In this work, I focused on the development and optimization of the green moulding clay sand process for producing railway steel castings, especially for coupler yokes and coupler knuckles. This technology is attractive because it offers high productivity, low material cost, short production cycles, and easy integration with mechanized and automated molding lines. However, green sand steel casting is also known to have a higher risk of producing casting defects such as blowholes, sand scabs, metal penetration, shrinkage cavities, and hot cracks. The key to success is a thorough understanding of the raw materials, precise control of the sand properties, and a robust gating and risering design. Therefore, I combined laboratory experiments, computer simulation, plant-scale trial production, and defect analysis to establish a reliable manufacturing route.
1. Introduction
Railway freight vehicles have been developed for higher speed and heavier axle loads. The coupler system is one of the most critical safety components because it transmits traction and buffing forces between cars. The coupler yoke and knuckle are subjected to repeated impact, bending, and wear. Casting defects in these components could lead to fatigue cracks, fracture, and severe economic loss. Consequently, the required quality is extremely high. Most modern railway steel castings are made by sand casting, and the selection of the molding process significantly influences both the quality and the cost.
Several molding processes are available for railway steel castings, including water-glass CO₂ hardened sand, ester-hardened alkaline phenolic resin sand, furan resin sand, and green moulding clay sand. Water-glass and resin sands generally produce good mold strength and collapsibility, but they are more expensive and less environmentally friendly. Green moulding clay sand, with bentonite as the binder, has the advantages of being continuously reusable, energy-saving, and much cheaper. Nevertheless, it has higher moisture content, lower green strength, and lower refractoriness compared with chemically bonded sands. These characteristics make the process more sensitive to environmental changes and more likely to generate casting defects if not carefully controlled.
To satisfy the requirement of high-quality railway steel castings, I first established suitable facing sand and backing sand formulations. I then used the finite-element software ProCAST to simulate mold filling and solidification of the original casting processes for the coupler yoke and the coupler knuckle. The simulation predicted the locations where casting defects such as erosion, slag entrapment, metal penetration, blowholes, shrinkage porosity, and hot tearing would probably occur. Based on these predictions, I optimized the casting process designs and verified them by trial production. Finally, I analyzed the major defects encountered during production and proposed corrective actions, which were applied to mass production.
2. Experimental Investigation of Green Sand Formulations
2.1 Raw Materials and Testing Methods
To develop reliable green sand recipes, I first tested the basic properties of the new sand and the reclaimed sand. The new sand was a natural silica sand with a 40/70 mesh size and a silicon dioxide content higher than 94%. The reclaimed sand was the return sand taken from the molding loop. The measured properties are summarized in Table 1.
| Property | New sand | Reclaimed sand |
|---|---|---|
| Main sieve fraction | 40/70 mesh | 50/70 mesh |
| Clay content (%) | <0.5 | 6.8 |
| Moisture content (%) | <1.0 | 0.7 |
| Active bentonite content (%) | — | 5.05 |
I used a commercial sodium bentonite as the primary binder and a mixture of α-starch and dextrin as organic additives. The ratio of α-starch to dextrin was fixed at 3:2 because this proportion gave the best combination of hot wet tensile strength, shatter index, and surface stability. The sand mixture was prepared in a laboratory sand mixer. I evaluated the following green sand properties: compactability, permeability, green compressive strength, hot wet tensile strength, shatter index, and surface stability index according to standard testing methods. The target properties required by the factory are listed in Table 2.
| Property | Facing sand | Backing sand |
|---|---|---|
| Moisture (%) | 3.1–3.6 | 2.8–3.6 |
| Compactability (%) | 50 ± 5 | 50 ± 5 |
| Permeability | ≥180 | ≥160 |
| Green compressive strength (kPa) | 80–100 | 70–80 |
| Hot wet tensile strength (kPa) | ≥3.0 | ≥2.5 |
| Surface stability index (%) | ≥95 | ≥95 |
2.2 Facing Sand Experiments
I first carried out single-factor experiments to determine the reasonable ranges of bentonite, starch additives, and water. Then a three-factor, three-level orthogonal experiment known as L9(3³) was designed. The three factors were bentonite addition, starch additive addition, and water addition. The total sand was kept as 100% new sand. Table 3 gives the test results for the facing sand formulations.
| Sample | Bentonite (%) | Additive (%) | Water (%) | Compactability (%) | Permeability | Green strength (kPa) | Hot wet tensile strength (kPa) | Shatter index (%) | Surface stability (%) |
|---|---|---|---|---|---|---|---|---|---|
| 1 | 7.5 | 0.1 | 3.0 | 53 | 435 | 85 | 4.17 | 67.5 | 95.6 |
| 2 | 7.5 | 0.3 | 3.3 | 56 | 440 | 82 | 4.38 | 70.2 | 96.7 |
| 3 | 7.5 | 0.5 | 3.6 | 55 | 445 | 79 | 4.15 | 74.1 | 96.8 |
| 4 | 8.0 | 0.1 | 3.6 | 57 | 431 | 82 | 4.19 | 71.0 | 97.0 |
| 5 | 8.0 | 0.3 | 3.0 | 51 | 442 | 86 | 4.18 | 70.4 | 97.0 |
| 6 | 8.0 | 0.5 | 3.3 | 51 | 444 | 84 | 4.42 | 71.5 | 97.1 |
| 7 | 8.5 | 0.1 | 3.3 | 52 | 449 | 96 | 3.98 | 67.2 | 96.3 |
| 8 | 8.5 | 0.3 | 3.6 | 54 | 449 | 94 | 4.44 | 70.1 | 96.9 |
| 9 | 8.5 | 0.5 | 3.0 | 44 | 460 | 103 | 4.62 | 66.3 | 97.0 |
The experimental results were analyzed by the range analysis method. For compactability, the most influential factor was water content, followed by additive content and then bentonite content. For permeability, the main factors were the contents of fine powders such as bentonite and additives. Green compressive strength was mainly governed by the bentonite addition. Hot wet tensile strength, shatter index, and surface stability index were strongly improved by the starch-based additives. Based on the verification tests, the optimized facing sand formula was:
\[
\text{Facing sand} = 100\% \text{new sand} + (8.0\text{–}8.5\%)\text{bentonite} + (0.3\text{–}0.5\%)\text{starch additive} + (3.0\text{–}3.6\%)\text{water}.
\]
The corresponding properties achieved in the laboratory are listed in Table 4.
| Property | Value |
|---|---|
| Compactability (%) | 51–54 |
| Permeability | 440–450 |
| Green compressive strength (kPa) | 85–95 |
| Hot wet tensile strength (kPa) | 4.1–4.5 |
| Shatter index (%) | 70–72 |
| Surface stability index (%) | 97.0–97.5 |
I also established empirical relationships between the sand ingredients and the sand properties. For example, the compactability \(C\) can be expressed by:
\[
C = C_0 + a \left(W – W_0\right) – b \left(B – B_0\right) – c \left(S – S_0\right),
\]
where \(W\) is the water content, \(B\) is the bentonite content, \(S\) is the starch additive content, and \(a\), \(b\), \(c\) are positive empirical coefficients. Water content increases compactability, whereas higher bentonite and additive contents tend to make the sand drier and slightly reduce compactability. These relationships are useful for adjusting the sand on the molding line.
2.3 Backing Sand Experiments
The backing sand was prepared from 100% reclaimed sand. Because the reclaimed sand already contained a significant amount of active bentonite and fine particles, only small additions of bentonite and starch additives were needed. I performed another L9(3³) orthogonal experiment with the same three factors but at lower concentration levels. Table 5 shows the backing sand test results.
| Sample | Bentonite (%) | Additive (%) | Water (%) | Compactability (%) | Permeability | Green strength (kPa) | Hot wet tensile strength (kPa) | Shatter index (%) | Surface stability (%) |
|---|---|---|---|---|---|---|---|---|---|
| 1 | 0.1 | 0.1 | 3.0 | 45 | 209 | 105 | 3.89 | 63.7 | 93.4 |
| 2 | 0.1 | 0.2 | 3.3 | 57 | 205 | 94 | 3.96 | 69.4 | 97.8 |
| 3 | 0.1 | 0.3 | 3.6 | 58 | 203 | 92 | 4.21 | 73.9 | 98.5 |
| 4 | 0.2 | 0.1 | 3.6 | 58 | 203 | 94 | 3.99 | 73.6 | 98.2 |
| 5 | 0.2 | 0.2 | 3.0 | 47 | 211 | 107 | 3.79 | 64.9 | 93.9 |
| 6 | 0.2 | 0.3 | 3.3 | 57 | 206 | 95 | 4.50 | 72.6 | 97.6 |
| 7 | 0.3 | 0.1 | 3.3 | 58 | 208 | 96 | 4.09 | 73.7 | 98.2 |
| 8 | 0.3 | 0.2 | 3.6 | 59 | 204 | 95 | 4.63 | 75.9 | 98.0 |
| 9 | 0.3 | 0.3 | 3.0 | 44 | 213 | 107 | 4.01 | 64.4 | 94.0 |
For the backing sand, water content was the most critical factor because the reclaimed sand contained a large amount of fines and dead clay. The optimized backing sand formula was:
\[
\text{Backing sand} = 100\% \text{reclaimed sand} + (0.2\text{–}0.3\%)\text{bentonite} + (0.1\text{–}0.3\%)\text{starch additive} + (3.3\text{–}3.6\%)\text{water}.
\]
The resulting properties are shown in Table 6.
| Property | Value |
|---|---|
| Compactability (%) | 55–60 |
| Permeability | 200–210 |
| Green compressive strength (kPa) | 90–95 |
| Hot wet tensile strength (kPa) | 3.8–4.5 |
| Shatter index (%) | 72–75 |
| Surface stability index (%) | 97.5–98.0 |
The backing sand showed a high water sensitivity because of the accumulation of micro-fines. In practice, the water addition must be adjusted according to the moisture content of the returned sand, the sand temperature, and the ambient humidity. A simple correction formula can be used:
\[
W_{\text{add}} = W_{\text{target}} – W_{\text{returned sand}} + \alpha \left(T_{\text{sand}} – T_{\text{ref}}\right) + \beta \left(\phi_{\text{air}} – \phi_{\text{ref}}\right),
\]
where \(\alpha\) and \(\beta\) are empirical coefficients, \(T_{\text{sand}}\) is the sand temperature, \(T_{\text{ref}}\) is a reference temperature, \(\phi_{\text{air}}\) is the relative humidity, and \(\phi_{\text{ref}}\) is a reference humidity. This equation helped the operators to keep the compactability within the target range despite seasonal variations.
2.4 Production Verification
After the laboratory work, the optimized sand formulas were implemented on the production molding line. The plant was equipped with a high-capacity sand preparation system, a multi-contact jolt-squeeze molding machine, and an online moldability controller. The online system automatically adjusted the amounts of water, bentonite, and starch additive for each batch. The actual sand recipes and properties measured during a representative production day are given in Table 7.
| Sand type | New sand (%) | Reclaimed sand (%) | Active bentonite in reclaimed sand (%) | Added bentonite (%) | Starch additive (%) | Water (%) | Compactability (%) | Permeability | Green strength (kPa) | Hot wet tensile strength (kPa) | Shatter index (%) | Surface stability (%) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Facing | 100 | 0 | — | 8.0 | 0.3 | 3.3 | 55 | 465 | 102 | 4.2 | 74 | 98 |
| Backing | 0 | 100 | 5.1 | 0.1 | 0.2 | 3.5 | 58 | 225 | 90 | 3.8 | 72 | 97.7 |
The production results agreed well with the laboratory predictions. This confirmed that the optimized green sand formulations could suppress the tendency for casting defects such as metal penetration, sand scab, and blowhole formation.
3. Numerical Simulation and Process Optimization
3.1 Simulation Setup
I used ProCAST, a finite-element based simulation package, to model the mold filling and solidification processes. The three-dimensional geometries of the castings and the gating systems were created in a CAD program and then transferred to MeshCAST for meshing. The material was a low-alloy E-grade steel, ZG25MnCrNiMo. Because this grade was not available in the standard ProCAST database, I calculated the thermophysical properties using the built-in thermodynamic database. The liquidus temperature was set to 1505 °C, the solidus temperature to 1433 °C, and the latent heat to 65 kJ/kg. The pouring temperature was 1585 °C, and the ambient temperature was 25 °C. The heat transfer coefficient between the steel and the green sand mold was assumed to be 500 W/(m²·K). The stress module was also activated to predict thermal stresses and hot tearing.
The governing equations for fluid flow and heat transfer are:
\[
\frac{\partial \rho}{\partial t} + \nabla \cdot \left(\rho \mathbf{u}\right) = 0,
\]
\[
\rho \frac{\partial \mathbf{u}}{\partial t} + \rho \left(\mathbf{u} \cdot \nabla \right) \mathbf{u} = -\nabla p + \nabla \cdot \left(\mu \nabla \mathbf{u}\right) + \rho \mathbf{g},
\]
\[
\rho c_p \frac{\partial T}{\partial t} + \rho c_p \mathbf{u} \cdot \nabla T = \nabla \cdot \left(k \nabla T\right) + \rho L \frac{\partial f_s}{\partial t},
\]
where \(\rho\) is the density, \(\mathbf{u}\) is the velocity vector, \(p\) is the pressure, \(\mu\) is the dynamic viscosity, \(\mathbf{g}\) is the gravitational acceleration, \(c_p\) is the specific heat, \(T\) is the temperature, \(k\) is the thermal conductivity, \(L\) is the latent heat, and \(f_s\) is the solid fraction. To evaluate shrinkage porosity, I used the Niyama criterion:
\[
N_y = \frac{G}{\sqrt{\dot{T}}},
\]
where \(G\) is the thermal gradient and \(\dot{T}\) is the cooling rate near the end of solidification. A low Niyama value indicates a higher risk of shrinkage porosity.
3.2 Coupler Yoke Simulation
The 17-type coupler yoke is a medium-sized steel casting with a mass of about 190 kg and dimensions of 975 mm × 574 mm × 496 mm. It has a complex hook head, a hollow shank, and a thick tail. The original process used two castings per mold, with the ingate placed directly opposite the tail pin hole. Three ordinary sand risers served the head and one riser served the tail. The simulation of the original process predicted several potential casting defects. High-velocity steel impinged on the sand core near the tail pin hole, which would cause erosion and promote sand inclusion. A large flat area on the shank was prone to trapped gas and blowhole defects. In addition, the sharp corners at the hook neck and around the internal traction seats created isolated liquid regions, leading to shrinkage porosity and hot tearing. The stress simulation indicated high effective stress at the riser roots, the hook-neck transition, and complex inner surfaces of the hook head.
To reduce these casting defects, I modified the process design as follows:
- The ingate was moved to one side of the shank to avoid direct impingement on the tail pin hole.
- Prefabricated refractory gating components were used to reduce sand erosion.
- The three ordinary risers on the head were replaced with cylindrical insulating risers, and an additional insulating riser was placed on the shank.
- A small blind riser was added near the ingate to improve feeding of the ingate region.
- The vents in the sand mold and those in the riser sleeves were staggered to prevent sand particles from falling into the cavity.
- Shaped chills were applied at the hot spots to promote simultaneous solidification or directional solidification toward the risers.
The optimized process was simulated again. The flow velocity around the pin hole was reduced and became much more uniform, which lowered the risk of erosion and sand entrapment. The isolated liquid regions at the critical corners were diminished, and the riser feeding channels remained open until the final stage of solidification. The stress level was significantly reduced, especially at the hook-neck transition and the inner corners. Table 8 compares the predicted defects of the original and optimized gating/risering systems.
| Defect type | Original process | Optimized process |
|---|---|---|
| Erosion / sand inclusion | High risk at tail pin hole and hook head | Low risk |
| Blowhole | Risk on the large flat area of the shank | Negligible |
| Shrinkage porosity | At hook neck, inner corners, traction seat roots | Minor at inner corners; mostly in risers |
| Hot tearing | High stress at riser roots and hook neck | Reduced stress, low risk |
3.3 Coupler Knuckle Simulation
The 16-type coupler knuckle is a small but complex steel casting weighing about 40 kg. The original mold had eight knuckles per flask, with a common blind riser located at the center of each group of four. The ingate was directed toward the knuckle pin hole area, causing severe erosion and subsequent sand defects in many trial castings. The simulation also showed isolated liquid regions at the inner pull face, at the traction lug, and near the pin hole, which could lead to shrinkage porosity and hot cracks.
The optimized process included the following changes:
- The number of castings per mold was changed to four, arranged symmetrically.
- The ingate was moved to the back side of the knuckle between the pin hole and the tail stop, so that the molten steel did not directly strike the pin-hole core.
- The large central riser was replaced with an insulating top blind riser placed directly above the pressure face, which was the thickest section requiring feeding.
- Prefabricated refractory gating components were adopted to reduce the chance of sand washing.
The simulation of the optimized knuckle process showed that the flow velocity near the pin hole was more uniform, reducing the likelihood of erosion and slag entrapment. The new top riser provided an open feeding path to the thick inner wall, and the shrinkage porosity was shifted away from the main load-bearing surfaces. The overall stress level was comparable to the original process, but the critical inner face showed no serious stress concentration. Table 9 summarizes the predicted defects for the coupler knuckle.
| Defect type | Original process | Optimized process |
|---|---|---|
| Erosion / sand inclusion | High risk at pin hole | Low risk |
| Shrinkage porosity | At inner pull face and traction lug | Reduced, shifted to less critical area |
| Hot tearing | At inner corners and riser roots | Moderate risk, less severe |
Because the optimized knuckle process required a completely new pattern, the plant continued to use the original pattern during early trial production. Nevertheless, the simulation results provided a clear direction for future tooling replacement.
4. Trial Production, Quality Inspection, and Defect Prevention
4.1 Trial Production Layout
The trial production facility was developed within the existing workshop of the rolling stock manufacturer. The production line included sand preparation, molding, core making, melting, pouring, shakeout, primary cleaning, pre-normalizing, magnetic particle inspection, welding repair, heat treatment, finishing, assembly, and final inspection. The steel was melted in a 10-ton electric arc furnace and refined in a 20-ton ladle refining furnace. This double-slag melting process helped to lower the oxygen and hydrogen contents in the steel and improved the cleanliness, thereby reducing casting defects related to gas and inclusions.
An automatic pouring line was introduced to stabilize the pouring temperature and pouring rate, which was essential for minimizing casting defects.
Molding was performed using a multi-contact jolt-squeeze molding machine, which gave a uniform mold hardness. The cores were produced by the cold-box amine process for the large integrated cores and by no-bake resin sand for small hand cores. A robot was used for setting the large integral core of the coupler yoke. The robot utilized three-dimensional laser scanning and position recognition, which greatly improved core-setting accuracy and reduced the misalignment that often contributes to casting defects.
4.2 Quality Inspection Results
The quality of the trial-produced yokes and knuckles was checked for chemical composition, mechanical properties, microstructure, specified section density, surface quality, and dimensional accuracy. The chemical composition met the E-grade steel standard, as shown in Table 10.
| Element | C | Si | Mn | P | S | Ni | Cr | Mo | Cu | Al | As |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Content (wt%) | 0.270 | 0.340 | 1.350 | 0.014 | 0.007 | 0.390 | 0.480 | 0.240 | 0.050 | 0.040 | 0.007 |
The tensile properties of the castings are shown in Table 11. All values exceeded the requirements of the standard.
| Component | Tensile strength (MPa) | Yield strength (MPa) | Elongation (%) | Reduction of area (%) | Impact energy at -40 °C (J) |
|---|---|---|---|---|---|
| 17-type yoke | 945 | 855 | 16 | 37 | 50 |
| 16-type knuckle | 940 | 860 | 16 | 35 | 31 |
Sectional density was evaluated by cutting the castings at specified locations, polishing the cross-sections, and comparing them with reference photographs. The rating levels are summarized in Table 12. The achieved levels were much lower than the maximum allowable levels, indicating that the optimized feeding system was effective in avoiding internal shrinkage defects.
| Section | Allowable level | Actual level |
|---|---|---|
| 17-type yoke | ≤2–5 depending on section | ≤1 |
| 16-type knuckle | ≤2–4 depending on section | ≤1 |
Hardness tests after quenching and tempering also met the specified values. Surface hardness on the traction surfaces was increased by local induction hardening. The results are shown in Table 13.
| Component | At 5 mm depth (HBW) | At 10 mm depth (HBW) | At 4.5 mm from tail sphere (HBW) |
|---|---|---|---|
| 17-type yoke | 429 | 306 | 455 |
| 16-type knuckle | 448 | 302 | 406 |
Dynamic tear tests and static tensile pull tests were carried out on randomly selected castings. The dynamic tear energy at -56 °C was 130 J for the yoke body and 165 J for the separately cast test bar. In the static pull test, the maximum permanent deformation after applying the specified proof load was only 0.1–0.2 mm, and the minimum breaking load was far above the acceptance value. The excellent test results demonstrated that the green sand casting process had essentially eliminated harmful casting defects in the critical sections.
4.3 Common Casting Defects in Green Sand Steel Casting
Although the final products were qualified, the trial production exposed some casting defects that required attention. In green sand steel casting, the most frequently encountered defects are blowholes, sand scabs, metal penetration, shrinkage porosity, and cracks. Each of these defects may be avoided by controlling the process parameters.
Blowholes
Blowholes are caused by gas that enters the liquid steel from the mold, core, or pouring stream. In green sand molding, the high moisture content is the main source of gas. The gas pressure in the mold can be estimated by the ideal gas law:
\[
P_{\text{gas}} = \frac{mRT}{M V},
\]
where \(m\) is the mass of gas, \(M\) is the molar mass, \(R\) is the gas constant, \(T\) is the temperature, and \(V\) is the gas volume. If the mold permeability is insufficient or the vents are blocked, the gas pressure can overcome the metal head and penetrate into the casting, forming a blowhole. To prevent blowholes, I recommended strict moisture control, adequate venting, and a stable pouring stream.
Sand scab and erosion
Sand scab is an expansion defect caused by the thermal expansion of silica sand. When the surface layer is heated rapidly by liquid steel, it expands while the underlying layers are still cool. The resulting shear stress can fracture the sand surface, and the liquid metal penetrates beneath the cracked layer. Starch additives improve the hot wet tensile strength and provide space for sand expansion, which greatly reduces the likelihood of sand scab. In addition, mold hardness should be uniform and not excessively high.
Metal penetration
Metal penetration occurs when liquid steel enters the interstices between sand grains. This is more likely in steel casting because of the high pouring temperature and high fluidity of steel. Factors that promote metal penetration include coarse sand grains, high pouring temperature, low mold compactness, and insufficient mold coating. Applying a high-quality refractory coating on the mold and core surfaces can effectively prevent this defect.
Shrinkage porosity and hot tears
Shrinkage porosity is caused by inadequate feeding during solidification. The Niyama criterion is often used to predict shrinkage porosity:
\[
N_y = \frac{G}{\sqrt{\dot{T}}}.
\]
A value below a critical threshold indicates that the local solidification structure will form porosity. Hot tears, on the other hand, occur when the accumulated tensile strain exceeds the strain capacity of the partially solidified metal. The condition can be written as:
\[
\varepsilon_{\text{local}} \geq \varepsilon_{\text{crit}},
\]
where \(\varepsilon_{\text{local}}\) is the local strain and \(\varepsilon_{\text{crit}}\) is the critical strain for tearing. Proper riser design, use of chills, increased fillet radii, and avoid of casting restrictions help to prevent both shrinkage porosity and hot tearing.
4.4 Analysis of Key Defects in Trial Production
Coupler yoke defects
During trial production, the main casting defects in the coupler yoke were hot cracks, blowholes, sand scab, and metal penetration. Hot cracks appeared at the roots of the ingate and risers, at the hook shoulder, and on the complex curved surfaces of the hook head. These cracks were typically intergranular with an oxidized dark fracture surface, which is characteristic of hot tearing. The main causes were insufficient fillet radii at the junction of the risers and the casting, non-uniform cooling due to the insulating riser sleeves, and premature shakeout. I recommended increasing the junction radii, applying an insulating coating on the mold surface near the junctions, and strictly controlling the shakeout time. A small steel chill was also added at specific hot spots to equalize the cooling rate.
Blowholes appeared primarily near the risers on the large upper surface of the shank. They had a smooth internal wall and a pear-like shape. The cause was the combined effect of high moisture in the green sand and insufficient venting. To reduce this defect, the water addition was adjusted to the lower acceptable limit, and additional vent holes were drilled in the upper mold.
Sand scabs were found on the upper internal curved surfaces of the hook head. The defect was managed by adding a small amount of starch additive, which increased the hot wet tensile strength and the surface stability. The operators were also instructed to ensure uniform mold compaction around the complicated core prints.
Metal penetration was observed inside the tail pin hole of the yoke. The pin-hole core was subjected to severe thermal attack from the liquid steel. The coating on the core was sometimes too thin or uneven. The solution was to use a more refractory coating with a higher solids content and to apply two thin coats instead of one thick coat. The pouring temperature was also reduced by about 10 °C to reduce the thermal attack.
Coupler knuckle defects
The coupler knuckle trial castings showed hot cracks at the gate root, on the upper middle part of the inner pull face, at the head opening, and near the anticreep lug. Similar to the yoke, these hot cracks were caused by high thermal strain during solidification and the presence of a weak hot spot. Because the optimized process could not be implemented immediately without a new pattern, I proposed several temporary measures: increasing the fillet radius at the gate-casting junction, reducing the pouring time slightly, and placing a small external chill at the corner of the protective lug. These measures reduced the number of cracks substantially.
Sand scab on the pin hole and on the upper surface near the pin hole was also common. This was caused by the impingement of molten steel on the core surface and the expansion of the sand. The temporary solution was to move the ingate direction by adjusting the gating system so that the steel entered tangential to the pin-hole core. The optimized pattern would completely solve this problem by moving the ingate to the back side.
Metal penetration appeared on the inner pull face of the knuckle. The inner core was difficult to coat uniformly because of its complex shape. I suggested using a brush coating with a zircon-rich paint and allowing it to dry completely before core assembly. In this way, the number of defect spots was reduced and the cleaning time was decreased.
4.5 Optimization of Processing Practices
Based on the defect analysis, the following practices were recommended for the production stage:
- Control the moisture of the facing sand between 3.0% and 3.6% depending on the weather. On humid days, the moisture should be reduced to avoid blowholes and sand scab.
- Maintain the compactability within the target range of 50–55% for the facing sand and 55–60% for the backing sand. The online compactability controller should be calibrated every day.
- Use a zircon or silica-based refractory coating on all cores that come into contact with steel for a long time, especially in thick sections.
- Design the gating system so that the molten steel does not directly strike the cores or the vertical walls of the mold. Use refractory sleeves and filters to reduce turbulence and slag entrapment.
- Set the riser-casting junction fillet radius to at least 15–20 mm for the yoke and 10–15 mm for the knuckle.
- Do not shake out the castings too early. The recommended shakeout time should be determined by simulation or by measuring the casting temperature near the thickest section.
- For welding repair, preheat the area to about 200–300 °C before welding and cool slowly afterward to avoid cold cracks and hydrogen-induced cracks.
- When cutting the risers and ingates, preheat the root region and cut step by step. Leave 2–4 cm of the riser for the first cut to reduce the thermal shock. After cutting, promptly place the casting in a furnace or in an insulated container.
These recommendations significantly reduced the incidence of casting defects during the subsequent trial runs. The castings used for the final certification were virtually free of cracks, and only small surface defects were found that could be removed by grinding and repair welding.
5. Batch Production Validation
5.1 Process Adjustments
After the company-level and ministry-level certifications, the foundry started batch production of the 17-type coupler yoke and the 16-type coupler knuckle. The process used for batch production was essentially the process that had been optimized during the trial production, with a few refinements. The fillet radii at the riser roots were increased, and the insulating coating was applied to the mold surfaces around the junctions. The cutting procedure for risers and ingates was standardized, and the weld repair procedure included preheating and slow cooling. The heat treatment cycle was also modified to ensure more uniform cooling after pre-normalizing.
5.2 Surface Inspection Results
In the initial inspection after shakeout and primary cleaning, I randomly selected several groups of products to analyze the surface quality. The results are shown in Table 14.
| Product | Number inspected | Number defective | Defect rate (%) | Cause of rejection |
|---|---|---|---|---|
| Yoke | 154 | 8 | 5.2 | Sand penetration through the wall |
| Knuckle | 223 | 6 | 2.7 | Sand penetration through the wall |
The only products that had to be scrapped were those with a complete sand penetration channel through the wall. The reason for this defect was related to improper molding around a complex core print, which left loose sand in the cavity. When the steel was poured, the loose sand was swept along and became embedded in the wall. This problem was effectively solved by improving the quality of the mold compaction near the core prints and by using a handheld air rammer to re-pack those areas. After this adjustment, the scrapped rate for this defect dropped to a very low level.
5.3 Crack Inspection Results
The castings were inspected twice with magnetic particle testing: once in the foundry shop after pre-normalizing, grinding, and repair welding, and once in the finishing shop after quenching, tempering, and final repair welding. The crack locations were recorded to evaluate the effectiveness of the process changes. Table 15 shows the crack statistics for the foundry shop inspection.
| Product | Crack location | Number of cracked castings | Percentage of inspected castings |
|---|---|---|---|
| Yoke | Head face (C, C1, C2) | 36 | 20.1 |
| Hook neck / tail portion | 30 | 16.8 | |
| Riser and gate roots | 112 | 62.6 | |
| Knuckle | Inner face (S) | 21 | 11.8 |
| Riser and gate roots | 8 | 4.5 |
From Table 15, I found that the majority of cracks detected in the foundry shop were located at the roots of the risers and gates. These cracks were shallow, with a depth of less than 10 mm, and could be completely removed by gouging and repair welding. The cracks on the yoke head face and hook neck were fewer but still noticeable. The knuckle inner face showed cracks in about 11.8% of the inspected castings, which was acceptable because the defects could be repaired. After the repair welding and heat treatment, the finishing-shop inspection showed a much lower crack rate, as shown in Table 16.
| Product | Number inspected | Crack location | Number cracked | Percentage (%) |
|---|---|---|---|---|
| Yoke | 690 | Hook ear, shank, hook neck, tail | 77 | 11.2 |
| Knuckle | 500 | Inner face / other | 15 | 3.0 |
The decrease in the crack rate from the foundry shop to the finishing shop was expected because many of the foundry-shop cracks had already been completely removed before heat treatment. The remaining cracks were mostly small, superficial, and concentrated in areas where the casting had been previously welded. This indicates that the thermal stress generated by the heat treatment and the repair welding process could still create minor fissures. Therefore, I recommended further training of the operators on the importance of preheating before welding and slow cooling after welding.
During the batch production period, the surface quality improved continuously. The number of products with no obvious surface defects increased from the initial inspection to the final inspection. The crack defects that did appear were within the control range of the machining and repair process, so the production line did not suffer major interruptions. The final products passed all the required checks, including dimensional inspection, hardness, magnetic particle inspection, and section density. This validated that the optimized green sand casting process was suitable for the mass production of railway steel castings.
6. Conclusions
In this thesis, I investigated the process optimization and defects prevention of railway steel castings produced in green moulding clay sand. The main conclusions are as follows:
First, I developed optimized formulations for both facing sand and backing sand. The facing sand formula used 100% new sand, 8.0–8.5% sodium bentonite, 0.3–0.5% starch additive, and 3.0–3.6% water. The resulting properties satisfied the factory target for compactability, permeability, green compressive strength, hot wet tensile strength, shatter index, and surface stability. The backing sand formula used 100% reclaimed sand with a small addition of bentonite and starch additive, and the optimum water content was 3.3–3.6%. The production verification proved that these formulas gave stable sand properties and significantly reduced casting defects such as sand scab and blowholes.
Second, I used ProCAST simulation to predict the locations of casting defects in the original coupler yoke and knuckle processes. The simulation identified high-risk zones for erosion, sand inclusion, gas entrapment, shrinkage porosity, and hot tearing. Based on the simulation results, I proposed optimized gating and risering systems that reduced the flow velocity in front of the cores, improved the feeding efficiency, and decreased the thermal stress in critical regions. The optimized processes were successfully applied to the trial production of the coupler yokes. For the knuckles, the optimized process was partially implemented, and the remaining defects were controlled by temporary process measures and improved welding practices.
Third, I analyzed the casting defects encountered during trial and batch production. The most important casting defects were hot cracks at the riser and gate roots, cold cracks caused by improper cutting, welding cracks, blowholes, sand scab, and metal penetration. The primary causes were related to insufficient fillet radii, non-uniform cooling, excessive moisture, poor coating quality, and improper post-casting operations. I proposed a set of practical solutions such as increasing fillet radii, applying insulating coating near the junctions, using high-quality refractory coatings, controlling the shakeout time, and standardizing the cutting and welding procedures. These solutions greatly reduced the incidence of casting defects.
Finally, the batch production data showed that the 17-type coupler yoke and the 16-type coupler knuckle had little visible surface damage, and the occasional crack defects were always within the range that could be repaired. The products passed all quality inspections including chemical analysis, tensile testing, impact testing, section density, hardness, dynamic tear, and static pull tests. This proves that green moulding clay sand is a reliable and economical process for manufacturing high-quality railway steel castings, provided that the materials, process parameters, and post-casting operations are strictly controlled.
In the future, more work is needed to automate the online sand testing system and to develop a comprehensive post-casting procedure that eliminates thermal cracks and welding cracks completely. I believe that the green sand process will be increasingly applied in the railway casting industry because of its economic and environmental benefits.