In the development of a newly introduced large marine diesel engine, our foundry encountered serious quality problems during the preproduction of a heavy-duty cylinder block made of spheroidal graphite cast iron. The material specification for this component follows European standards and demands a very high level of internal soundness. During the initial trial production, several recurring defects were observed, including metallic splash inclusions (commonly known as “iron peas”), gas-slag laps, and blowholes. These defects not only reduced the yield but also threatened the integrity of the casting. As a result, a comprehensive investigation was carried out to identify the root causes, and later effective countermeasures were implemented. This paper describes the entire problem-solving process from the perspective of a process engineer directly involved in the project. The focus is on the analysis of the filling behavior, thermal field evolution, and the systematic improvement of the gating system and cooling strategies for ductile iron castings.
The cylinder block is one of the largest and most critical castings in the engine. It weighs approximately 27 tons with a green casting weight exceeding 25 tons. The casting is produced using alkaline phenolic resin self-hardening sand for both molds and cores. The manufacturing route was established with the bore surface facing downward and the base deck facing upward in a vertical pouring configuration. This orientation was selected to ensure that the critical regions, such as the cylinder bores, main bearing saddles, and camshaft bores, are located in the lower portion of the mold where higher metallostatic pressure exists and where feeding and solidification are more favorable. However, despite the rational placement of the casting relative to the feeding system, a number of unacceptable defects appeared after machining.
Technical Requirements for the Cylinder Block
The material designation is EN-GJS-100-15U, which is closely related to the Chinese grade QT400-15. For ductile iron castings of this class, the following mechanical properties are mandatory at room temperature. The values are listed in Table 1. Additionally, the matrix microstructure must be predominantly ferritic with a nodularity of at least 90%. The graphite nodule count should be within the range typically expected for heavy-section ductile iron. Every cylinder block must be subjected to ultrasonic testing at critical locations, including the cylinder bores, bearing seats, side pull bolt holes, camshaft bores, and inspection window openings. Magnetic particle inspection is required for the areas around the observation windows. Any linear indication is considered unacceptable. This requirement imposes a strict limit on the presence of oxide films, gas pores, and inclusions.
| Property | Value | Unit |
|---|---|---|
| Tensile strength (ultimate) | ≥ 400 | MPa |
| Yield strength (0.2% offset) | ≥ 250 | MPa |
| Elongation at break | ≥ 15 | % |
| Brinell hardness (body) | 135 – 185 | HBW |
| Nodularity | ≥ 90 | % |
| Ferrite content | ≥ 80 | % |
The production of these ductile iron castings is extremely challenging because any defect that affects the mechanical properties may lead to scrapping of the entire block, resulting in significant financial loss. Therefore, the process window must be very narrow, and the filling system must be designed to avoid any turbulent flow, gas entrapment, or premature solidification. The initial process parameters are summarized in Table 2.
| Parameter | Value |
|---|---|
| Molding process | Alkaline phenolic resin self-hardening sand |
| Core making | Manual core making, same resin system |
| Pouring temperature | 1330 – 1340 °C |
| Total pouring weight | 27 t |
| Pouring orientation | Bore face down, base deck up |
| Feeding system | Top risers (insulated) |
| Chills | Placed at heavy sections and critical areas |
| Pouring time | Approximately 150 – 180 s |
During the first validation batch, the castings were machined to a rough finish and then carefully inspected. The dominant defect types are illustrated in Figure 1. These defects were observed particularly on the side opposite to the ingates. This location dependence strongly suggested that the filling pattern was nonuniform and that the liquid metal front was becoming unstable as it traveled across the mold.
The defects shown in the image are representative of three categories: (a) iron peas, (b) gas-slag laps, and (c) blowholes. Iron peas are small metallic globules embedded in the casting surface or slightly below the surface. They are not fused with the base metal but mechanically attached. Gas-slag laps appear as elongated discontinuities covered with a thin oxide film. Blowholes are internal cavities with smooth walls, often filled with gas that has a strong reflecting surface for ultrasonic waves. The occurrence of these defects is closely related to the hydrodynamic and thermal behavior of the liquid metal during filling.
Root Cause Analysis Using Numerical Simulation
To understand the origin of these defects, we used the MAGMA software to simulate the filling and solidification of the original process. The simulation provided detailed information about the velocity field, temperature distribution, and air entrainment index throughout the pouring operation. The results clearly showed several problematic features. First, during the initial stage of pouring, the metal exiting the ingates had a very high velocity and a strong tendency to splash. The splashed droplets became oxidized and rapidly formed solid films. Because the pouring temperature was relatively low (1330–1340 °C), these oxidized particles could not be re-melted by the subsequent liquid metal and were entrapped in the solidifying shell, forming iron peas.
The degree of turbulent flow can be characterized by the Reynolds number of the liquid metal in the gating system:
$$ Re = \frac{\rho v d}{\mu} $$
where \( \rho \) is the density of the molten iron (approximately 7000 kg/m³), \( v \) is the local velocity, \( d \) is the characteristic channel diameter, and \( \mu \) is the dynamic viscosity (approximately 0.006 Pa·s for ductile iron at 1350 °C). For a typical ingate velocity of 1.5 m/s and a hydraulic diameter of 0.05 m, the Reynolds number is:
$$ Re = \frac{7000 \times 1.5 \times 0.05}{0.006} = 87{,}500 $$
This value is far above the critical Reynolds number for laminar flow (about 2300). Hence the flow is highly turbulent. A turbulent flow promotes surface wave formation, air entrainment, and droplet ejection. The air entrainment rate is proportional to the turbulent kinetic energy, which can be expressed as:
$$ G = C_D A \sqrt{2 g H_p} $$
where \( G \) is the flow rate, \( C_D \) is the discharge coefficient, \( A \) is the total cross-sectional area of the ingates, \( g \) is the gravitational acceleration, and \( H_p \) is the average metallostatic pressure head. In the initial design, the system was closed, meaning that the total cross-sectional area of the sprue was smaller than that of the ingates. This arrangement led to a pressure buildup and excessive velocity at the gates. The simulation indicated that the flow velocity in the ingate region exceeded 2.0 m/s, which is considered too high for a large ductile iron casting.
Figure 2 illustrates the air entrainment tendency predicted by the simulation at different filling stages. The degree of entrainment is expressed as a dimensionless index ranging from 0 to 1, where values above 0.3 indicate severe air entrapment. The simulation showed that from the beginning to the end of filling, air entrainment was present throughout the mold cavity. In particular, the regions near the upper deck and the opposite side of the ingates exhibited high entrainment values. The reason is that the liquid metal traveled a long distance across the mold and lost its kinetic energy, causing the front to break up and fold over. The folded metal captured air bubbles and oxide films, leading to gas-slag laps.
The thermal field evolution is also a key factor. The temperature of the liquid metal at different positions and times was extracted from the simulation. The results showed a nonuniform temperature distribution across the casting. Near the ingates, the temperature remained relatively high, but on the opposite side, especially in the lower observation window area, the temperature dropped by nearly 30 °C by the end of filling. This temperature drop increased the viscosity of the liquid metal and reduced its ability to float out non-metallic inclusions. Additionally, the upper regions of the casting, near the base deck, showed a rapid temperature decrease because the surface-to-volume ratio was high and the chills acted as efficient heat sinks. This sharp thermal gradient caused the metal to solidify prematurely, trapping gases that could not escape.
Another important issue was the leakage of molten metal from the pouring box before the stopper was lifted. In our pouring arrangement, a heated tundish with a stopper rod was used. Due to the mismatch between the stopper head and the seat, a small gap existed, allowing a thin stream of liquid metal to leak into the mold during the pre-filling stage. This leaked metal entered the cavity in a fragmented form and solidified as iron peas at the lower part of the casting, particularly around the lower observation window. The amount of leaked metal was small compared to the total pour, but its effect was significant because it oxidized and did not fuse with the main metal.
Let us also consider the role of gases dissolved in the liquid metal. The solubility of hydrogen in liquid iron follows Sieverts’ law:
$$ \sqrt{S_H} = K_H \cdot p_{H_2}^{1/2} $$
where \( S_H \) is the hydrogen content, \( K_H \) is the equilibrium constant, and \( p_{H_2} \) is the partial pressure of hydrogen in the atmosphere above the melt. During the pouring of ductile iron castings, the melt absorbs hydrogen from the surrounding atmosphere, especially if the atmosphere is humid. The phenolic resin sand generates gases during casting, including hydrogen, nitrogen, and carbon monoxide. If these gases cannot escape rapidly enough, they will be entrapped in the solidifying metal and form blowholes. In our case, the presence of many chills on the mold surface acted as barriers to gas flow. The chills prevented the upward movement of gases, and the gas pockets were trapped at the interface between the chill and the liquid metal. If the chill surface was rusty or coated with moisture, the gas generation was even more pronounced.
To quantify the thermal effect of the chills, we used the modulus concept. The local solidification time is proportional to the square of the modulus:
$$ t_s = \frac{M^2}{\alpha} $$
where \( M = V/A \) is the local modulus (volume divided by surface area) and \( \alpha \) is a constant depending on the material and heat transfer coefficient. Placing a chill adjacent to a heavy section reduces the local modulus artificially and increases the cooling rate. However, if the chill is placed above a region where gas accumulates, the chill creates a cold spot that prevents gases from escaping. In the original design, two chills (labeled 27# and 18#) were placed in the core above the upper deck. These chills blocked the natural upward escape route for gases, causing the formation of gas-slag laps at that location.
In summary, the root causes of the defects in these ductile iron castings were identified as follows:
- Unstable filling with high velocity and splashing, leading to iron peas.
- Air entrainment during the entire filling process, especially in the later stages, leading to gas-slag laps.
- Nonuniform temperature field with a sharp temperature drop at the upper and far-side regions, reducing the ability of the metal to clean itself.
- Leakage of melt from the stopper before the main pour, causing early iron peas.
- Chills placed in the path of escaping gases, blocking venting and promoting blowholes.
| Defect | Location | Root Cause |
|---|---|---|
| Iron peas | Lower inspection window, oppposite ingates | Initial splashing, stopper leakage, low pouring temperature |
| Gas-slag laps | Upper base deck area | Air entrainment during filling, gas blockage by chills, rapid temperature drop |
| Blowholes | Near chills, heavy sections | Gas evolution from core, inability to escape, oxidized chill surfaces |
Process Improvement Measures
Based on the analysis, we implemented a series of corrective actions aimed at improving the filling stability, balancing the temperature field, and facilitating gas removal. The following measures were applied to all subsequent production runs of ductile iron castings.
1. Revamping the Gating System
The original gating system was single-sided with a closed design. The new system was made double-sided, meaning that liquid metal was introduced from both sides of the mold. This arrangement reduces the travel distance of the metal front and balances the temperature between the two halves. The cross-sectional areas of the runner and the ingates were increased. The system was converted from a pressurized (closed) system to a non-pressurized (open) system. The new area ratios are given in Table 4.
| Element | Original ratio | Modified ratio |
|---|---|---|
| Sprue / runner | 1 : 2.0 | 1 : 2.5 |
| Runner / ingates | 1 : 1.5 | 1 : 2.0 |
| Total ingate area | 0.012 m² | 0.020 m² |
| Number of ingates | 4 | 8 (4 on each side) |
The open system ensures that the runner remains full but the metal enters the mold at a velocity below 0.5 m/s. The lower velocity significantly reduces the Reynolds number, and the flow regime approaches a more laminar state. The reduced turbulence also minimizes air entrainment and the formation of oxidized droplets.
2. Increasing the Pouring Temperature
The pouring temperature was raised from 1330–1340 °C to 1360–1370 °C. A higher pouring temperature provides several benefits. First, it lowers the viscosity of the molten iron, which improves the fluidity and allows gases and inclusions to float out more easily. Second, any metal droplets that are formed during filling have a higher chance of being re-melted because the surrounding melt is hotter. This directly suppresses the formation of iron peas. Third, a higher initial temperature compensates for the temperature loss during the long fill time of approximately 150 seconds, thereby reducing the thermal gradient across the casting. However, care must be taken not to overheat the melt, as this can lead to excessive shrinkage porosity and increased gas pickup. The selected range was found to be optimal for this casting.
3. Improving the Stopper Mechanism
The original stopper was lifted using a chain, which allowed slight lateral movement and caused misalignment between the stopper head and the seat. We replaced this with a lever-type (press-down) mechanism. In the new design, the stopper head is pressed down by a lever arm, and the weight of the stopper itself ensures a tight seal. This prevents any leakage of metal into the mold before the intentional start of pouring. The improved mechanism is shown schematically in Figure 3 (not reproduced here for brevity). The stopper is made of low-carbon steel with a refractory coating, and the seat is machined to a tolerance of ±0.05 mm. This eliminated the source of early metal leakage.
4. Modified Pulling Sequence of the Stoppers
We also adopted a controlled “two first, one later” pulling sequence. Instead of lifting all three stoppers simultaneously, we removed two stoppers at the beginning, and the third stopper was lifted after 100 seconds of pouring. This controlled the initial filling rate and reduced the peak flow velocity. The pouring time was extended slightly, but the overall filling remained stable. The effect was a reduction in the turbulence kinetic energy during the initial stage, which is the most critical period for splashing. The flow rate can be related to the pouring time by:
$$ t_p = \frac{W}{\rho \cdot Q} $$
where \( W \) is the total casting weight (27 t), \( \rho \) is the iron density, and \( Q \) is the volumetric flow rate. With a reduced initial flow, the front advanced more uniformly, and the probability of air entrapment decreased considerably.
5. Optimization of Chill Placement
The original design used internal chills in the core above the top deck, which blocked the escape of gases. We removed the 27# and 18# chills from the core and placed two larger contoured chills (30# and 31#) in the bottom mold at corresponding locations. The new chill arrangement provides the same chilling effect on the heavy sections but does not obstruct the upward venting path. The gas generated from the core and mold can now flow freely through the vents and risers. A comparison of the chill layout is given in Table 5.
| Chill ID | Original position | New position | Function |
|---|---|---|---|
| 27# | Upper core (above deck) | Removed | — |
| 18# | Upper core (near camshaft) | Removed | — |
| 30# | — | Bottom mold (opposite ingates) | Cool critical bearing area |
| 31# | — | Bottom mold (near bore) | Cool cylinder bore region |
In addition to repositioning the chills, we introduced a strict maintenance program for all chills. They were blasted to remove any rust, preheated to 80 °C before assembly, and coated with a refractory wash to prevent sand adhesion and gas generation. The elimination of moisture and oxide layers on the chill surfaces significantly reduced the formation of blowholes at the chill-metal interface.
Theoretical Considerations for Improved Filling
To better understand why the new process works, let us analyze the filling behavior using a simplified model. The liquid metal flow in the gating system can be described by Bernoulli’s equation:
$$ \frac{v_1^2}{2g} + \frac{p_1}{\rho g} + z_1 = \frac{v_2^2}{2g} + \frac{p_2}{\rho g} + z_2 + h_f $$
where \( h_f \) is the head loss due to friction and turbulence. In the open gating system, the larger cross-sectional area reduces the velocity in the runner, and the pressure at the gates is lower, thus reducing jetting. The head loss in a gating channel can be approximated by Darcy’s equation:
$$ h_f = f \frac{L}{D} \frac{v^2}{2g} $$
where \( f \) is the friction factor, \( L \) is the channel length, and \( D \) is the hydraulic diameter. By increasing \( D \) and reducing \( L \) (through double-sided pouring), we reduced \( h_f \) and also the associated turbulence generation. The Reynolds number in the modified ingates was calculated to be approximately 8,000–10,000, which is still technically turbulent but much less violent than 87,500. The air entrainment index dropped from 0.7 to less than 0.2 in the critical regions, as confirmed by a subsequent simulation.
The temperature equilibration can be assessed by considering the thermal diffusivity of the mold material. The heat transfer from the casting to the mold is:
$$ \frac{\partial T}{\partial t} = \alpha_m \nabla^2 T $$
where \( \alpha_m \) is the thermal diffusivity of the mold sand. For alkaline phenolic resin sand, \( \alpha_m \) is about \( 0.3 \times 10^{-6} \,\mathrm{m^2/s} \). The higher pouring temperature effectively increases the superheat, so that the temperature at the end of filling is still above the liquidus temperature (about 1150 °C for this grade). The temperature difference between the near and far sides was reduced from 32 °C to approximately 12 °C. This was verified by thermocouple measurements in a validation run.
Another important aspect is the removal of gases from the mold cavity. The venting system was also improved. We added additional vent holes in the upper core and increased the diameter of the existing vents from 10 mm to 16 mm. The resistance to gas flow through a vent can be expressed by Poiseuille’s law for laminar flow:
$$ Q_g = \frac{\pi \Delta P D_v^4}{128 \mu_g L_v} $$
where \( Q_g \) is the gas flow rate, \( \Delta P \) is the pressure difference between the cavity and the atmosphere, \( D_v \) is the vent diameter, \( \mu_g \) is the gas viscosity, and \( L_v \) is the vent length. Increasing the vent diameter by a factor of 1.6 increases the gas flow rate by a factor of \( 1.6^4 = 6.55 \), which greatly improves the venting capacity.
We also modified the pouring basin design to minimize the aspiration of air into the sprue. A tapered sprue was used to maintain a full cross-section and to avoid negative pressure. The critical height of the sprue was calculated to prevent aspiration:
$$ h_c = \frac{1}{2} \left( \frac{A_s}{A_r} \right)^2 \frac{v_s^2}{g} $$
where \( A_s \) and \( A_r \) are the cross-sectional areas of the sprue and the runner, and \( v_s \) is the sprue inlet velocity. The design ensured that the pressure in the sprue never fell below atmospheric pressure, thus preventing air aspiration.
Results and Discussion
After implementing the above improvements, we produced another 20 cylinder blocks under the same production conditions as in the original trials. The visual inspection immediately showed a dramatic improvement. No iron peas or gas-slag laps were observed on any of the castings. The surfaces were clean, and the machined faces showed no linear indications. Ultrasonic testing of all critical areas passed without any report of unacceptable defects. Magnetic particle inspection of the observation windows also showed no cracks or laps.
Table 6 summarizes the defect incidence before and after the improvement.
| Defect Type | Original process (batch of 5) | Improved process (batch of 20) |
|---|---|---|
| Iron peas | Present in 4 castings | 0 |
| Gas-slag laps | Present in 3 castings | 0 |
| Blowholes | Present in 2 castings | 0 |
| Overall rejection rate | 60% | 0% |
The improvement in quality is attributed to the combined effect of reduced turbulence, better thermal balance, and superior venting. The Reynolds number in the gating system was lowered by a factor of 8, and the air entrainment index was reduced by more than 70%. The pouring temperature increase ensured that any droplets formed were re-melted. The new stopper mechanism eliminated the early leakage. The repositioning of chills opened up the gas escape paths. All these measures together created a stable filling condition for ductile iron castings.
We also monitored the microstructure of the test bars attached to the castings. The ferrite content remained within specification, and the graphite nodules were uniformly distributed with a nodularity of 92–95%. The mechanical properties of the attached test specimens are listed in Table 7.
| Sample No. | Tensile strength (MPa) | Yield strength (MPa) | Elongation (%) | Hardness (HBW) |
|---|---|---|---|---|
| 1 | 418 | 275 | 18.5 | 155 |
| 2 | 425 | 282 | 19.2 | 152 |
| 3 | 409 | 268 | 17.8 | 160 |
| 4 | 421 | 278 | 18.1 | 158 |
All values comfortably meet the minimum requirements. The hardness is within the specified range, and the variation is small, indicating a consistent cooling rate. The absence of internal defects is further confirmed by the ultrasonic testing results. In comparison with the original process, the improved castings showed a much cleaner microstructure, with no evidence of oxide films or gas pores at high magnification.
The success of this campaign demonstrates that a systematic approach, combining numerical simulation with careful process control, is essential for producing high-integrity ductile iron castings. The principles applied here are not limited to cylinder blocks but can be extended to other large and complex ductile iron castings. The key is to understand the filling dynamics and thermal history of the liquid metal. Every casting has its unique geometry, and the gating system must be designed accordingly. In our foundry, we now follow a standard procedure for the development of ductile iron castings, which includes the following steps:
- Preliminary gating system design based on traditional empirical rules.
- Numerical simulation of filling and solidification to identify potential turbulence, air entrapment, and hot spots.
- Optimization of the gating ratio, pouring temperature, and chill placement.
- Validation through instrumented trials with thermocouples and flow meters.
- Final inspection using ultrasonic and magnetic particle methods.
The present case taught us that even small details, such as the stopper sealing or the surface condition of a chill, can have a decisive impact on the quality of ductile iron castings. We now carry out routine maintenance of all stoppers and chills. The pouring boxes are preheated to avoid any condensation. The core assembly is vented with additional channels, and the mold coating is applied with a controlled thickness to minimize gas generation.
Furthermore, we established a stringent control of the pouring process. The pouring temperature is measured automatically before each pour, and the pouring time is recorded. The stopper sequence is automated to ensure repeatability. This level of process automation has greatly reduced the variability between castings. The result is a robust production process that consistently delivers defect-free ductile iron castings.
In conclusion, the defects encountered during the preproduction of the large cylinder block were successfully eliminated by modifying the gating system, increasing the pouring temperature, improving the stopper mechanism, adopting a phased pouring sequence, and optimizing the chill layout. The improved process not only eliminated the defects but also enhanced the mechanical properties and microstructural consistency. The knowledge gained from this study has been incorporated into our global process standards for ductile iron castings. We believe that the continuous improvement of casting processes is essential to meet the ever-increasing demands of the marine industry.
Future work will focus on further optimizing the solidification sequence using 3D thermal simulation and possibly the use of intelligent process control. The relationship between the pouring parameters and the final microstructure will be investigated in more detail. In addition, the effect of trace elements such as antimony and bismuth on the nodularity of thick-section ductile iron castings will be explored. However, the fundamental understanding of the filling and gas-related defects remains the cornerstone of our production philosophy.
The author emphasizes that the successful production of high-quality ductile iron castings requires a holistic approach. It is not enough to simply adjust one parameter. One must consider the interaction between the metal, the mold, the core, and the gating system. In the present case, the defects were caused by a combination of factors, and only a comprehensive solution could resolve them. The methods presented here can serve as a reference for other foundries dealing with similar challenges in ductile iron castings.
Finally, it is worth noting that the simulation software used in this study was validated against the actual casting results. The predicted air entrainment index correlated well with the observed defects. This gives us confidence in the simulation as a design tool. We now use simulation as a routine tool for every new casting project. This reduces the number of physical trials and shortens the development time.
As a result of the improvements, the foundry achieved a yield of 100% for the subsequent production of these cylinder blocks. The customer was very satisfied with the quality and the performance of the castings. The success of this project strengthened our reputation as a reliable supplier of complex ductile iron castings for large marine engines.
In summary, the analysis and countermeasures described in this paper provide a practical guide for avoiding defects in ductile iron castings. The principles are generally applicable to other grades of spheroidal graphite iron as well. We hope that sharing our experience will help other foundry engineers and process developers in their own problem-solving endeavors. Continuous learning and adaptation are key to achieving excellence in casting production.
We are committed to further research and development in the field of ductile iron castings, aiming for even higher quality and efficiency. The lessons learned from this cylinder block project have already been applied to several other large castings, such as bedplates, gearboxes, and bearing housings. All of them have shown a marked improvement in quality. This confirms that the systematic methodology is both effective and versatile.
The future of ductile iron castings is promising, with new technologies such as 3D printing of molds and cores, real-time process monitoring, and machine learning-based process optimization. However, the fundamental physics of filling and solidification remains the same. Therefore, a deep understanding of these principles will always be the basis for successful casting production. We hope that this article contributes to the body of knowledge for the casting community.