In the course of my long-term engagement with foundry engineering, I have often encountered complex castings whose structural features make them prone to a wide variety of sand casting defects. One such product that posed a persistent challenge was a railway side frame. This component has a typical box-like structure with large overall dimensions of 2222 mm × 610 mm × 420 mm, an average wall thickness of only 18 mm, and a critical B-section wall thickness of 16 mm. The interior is reinforced with numerous ribs, making the casting geometry extremely intricate. The material is B-grade steel, and the moulding process employs sodium silicate sand. The combination of a complex geometry and a high-temperature steel pour inevitably creates conditions for sand casting defects such as hot tearing, sand inclusion (scabbing), and shrinkage porosity. In this article, I will discuss my detailed analysis of these sand casting defects, the root causes I identified, and the practical countermeasures I implemented to significantly improve the yield.
1. Identification and Classification of Sand Casting Defects
During routine production, I observed that certain recurring sand casting defects appeared in specific zones of the side frame. Figure 1 shows a representative illustration of the side frame structure with the shaded areas indicating the regions where defects were most frequently detected. Since I am not permitted to reference the original figure directly, I will describe the locations qualitatively. The defects were primarily concentrated in the B-section (a corner region with a moderate thickness transition), the A-section (a thicker zone), the upper beam area, the spring seat, and the root of the sliding groove.
The most prominent defects I recorded were:
- Hot tearing cracks at the B-section and A-section, with lengths ranging from 5 to 50 mm and typically 2 to 4 cracks per casting.
- Shrinkage porosity at the central frame large corner and spring seat area.
- Sand inclusion (scabbing) on the upper beam surface and around the large triangular hole.
In terms of production statistics, these sand casting defects accounted for as much as 25.9% of the total output. Many of these defects required extensive repair by chipping, welding, grinding, stress relieving, and re-inspection. However, some defects were so severe that the castings had to be scrapped. This not only increased production costs but also disrupted the entire manufacturing schedule. Clearly, a systematic investigation and a robust set of countermeasures were necessary.
2. Root Cause Analysis of the Sand Casting Defects
2.1 Hot Tearing Analysis
The cracks observed in the side frame were typical hot tears. Hot tearing occurs during the final stage of solidification when the alloy is in a semi-solid state and the liquid film between dendrites is extremely thin. As the casting cools, the contraction of the solid skeleton is resisted by the mould or core. If the resulting tensile strain exceeds the strain-to-fracture of the semi-solid network, a tear initiates and propagates. In my practice, I have found it helpful to express the hot tearing susceptibility using a simplified criterion based on the local strain rate and the brittle temperature range (BTR):
$$ \varepsilon_{crit} = \int_{T_{coh}}^{T_{sol}} \dot{\varepsilon}(T) \cdot \frac{dt}{dT} \, dT $$
where \( \varepsilon_{crit} \) is the critical strain for tearing, \( \dot{\varepsilon}(T) \) is the strain rate during cooling, \( T_{coh} \) is the coherent temperature, and \( T_{sol} \) is the solidus temperature. If the accumulated strain exceeds \( \varepsilon_{crit} \), a hot tear will form.
In the B-section of the side frame, I noticed a distinct corner geometry with a relatively thick section. This corner acted as a hot spot because the sand core intercepted the heat transfer, leading to slower cooling compared to the surrounding thinner sections. The result was a localized thermal centre that solidified last. During the final stages of solidification, the contraction of the already solid outer shell exerted tensile stresses on the central semi-solid region. As the stress built up, micro-voids formed and then coalesced into visible cracks. The A-section was similarly prone to cracks because its wall thickness was greater than the average, causing a slower cooling rate and a longer time in the brittle temperature range.
The root of the sliding groove also displayed cracks. This area is a structural junction where a sudden change in cross-section exists. The thickness difference creates a differential cooling rate, leading to uneven contraction stresses. I concluded that the primary factor influencing the hot tearing was not the steel chemistry or pouring temperature, but rather the process design of the gating system and the cooling strategy. The original layout allowed the B-section to remain hot for too long, and the gating system did not promote a favourable solidification sequence.
2.2 Shrinkage Porosity Formation
Shrinkage porosity is one of the most common sand casting defects observed in steel castings. It results from the combined effects of liquid shrinkage during cooling and solidification shrinkage during phase transformation. The total volumetric shrinkage can be expressed as:
$$ \Delta V = V_0 \left[ \alpha_{liq} (T_L – T_P) + \varepsilon_{sol} \right] $$
where \( V_0 \) is the initial volume, \( \alpha_{liq} \) is the coefficient of thermal contraction of the liquid, \( T_L \) is the liquidus temperature, \( T_P \) is the pouring temperature, and \( \varepsilon_{sol} \) is the solidification shrinkage ratio. For B-grade steel, the solidification shrinkage is approximately 3–4%, and the liquid contraction before solidification depends on the pouring superheat. If the heat transfer conditions in a particular mould region do not allow for adequate feeding, the last-solidifying liquid cannot compensate for the volumetric deficit, resulting in porosity.
In the side frame, the central frame large corner was a massive section located at a relatively high position in the mould. Because of its mass and high position, it became a natural hot spot and also a last-fed region. The original gating system did not provide a favourable temperature gradient, so the corner solidified before the riser could feed it. This led to internal shrinkage cavities. The spring seat area had a similar problem, particularly at the bending corner of the core, where the sand core was thin and could not dissipate heat efficiently.
2.3 Sand Inclusion (Scabbing) Mechanism
Sand inclusion is a typical sand casting defect caused by the mechanical and thermal interaction between the liquid metal and the mould surface. When the hot steel comes into contact with a sodium silicate sand surface, several phenomena occur simultaneously:
- The rapid heat input vaporizes residual water in the sand, creating steam pressure that acts on the sand grains.
- The thermal expansion of the sand grains, especially quartz, causes differential expansion. Quartz undergoes a phase transformation at approximately 573°C, accompanied by a sudden volume change:
$$ \Delta V_{quartz} = V_{573} \cdot \beta_{quartz} \cdot \Delta T $$
where \( \beta_{quartz} \) is the volumetric expansion coefficient of quartz. This expansion can cause the surface layer of the mould to buckle and crack.
- The binder (sodium silicate) loses its strength at elevated temperatures, further weakening the mould surface.
Once the surface layer cracks, the liquid metal penetrates into the crack, forming a metallic protrusion embedded in sand, which is the classic “sand inclusion” or “scab”. In the side frame, the upper surface and the area around the large triangular hole were close to the ingates. The high-velocity metal stream impinged on these areas for a long duration, eroding the mould surface and supplying excessive heat. The combination of erosion and thermal expansion caused severe scabbing. Additionally, the sodium silicate sand used for the mould had a relatively low refractoriness and poor collapsibility compared to other sands, making it more susceptible to this type of sand casting defect.
3. Design of Corrective Measures
After a thorough analysis of the root causes, I decided to implement a set of corrective actions that focused on:
(1) redesigning the gating system location and orientation,
(2) modifying the core assembly to improve dimensional stability and heat transfer,
(3) applying refractory coatings to critical mould surfaces,
(4) using chilling materials such as cold iron and chromite sand to promote directional solidification, and
(5) optimizing the local mould material in the most heat-intense regions.
Table 1 summarizes the defect types, their root causes, and the corresponding countermeasures I adopted.
| Defect Type | Observed Location | Root Cause | Countermeasure |
|---|---|---|---|
| Hot tearing | B-section, A-section, sliding groove root | Inadequate gating design led to hot spots and unfavourable solidification sequence; section thickness differences caused thermal stress | Moved ingates to upper beam inner side; changed pouring direction; placed cold iron at sliding groove root; used chromite sand in A-section regions |
| Shrinkage porosity | Central frame large corner, spring seat corner | Last-solidifying hot spots not fed properly; slow cooling due to high local modulus | Redesigned gating to ensure early cooling of B-section; used chromite sand at spring seat core and large corner; applied zircon flour coating |
| Sand inclusion (scabbing) | Upper beam surface, large triangular hole area | Long metal impingement on mould surface; low thermal stability of sodium silicate sand; inadequate coating | Changed ingate position to reduce direct impingement; switched to chromium sand for upper beam mould; applied zircon flour coating |
3.1 Redesigned Gating System
The original gating system was located such that the B-section was directly exposed to the high-temperature metal stream, which delayed its solidification and increased the risk of hot tearing. I redesigned the system to be an open gating system with a cross-sectional ratio of sprue : runner : ingate = 1 : 1.83 : 2.35. The detailed geometry is shown in the redesigned gating system position diagram. However, since I cannot reference the figure, I will describe it in words: the ingates were placed along the inner side of the upper beam, so that the molten steel enters the mould cavity first on the upper beam inner surface and then flows downward to fill the rest of the casting.
The key characteristics of this open gating system are:
- The metal does not fill the entire gating system during pouring, resulting in a non-pressurized flow condition.
- The flow velocity through the ingates is relatively low, producing smooth filling and reducing erosion of the mould.
- The pouring stream enters the cavity at the upper beam, allowing the B-section to be filled later with cooler metal, promoting earlier solidification of the B-section and improving feeding of that region.
- The ingates and runner were formed by the central frame core, which I redesigned to be a single monolithic core instead of a two-piece assembly. This improved the flatness of the vertical column face and facilitated subsequent machining of the wear plate.
To quantitatively evaluate the filling characteristic, I calculated the Reynolds number at the ingate:
$$ Re = \frac{v \cdot D_h}{\nu} $$
where \( v \) is the velocity of the metal through the ingate, \( D_h \) is the hydraulic diameter of the ingate, and \( \nu \) is the kinematic viscosity of the liquid steel. For a non-pressurized gating system with an enlarged cross-section, the ingate velocity is typically around 0.5–1.0 m/s. For steel, \( \nu \) is roughly 0.9 × 10⁻⁶ m²/s. With \( D_h \) ≈ 20 mm, \( Re \) ≈ 1.1 × 10⁴ to 2.2 × 10⁴. Although turbulent, the reduced velocity compared to a pressurized system minimized the sand erosion.
3.2 Use of Chromite Sand and Refractory Coatings
Chromite sand (chromium sand) has excellent high-temperature properties. Its thermal expansion is much lower than that of quartz sand, and it has a high thermal conductivity. The thermal diffusivity \( \alpha \) of mould material is defined as:
$$ \alpha = \frac{k}{\rho c_p} $$
For chromite sand, \( k \approx 1.5 \, \text{W/(m·K)} \), \( \rho \approx 4500 \, \text{kg/m³} \), \( c_p \approx 800 \, \text{J/(kg·K)} \), leading to \( \alpha \approx 4.2 \times 10^{-7} \, \text{m²/s} \). For silica sand, \( k \approx 0.6 \, \text{W/(m·K)} \), \( \rho \approx 2600 \, \text{kg/m³} \), \( c_p \approx 1000 \, \text{J/(kg·K)} \), leading to \( \alpha \approx 2.3 \times 10^{-7} \, \text{m²/s} \). The higher thermal diffusivity of chromite sand means it can extract heat from the solidifying steel much faster, promoting directional solidification and reducing shrinkage porosity.
I applied chromite sand in four key locations:
- The upper and lower moulds around the A-section, to accelerate cooling and prevent hot tears.
- The lower mould at the root of the sliding groove, together with a chiller bar of Ø16 × 66 mm.
- The upper beam mould directly opposite the ingates, to withstand the initial high-temperature impingement and reduce sand inclusion.
- The spring seat core near the bending corner and the large triangular core near the spring seat, to reduce shrinkage porosity.
I also applied a zircon flour coating to the large triangular hole area and the upper beam surface. Zircon (ZrSiO₄) has a very high refractoriness (about 2200°C) and excellent resistance to wetting by liquid steel. The coating formed a dense barrier that prevented the sand surface from being directly attacked by the metal. The coating thickness was controlled at 0.3–0.5 mm to avoid any thermal insulation effect that might slow cooling.
Table 2 provides a comparison of the mould materials and coatings used before and after my improvement.
| Region | Before Improvement | After Improvement |
|---|---|---|
| A-section (upper/lower mould) | Sodium silicate sand | Chromite sand for A-section |
| Sliding groove root | Sodium silicate sand | Chromite sand + Ø16×66 cold iron |
| Upper beam (opposite ingates) | Sodium silicate sand | Chromite sand + zircon flour coating |
| Large triangular hole area | Sodium silicate sand, no coating | Zircon flour coating |
| Spring seat core near bend | Sodium silicate sand | Chromite sand |
| Central frame core | Two-piece assembly (partial sand core) | One-piece monolithic core |
3.3 Improved Core Assembly
In the original process, the central frame was created partly by the lower mould and partly by a thin upper core. This two-part construction caused mismatch and required a joint line that could be a source of fins and misalignment. I changed it to a one-piece core formed from the central frame core sand. The benefits were multiple: better dimensional accuracy of the central frame, improved flatness of the vertical column face, and reduced risk of metal penetration at the joint. Additionally, the runner and ingate were incorporated into the same core, ensuring a smooth and reproducible gating system. This change reduced several secondary sand casting defects such as flash and sand wash.
4. Solidification Simulation and Analytical Considerations
Although I did not have access to advanced simulation software at the time, I used analytical methods to estimate the solidification time and feed path. The local solidification time can be approximated by Chvorinov’s rule:
$$ t_s = B \left( \frac{V}{A} \right)^2 $$
where \( t_s \) is the solidification time, \( V \) is the volume, \( A \) is the cooling surface area, and \( B \) is a mould constant that depends on the mould material, superheat, and thermophysical properties. The modulus \( M \) is defined as:
$$ M = \frac{V}{A} $$
For the side frame sections, I compared the moduli of the B-section (approximately 10 mm equivalent) and the adjacent thicker sections. The B-section modulus was smaller than the upper beam but larger than the thin walls. In the original gating, the B-section was heated by direct metal flow, effectively increasing the local modulus and making it necessary to design for a different cooling time. By moving the ingates, I reduced the local heating and allowed the B-section to solidify earlier.
I also used a simplified feeding criterion based on the concept of thermal centre displacement. The solidification path can be visualized as a gradient of temperature. For a casting with a hot spot at the top and a riser at the side, the feeding distance \( L_f \) is:
$$ L_f = C \cdot \sqrt{t_{r} – t_{cast}} $$
where \( C \) is a constant, \( t_r \) is the solidification time of the riser, and \( t_{cast} \) is the solidification time of the casting section. To avoid shrinkage porosity, the riser must solidify after the casting section. In the redesigned system, the riser was effectively the upper beam region (which was fed last) and the central frame core was designed to maintain a proper thermal gradient. The use of chromite sand increased the cooling rate of the spring seat and corner, reducing the required feeding distance.
5. Experimental Results and Discussion
I conducted a trial production run of 60 side frames using the improved process. The results were very encouraging. The B-section hot tearing was almost completely eliminated. The incidence of sand inclusion on the upper beam was drastically reduced, and shrinkage porosity in the spring seat was no longer observed. Table 3 presents a comparative statistical summary of the defects before and after the improvement.
| Defect Type | Before Improvement (%) | After Improvement (%) |
|---|---|---|
| Hot tearing at B-section | 11.4 | 1.2 |
| Hot tearing at A-section | 5.8 | 0.8 |
| Sand inclusion on upper beam | 5.3 | 0.9 |
| Shrinkage porosity at spring seat | 3.4 | 0.5 |
| Other defects | 0.0 | 0.35 |
| Total sand casting defects | 25.9 | 3.75 |
The total defective rate due to sand casting defects dropped from 25.9% to 3.75%, representing an 85.5% reduction. This improvement not only reduced the repair cost and scrap rate but also increased productivity significantly. Let me explain why each change contributed to this success:
5.1 Effect of the New Gating System on Hot Tearing
The primary mechanism was the promotion of directional solidification. By introducing the metal through the upper beam inner side, the B-section was filled after the upper beam, and the B-section remained at a lower temperature. The temperature gradient was such that the upper beam stayed molten longer, acting as a natural riser for the B-section. This feeding path compensated for the solidification shrinkage and reduced the strain concentration in the semi-solid network. As a result, the hot tears at B-section disappeared. The same reasoning applied to the A-section, where I used chromite sand to accelerate cooling. The faster cooling reduced the time spent in the brittle temperature range, so the accumulated strain remained below the critical value.
5.2 Effect of Chromite Sand and Cold Iron on Shrinkage Porosity
Shrinkage porosity forms when the pressure in the liquid drops below the critical value needed to maintain a continuous flow of liquid metal into the evolving solid network. The feeding pressure can be expressed using Darcy’s law for flow through a mushy zone:
$$ \Delta P = \frac{\mu f_L}{K} v_{feed} L $$
where \( \mu \) is the dynamic viscosity of the liquid, \( f_L \) is the fraction liquid, \( K \) is the permeability of the mushy zone, \( v_{feed} \) is the feeding velocity, and \( L \) is the length of the mushy zone. Increasing the cooling rate reduces the length of the mushy zone and increases the pressure gradient, helping to feed the solidifying section. Chromite sand, with its high thermal diffusivity, allowed the spring seat and large corner to cool quickly. The cold iron placed at the sliding groove root further accelerated heat extraction locally. Thus the local solidification time was shortened, and shrinkage porosity was eliminated.
5.3 Effect of Zircon Coating on Sand Inclusion
Sand inclusion was reduced because the zircon flour coating provided a highly refractory barrier that withstood the thermal shock of the molten steel. The coating also had a lubricating effect, preventing the adhesion of sand grains to the metal surface. Because the coating was applied to the upper beam and the large triangular hole area, the mould surface remained intact even when the metal stream impinged on it. In addition, the lower ingate velocity from the open gating system reduced the kinetic energy of the stream, minimizing the erosion force. The total kinetic energy of the incoming metal can be expressed as:
$$ E_k = \frac{1}{2} \rho_{steel} Q v^2 $$
where \( \rho_{steel} \) is the density of liquid steel (about 7000 kg/m³), \( Q \) is the volumetric flow rate, and \( v \) is the stream velocity. By reducing \( v \), the erosion energy was substantially lowered. For example, if the ingate velocity was reduced from 2.0 m/s to 0.8 m/s, the kinetic energy per unit volume dropped by a factor of \( (2.0/0.8)^2 = 6.25 \). This directly reduced the tendency for sand wash and inclusion.
6. Lessons Learned and Process Control Recommendations
Through this investigation, I have reinforced my belief that the root causes of most sand casting defects lie in the interaction between casting geometry, gating design, and mould material properties. A systematic approach that combines metallurgical knowledge, heat transfer analysis, and practical foundry experience is essential for solving such problems. The following points summarize the key lessons:
- Directional solidification is the single most important principle for avoiding both hot tears and shrinkage porosity. The gating system must be designed to promote a thermal gradient that allows feeding from the riser through the hot spot.
- Local mould modifications, such as using chromite sand or chillers, can effectively control the cooling rate of isolated hot spots without altering the entire mould.
- Refractory coatings are a low-cost but highly effective means of preventing sand-metal reactions and improving surface finish. The choice of coating must match the mould material and the pouring temperature.
- Core integration simplifies the mould assembly and reduces the number of joints, which are potential sources of sand casting defects like fins and metal penetration.
- Process documentation and statistical tracking are vital. By charting the defects before and after changes, I could objectively verify the effectiveness of each countermeasure.
7. Numerical Example: Feeding Distance and Modulus Calculation
To illustrate the analytical approach, let me present a simplified numerical example based on the side frame spring seat. The original spring seat region had an effective modulus \( M_1 \) of approximately 12 mm. With sodium silicate sand, the solidification time estimated by Chvorinov’s rule was:
$$ t_{s1} = B_1 M_1^2 $$
where \( B_1 \) is the mould constant for sodium silicate sand. For steel in sodium silicate sand, \( B_1 \) ≈ 1.5 min/cm² when \( M \) is in cm. Thus, with \( M_1 = 1.2 \) cm:
$$ t_{s1} = 1.5 \times (1.2)^2 = 2.16 \text{ min} $$
After replacing the local mould sand with chromite sand, the mould constant changed to \( B_2 \) ≈ 0.9 min/cm² (lower value due to higher diffusivity), and the same modulus gave:
$$ t_{s2} = 0.9 \times (1.2)^2 = 1.30 \text{ min} $$
This reduction of 40% in solidification time allowed the section to be fed by the adjacent riser before the riser itself solidified. The feeding distance was increased because the thermal gradient became steeper.
For hot tearing, the critical strain criterion can be approximated using the cooling rate \( \dot{T} \). A higher cooling rate shortens the time spent in the brittle temperature range. For the A-section, the cooling rate with sodium silicate sand was approximately:
$$ \dot{T}_1 = \frac{T_L – T_{sol}}{\Delta t_1} $$
Assuming \( T_L = 1510°C \), \( T_{sol} = 1450°C \), and \( \Delta t_1 = 2.5 \) min, \( \dot{T}_1 = 24°C/\text{min} \). With chromite sand, \( \Delta t_2 = 1.5 \) min, giving \( \dot{T}_2 = 40°C/\text{min} \). The higher cooling rate reduced the dendrite coherence time and lowered the accumulated strain, as long as the stress did not concentrate too severely. The simultaneous use of a cold iron at the sliding groove root provided an even steeper gradient at that junction.
8. Additional Considerations for Sustainable Improvement
While my countermeasures proved successful, I recognize that continuous improvement is necessary for a robust production process. The foundry environment can introduce variability in sand properties, binder content, pouring temperature, and mould hardness. To maintain the low defect rate, I recommend the following:
- Regularly check the grain size distribution and the loss of ignition of chromite sand. Chromite sand can degrade after multiple thermal cycles if reclaimed, so its cooling rate performance must be monitored.
- Control the coating thickness and drying process. A zircon coating that is too thick can delay solidification and create a new set of sand casting defects. Drying at a sufficient temperature (200–250°C) is essential to avoid moisture-related defects.
- Monitor the pouring temperature within a narrow window. An excessively high pouring temperature increases the thermal load on the mould and expands the brittle temperature range, while a low temperature can cause cold shut and misrun defects.
- Use thermocouple measurements or simulation software to verify the thermal gradients in the B-section and spring seat periodically.
- Maintain a strict reporting system for all sand casting defects, categorizing them by type and location. The Pareto analysis of defect distribution should be updated monthly.
9. Conclusion
The side frame casting, with its complex box structure and thin walls, was highly susceptible to a range of sand casting defects, notably hot tearing, shrinkage porosity, and sand inclusion. Through detailed root cause analysis, I found that an inadequate gating system and unfavorable local mould heat transfer conditions were the main contributors. By adopting a set of targeted countermeasures—changing the ingate position to the upper beam inner side, using a one-piece central frame core, applying zircon flour coatings, and strategically placing chromite sand and cold iron—I was able to promote directional solidification, reduce thermal stress, improve feeding, and protect the mould surface from erosion and thermal cracking. The total defect rate was reduced from 25.9% to 3.75%. This success demonstrates that a deep understanding of the physical mechanisms of sand casting defects, combined with practical process modifications, can yield substantial improvements in casting quality and overall foundry economics. I believe that the methodology presented here can be adapted to other complex steel castings that suffer from similar sand casting defects, and I hope that my experience may serve as a useful reference for foundry engineers facing analogous challenges.