In 2005, my foundry received a trial order from a Japanese company to produce a nodular iron rotator seat for high‑speed elevator thin‑type motors. The casting had a contour size of φ560×190 mm, a main wall thickness of 45 mm, a weight of 97 kg, and required material grade QT450‑10. As a high‑speed rotating part, it had to undergo dynamic and static balancing after full machining, and no shrinkage cavity, shrinkage porosity, or crack was allowed. Critical surfaces were forbidden to contain any gas hole or sand inclusion; other surfaces could have at most two defects of size ≤1.5×1×1 mm. This was an extremely stringent specification. Over the course of several months, I led the effort to eliminate sand casting defects, particularly shrinkage porosity. Through systematic trials, I discovered that the key to success lay in proper riser design, ingate geometry, and the correct use of the graphite expansion of nodular iron. This paper presents the entire journey, including detailed tables, formulas, and explanations.
Production Conditions and Initial Challenge
My foundry’s iron casting workshop produced multiple types of engine cylinder blocks using a green sand air‑impact moulding line with flask dimensions of 1200×800×350 mm. Melting was done in a 6‑ton medium‑frequency induction furnace. This was the first time we attempted to produce nodular iron castings on a green sand line, and the biggest concern was how to avoid shrinkage‑related sand casting defects. Nodular iron exhibits significant graphitic expansion during solidification, which can provide self‑feeding if the mould is rigid. However, green sand moulds, even with high‑pressure compaction, can expand under the metallostatic pressure and graphitic force, leading to cavity enlargement and subsequent shrinkage. The mould surface hardness ranged from 75 to 95 on the wet‑mould hardness tester, and moisture control was critical to prevent gas defects.
First Attempt: Bottom Gating with Chills (Scheme 1)
Following the “balanced solidification” principle, I initially designed a bottom‑gating system with six fan‑shaped ingates distributed around the casting. No riser was used, relying on the self‑feeding of nodular iron with the addition of two rows of external chills at the φ480 and φ290 root corners. The pouring system was bottom‑gated to avoid sand erosion and inclusions. The moulding sand was compacted by air‑impact, which gave a reasonably stiff mould. Nevertheless, after the first trial, the castings showed no surface defects but, upon machining, revealed visible shrinkage porosity in the root fillet areas at φ480 and φ290, extending over nearly one‑third of the circumference. The chills were then repositioned and increased, and the porosity disappeared, but the use of chills brought new problems. The chills required extensive preparation: casting, cleaning, grinding, rust removal, welding, coating, and drying. On a high‑speed mechanised line, placing chills for every mould caused line stoppages and increased scrap due to misplaced chills. Furthermore, chill placement increased the occurrence of subsurface blowholes on the top face of the casting. The cleaning and inspection effort also rose dramatically. Clearly, a chill‑based solution was not sustainable for mass production.
Table 1 summarises the key parameters of Scheme 1 and its outcomes.
| Parameter | Details |
|---|---|
| Gating type | Bottom gating, 6 fan ingates (30×8 mm each) |
| Riser type | None (no riser) |
| Chills | Two rings of external chills (φ480 and φ290 root) |
| Pouring temperature | 1380–1420 °C |
| Mould hardness (green sand) | 75–95 (Wet mould hardness tester) |
| Result | Shrinkage porosity at φ480/φ290 root after machining; nearly 1/3 arc |
| Issue with chills | Production line stoppage; chill preparation costs; increased blowholes |
The fundamental reason for the shrinkage was the expansion of the green sand mould cavity during solidification, which consumed the volumetric expansion of graphite and left no compensation for the liquid shrinkage. The chills did locally accelerate cooling but could not counteract the overall cavity growth. I realised that a more robust feeding system was needed to overcome the mould compliance.
Theoretical Consideration of Mould Expansion and Feeding
To quantify the problem, I revisited the basic feeding theory for nodular iron. The total volume change during solidification can be expressed as:
$$ \Delta V_{\text{total}} = V_{\text{liquid}} \cdot \varepsilon_{\text{l}} + V_{\text{graphite}} \cdot \varepsilon_{\text{gr}} $$
where \(\varepsilon_{\text{l}}\) is the liquid contraction coefficient (~1.5–2.0% for nodular iron) and \(\varepsilon_{\text{gr}}\) is the expansion coefficient from graphite precipitation (~2.5–3.5%). In a rigid mould, the graphite expansion compensates the liquid contraction, allowing “self‑feeding”. However, in a green sand mould, the mould walls can yield under the internal pressure. The cavity expansion volume \(\Delta V_{\text{cavity}}\) must be added to the feeding demand. The net feeding requirement becomes:
$$ F_{\text{req}} = V_{\text{liquid}} \cdot \varepsilon_{\text{l}} + \Delta V_{\text{cavity}} – V_{\text{graphite}} \cdot \varepsilon_{\text{gr}} $$
If the mould expansion is sufficiently large, the net requirement becomes positive, and an external riser is needed. In our first scheme, the absence of riser meant that no extra liquid was available, so shrinkage porosity appeared at the last‑solidifying regions (the heavy root areas). The chills only delayed solidification locally but did not provide liquid metal.
I also considered the modulus of the casting sections. The modulus \(M\) is defined as volume divided by cooling surface area:
$$ M_c = \frac{V_c}{A_c} $$
The main wall (45 mm thick) has a modulus of roughly 22.5 mm. The heavy root areas have a larger modulus due to the corner effect. For nodular iron in green sand moulds, the necessary riser modulus \(M_r\) should be at least 1.1–1.2 times the casting modulus to ensure directional solidification. This principle guided my later designs.
Second Attempt: Top Gating with Hot and Cold Kiss Risers (Scheme 2)
Based on the lessons from Scheme 1, I changed the gating to top gating, flipping the cope and drag. This allowed larger risers to be placed on the top surface. I used four kiss risers (also known as pressure‑edge risers) with a kiss width of 6 mm. Two of these risers were “hot” (the metal entered the cavity through them) and two were “cold” (overflow risers, with the metal entering through a separate gating system that bypassed the riser). A ceramic foam filter was placed below the sprue to reduce turbulence. The ingates under the two hot risers were thin and wide (40×8 mm) to promote early closure. The aim was to use the hot risers as feeding sources and the cold risers as overflow to remove cold and contaminated metal. The casting modulus was still not large enough to guarantee self‑feeding, so external chills were removed.
The trial castings showed no surface defects, and machining revealed that the φ290 root area was sound, but the φ480 root area directly opposite the hot risers and ingates contained 2–3 concentrated shrinkage cavities and severe porosity. This was puzzling because the hot risers should have provided feeding. Upon analysis, I identified two problems:
- The kiss width of only 6 mm solidified very quickly after pouring. The hot riser’s feeding channel closed before the casting had fully solidified. Meanwhile, the mould cavity continued to expand due to the metallostatic pressure and graphitic expansion, creating a demand for liquid that could not be met.
- The cold risers (overflow) had colder metal than the casting because the first metal flowed through the gating system and then into the riser before entering the cavity. That cold metal actually solidified earlier than the casting, so instead of feeding, it acted as a heat sink that could even suck liquid from the casting, exacerbating shrinkage.
Table 2 outlines the details of Scheme 2.
| Parameter | Details |
|---|---|
| Gating type | Top gating; metal enters through two hot kiss risers |
| Number of risers | 4 kiss risers (2 hot, 2 cold) |
| Kiss width | 6 mm (all) |
| Ingate under hot riser | 40×8 mm (thin and wide) |
| Cold riser design | Overflow only; metal enters via separate gating |
| Filter | Ceramic foam filter below sprue |
| Result | φ290 root sound; φ480 root opposite hot risers had 2–3 shrinkage cavities |
| Reason | Kiss width too small → early closure; cold risers solidified earlier → no feeding |
I measured the solidification time of the kiss using the modulus calculation. For a rectangular plate‑shaped connection between riser and casting, the modulus \(M_{\text{kiss}}\) is approximately the thickness (kiss width) if the area is large compared to the lateral faces. With a 6 mm thickness, \(M_{\text{kiss}} \approx 3\) mm. The casting’s main wall modulus was 22.5 mm, and the heavy root modulus was even larger. Using Chvorinov’s rule:
$$ t = k \cdot M^2 $$
where k is a constant (for green sand moulds, typical k ≈ 2–3 min/cm² for nodular iron). The solidification time of the kiss would be at least (3/22.5)² ≈ 1/56 of the casting solidification time, i.e., the kiss solidified almost immediately after pouring. Even if the riser contained hot liquid, the channel was blocked. The solution was to increase the kiss width dramatically to keep it open until the casting was almost completely solid.
Third Attempt: Enlarged Hot Kiss Risers, No Cold Risers (Scheme 3)
I made three critical changes for the final design:
- Eliminated the cold overflow risers completely. They provided no feeding benefit and only generated problems. Two small 8‑mm vent holes were used instead to expel air.
- Increased the kiss width of the two hot risers from 6 mm to 20 mm. This ensured that the feeding channel remained open until the casting had substantially solidified. The increased width also increased the modulus of the riser neck itself, making it solidify later than the casting’s heavy sections.
- Made the ingates entering the riser very thin and wide (40×15 mm). These thin ingates would freeze quickly after pouring, thereby isolating the riser from the casting. In effect, the riser became a part of the casting that solidified last, drawing liquid from itself, not from the casting. The combination of a thick kiss and thin ingates created a directional solidification path from the casting toward the riser.
The gating system remained top‑gating through the two hot risers, with ceramic filter. The mould was still green sand air‑impact. I ran three successive heats and randomly selected four castings for destructive sectioning. No shrinkage cavity or porosity was found. Eighteen additional castings were machined and passed the Japanese customer’s inspection, including dynamic and static balancing.
Table 3 summarises the critical parameters and results of the winning scheme.
| Parameter | Details |
|---|---|
| Gating type | Top gating through two hot kiss risers |
| Number of risers | 2 (both hot, no cold risers) |
| Kiss width | 20 mm (increased from 6 mm) |
| Ingate under riser | 40×15 mm (thin and wide; freeze quickly) |
| Vents | 6× φ8 mm (only for air escape) |
| Filter | Ceramic foam filter below sprue |
| Result | No shrinkage defects in any section; all castings passed Japanese inspection |
The success can be understood through the modulus balance. The casting’s heavy root had a modulus of about 32 mm (estimated from geometry). The riser neck (kiss) with 20 mm thickness had a modulus of 10 mm, which is still lower than the casting. However, because the riser body itself had a much larger modulus (~40 mm), the overall feeding system was designed so that the casting solidified directionally toward the riser. The thin ingates (modulus ~7.5 mm) solidified first, sealing the riser from the gating system. The kiss remained open due to its large area, allowing the riser’s hot metal to compensate for shrinkage until the casting’s heavy sections were fully solid. The graphite expansion then filled the remaining spaces within the casting, and since the mould cavity had already stabilised (the expansion was accommodated by the liquid feeding), no final shrinkage occurred.
General Formulas for Feeding Design in Green Sand Nodular Iron Castings
Based on the three trials, I derived a set of practical guidelines. The required volume of the hot riser can be approximated by:
$$ V_r = \varepsilon_{\text{net}} \cdot V_c $$
where \(\varepsilon_{\text{net}} = \varepsilon_{\text{l}} + \varepsilon_{\text{mould}} – \varepsilon_{\text{gr}}\). In our case, with a green sand mould, I estimated \(\varepsilon_{\text{mould}}\) from cavity expansion measurements to be about 0.5–1.0%. For our casting, \(V_c = 97\) kg / 7.1 g/cm³ ≈ 13,660 cm³. Using \(\varepsilon_{\text{l}} = 1.8\%\), \(\varepsilon_{\text{gr}} = 3.0\%\), and \(\varepsilon_{\text{mould}} = 0.8\%\), we get \(\varepsilon_{\text{net}} = 1.8+0.8-3.0 = -0.4\%\). The negative value suggests that theoretically no external feeding is needed if the mould were perfectly rigid. However, because the mould is not rigid, the cavity expansion exceeds the graphite expansion in the early stages. In practice, a small net feeding requirement exists, which is why the hot riser worked—it provided a reservoir of liquid to counteract the initial mould expansion.
I also developed a simple criterion for the kiss width w based on the modulus of the casting section it must feed:
$$ w \ge 0.6 \cdot M_{\text{section}} $$
In our case, the heavy root had \(M_{\text{section}} \approx 32\) mm, so the required kiss width was at least 19.2 mm. The final width of 20 mm satisfied this. The earlier width of 6 mm corresponded to a section modulus of only 10 mm, which was far too small.
The cross‑sectional area of the ingate connecting the gating system to the riser should be designed to freeze quickly. The modulus of the ingate \(M_{\text{ig}}\) should be less than half of the kiss modulus. For our thin ingate (40×15 mm), the modulus was 7.5 mm, which was indeed less than half of the kiss modulus of 10 mm. This ensured that the ingate solidified before the kiss, isolating the riser body as a “last freezing” reservoir.
Illustration of Typical Sand Casting Defects
Throughout this project, I encountered various sand casting defects including shrinkage porosity, blowholes, and sand inclusions. The image below shows common sand casting defects that I studied to improve my understanding. It highlights the importance of proper riser and gating design, especially for ductile iron in green sand moulds.
Comparison of All Three Schemes
To give a complete overview, Table 4 compares the key features and outcomes of all three casting methods.
| Parameter | Scheme 1 (Bottom gating + chills) | Scheme 2 (Top gating + hot/cold kiss risers) | Scheme 3 (Top gating + enlarged hot kiss risers) |
|---|---|---|---|
| Gating location | Bottom | Top (through hot risers) | Top (through hot risers) |
| Number of risers | 0 | 4 (2 hot, 2 cold) | 2 (hot only) |
| Riser type | None | Kiss riser, width 6 mm | Kiss riser, width 20 mm |
| Ingate under riser | N/A | 40×8 mm (thin) | 40×15 mm (thin) |
| Use of chills | External chills (two rings) | None | None |
| Filter | No | Ceramic foam below sprue | Ceramic foam below sprue |
| Venting | 10 vent rods | 8 φ8 vents | 6 φ8 vents |
| Shrinkage defects | Yes (at φ480/φ290 root) | Yes (at φ480 root opposite risers) | None |
| Other defects | Subsurface blowholes near chills | Few sand inclusions | No significant defects |
| Production issues | Line stoppages; chill preparation costs | Cold risers ineffective | None |
| Customer acceptance | Rejected (after remachining) | Rejected | Accepted (all 18 pieces) |
Conclusion and Lessons Learned
Through this intensive trial‑and‑error process, I learned several critical lessons for preventing sand casting defects in nodular iron castings produced with green sand moulds:
- Sequence solidification is superior to balanced solidification when the mould has limited rigidity. Even though the “balanced solidification” theory promotes self‑feeding, the mould expansion in green sand nullifies the graphite expansion benefit. Directional solidification with hot risers placed at the top ensures a positive feeding pressure.
- The kiss (pressure‑edge) width must be sufficiently large to keep the feeding channel open until the casting has completed its initial solidification. A width of at least 0.6 times the modulus of the fed section proved reliable.
- Cold risers (overflow risers) are detrimental. Metal entering the cavity via a separate path into a cold riser makes the riser solidify earlier than the casting, leading to reverse feeding and increased shrinkage. They should be avoided or replaced by simple vents.
- Ingates connecting the gating system to the riser should be thin so that they freeze quickly after pouring, isolating the riser body. This makes the riser the last part to solidify, allowing it to feed the casting without itself being fed.
- External chills can be useful but are impractical for high‑volume production due to preparation time and process instability. They also increase the risk of gas defects.
- Ceramic foam filters help reduce slag and sand inclusions, which are common sand casting defects in top‑gated systems.
These findings have been applied to other nodular iron castings in my foundry, consistently reducing scrap related to shrinkage. The successful production of the rotator seat demonstrated that green sand moulding, if properly designed, can produce high‑integrity ductile iron castings that meet the most demanding requirements. I hope that the detailed tables and formulas provided here will assist other foundry engineers facing similar sand casting defect challenges.
Note: All formulas are based on practical engineering approximations and should be validated for specific casting geometries and foundry conditions.