This paper presents a comprehensive study on the development of a reliable casting process for a BOSS component, which is the critical connection between the large arm of an excavator and its frame. The component, originally produced by forging and welding, was redesigned as a single steel casting to reduce cost and eliminate weld-related cracking risks. However, conventional casting practices led to hot tearing near the root of the boss and caused serious difficulties in removing core sand from the internal axle bore. In this work, I combined numerical simulation using PROCAST with physical tryout experiments to understand the root causes of the observed defects and to establish an optimized process. The study focused on the effects of external chills on hot tearing tendency, the influence of core material on casting quality, and the development of a practical solution for sand removal from the deep bore. The final process uses a gravity casting method with a silica-sand mold, a limestone-sand core, no external chill, a pouring temperature of 1565°C, and a filling time of 25 seconds. The results demonstrate that the proposed process successfully eliminates hot cracks and significantly improves the cleanability of the bore without compromising the mechanical integrity of the casting.
The BOSS part belongs to the family of thin-walled steel castings with complex geometry. As shown in the part drawing, the component has a long cylindrical bore with a diameter of 90 mm and a total length of 754 mm. The local wall thickness varies from 15 mm in the central web to 52 mm at the thickened bosses that support the axle holes. This large thickness difference, combined with the high restraint imposed by the mold and core, makes the casting prone to hot tearing. The technical requirements are extremely stringent: no crack-like indications larger than 2 mm are allowed in the fluorescent magnetic particle inspection, and the internal bore must be free of adhered sand after cleaning. The material specified by the customer is a medium-carbon cast steel with a composition close to ZG25, but with slightly higher manganese to improve fluidity and to counteract the harmful effects of sulfur. This composition provides acceptable castability, but the susceptibility to hot tearing remains significant, especially when the cooling is non-uniform and when external chills are used in an aggressive manner.
I began the investigation by designing a conventional casting process that followed the traditional rules of directional solidification. The parting line was placed along the horizontal mid-plane of the casting, allowing the central bore core to be supported at both ends and permitting easy dimensional inspection after closing the mold. A tapered sprue, runner, and two ingates were designed to feed liquid steel toward the risers. Two cylindrical risers with a diameter of 150 mm and a height of 300 mm were located on the thickened bosses to compensate for liquid shrinkage. The mold and core were originally made of CO2-hardened sodium silicate-bonded silica sand. This mold material is inexpensive and provides good surface quality, but it suffers from poor collapsibility and is extremely difficult to remove from deep cavities. In the first process version, I followed the common industrial practice of placing a large external chill at the bottom of the casting to accelerate cooling of the thick sections. The chill was a single flat plate of Q235 steel with dimensions of 32 × 40 × 750 mm, placed parallel to the parting line, directly under the main body. That arrangement was actually used by a previous supplier, but it produced severe cracks in the transition zone between the thick boss and the thin central web. To understand the cause and to evaluate alternatives, I formulated three chill arrangements, as summarized in Table 1.
| Scheme | Chill location | Chill dimensions (mm) | Remark |
|---|---|---|---|
| 1 | Entire bottom face | 32 × 40 × 750 | Original production practice |
| 2 | Bottom of thick bosses only | 32 × 30 × 120 (two pieces) | Reduced chill size |
| 3 | No chill | — | Alternative proposed by simulation |
For all three schemes, the mold and core were made of water-glass-bonded silica sand, and the same gating and risering system was employed. The pouring temperature for the simulation was set to 1565°C, and the pouring time was 25 seconds, corresponding to a mass flow rate of approximately 9.5 kg/s. The initial mold temperature was 20°C, and the ambient temperature was 20°C. The simulation was terminated when the casting reached 1132.8°C, which is 10°C below the solidus temperature of the alloy. I used PROCAST’s casting process simulation platform with the thermo-mechanical stress module to predict the risk of hot tearing. The hot tearing indicator (HTI), which is based on the strain-driven model and evaluates the accumulated plastic strain in the semisolid region, was used to quantify the cracking tendency. The stress fields were computed from the thermal history using an elastic-plastic constitutive model. The material properties for the steel were derived from the chemical composition using the thermodynamic database. The mesh consisted of about 830,000 elements, with a finer mesh of 10 mm in the casting and chill areas, in order to capture the steep thermal gradients accurately.
The filling sequence showed a smooth and tranquil advancement of the liquid metal front. However, the gas entrapment analysis indicated that the two ears at the upper part of the casting were the last to fill and could trap gas if no venting holes were provided. In the actual production, I added small vent holes at those positions. The temperature distribution during the solidification showed that the risers were effective in feeding the thickened portions of the casting. Figure 1 illustrates the solidification sequence obtained from the simulation. The macro-porosity prediction indicated that shrinkage porosity was mainly located inside the risers, with a slight porosity below the neck of the right riser. The Niyama criterion map, computed at a threshold value of 7.0, revealed that the central thin wall section contained a considerable amount of micro-porosity, but the porosity level was less than 2 vol. %, which met the acceptance criteria. This micro-porosity was attributed to the wide solidification range of the steel (the liquidus was 1501.1 °C and the solidus was 1142.8 °C, giving a difference of 358 °C), which leads to a large mushy zone and impedes the flow of residual liquid through the dendritic network.
The most important finding from the simulation was the influence of the external chill. Table 2 summarizes the key filling and solidification data for the three chill schemes. The total process time from pouring to the completion of cooling to 1132.8 °C was slightly shorter (5154 s) for scheme 1 than for schemes 2 and 3, because the large chill accelerated the cooling of the entire casting. The filling time was essentially the same for all schemes, as expected, because the gating system was unchanged. However, the average solid fraction at the end of filling was 15.1% for scheme 1, compared to 11.3% and 11.5% for schemes 2 and 3, respectively. This higher solid fraction in scheme 1 is a warning sign because the melt becomes more viscous during filling, increasing the risk of incomplete filling and cold shut. In practice, this would require a higher pouring temperature or a modified gating system to ensure a complete fill. As for the chill temperatures, the surface of the single large chill in scheme 1 reached a maximum of 1178.3 °C at the moment when the solid fraction of the casting was 84%. This is still below the solidus of Q235 steel (1493 °C), so there was no risk of chill melting and welding to the casting. In scheme 2, the smaller chills reached 1257.0 °C, which is also safe. However, the core surface temperature differed substantially between the schemes: in schemes 1 and 2, the maximum core surface temperature was about 1421 °C, reached after 85% solidification, whereas in scheme 3, the core surface temperature peaked at 1594.7 °C at a much earlier stage (32% solidification). This high temperature would create a significant risk of sand adhesion and mechanical penetration, so the use of a coating on the core would be necessary if no chill is used.
| Parameter | Scheme 1 (one large chill) | Scheme 2 (two small chills) | Scheme 3 (no chill) |
|---|---|---|---|
| Total process time (s) | 5154 | 5290 | 5327 |
| Filling time (s) | 24.27 | 24.25 | 24.14 |
| Average solid fraction at end of filling (%) | 15.1 | 11.3 | 11.5 |
| Time to complete solidification (s) | 4916 | 5019 | 5040 |
| Maximum chill surface temperature (°C) | 1178.3 (at 84% solid) | 1257.0 (at 78% solid) | — |
| Maximum core surface temperature (°C) | 1420.8 (at 85% solid) | 1421.1 (at 87% solid) | 1594.7 (at 32% solid) |
The hot tearing indicator was computed at the nodes with the maximum value when the local solid fraction reached 99%. For scheme 1, the maximum HTI was 0.0042, and the location of the predicted crack was exactly the same as the crack that appeared in the real casting from the supplier. The stress components in the potential crack region are listed in Table 3. In scheme 1, the Y-direction stress (parallel to the casting axis) was 316.5 MPa, far larger than the X and Z components. This indicates that if a hot crack formed, it would propagate perpendicular to the casting axis, which matches the actual crack morphology. For scheme 2, the maximum HTI dropped to 0.0029, and for scheme 3 it was 0.0030. The stress components were also lower in schemes 2 and 3 than in scheme 1. The large difference in the Y-stress is particularly notable: 316.5 MPa in scheme 1 versus 150.4 MPa in scheme 2 and 222.6 MPa in scheme 3. The reason for the increased stress in scheme 1 is the combined effect of the large cooling surface of the chill and the friction at the casting/chill interface. The chill prevents the casting from shrinking freely, while the solid shell of the thin section already has sufficient strength at high temperature, causing a high tensile stress in the hot spot. The solidification of the thick section, which is still partially liquid, cannot withstand this imposed deformation and thus cracks along the intergranular regions. Therefore, the conventional design of a large, single external chill is highly unfavorable for this steel casting.
| Parameter | Scheme 1 | Scheme 2 | Scheme 3 |
|---|---|---|---|
| HTI (−) | 0.0042 | 0.0029 | 0.0030 |
| σ eq (MPa) | 234.3 | 199.2 | 210.8 |
| σ x (MPa) | 29.1 | 6.3 | 2.0 |
| σ y (MPa) | 316.5 | 150.4 | 222.6 |
| σ z (MPa) | 19.2 | 25.1 | 11.8 |
Although scheme 2 showed a slightly lower HTI than scheme 3, I decided not to use any chill in the final process. The reason is that the benefit of the chill in reducing the hot tearing tendency is marginal, while the presence of chills always carries a risk of creating gas porosity and cold shut defects. In addition, the use of chills requires additional labor in the molding process and increases the variability of the process. Since the simulation clearly showed that the casting without chill is safe, the chill was completely eliminated from the final design.
Another major challenge was the removal of the sand core from the 90 mm bore. The original silica sand core was bonded with sodium silicate and hardened by CO2. After pouring, this sand core sinters into a hard block that cannot be easily removed from the bore by conventional tools. The only available tool in the workshop was a pneumatic chipping hammer, which is inefficient for such a long and narrow cavity. Therefore, I decided to replace the core material with limestone sand. Limestone sand has a much better collapsibility because its main component, calcium carbonate, decomposes at the temperature of the steel casting, releasing CO2 gas and producing a porous layer of calcium oxide. The reactions can be expressed as:
\[
\mathrm{CaCO_3 \longrightarrow CaO + CO_2\uparrow}
\]
\[
\mathrm{CO_2 + 2Fe \longrightarrow 2FeO + CO\uparrow}
\]
\[
\mathrm{3FeO + CO_2 \longrightarrow Fe_3O_4 + CO\uparrow}
\]
\[
\mathrm{2Fe_3O_4 + CO_2 \longrightarrow 3Fe_2O_3 + CO\uparrow}
\]
These reactions create a reaction zone on the casting surface that is easily detached during knockout. The limestone core also shrinks and cracks after decomposition, so it can be removed much more easily. However, the decomposition of CaCO3 produces a large amount of gas, which must be safely vented from the mold. The gas evolution per gram of limestone sand is roughly 40 to 50 mL, much higher than for silica sand. If the gas is not vented, it can be forced into the liquid steel and cause blowholes, particularly in the thick-to-thin transition regions. In the first physical tryout, I used four ventilation tubes in the limestone core, but one casting still showed a blowhole in the bore. The problem was traced to the insufficient venting capacity of the tubes, which were partially collapsed during molding. To overcome this problem, I redesigned the core to have an axial through-hole with a diameter of 20 mm, which was connected to a corresponding hole in the mold and to a 40 mm vent hole at the top of the ears. This design allowed the gas to escape freely and eliminated the blowhole defect in subsequent production.
I performed two series of physical tryouts. The first series produced four castings with two silica-sand cores and two limestone-sand cores. All other conditions were identical. The pouring temperature was between 1554°C and 1578°C, and the pouring time was between 24 and 28 seconds. The chemical composition of the castings was within the specified ranges. After solidification, the castings were knocked out and cleaned. The cleaning time for the bore was measured for each casting using the same worker to avoid skill differences. The results are given in Table 4. The average cleaning time for the silica cores was 9.5 minutes, while for the limestone cores it was only 3.5 minutes. Moreover, after shot blasting, the limestone-core castings had clean and smooth inner surfaces, while the silica-core castings showed adhered sand in the middle portion of the bore. The adhesion was confirmed by cutting the casting in half and re-blasting. This demonstrates the clear advantage of limestone sand for the internal core of this steel casting.
| Casting No. | Core material | Cleaning time (min) |
|---|---|---|
| 1 | Silica sand | 8 |
| 2 | Silica sand | 11 |
| 3 | Limestone sand | 4 |
| 4 | Limestone sand | 3 |
Based on these results, I replaced the core material for the bore with limestone sand in the final process. Because the limestone sand has a lower thermal conductivity and a higher specific heat than silica sand, its chilling effect on the casting is stronger. This can be beneficial for refining the microstructure and reducing the mushy zone. The numerical simulation confirmed that the limestone-sand core produced a faster cooling rate in the bore area, which improved the temperature gradient and helped to feed the solidification shrinkage. Table 5 shows the comparison of the casting behavior between silica-sand cores and limestone-sand cores without chills. Both filling time and average solid fraction at the end of filling were almost identical. The total process time and the time to complete solidification were slightly shorter with the limestone core (about 2 minutes difference). The maximum core surface temperature, however, was substantially lower: 1388.6 °C for the limestone core versus 1594.7 °C for the silica core. This lower surface temperature reduces the tendency for metal penetration and sand adhesion, which is in agreement with the physical tryout results.
| Parameter | Silica sand core | Limestone sand core |
|---|---|---|
| Total process time (s) | 5327 | 5205 |
| Filling time (s) | 24.14 | 24.14 |
| Average solid fraction at end of filling (%) | 11.5 | 11.2 |
| Time to complete solidification (s) | 5040 | 4936 |
| Maximum core surface temperature (°C) | 1594.7 (at 32% solid) | 1388.6 (at 87% solid) |
I also extracted the temperature, solid fraction, solidification time, and thermal modulus at a representative point in the thick section of the casting when the average solid fraction was 90%. The data are presented in Table 6. At this instant, the solid fraction at the data point was 79.8% for the limestone core, compared to 70.4% for the silica core. The solidification temperature was lower (1450.3 °C vs. 1466.1 °C), the solidification time was shorter (3410.9 s vs. 3908.4 s), and the thermal modulus was smaller (1.502 cm vs. 1.550 cm). All of these results indicate that the limestone core improves the directional solidification and feeding conditions, which in turn helps to reduce shrinkage porosity and also reduces the hot tearing tendency.
| Parameter | Silica sand core | Limestone sand core |
|---|---|---|
| Solid fraction (%) | 70.4 | 79.8 |
| Temperature (°C) | 1466.1 | 1450.3 |
| Solidification time (s) | 3908.4 | 3410.9 |
| Thermal modulus (cm) | 1.550 | 1.502 |
The hot tearing risk was also evaluated for the limestone core without chill. Table 7 compares the maximum HTI and stress components for the two core materials without the chill. The limestone core produced a lower HTI (0.0027 vs. 0.0030), a lower equivalent stress (179.0 MPa vs. 210.8 MPa), and much lower x and y stress components. The z stress component was slightly higher (13.1 MPa vs. 11.8 MPa), but the overall tendency was clearly favorable. The location of the maximum HTI in the limestone-core casting was still in the thin-wall transition region, but the magnitude was below the threshold for crack initiation in practice. The physical tryout of the final process confirmed that no cracks were detected by fluorescent magnetic particle inspection.
| Parameter | Silica sand core | Limestone sand core |
|---|---|---|
| HTI (−) | 0.0030 | 0.0027 |
| σ eq (MPa) | 210.8 | 179.0 |
| σ x (MPa) | 2.0 | 0.7 |
| σ y (MPa) | 222.6 | 184.8 |
| σ z (MPa) | 11.8 | 13.1 |
In the second physical tryout, I produced four additional castings using the limestone core with the improved venting design. The cores were made with a through-hole of 20 mm diameter in the axial direction, and the mold included a matching vent hole and a 40 mm vent at the top of the ears. The pouring conditions were the same as in the first tryout. After cooling, the castings were knocked out and cleaned. The bores were clean and required minimal effort. The castings were then heat-treated by normalizing to eliminate any residual stress and to refine the grain structure. The normalizing cycle was as follows: heating at a rate of at least 80°C/h to 880±10°C, holding for 2 hours, and then cooling in air. The mechanical properties obtained from separately cast test bars are listed in Table 8. The yield strength, tensile strength, elongation, impact toughness, reduction of area, and hardness all met the specified requirements. The castings were dimensionally inspected and found to be within the CT10 tolerance. The surface roughness met the 100S target for the external surfaces. The most critical test, fluorescent magnetic particle inspection, was performed on all four castings, and no crack-like indications larger than 2 mm were found. Thus, the final process was validated.
| Property | Value | Requirement |
|---|---|---|
| Yield strength (MPa) | 325 | ≥ 240 |
| Tensile strength (MPa) | 570 | ≥ 450 |
| Elongation (%) | 25 | ≥ 18 |
| Impact toughness (J/cm²) | 37 | ≥ 25 |
| Reduction of area (%) | 26 | ≥ 20 |
| Hardness (HB) | 197 | 170–220 |
The key learnings from my work can be summarized as follows. First, the conventional use of a large external chill on the bottom of a steel casting that has a thin thick-to-thin transition creates a serious hot tearing risk. The chill increases the cooling rate in the thin sections, which already solidify quickly, and creates a strong interface friction that restricts the free contraction of the casting. As a result, the solidified shell in the thin sections subjects the hot spot in the thick section to a high tensile stress at a time when the alloy has low strength and ductility. The removal of the chill eliminates this stress and is a safe solution for this particular component, as proven by both simulation and production. Second, the use of limestone sand for the internal core of a steel casting with a deep bore provides excellent collapsibility and prevents sand adhesion. The decomposition of calcium carbonate creates a porous, weak layer that can be easily removed, reducing the cleaning time by about 60% and yielding a cleaner bore surface. Third, the high gas evolution of limestone sand must be addressed by providing adequate venting channels in the core and mold. A simple axial through-hole in the core with a diameter of 20 mm was sufficient to vent the gases and eliminate blowholes in this casting. Fourth, numerical simulation with PROCAST is a powerful tool that can predict hot tearing risk and guide the process design, thus reducing the number of physical tryouts and saving both time and material cost. In my case, the simulation allowed me to eliminate the chill without additional physical trials, even though the initial guess was to use a chill. This demonstrates the value of integrating simulation into the process development of steel castings.
To further illustrate the thermal stress analysis, I can state that the thermo-mechanical model used in PROCAST solves the heat transfer equation coupled with a small-strain elastic-plastic formulation. The equilibrium equation in the solid is:
\[
\nabla \cdot \boldsymbol{\sigma} + \rho \mathbf{b} = 0
\]
where \( \boldsymbol{\sigma} \) is the Cauchy stress tensor, \( \rho \) is the density, and \( \mathbf{b} \) is the body force. The constitutive relation for the elastic-plastic material is written in rate form:
\[
\dot{\boldsymbol{\sigma}} = \mathbf{C} : \left( \dot{\boldsymbol{\varepsilon}} – \dot{\boldsymbol{\varepsilon}}^p – \dot{\boldsymbol{\varepsilon}}^{th} \right)
\]
where \( \mathbf{C} \) is the fourth-order elasticity tensor, \( \dot{\boldsymbol{\varepsilon}} \) is the total strain rate, \( \dot{\boldsymbol{\varepsilon}}^p \) is the plastic strain rate, and \( \dot{\boldsymbol{\varepsilon}}^{th} = \alpha \dot{T} \mathbf{I} \) is the thermal strain rate. The plastic strain rate is governed by the associated flow rule:
\[
\dot{\boldsymbol{\varepsilon}}^p = \dot{\lambda} \frac{\partial f}{\partial \boldsymbol{\sigma}}
\]
with the von Mises yield function
\[
f = \sqrt{\frac{3}{2} \mathbf{s} : \mathbf{s}} – \sigma_y(T)
\]
where \( \mathbf{s} \) is the deviatoric stress and \( \sigma_y(T) \) is the temperature-dependent yield stress. The accumulated plastic strain is used in the hot tearing model. The maximum HTI is calculated along the solidification path once the temperature falls below the coherency point until the solid fraction reaches 99%. The indicator is given by:
\[
\mathrm{HTI} = \int_{t_c}^{t_s} \dot{\varepsilon}_p \, dt
\]
where \( t_c \) is the time when grains begin to contact, \( t_s \) is the time when the solid fraction reaches 99%, and \( \dot{\varepsilon}_p \) is the plastic strain rate. The higher the HTI value, the greater the hot tearing tendency. Using this model, I was able to compare the three chill arrangements and to confirm that the no-chill design is viable. This saved at least two mold tryouts that would have been needed to test the chill alternatives experimentally.
In conclusion, the final casting process for the excavator large-arm root BOSS part is a gravity-cast steel casting made with a silica-sand mold and a limestone-sand core, without external chills. The pouring temperature is 1565°C and the filling time is 25 seconds. This process is now in production and has delivered castings with no cracks, no sand adhesion in the bore, and acceptable mechanical properties. The combination of numerical simulation and targeted physical verification proved to be an efficient and reliable methodology for developing a robust casting process for a challenging thin-walled steel casting with a deep cavity.