser to a hot spot passes through a thin section, the path can close before the hot spot is fed. That is exactly the kind of failure I later observed.
| Region type | Approximate modulus relationship | Solidification behavior | Sand casting risk |
|---|---|---|---|
| Thin shell | M roughly t/2 | Fast freezing | Cold shut, incomplete fill, early feeding blockage |
| Medium wall | M between thin and heavy values | Moderate freezing | Transition hot spots |
| Heavy section | M greater than thin shell | Slow freezing | Shrinkage cavity and porosity |
| Curved transition | Variable and asymmetric | Complex thermal gradient | Isolated liquid pocket |
| Core cavity | Local hot spot at core surface | Retarded cooling | Core-related shrinkage and distortion |
I also considered the feeding distance. A practical form is:
$$L_f = f(M_r, M_c, \Delta T, k)$$
where Lf is the allowable feeding distance, Mr is the riser modulus, Mc is the casting modulus, Delta T is the superheat or thermal gradient, and k represents the thermal properties of the mold and metal. The exact function is empirical, but the principle is constant: the riser must remain liquid longer than the feeding path, and the path must remain open long enough to transfer liquid metal into the last solidifying region.
For a riser to feed a casting section, I used the standard modulus requirement:
$$M_r \ge 1.3 M_c$$
I also estimated the required riser volume from the solidification contraction:
$$V_r = \frac{\beta V_c}{\eta_f}$$
where beta is the volumetric solidification contraction, Vc is the casting volume to be fed, and eta_f is the feeding efficiency of the riser. A riser can only be effective if it remains liquid, has a sufficient hot zone, and is connected to the casting through an open feed path. Padding and chills are used to modify the thermal gradient so that the feed path stays open.
4. Initial Sand Casting Process Design
My initial sand casting design used a rotated molding orientation to reduce the overall height and improve feeding. I selected a sand box plus core assembly method. The wooden pattern was made with core prints, and the internal cavities were formed by multiple sand cores. Because the stem post behaves like a long bar with variable cross-section, I used six risers along the length to provide feeding. I placed external chills between risers to create cooling zones and to increase the effective feeding distance.
The initial riser design followed the rule that the riser modulus should be at least 1.3 times the local casting modulus. I added padding from the hot spot toward the riser using a circular rolling method. The purpose of the padding was to create a tapered feed channel that would remain liquid after the surrounding thin wall had frozen. I also applied separate shrinkage allowances in the length, width, and height directions. To counteract distortion, I introduced gradual camber correction. The intention was that the casting would distort during cooling and heat treatment but would end up within the required dimensional tolerance after processing.
| Initial sand casting decision | Design intent | Assumed benefit |
|---|---|---|
| Rotate the casting for molding | Reduce height and improve feeding | Better riser placement and easier molding |
| Use sand box plus core assembly | Form complex internal cavities | Accurate core location and cavity shape |
| Use six risers | Feed the long variable section | Multiple feeding centers along the length |
| Use external chills | Create directional solidification | Increase feeding distance between risers |
| Use padding near hot spots | Maintain an open feed channel | Prevent isolated liquid regions |
| Apply directional shrinkage allowances | Compensate thermal contraction | Dimensional control after cooling |
| Apply camber correction | Counteract straightening tendency | Final line type within tolerance |
I used MAGMA for the initial sand casting simulation. The process components were imported as STL files, meshed, and solved with the sand casting module. The mesh contained approximately 20 million tetrahedral cells. I set the sand mold properties, alloy properties, interfacial heat transfer coefficients, initial temperatures, and filling conditions. The initial simulation predicted that the casting body had no major shrinkage cavity and reported a minimum safe height of 164 mm.
| Initial simulation item | Value or setting |
|---|---|
| Software | MAGMA |
| Process | Sand casting |
| Mesh type | Tetrahedral |
| Total mesh elements | Approximately 20 million |
| Initial prediction | No major shrinkage cavity in the body |
| Minimum reported safe height | 164 mm |
| Feeding design | Six risers, padding, external chills |
At that stage, I believed the sand casting design was conservative enough. The simulation did not show a critical defect in the location that later failed. However, I later learned that the boundary conditions used in the first simulation did not capture the local thermal resistance created by the curved transition, the actual padding geometry, and the competing feeding behavior between adjacent risers.
5. First Production Casting and Defect Discovery
The first production casting was poured, insulated, heat treated, and inspected. The mechanical test blocks passed. The problem appeared during finishing, after the residual riser and padding were removed by gas cutting. I found a visible shrinkage cavity on the side of a blind riser. I removed material carefully by carbon arc gouging to determine the extent. The defect was approximately 350 mm long and 60 mm deep, and it extended obliquely upward.
| Observation | Measured or observed condition | Interpretation |
|---|---|---|
| Location | Side of a blind riser | Feeding path failure near riser contact |
| Visible defect | Shrinkage cavity | Isolated liquid region during solidification |
| Length | Approximately 350 mm | Not a local pinhole; a connected feeding defect |
| Depth | Approximately 60 mm | Required removal and repair evaluation |
| Direction | Oblique upward | Followed the last solidifying channel |
| Mechanical blocks | Passed before discovery | Defect was local and not represented by test blocks |
I reviewed the entire sand casting chain: design, pattern, molding, core making, melting, pouring, shakeout, heat treatment, and cleaning. I checked people, machines, materials, methods, environment, and measurement. The melt chemistry was within specification. The molding, pouring, shakeout, and heat treatment records showed no abnormal events. The operations followed the approved sand casting procedure. Therefore, I returned to the thermal design and the simulation assumptions.
I modified the boundary conditions in the sand casting simulation to represent the actual local conditions more closely. When I did that, I reproduced shrinkage porosity at the same location where the production casting failed. The simulation showed a red shrinkage region at the blind riser side. It also revealed that the padding next to that riser solidified before the hot spot could be fed. The adjacent larger riser did not have enough feeding capacity because the feed path was not open. The result was an isolated liquid region that formed a secondary shrinkage cavity extending obliquely upward.
The root cause was not a single operator error. It was a sand casting design error. The original concept relied on the blind riser to feed upward and sideways into a curved transition. In reality, the curved transition had a larger local modulus than the original model assumed. The padding solidified too early. The adjacent riser was too small and too far away to compensate. The external chill arrangement did not create the intended directional gradient in that local region. I summarized the cause as a blocked feeding channel caused by incorrect riser-to-riser feeding logic.
| Potential cause category | Finding | Effect on sand casting quality |
|---|---|---|
| Design | Feeding direction was incorrect for the curved hot spot | Isolated liquid region |
| Simulation boundary condition | Local thermal resistance was underestimated | False confidence in initial result |
| Padding | Padding solidified before the hot spot | Feed channel closed early |
| Adjacent riser | Riser size and distance were insufficient | Inadequate compensatory feeding |
| Chill placement | Chill effect did not dominate the local gradient | No directional solidification at the defect site |
| Geometry | Curved transition created a hidden hot spot | Higher local modulus than expected |
6. Thermal Reasoning Behind the Defect
I used solidification fraction to describe the sequence:
$$f_s = \frac{T_L – T}{T_L – T_S}$$
where f_s is solid fraction, T_L is liquidus temperature, T_S is solidus temperature, and T is local temperature. A feeding path remains useful only while the path is above a critical solid fraction and the riser is still liquid. The Niyama criterion is useful for porosity risk:
$$N_y = \frac{G}{\sqrt{\dot{T}}}$$
where G is the temperature gradient and dot T is the cooling rate. Low Niyama values indicate a high risk of shrinkage porosity. In the failed location, the gradient toward the original blind riser was not strong enough, and the cooling rate was not high enough to compensate. The liquid pocket therefore remained isolated.
The heat conduction equation for the sand casting solidification problem can be written as:
$$\rho c_p \frac{\partial T}{\partial t} = \nabla \cdot \left(k \nabla T\right) + \rho L \frac{\partial f_s}{\partial t}$$
where rho is density, c_p is specific heat, k is thermal conductivity, L is latent heat, and f_s is solid fraction. The sand mold, chill, padding, and riser all change the boundary conditions of this equation. If the local boundary condition is wrong, the predicted feeding behavior is wrong even if the mesh is fine and the material data are reasonable.
The feeding condition can be expressed conceptually as:
$$t_{path} < t_{hotspot}$$
where t_path is the time during which the feed channel remains open and t_hotspot is the time required for the hot spot to finish solidifying. In the failed casting, the padding path solidified first, so t_path became shorter than t_hotspot. The liquid metal in the hot spot could not be replenished.
I also considered the riser volume requirement. The riser must contain enough liquid to compensate for the solidification contraction of the fed region:
$$V_r \ge \frac{\beta V_c}{\eta_f}$$
When the feed path is blocked, the effective fed volume Vc decreases to near zero for the isolated region, even if the riser itself contains liquid metal. This explains why the presence of a riser does not guarantee feeding. The riser, padding, chill, and casting geometry must function as a connected thermal system.
7. Optimized Sand Casting Design
I redesigned the sand casting process by changing the feeding direction. Instead of relying on the smaller blind riser to feed sideways and upward into the curved transition, I used the larger adjacent riser to feed downward from above. This changed the gravity-assisted feeding direction and placed the largest liquid reservoir directly above the hot spot.
I enlarged the upper riser from a diameter of 400 mm and height of 500 mm to a diameter of 700 mm and height of 880 mm. I increased the padding below that riser and changed it to a trapezoidal gradient with a width of 400 mm. The new padding fully covered the previously defective location. I also added eleven external chills between the smaller riser and the enlarged riser to create a stronger cooling partition and to promote directional solidification.
I enlarged the smaller riser from a diameter of 550 mm to 600 mm. I moved it closer to the adjacent riser. The center distance between the smaller riser and the neighboring riser was reduced from 1450 mm to 1028 mm. This increased the local feeding liquid volume and reduced the comprehensive feeding burden on the enlarged riser. The combination of a larger upper riser, a wider trapezoidal padding path, additional chills, and a shorter riser spacing created a more robust sand casting feeding system.
| Optimization action | Original condition | Optimized condition | Purpose |
|---|---|---|---|
| Feeding direction | Side and upward from smaller blind riser | Downward from enlarged upper riser | Use gravity and a larger liquid reservoir |
| Upper riser size | 400 mm diameter by 500 mm height | 700 mm diameter by 880 mm height | Increase feeding volume and modulus |
| Padding below upper riser | Smaller and local | 400 mm wide trapezoidal gradient | Keep the feed channel open |
| External chills | Original arrangement | Eleven chills between two risers | Create directional solidification |
| Smaller riser size | 550 mm diameter | 600 mm diameter | Add local feeding capacity |
| Riser center distance | 1450 mm | 1028 mm | Improve interaction between risers |
| Defect area coverage | Partial | Full coverage by padding and chill zone | Eliminate isolated liquid pocket |
I recalculated the riser modulus after the change. For the enlarged upper riser, the modulus can be approximated as:
$$M_r \approx \frac{V_r}{A_r}$$
For a cylindrical riser with diameter D and height H, excluding the bottom contact area, the modulus is approximately:
$$M_r \approx \frac{D H}{4H + D}$$
For the original riser, D = 400 mm and H = 500 mm:
$$M_{r,original} \approx \frac{400 \times 500}{4 \times 500 + 400} = \frac{200000}{2400} \approx 83.3 \text{ mm}$$
For the optimized riser, D = 700 mm and H = 880 mm:
$$M_{r,optimized} \approx \frac{700 \times 880}{4 \times 880 + 700} = \frac{616000}{4220} \approx 146.0 \text{ mm}$$
The modulus increase is substantial. This supports the observed improvement in feeding. I also verified that the local casting modulus at the curved transition was lower than the optimized riser modulus divided by 1.3. The requirement is:
$$M_r \ge 1.3 M_c$$
If the local casting modulus Mc is estimated at 100 mm, the required riser modulus is 130 mm. The original riser modulus was only about 83 mm, which is insufficient. The optimized riser modulus is about 146 mm, which satisfies the requirement. This simple calculation explains why the original design failed and why the optimized design is more reliable.
| Riser | Diameter, mm | Height, mm | Approximate modulus, mm | Assessment |
|---|---|---|---|---|
| Original upper riser | 400 | 500 | 83.3 | Insufficient for a 100 mm local modulus |
| Optimized upper riser | 700 | 880 | 146.0 | Sufficient for a 100 mm local modulus |
| Original smaller riser | 550 | Not listed | Not controlling | Limited compensatory feeding |
| Optimized smaller riser | 600 | Not listed | Increased | Improves local feeding and interaction |
I also considered the feeding distance between risers. By reducing the center distance from 1450 mm to 1028 mm, I reduced the unsupported length between feeding centers. The required feeding distance can be described as:
$$L_{required} = L_{center} – L_{riser, effective}$$
By reducing L_center and increasing the effective riser zone, I reduced L_required. The eleven chills further split the thermal field into smaller zones. Each zone solidifies with a stronger gradient toward the nearest riser or chill. This is a classic sand casting strategy: use chills to create local heat extraction, use padding to maintain a feed channel, and use risers to supply liquid metal.
8. Verification of the Optimized Sand Casting Design with MAGMA
I simulated the optimized sand casting design with MAGMA using the same alloy, sand, and interfacial conditions as the refined model. The previously predicted shrinkage region at the blind riser side disappeared. The new solidification sequence showed that the enlarged upper riser remained liquid after the curved transition began to freeze. The padding stayed open. The chills created a stronger gradient. The smaller riser contributed to local feeding but did not compete with the enlarged riser.
| Simulation result | Initial sand casting design | Optimized sand casting design |
|---|---|---|
| Shrinkage at defect location | Predicted after boundary correction | Not predicted |
| Feeding path condition | Blocked by early padding solidification | Open through trapezoidal padding |
| Upper riser liquid life | Insufficient | Extended |
| Directional solidification | Weak at curved transition | Improved by chills and riser enlargement |
| Isolated liquid region | Present | Eliminated in the model |
| Overall defect risk | High | Significantly reduced |
I did not rely on a single simulation result. I cross-validated the optimized sand casting process with ProCAST. ProCAST has strong mesh repair, fluid flow, thermal, and shrinkage prediction capabilities. I imported the same geometry components, generated a surface mesh of 194680 elements and a volume mesh of 1480627 elements, and used material and heat transfer parameters consistent with the MAGMA model.
| Cross-validation item | MAGMA | ProCAST |
|---|---|---|
| Process | Sand casting | Sand casting |
| Mesh type | Tetrahedral | Surface and volume mesh |
| Mesh count | Approximately 20 million | 194680 surface, 1480627 volume |
| Material data | Same alloy and sand properties | Same alloy and sand properties |
| Heat transfer | Initial and interfacial coefficients | Consistent coefficients |
| Output | Temperature, solid fraction, shrinkage | Temperature, solid fraction, shrinkage |
The ProCAST temperature field showed that the optimized casting solidified in a more acceptable order. At 80% solidification, the liquid regions were still connected to the feeding channels. At 100% solidification, the last liquid was located in the upper riser rather than in the curved transition. The solid fraction comparison confirmed that the feed path remained open longer than in the initial design.
The shrinkage comparison was equally important. The initial design showed shrinkage porosity risk at the defect location. The optimized design did not show that risk. The Niyama criterion values in the optimized case were higher than the critical threshold in the previously defective region. The predicted temperature gradient was stronger, and the cooling rate was more favorable. This cross-validation gave me confidence that the optimized sand casting route was not a numerical artifact from one software package.
| Solidification stage | Initial design behavior | Optimized design behavior |
|---|---|---|
| 80% solid fraction | Local isolation began near curved transition | Feed path remained connected |
| 100% solid fraction | Shrinkage risk at blind riser side | Last liquid in upper riser |
| Temperature gradient | Weak toward original riser | Strong toward enlarged riser and chills |
| Niyama risk | Low in defect region | Above critical threshold |
| Shrinkage display | Present | Absent or negligible |
9. Second Manufacturing Campaign and Sand Casting Process Control
For the second manufacturing campaign, I converted the optimized design into shop-floor controls. I treated pattern accuracy, core assembly, sand properties, gating, pouring, and riser removal as critical control points. The sand casting process window had to be maintained consistently because the simulation assumed specific thermal conditions.
The wooden pattern was machined to control dimensions within plus or minus 2 mm. After assembly, I required three-dimensional scanning and fitting software comparison. Any area outside the sand casting process requirement was corrected before molding. I paid particular attention to the new riser sizes, the new padding geometry, and the positions of the eleven external chills. If the pattern or chill location deviated from the optimized design, the thermal balance would change.
| Control point | Requirement | Reason for sand casting quality |
|---|---|---|
| Pattern machining | Plus or minus 2 mm | Ensures geometry and feeding geometry |
| Three-dimensional scanning | Full pattern comparison | Detects deviations before molding |
| Riser position | Match optimized design | Maintains feeding volume and path |
| Chill position | Match optimized design | Maintains directional solidification |
| Padding geometry | 400 mm trapezoidal gradient | Keeps feed channel open |
| Core location | Use core prints and checking templates | Controls wall thickness and cavity shape |
The molding sand was furan resin sand. I used chromite resin sand for the facing sand and silica sand for the backing sand. To prevent burn-on and penetration in the internal cavity, I used an alkali phenolic binder with new chromite sand for the core surfaces. I improved the collapsibility of the sand to reduce restraint on the contracting casting. This helps control distortion and reduces the risk of hot tearing. I cleaned the internal runners and core surfaces before closing the mold, removed loose sand and dust, and sealed core gaps to prevent metal penetration.
I controlled the core setting sequence to avoid a situation where the last core could not be installed. I used locating core prints and checking templates to confirm core position. The internal cavity geometry and wall thickness depend on core stability. In a long sand casting, a small core shift can produce a large dimensional error or a local thin wall.
| Molding and core item | Specification or practice | Effect |
|---|---|---|
| Molding sand | Furan resin sand | Strength and mold stability |
| Facing sand | Chromite resin sand | Chill effect and burn-on resistance |
| Backing sand | Silica sand | Economical mold support |
| Core surface | Alkali phenolic binder with chromite sand | Improved cavity surface quality |
| Collapsibility | Adjusted sand formulation | Reduced contraction restraint |
| Core setting | Sequence control and templates | Accurate cavity and wall thickness |
| Mold closing | Clean runners, seal gaps | Prevent sand inclusions and metal penetration |
The gating system used two levels and six side ingates. After mold closing, I baked the mold with a hot air furnace. The outlet temperature was controlled between 180 and 220 degrees C, and hot air was supplied for at least 12 hours. Pouring began within 1 hour after hot air was stopped. This reduced moisture and improved mold surface conditions. Before pouring, argon was introduced into the mold cavity for 30 minutes. During pouring, argon protection was maintained through a protective ring and a blowing pipe. This was intended to reduce oxidation and gas-related defects in the sand casting.
The liquid metal weight was 20.4 tonnes. Based on production experience, I set the pouring temperature between 1550 and 1560 degrees C. The pouring time can be estimated from:
$$t_p = \frac{W}{Q}$$
where t_p is pouring time, W is liquid metal weight, and Q is average pouring rate. I also used the practical relationship:
$$t_p = K \sqrt{W}$$
where K is an empirical coefficient that depends on alloy, casting geometry, and gating design. The exact values were adjusted on the shop floor to ensure smooth filling without turbulence. A smooth fill is important because turbulence can cause reoxidation, gas entrapment, and sand erosion.
| Pouring and gating parameter | Value or practice | Sand casting purpose |
|---|---|---|
| Gating system | Two levels, six side ingates | Controlled filling and reduced turbulence |
| Mold baking temperature | 180 to 220 degrees C | Reduce moisture and improve surface quality |
| Mold baking time | At least 12 hours | Stabilize mold thermal condition |
| Pouring start | Within 1 hour after hot air stop | Avoid moisture re-absorption |
| Argon prefill | 30 minutes before pouring | Displace air and reduce oxidation |
| Argon protection during pour | Protective ring and blowing pipe | Reduce oxidation and gas defects |
| Liquid metal weight | 20.4 tonnes | Heat and pouring rate control |
| Pouring temperature | 1550 to 1560 degrees C | Fluidity and feeding balance |
After pouring, I controlled shakeout temperature to not more than 250 degrees C. To avoid cracking and distortion during riser removal, the casting was cleaned and then heated to 450 plus or minus 20 degrees C and held for 4 hours before hot cutting of risers and padding. This was a critical step. If the risers are cut cold, the local thermal shock and restraint can initiate cracks or cause movement. Hot cutting allows the casting to relax more gradually.
10. Actual Production Results After Optimization
The second casting was produced with the optimized sand casting process. The final component met the line-type requirements. The section profiles matched the checking templates. The key dimensions and shapes were controlled within the required tolerances. The test blocks passed mechanical property testing after heat treatment. The machined surfaces and as-cast surfaces passed inspection. Ultrasonic testing and magnetic particle testing did not reveal shrinkage porosity, cracks, or other harmful casting defects.
| Verification item | Result | Acceptance status |
|---|---|---|
| Line type and section profile | Matched checking templates | Accepted |
| Key dimensions | Controlled within drawing and standard requirements | Accepted |
| Mass deviation | Within 0 to +3% target | Accepted |
| Out-of-plane deviation | Within 3 mm requirement | Accepted |
| Magnetic particle testing | No harmful surface indications | Accepted |
| Ultrasonic testing | No harmful internal indications | Accepted |
| Mechanical properties | Met all specified requirements | Accepted |
| Final visual and dimensional inspection | Passed multi-party review | Accepted |
The measured mechanical properties from the production test blocks were consistent and acceptable. I recorded the yield strength, tensile strength, elongation, reduction of area, and low-temperature impact energy. The values demonstrated that the optimized sand casting process, together with the heat treatment, delivered the required strength and toughness.
| Test block | Yield strength, MPa | Tensile strength, MPa | Elongation, % | Reduction of area, % | Impact at minus 40 degrees C, J |
|---|---|---|---|---|---|
| 1 | 437 | 581 | 29.0 | 69 | 101 / 123 / 133 |
| 2 | 444 | 578 | 30.0 | 72 | 167 / 169 / 151 |
| 3 | 426 | 562 | 27.0 | 69 | 162 / 184 / 171 |
I compared the measured values with the minimum requirements. The yield strength exceeded 370 MPa in all three test blocks. The tensile strength exceeded 490 MPa in all three test blocks. The elongation exceeded 20%, the reduction of area exceeded 40%, and the low-temperature impact energy exceeded 27 J. The margin above the minimum requirements was comfortable. This confirmed that the sand casting process produced a clean, well-fed casting that responded correctly to heat treatment.
| Property | Minimum requirement | Measured range | Acceptance |
|---|---|---|---|
| Yield strength, MPa | 370 | 426 to 444 | Pass |
| Tensile strength, MPa | 490 | 562 to 581 | Pass |
| Elongation, % | 20 | 27.0 to 30.0 | Pass |
| Reduction of area, % | 40 | 69 to 72 | Pass |
| Impact at minus 40 degrees C, J | 27 | 101 to 184 | Pass |
11. Discussion of Why the Optimized Sand Casting Worked
The optimized sand casting worked because it corrected the thermal logic of the feeding system. In the initial design, the smaller blind riser was expected to feed a curved transition sideways and upward. That assumption did not account for the actual solidification sequence. The padding cooled too quickly, the feed channel closed, and the adjacent riser could not compensate. In the optimized design, the largest riser was placed above the hot spot and supplied liquid metal downward. Gravity-assisted feeding, a larger riser modulus, a wider trapezoidal padding, and additional chills all worked together to keep the feed channel open.
The most important change was the increase in riser modulus. The original riser modulus was approximately 83.3 mm. The optimized riser modulus was approximately 146.0 mm. For a local casting modulus near 100 mm, the required riser modulus is 130 mm. The original riser did not meet the requirement. The optimized riser did. This alone explains much of the improvement. The additional chills and reduced riser spacing improved the thermal gradient and reduced the unsupported feeding distance.
| Mechanism | Initial sand casting condition | Optimized sand casting condition | Result |
|---|---|---|---|
| Riser modulus | 83.3 mm | 146.0 mm | Meets 1.3 times casting modulus |
| Feeding direction | Side and upward | Downward from above | Better gravity assistance |
| Padding | Narrow and early freezing | 400 mm trapezoidal gradient | Feed path remains open |
| Chill effect | Weak at defect location | Eleven chills between risers | Stronger directional gradient |
| Riser spacing | 1450 mm | 1028 mm | Shorter unsupported feeding distance |
| Liquid isolation | Present | Eliminated | No shrinkage cavity at the defect site |
I also learned that sand casting simulation must be calibrated with production feedback. The first MAGMA result was not sufficient because the boundary conditions did not represent the local thermal resistance and the actual behavior of the padding. After I corrected the boundary conditions, the simulation reproduced the production defect. This is a valuable lesson. A simulation is not simply a yes-or-no check. It is a thermal model that must be adjusted when reality disagrees. The agreement between MAGMA and ProCAST after optimization increased my confidence because two different numerical approaches predicted the same improvement.
The use of two simulation packages also helped me examine the optimized design from different numerical perspectives. MAGMA provided a robust finite difference solution for solidification and feeding. ProCAST provided a detailed finite element view of temperature, solid fraction, and shrinkage. The agreement between the two packages reduced the risk of a software-specific false positive. In large sand casting projects, cross-validation is especially useful when the cost of a failed casting is high and the production lead time is long.
12. Key Process Equations Used in the Optimization
For clarity, I summarize the main equations that guided the sand casting optimization. These equations were not used blindly. I combined them with shop-floor experience, simulation, and inspection feedback.
| Equation name | Expression | Use in sand casting |
|---|---|---|
| Modulus | $$M = \frac{V}{A}$$ | Compare local cooling capacity and feeding demand |
| Chvorinov’s rule | $$t_s = k M^n$$ | Estimate solidification time |
| Riser modulus requirement | $$M_r \ge 1.3 M_c$$ | Ensure riser remains liquid longer than casting |
| Riser volume requirement | $$V_r = \frac{\beta V_c}{\eta_f}$$ | Estimate liquid metal needed for contraction |
| Solid fraction | $$f_s = \frac{T_L – T}{T_L – T_S}$$ | Track freezing progress |
| Niyama criterion | $$N_y = \frac{G}{\sqrt{\dot{T}}}$$ | Assess shrinkage porosity risk |
| Heat conduction | $$\rho c_p \frac{\partial T}{\partial t} = \nabla \cdot \left(k \nabla T\right) + \rho L \frac{\partial f_s}{\partial t}$$ | Model sand casting solidification |
| Pouring time | $$t_p = \frac{W}{Q}$$ | Estimate filling duration |
| Empirical pouring time | $$t_p = K \sqrt{W}$$ | Adjust gating for smooth fill |
The modulus of a cylindrical riser is particularly useful in sand casting design. For a cylinder with diameter D and height H, the volume is:
$$V = \frac{\pi D^2 H}{4}$$
The effective cooling surface area, excluding the bottom contact, is approximately:
$$A = \frac{\pi D^2}{4} + \pi D H$$
Therefore:
$$M = \frac{V}{A} = \frac{\frac{\pi D^2 H}{4}}{\frac{\pi D^2}{4} + \pi D H}$$
After simplification:
$$M = \frac{D H}{D + 4H}$$
I used this expression to compare the original and optimized risers. The result showed a clear increase in feeding modulus and confirmed that the optimized riser was better matched to the local hot spot.
13. Dimensional and Distortion Control in the Sand Casting
The stem post has a strong tendency to straighten during cooling and heat treatment. I controlled distortion through a combination of pattern allowance, camber correction, sand collapsibility, core stability, and controlled riser removal. I set separate shrinkage rates in the length, width, and height directions because the thermal gradients and restraint conditions are not isotropic. I also used gradual camber correction rather than a single uniform allowance. This means the pattern was intentionally curved in the opposite direction of the expected distortion.
| Distortion control method | Implementation | Expected effect |
|---|---|---|
| Directional shrinkage allowance | Separate length, width, and height factors | Compensate anisotropic contraction |
| Camber correction | Gradual reverse curvature | Counteract straightening |
| Sand collapsibility | Optimized binder and sand composition | Reduce contraction restraint |
| Core stability | Locating prints and checking templates | Maintain wall thickness and shape |
| Hot riser removal | Heat to 450 plus or minus 20 degrees C for 4 hours | Reduce cutting stress and cracking |
| Controlled shakeout | Not more than 250 degrees C | Avoid early thermal shock |
The final dimensional inspection confirmed that these measures were effective. The line type followed the checking templates. The key dimensions were within tolerance. The out-of-plane deviation was within the required 3 mm limit. The mass deviation was within the 0 to +3% requirement. This is important because a casting can be internally sound but still be rejected if its dimensions or line type are outside acceptance.
14. Non-Destructive Testing and Final Acceptance
After heat treatment and finishing, I performed the required non-destructive testing. Magnetic particle testing covered 100% of the accessible surfaces. Ultrasonic testing covered all external surfaces with a sensitivity equivalent to a 3 mm flat-bottom hole. The optimized sand casting passed both inspections. No shrinkage porosity, cracks, or harmful inclusions were found in the previously defective region or elsewhere.
| Inspection | Coverage | Result | Final disposition |
|---|---|---|---|
| Magnetic particle testing | 100% accessible surfaces | No harmful indications | Accepted |
| Ultrasonic testing | All external surfaces | No harmful internal indications | Accepted |
| Visual inspection | Full casting | Smooth transition and no harmful surface defects | Accepted |
| Dimensional inspection | Key sections and interfaces | Line type and dimensions accepted | Accepted |
| Mechanical testing | Representative test blocks | All properties above minimum | Accepted |
| Multi-party review | Full documentation and hardware | No open non-conformities | Accepted |
I considered the final acceptance to be a confirmation of the entire sand casting route, not just the riser change. The optimized design had to be supported by accurate pattern making, controlled sand properties, correct core assembly, clean melting, proper pouring temperature, argon protection, controlled shakeout, and hot cutting of risers. If any of these steps had deviated, the thermal balance assumed by the simulation could have changed. The successful result therefore validated both the design and the manufacturing control system.
15. Lessons from the First Failure and the Optimized Sand Casting
The first failure taught me that a sand casting simulation is only as good as its boundary conditions. The original model predicted no major shrinkage because it did not capture the local thermal resistance and the actual padding behavior. When I adjusted the boundary conditions to match the real sand casting, the defect appeared in the simulation. This was a turning point. It showed that the production defect was not random. It was the predictable result of a blocked feeding channel.
The second lesson is that riser modulus and feeding direction cannot be separated. A riser can be large enough by volume but still fail if the feed path closes or if the feeding direction is unfavorable. In the optimized sand casting, the enlarged upper riser was placed above the hot spot, and the padding was widened to maintain an open channel. The eleven chills between risers created a strong local gradient. The reduced center distance increased the interaction between adjacent risers. The combined effect was more important than any single change.
The third lesson is that cross-software validation is valuable for high-value sand castings. MAGMA and ProCAST use different numerical formulations. When both predicted the same improvement, I had greater confidence. Cross-validation does not replace production experience, but it reduces the risk of relying on a single model.
| Lesson | Initial sand casting issue | Optimized sand casting response |
|---|---|---|
| Boundary conditions matter | Initial model missed the local thermal resistance | Refined model reproduced the defect |
| Feeding direction matters | Side and upward feeding failed | Downward feeding from a larger upper riser succeeded |
| Padding must remain open | Padding solidified too early | 400 mm trapezoidal padding maintained the channel |
| Chills must create a gradient | Original chill effect was insufficient | Eleven chills created a stronger directional gradient |
| Riser interaction matters | Risers were too far apart | Center distance reduced from 1450 mm to 1028 mm |
| Validation matters | Single-software confidence was insufficient | MAGMA and ProCAST cross-validation confirmed the design |
16. Summary of Optimized Sand Casting Parameters
I summarize the final sand casting parameters that were used for the successful production campaign. These values are specific to this stem post, but the logic can be applied to other large, variable cross-section cast steel components produced by sand casting.
| Parameter | Final condition |
|---|---|
| Casting method | Sand casting with sand box and core assembly |
| Molding orientation | Rotated to reduce height and improve feeding |
| Number of main risers | Six along the length |
| Critical upper riser size | 700 mm diameter by 880 mm height |
| Critical smaller riser size | 600 mm diameter |
| Padding below critical riser | 400 mm wide trapezoidal gradient |
| External chills between critical risers | Eleven chills |
| Critical riser center distance | 1028 mm |
| Feeding direction at defect site | Downward from enlarged upper riser |
| Pouring temperature | 1550 to 1560 degrees C |
| Liquid metal weight | 20.4 tonnes |
| Mold baking | 180 to 220 degrees C for at least 12 hours |
| Argon prefill | 30 minutes before pouring |
| Shakeout temperature | Not more than 250 degrees C |
| Hot cutting condition | 450 plus or minus 20 degrees C for 4 hours |
I also developed a risk matrix for the optimized sand casting. The matrix helped the production team focus on the most important controls. The highest residual risks were pattern accuracy, chill position, pouring temperature, and core stability. These were controlled through scanning, templates, thermocouples, and strict process records.
| Risk item | Probability after optimization | Impact if uncontrolled | Control action |
|---|---|---|---|
| Shrinkage porosity at curved transition | Low | High | Enlarged riser, trapezoidal padding, chills |
| Distortion and straightening | Medium | High | Camber correction, collapsible sand, controlled cooling |
| Core shift | Low to medium | Medium | Core prints, checking templates, sequence control |
| Sand inclusion | Low | Medium | Clean runners, seal gaps, controlled filling |
| Gas porosity | Low | Medium | Mold baking, argon protection, clean melt |
| Cracking during riser removal | Low | High | Hot cutting after controlled heating |
| Dimensional deviation | Low | High | Pattern scanning, shrinkage allowance, final inspection |
17. Final First-Person Assessment
I judge the optimized sand casting route to be successful because it delivered a casting that satisfied internal quality, dimensional, and mechanical requirements. The key technical achievement was the correct reorientation of feeding at the curved transition. The original design relied on a smaller blind riser to feed sideways and upward. The optimized design used an enlarged upper riser to feed downward. This change was supported by a wider trapezoidal padding and eleven external chills. The riser modulus increased from approximately 83.3 mm to approximately 146.0 mm, which satisfied the 1.3 times casting modulus requirement for the local hot spot.
I also confirmed the importance of cross-validation. MAGMA and ProCAST both showed that the optimized sand casting would solidify with a more favorable temperature gradient and solid fraction sequence. The shrinkage risk at the previously defective location disappeared. The production casting then verified the simulation. Mechanical properties exceeded the minimum requirements, and non-destructive testing found no harmful defects.
From my perspective, the project demonstrated that large marine cast steel stem posts can be produced reliably by sand casting when the process design is based on thermal logic, not only on geometric intuition. The most dangerous assumption is that a riser will feed a hot spot simply because it is nearby. The feed path must remain open, the riser modulus must be sufficient, the feeding direction must be favorable, and the chilling arrangement must create a directional gradient. When these conditions are met, sand casting can deliver a complex, variable cross-section casting with sound internal quality and controlled dimensions.
I would apply the same approach to future large sand casting projects: define the local modulus, calculate the required riser modulus, check the feeding distance, simulate the solidification sequence, cross-validate with a second software package when the risk is high, and control the shop-floor process window tightly. This combination of analysis and production discipline is what turned a defective first sand casting into a qualified final product.