In this paper, I present a comprehensive quality analysis and design optimization study for a large marine cast steel stem post produced in a sand casting foundry. The component exhibited shrinkage cavities, porosity, and deformation risks due to its complex variable cross-section geometry. I first carried out solidification simulations using MAGMA software on the initial process design, which revealed that inadequate feeding channels led to shrinkage porosity in specific regions. Subsequently, I optimized the riser and padding geometry, their positions, and the arrangement of external chills. Cross-validation using both MAGMA and ProCAST confirmed that the optimized process achieved directional solidification and significantly reduced defect risk. Actual production trials verified that the optimized process yielded castings with sound internal quality, dimensional accuracy, and mechanical properties meeting all technical requirements. This study demonstrates the effectiveness of combining numerical simulation with rigorous root-cause analysis in a sand casting foundry environment, providing reliable guidance for manufacturing similar large complex steel castings.
1. Introduction and Background
The stem post is one of the most critical structural components of a ship. It connects the side shell plating, deck, and keel ends, and reinforces the bow region to maintain the shape of the ship’s forward end. Since the stem post is located at the very front of the hull, it must withstand hydrostatic pressure, wave impact forces, and potential collision loads. Sufficient stiffness and strength are therefore essential. Among various manufacturing routes, cast steel stem posts benefit from the flexibility of casting processes to produce complex cross-sections that match the intricate hull lines and satisfy strength requirements. However, the stem post is a slender, variable cross-section steel casting that is highly prone to deformation, dimensional instability, and internal defects such as shrinkage cavities and porosity.
I encountered a specific case where a large marine stem post produced in a sand casting foundry exhibited shrinkage problems during its first manufacturing attempt. This paper describes how I identified the root cause, optimized the casting process, and verified the improvements through numerical simulation and actual production. The complete workflow, from initial design to final qualification, is presented with the aim of providing a methodological reference for similar large complex steel castings.
2. Structural Characteristics and Technical Requirements
2.1 Structural Analysis
The stem post under investigation had a length exceeding 3500 mm, with a curved length close to 4000 mm. Its width varied from a maximum of 2570 mm to a minimum of 290 mm, and the wall thickness ranged from 300 mm down to 50 mm. This represents an ultra-long, thin-wall casting with abrupt changes in cross-sectional dimensions and wall thickness. The overall structure tended to “straighten” during solidification, creating a high risk of distortion. The component was required to be free from cracks, inclusions, shrinkage porosity, gas holes, sand defects, and any other discontinuities that could impair mechanical performance. The as-delivered surfaces had to be smooth and gradually blended, with flowing lines. The weight deviation was limited to 0 to +3%, and length and width tolerances had to comply with DCTG 10 and DCTG 12 grades of GB/T 6414—2017, respectively. The shape deviation was limited to 3 mm, and the surface flatness deviation also to 3 mm. These stringent requirements made both dimensional control and internal quality assurance extremely challenging.
2.2 Material Analysis
The stem post was made of a newly developed cast steel with a minimum yield strength of 370 MPa. The chemical composition requirements are summarized in Table 1. Notably, phosphorus and sulfur were each limited to less than 0.01%, which demands strict control during melting and refining.
| C | Si | Mn | Ni | Cr | V | Cu |
|---|---|---|---|---|---|---|
| 0.10–0.13 | 0.10–0.60 | 0.40–0.80 | 1.30–1.70 | ≤0.040 | ≤0.20 | 0.50–0.80 |
Mechanical property test blocks were required to be cast with the same heat as the casting and subjected to the same heat treatment. The specifications are listed in Table 2. Samples were taken from near the surface, at one-quarter wall thickness, and at one-half wall thickness to ensure through-thickness consistency.
| Rm (MPa) | ReL (MPa) | A (%) | Z (%) | KV2 at −40°C (J) |
|---|---|---|---|---|
| ≥490 | ≥370 | ≥20 | ≥40 | ≥27 |
2.3 Non-Destructive Testing
All surfaces were subjected to 100% magnetic particle inspection (MT) according to GB/T 9444—2007 using the wet method with a magnetic yoke. Ultrasonic testing (UT) was performed on all external surfaces according to GB/T 7233.1—2009 with a sensitivity of ∅3 mm.
3. Castability and Process Design Challenges
Because of the abrupt transitions in cross-section and the large wall thickness differences, the stem post was extremely susceptible to distortion during both casting and heat treatment. The external surface was largely uniform with a nominal wall thickness of 80 mm, as shown in the outer surface structure diagram. This uniform thin-wall configuration made self-feeding difficult. The modulus of the thin sections was not sufficient to promote a natural directional solidification from the casting into the risers. Without proper chilling or padding, the solidification front would not be able to feed the thicker sections effectively, leading to shrinkage defects. This is a common challenge in a sand casting foundry when dealing with large plate-like steel castings connected to heavy bosses or flanges.
To overcome these issues, I relied on the combined use of risers, external chills, and padding. Risers compensate for the volumetric shrinkage during solidification. External chills accelerate cooling in selected zones, increasing the feeding distance of adjacent risers. Padding creates a thermal gradient between the hot spot and the riser, opening a feeding path. The correct application of these elements is the key to achieving sound castings.
4. Initial Casting Process Design
4.1 Process Layout
In the initial design, I rotated the stem post to reduce the overall height difference. This helped not only to improve feeding but also to facilitate molding and mold closing. The riser size was determined based on the modulus factor: the riser modulus was at least 1.3 times the modulus of the casting or the largest hot spot. When padding was required, I used the rolling circle method to create an artificial feeding path from the hot spot to the main casting body. External chills were placed in proportion to the wall thickness between risers to create a segmented chilling effect and to increase the feeding distance. Separate linear contraction allowances were set in the length, width, and height directions. In addition, a gradual correction amount (also known as pattern allowance for distortion) was included to compensate for the expected straightening deformation during solidification and after heat treatment.
The molding method adopted was a “flask + core assembly” approach. Due to the multiple internal cavities, each cavity was formed by a separate sand core. Since the stem post is similar to a long bar-like component, a total of six risers were placed to maintain an adequate feeding distance. The initial casting process is schematically shown in Figure 1 (note: this is a representative illustration from a sand casting foundry setup).

4.2 Numerical Simulation of the Initial Process
I used MAGMA software, a finite-difference-based casting simulation package, to analyze the solidification behavior. The process components were imported in STL format and meshed into approximately 20 million tetrahedral elements. The simulation results for the initial process showed no shrinkage cavities in the main body, and the minimum safe height was recorded as 164 mm. However, closer inspection of the porosity prediction revealed a potential risk area near riser No. 3, as shown in the porosity distribution output. This was initially considered acceptable, but subsequent actual production revealed a real defect at that exact location.
5. Quality Problem Observed in First Manufacturing
5.1 Defect Description
After pouring, holding, and heat treatment, the mechanical properties of the test blocks passed. During post-heat-treatment fettling, I noticed a visible shrinkage cavity on the side of riser No. 3 after removing the remaining riser and padding. To accurately determine the extent of the defect, carbon arc gouging was performed. The defect was measured to be approximately 350 mm in length and 60 mm in depth, extending diagonally upward. This indicated that the riser padding beneath riser No. 3 had solidified prematurely, while riser No. 4 was unable to provide sufficient feeding to that region. As a result, an isolated liquid volume formed and ultimately produced the shrinkage cavity.
5.2 Root Cause Analysis
I systematically evaluated the design, manufacturing, and inspection processes using the man-machine-material-method-environment-measurement approach. All operational controls, including melting, molding, pouring, shakeout, and heat treatment, were within specification. Thus, I concluded that the root cause lay in the casting process design rather than in execution. The main issue was that the large curvature in the bow region had not been fully accounted for in the thermal boundary conditions. My initial design relied on riser No. 3 to feed the upper side area, but the geometric configuration prevented a sound feeding channel.
To confirm this hypothesis, I re-ran the MAGMA simulation with refined boundary conditions and obtained a clear indication of shrinkage porosity in the same area where the actual defect had occurred. This correlation between simulation and reality validated the need for process modification.
6. Optimization of the Casting Process
Based on the root cause analysis, I revised the process in the following ways:
- The feeding direction for the upper side area was changed from riser No. 3 to riser No. 4, which meant that riser No. 4 would feed from top to bottom. Riser No. 4 was enlarged from ∅400 mm × 500 mm to ∅700 mm × 880 mm. The padding below riser No. 4 was also redesigned as a trapezoidal transition with a width of 400 mm to completely cover the defective zone.
- Eleven external chills were added between risers No. 3 and No. 4 to create a segmented chilling effect and to increase the feeding distance.
- Riser No. 3 was slightly enlarged from ∅550 mm × 600 mm to ∅600 mm × 650 mm, and its position was adjusted. The center distance between risers No. 3 and No. 2 was reduced from 1450 mm to 1028 mm, increasing the feeding liquid volume and relieving the overall feeding demand on riser No. 4.
The optimized casting process is illustrated in Figure 2 (wall-type padding and riser arrangement). This revision was specifically designed to ensure that all hot spots would be within the feeding range of the nearest riser, and that the solidification front would move directionally from thinner to thicker sections.
6.1 Verification with MAGMA
I re-simulated the optimized process using MAGMA. The shrinkage cavity and porosity predictions for the problematic area were now eliminated. The solidification sequence showed a clear progression toward the risers, with no isolated liquid regions.
6.2 Cross-Validation with ProCAST
To further verify the robustness of the optimized design, I used ProCAST software, which employs a finite-element solver with advanced criteria for porosity, distortion, and stress. The same material properties, heat transfer coefficients, and boundary conditions were applied. I meshed the geometry with 194,680 surface elements and 1,480,627 volume elements. The temperature field and solidification fraction results were extracted at 80% and 100% solidification.
The temperature field comparison between the initial and optimized processes showed a more uniform solidification pattern in the optimized case. The last solidified region was consistently located inside riser No. 4, confirming that the feeding channel remained open until the end. The solidification fraction results also agreed with the temperature field: in the optimized case, the riser No. 4 solidified last, providing complete feeding to the previously defective region.
Finally, the shrinkage porosity criterion in ProCAST indicated a high risk of porosity in the initial process at the exact location of the actual defect, while the optimized process showed no porosity risk. This cross-validation from two independent simulation codes gave me high confidence that the optimized process would produce a sound casting.
7. Quantitative Design Formulas and Parameters
In the design process, I used several quantitative relationships to determine riser dimensions and feeding distances. The modulus of the casting hot spot, \( M_c \), is given by:
$$ M_c = \frac{V_c}{A_c} $$
where \( V_c \) is the volume and \( A_c \) is the surface area of the hot spot. For steel castings, the riser modulus \( M_r \) must satisfy:
$$ M_r \geq 1.3 \, M_c $$
The riser volume \( V_r \) was determined from the solidification shrinkage requirement:
$$ V_r = \frac{\varepsilon \, V_f}{1 – \varepsilon} $$
where \( \varepsilon \) is the volumetric solidification shrinkage fraction (approximately 3–6% for steel) and \( V_f \) is the volume of the casting section to be fed. For the specific geometry of the stem post, the maximum feeding distance \( L_f \) for a plate section can be approximated by:
$$ L_f = 2 \, t + 4 \, C $$
where \( t \) is the plate thickness and \( C \) is a chill-effect factor. I used external chills to increase the effective feeding distance and to split the long feeding path into shorter segments.
For the padding design, I applied the rolling circle method, which ensures that the taper angle \( \theta \) of the padding satisfies:
$$ \tan \theta = \frac{t_{\text{max}} – t_{\text{min}}}{L_p} $$
where \( L_p \) is the length of the padding section. In this case, the padding width was set to 400 mm, with a gradual taper that completely covered the defect region.
These formulas were implemented manually to design the initial process, and then the numerical simulation provided the necessary validation. The final optimized riser dimensions are summarized in Table 3.
| Riser No. | Initial Size (mm) | Optimized Size (mm) | Center distance change (mm) |
|---|---|---|---|
| 3 | ∅550×600 | ∅600×650 | 1450 → 1028 (with riser 2) |
| 4 | ∅400×500 | ∅700×880 | — |
The number of external chills was increased from 0 to 11 between risers No. 3 and No. 4. Each chill had a thickness proportional to the local wall thickness, approximately 80% of the wall thickness, and a width of 150 mm, placed with an edge distance of about 100 mm. This arrangement created a segmented solidification pattern that enhanced the feeding capability of the adjacent risers.
8. Manufacturing Quality Control during the Second Trial
8.1 Pattern and Mold Quality
To ensure dimensional accuracy in the second production run, the wooden pattern was CNC-machined with a dimensional tolerance of ±2 mm. After machining, the pattern was subjected to three-dimensional laser scanning, and the scanned data were compared with the CAD model using best-fit alignment software. Any deviations exceeding the allowable limit, especially around the new riser and chill positions, were corrected before proceeding.
For molding, I used furan resin sand for the back-up sand and chromite sand for the facing sand. The internal cavities were made with alkali phenolic resin mixed with fresh chromite sand to prevent metal penetration. All cores were placed carefully, ensuring no loose sand or dust remained in the mold cavity. The core gaps were sealed properly. To reduce resistance to casting contraction and avoid distortion, the sand had good collapsibility. Core positioning was verified using core prints and setting gauges.
8.2 Pouring and Shakeout
The gating system consisted of two layers with a total of six ingates, all side-fed. After mold closing, the mold was dried with a hot air furnace at an outlet temperature between 180°C and 220°C for at least 12 hours. Pouring was performed within 1 hour after stopping the hot air. The mold cavity was purged with argon gas for 30 minutes before pouring, and argon shielding was maintained during the entire pour using a gas protection ring and a purging lance.
The total liquid steel weight was 20.4 tons. Based on previous experience, I set the pouring temperature to 1550–1560°C. The shakeout temperature was kept below 250°C to minimize distortion and cracking risk. After shakeout and sand removal, the casting was reheated to 450°C ± 20°C, held for 4 hours, and then hot-cut to remove the risers and padding. This procedure reduced the risk of distortion and cracking caused by thermal stress.
8.3 Mechanical Properties and Non-Destructive Testing of the Final Casting
The final product obtained after the optimized process is shown in Figure 3. The casting demonstrated that all cross-sections matched the inspection templates, and the key dimensions and shapes were within tolerance. The mechanical property test results from the integral test block are summarized in Table 4. All values exceeded the requirements.
| Specimen | Rm (MPa) | ReL (MPa) | A (%) | Z (%) | KV2 at −40°C (J) |
|---|---|---|---|---|---|
| 1 | 581 | 437 | 29.0 | 69 | 101, 123, 133 |
| 2 | 578 | 444 | 30.0 | 72 | 167, 169, 151 |
| 3 | 562 | 426 | 27.0 | 69 | 162, 184, 171 |
Ultrasonic and magnetic particle inspections were performed on all required surfaces. No shrinkage cavities, cracks, or other casting defects were found. The results were fully consistent with the simulation predictions. The casting passed the multi-party inspection and was accepted by the customer.
9. Discussion
This case underscores the importance of combining robust process design with numerical simulation in a sand casting foundry. The initial process passed a conventional simulation but still produced a defect in a localized area. The reason was that the thermal boundary conditions near the high-curvature region had not been sufficiently resolved, and the feeding path from riser No. 3 was insufficient to counteract the premature solidification of the padding. By re-examining the defect morphology (diagonal upward extension) and correlating the simulation porosity results with the actual defect location, I was able to identify the true feeding deficiency.
The modification shifted the feeding responsibility to riser No. 4, which had a much enhanced size and a properly designed padding profile. The addition of exterior chills between risers No. 3 and No. 4 subdivided the long feeding zone into smaller segments, each adequately fed by its nearest riser. The adjustment of riser No. 3’s position and diameter further balanced the feeding load across the entire length of the casting.
Cross-validation with ProCAST was particularly valuable. Although the two simulation codes use different numerical methods (finite difference vs. finite element), both predicted the same outcome: the optimized process produced a sound casting. This convergence confirms that the process design was robust and not merely a numerical artifact.
From a practical standpoint, the optimized process also helped to control distortion. The provision of a gradual correction amount, combined with careful control of shakeout and hot cutting, allowed the final casting to meet the strict dimensional tolerances. The first trial had not failed in terms of mechanical properties, but the internal defect rendered it unacceptable. The second trial achieved both sound internal quality and dimensional accuracy, demonstrating that the systematic quality analysis method was effective.
10. Conclusions
Through this study, I drew the following conclusions:
- Systematic root-cause analysis using the man-machine-material-method-environment-measurement framework allowed me to pinpoint that the shrinkage defect was caused by inadequate feeding channels in the original riser system, rather than by operational errors.
- Optimizing the size and position of risers, redesigning the padding shape, and adding external chills successfully established a directional solidification profile. This ensured effective feeding of all hot spots and eliminated the shrinkage porosity risk.
- The reliability of the optimized process was verified by two independent casting simulation software packages, MAGMA and ProCAST. Both predicted a sound casting with no porosity. Actual production confirmed the accuracy of these predictions.
- The anti-distortion measures, including pattern correction allowances, controlled shakeout temperature, and hot cutting procedures, were effective in keeping all key dimensions within the required tolerances.
- This work provides a valuable technical reference for producing large complex steel castings with abrupt thickness transitions in a sand casting foundry, particularly when tight internal quality and dimensional specifications are imposed.
In conclusion, the integration of numerical simulation, rigorous defect analysis, and carefully controlled manufacturing practices enabled the successful production of a high-quality marine stem post in a sand casting foundry. The same methodology can be extended to other large cast steel components with similar challenges, ensuring both first-pass quality and reliable service performance.
