I carried out a systematic investigation on the solidification and shrinkage behaviour of a diffuser casting made of ZG0Cr13Ni4Mo, a low-carbon martensitic stainless steel, with particular focus on the prevention of sand casting defects. The diffuser is a key component of the containment spray pump used in nuclear power plants. Because the component belongs to nuclear safety-related equipment, its quality requirements are very strict. The finished casting must pass tensile testing, impact testing, penetrant examination, radiographic examination and dimensional inspection. Surface defects such as peeling, cracks, sand holes, shrinkage cavities and shrinkage porosity are not permitted. In this paper I present the complete procedure that I followed to eliminate sand casting defects and to bring the casting dimensions within the required tolerances.
The diffuser is a complex stainless steel casting with a contour size of about 740 mm in diameter and 620 mm in height. The wall thickness varies from only 12 mm to about 55 mm, and the internal spiral flow passage creates a difficult feeding situation. The material ZG0Cr13Ni4Mo has a relatively large solidification shrinkage, and it is also affected by alloying elements that increase the volumetric contraction. These features make the casting prone to shrinkage cavities and shrinkage porosity. In addition, the phase transformation from austenite to martensite during cooling produces a volumetric expansion of approximately 0.54%. This expansion, together with the large temperature gradient, often causes distortion and dimensional errors. Therefore, the control of solidification shrinkage is the most important task in the design of the casting process.
The traditional procedure for developing a casting process depends strongly on the experience of the process engineer. For a complicated casting like the diffuser, trial castings are normally required to verify the design and to expose hidden sand casting defects. This trial-and-error approach is costly, time-consuming and not always reliable. To avoid this problem, I used a combination of CAD, numerical simulation and statistical experiment design. I built a three-dimensional model of the diffuser with the aid of a commercial CAD system. I designed an initial casting process that followed the accepted rules for risers, gating and chills. I then simulated the filling and solidification stages with a commercial casting simulation package. The simulation predicted shrinkage defects that matched the actual defects observed in production. Based on this diagnosis, I improved the process by adding external chills and an additional blind riser. I also used an orthogonal array to optimize the pouring temperature, pouring time and initial mold temperature. Finally, I analysed the dimensional deviations of the castings by three-dimensional scanning and corrected the core box dimensions. The final production results showed that the sand casting defects were effectively eliminated and that the dimensional accuracy of the casting met the acceptance criteria.
1. Theoretical Basis of Solidification and Shrinkage
1.1 Heat Transfer During Casting Solidification
The solidification process is governed by heat transfer. Heat is transferred by conduction, convection and radiation. In sand casting, the metal is poured into a sand mold, and the heat of the molten metal is removed through the mold. Since the thermal conductivity of the sand mold is much smaller than that of the metal, the cooling rate is relatively slow and the temperature difference within the casting cross-section is small. For the mould itself, the temperature difference between its inner surface and outer surface is large, and this gradient controls the heat flow. The fundamental differential equation of heat conduction is:
$$
\frac{\partial}{\partial x}\left(\lambda \frac{\partial T}{\partial x}\right)
+\frac{\partial}{\partial y}\left(\lambda \frac{\partial T}{\partial y}\right)
+\frac{\partial}{\partial z}\left(\lambda \frac{\partial T}{\partial z}\right)
+\dot{q}
=
c \rho \frac{\partial T}{\partial t}
$$
where T is the temperature, λ is the thermal conductivity, c is the specific heat capacity, ρ is the density, t is the time, and q̇ is the internal heat generation rate. When the thermal conductivity is assumed to be constant, the equation can be simplified to:
$$
\lambda \left( \frac{\partial^2 T}{\partial x^2}
+\frac{\partial^2 T}{\partial y^2}
+\frac{\partial^2 T}{\partial z^2} \right)
+\dot{q}
=
c \rho \frac{\partial T}{\partial t}
$$
During solidification, two interfaces move simultaneously. One is the solid-liquid interface which advances into the remaining liquid, and the other is the metal-mold interface which changes with temperature. Because the metal and the mold are not in perfect contact, an interfacial heat resistance appears. The heat transfer at the interface is a combination of conduction, convection and radiation. In the numerical simulation of sand casting defects, it is important to define the heat transfer coefficient as a function of temperature rather than as a constant value.
1.2 Shrinkage of Castings
Shrinkage occurs in three distinct stages during cooling. The first stage is liquid shrinkage, which takes place from the pouring temperature to the liquidus temperature. The second stage is solidification shrinkage, which occurs between the liquidus and solidus temperatures. The third stage is solid shrinkage, which takes place from the solidus temperature to room temperature. The first two stages are responsible for shrinkage cavities and shrinkage porosity, while the third stage is mainly responsible for dimensional changes and residual stresses.
The volumetric shrinkage and linear shrinkage are defined as:
$$
\varepsilon_V = \frac{V_0 – V_1}{V_0} \times 100\%
$$
$$
\varepsilon_l = \frac{l_0 – l_1}{l_0} \times 100\%
$$
where V₀ and V₁ are the volumes at temperatures t₀ and t₁, and l₀ and l₁ are the corresponding lengths. For cast steel, the solidification shrinkage depends on the carbon content and on the alloying elements. Table 1 lists the influence coefficients of different alloying elements on the solidification shrinkage of cast steel. Chromium, silicon and aluminium increase the shrinkage, while nickel and tungsten reduce it.
| Element | Mn | Cr | Si | Al | W | Ni |
|---|---|---|---|---|---|---|
| Influence coefficient | +0.0585 | +0.12 | +1.03 | +1.70 | -0.53 | -0.0354 |
Because ZG0Cr13Ni4Mo contains chromium and silicon, its solidification shrinkage is larger than that of plain carbon steel. The large shrinkage, combined with a wide solidification range, creates a strong tendency for sand casting defects. The diffuser solidifies between the liquidus temperature of 1490°C and the solidus temperature of 1455°C. In this temperature interval, dendrites grow and can easily block the feeding channels. If the riser pressure is not sufficient to overcome the intergranular resistance, shrinkage porosity appears in the centre of the thick sections or at the last-solidifying regions.
1.3 Shrinkage Cavities and Shrinkage Porosity
Shrinkage cavities are large voids formed when liquid metal is not available to compensate for the volumetric contraction during solidification. Shrinkage porosity, on the other hand, consists of a network of small voids that appear in the intergranular regions. Both are typical sand casting defects. They can be prevented by promoting directional solidification, so that the last solidifying regions are located in the risers. This can be achieved by proper riser design, by the use of chills, and by careful control of the pouring temperature and pouring time.
In a simple cylindrical casting, the outer surface solidifies first and forms a solid shell. As the temperature decreases, the inner liquid level drops and a shrinkage cavity forms. The condition for a riser to feed a casting can be expressed by the modulus method. The modulus of a casting section is the ratio of its volume to its cooling surface area. For the riser to solidify later than the casting, the riser modulus must be larger than the casting modulus:
$$
M_r = f M_c
$$
where Mr is the riser modulus, Mc is the modulus of the casting section, and f is a safety factor, usually between 1.1 and 1.3. For a side riser, the modulus of the riser neck should also satisfy the relationship:
$$
M_c : M_n : M_r = 1 : 1.1 : 1.2
$$
In addition, the maximum volume that a riser can feed is limited by its feeding efficiency. The limiting condition is:
$$
V_{Cmax} = \frac{\eta – \varepsilon}{\varepsilon} V_R
$$
where VCmax is the maximum casting volume that can be fed, η is the riser efficiency, ε is the solidification shrinkage of the alloy, and VR is the riser volume. If the required feeding volume is larger than this maximum value, the riser must be enlarged or additional risers must be added.
2. Diffuser Casting and Initial Process Design
2.1 Description of the Diffuser Casting
The diffuser is an internal component of the containment spray pump. It is made of ZG0Cr13Ni4Mo, which is a low-carbon martensitic stainless steel. This steel has excellent underwater fatigue strength, good static and dynamic fracture toughness, and good resistance to cavitation. It can tolerate small amounts of chloride ions, ammonia and carbon dioxide. It has also good casting and welding properties. The required chemical composition is summarized in Table 2.
| Element | C | Si | Mn | S | P | Cr | Ni | Mo |
|---|---|---|---|---|---|---|---|---|
| Content (wt%) | ≤0.06 | 0.30–0.80 | ≤1.00 | ≤0.020 | ≤0.030 | 12.00–13.50 | 3.50–4.50 | 0.40–0.70 |
The mechanical properties required after heat treatment are given in Table 3.
| Yield strength σs (MPa) | Tensile strength σb (MPa) | Elongation δ (%) |
|---|---|---|
| 550 | 760 | 15 |
2.2 Initial Casting Process
The initial casting process was designed according to the conventional rules for steel castings. The material was considered to be difficult to feed because of its high shrinkage, so I adopted the principle of directional solidification. The important surface is located on the large end of the diffuser. For this reason, I placed the large end at the bottom and the small end at the top of the mold. The casting position and the parting surfaces had to be chosen with care. Because the diffuser geometry is complex, a three-part mold was used. The pattern was split into several parts to facilitate molding.
The gating system was a bottom-gate system. A bottom gate is preferred for large steel castings because it allows a calm filling of the mold cavity and a convenient escape of gas. The pouring basin and the gating channels were made from ceramic tubes in order to avoid entrainment of sand and to prevent erosion of the mold. The ceramic tubes also maintain a constant cross-section during pouring, which improves the reproducibility of the filling process. The inner gate was placed at the bottom of a blind riser. This arrangement avoids the so-called “neck rot” effect, which is a typical sand casting defect near the ingate when the steel enters directly into the casting without passing through a riser.
The risers were initially designed by the modulus method. Three open oval risers were placed at the top of the casting to feed the large upper flange. An insulated riser was placed above the central hole to feed the thick central region. Eight blind risers were arranged around the bottom flange. The original riser arrangement was intended to compensate for the larger contraction of the steel. However, because the wall thickness of the diffuser changes abruptly, the feeding distances were not sufficient in some areas. Later numerical simulation showed that isolated liquid regions appeared at the bottom flange and at the lower surface of the inner ring. These regions produced sand casting defects after solidification.
Two external chills were used in the initial process. They were placed on two protruding steps of the outer contour of the casting. The size of each chill was 60 mm × 60 mm × 30 mm. The chill material was the same grade as the casting, namely ZG0Cr13Ni4Mo. The intention was to accelerate the cooling of these local hot spots and to extend the feeding distance of the adjacent risers. The initial process parameters are listed in Table 4.
| Parameter | Value |
|---|---|
| Pouring temperature | 1570 °C |
| Pouring time | 25 s |
| Pouring nozzle diameter | 40 mm |
| Mold material | Resin-bonded sand |
| Core material | Resin-bonded sand |
The sand molds and cores were manufactured manually because the production volume was small. The mold surface was coated with a water-based or alcohol-based refractory coating. All risers and gates were assembled on the pattern and later transferred to the mold. The total mass of the poured metal, including the gating and riser system, was about 750 kg.
3. Numerical Simulation of the Initial Process
3.1 Pre-processing and Mesh
I used the three-dimensional solid models of the casting, runners, risers, chills and core assembly and exported them in the STL format. These STL files were imported into the pre-processor of the casting simulation software. I assigned the correct entity type to each STL body, that is, casting, runner, riser, chill, core, or mold. The material database of the simulation software did not contain ZG0Cr13Ni4Mo, so I created a user-defined material database and inserted the thermophysical properties of this steel.
The finite difference method was used to discretize the computational domain. A non-uniform mesh was created to refine the thin sections and the regions near the chills and risers. The final mesh contained 3,334,992 elements. This mesh density was sufficient to ensure that the thin wall of 12 mm was represented by at least two elements. The initial process parameters used in the simulation are shown in Table 5.
| Simulation setting | Value |
|---|---|
| Pouring temperature | 1570 °C |
| Pouring time | 25 s |
| Nozzle diameter | 40 mm |
| Mold type | Sand mold |
| Heat transfer coefficient | Function of temperature |
3.2 Filling Simulation
The filling simulation showed that the metal rose smoothly in the mold cavity. The bottom-gating system prevented turbulence and helped to keep the liquid front intact. The air in the cavity was expelled through the risers and through the vents at the top of the mold. During the initial stage of filling, some vortex flow appeared in the four blind risers that were not directly connected to the inner gate. The vortex was confined to the riser cavity, so it did not create sand inclusions or gas defects in the casting. The total filling time was about 25 seconds.
From the velocity distribution, I could see that the metal front was stable in the main body of the casting. The maximum velocity occurred near the gate, but the ceramic gating system prevented sand erosion. The filling simulation indicated that the initial gating system was adequate and that no filling-related sand casting defects were expected. The gas porosity risk was low because the exhaust path through the risers was unobstructed.
3.3 Solidification Simulation
The solidification analysis was performed after the filling simulation. The total solidification time was about 4433 seconds. The thin outer walls, the gates and the chill regions solidified first. The risers remained liquid for a long time and were expected to feed the adjacent casting sections. However, the simulation showed the presence of isolated liquid islands in two critical locations. The first isolated liquid region appeared inside the lower flange at about 519 seconds. The second one appeared on the lower surface of the inner ring at about 1359 seconds. These isolated regions could not be reached by the risers because the feeding channels between the riser and the liquid islands were already blocked by solid dendrites. As a result, the simulation predicted shrinkage cavities and shrinkage porosity at these two locations.
The predicted sand casting defects were in good agreement with the defects actually observed in the diffuser casting. After machining the bottom flange, the foundry workers found small shrinkage cavities and dispersed porosity in the same locations that the simulation had indicated. The lower surface of the inner ring also showed shrinkage porosity. These defects could not be fully removed by machining and caused the rejection of the casting. This confirmed that the numerical simulation was reliable and that the initial process design had to be improved.
4. Process Optimization
4.1 Improvement of the Feeding System
Based on the simulation results, I proposed two major improvements. The first improvement was to add a blind riser to the bottom flange. The bottom flange is thick and far from the top risers. The wall above the flange is relatively thin, and it solidifies earlier than the flange, blocking the feeding path. In the original design, the bottom flange was fed only by the eight small blind risers, but these were not sufficient. In the improved design, I added an additional blind riser at the centre of the flange. The modulus of the flange section was about 1.25 cm. Using a safety factor of 1.2, the riser modulus was calculated as:
$$
M_r = 1.2 \times 1.25 = 1.50 \text{ cm}
$$
The riser was designed with a diameter of 100 mm and a height of 150 mm. It was connected to the flange through three relatively small necks. The total feeding volume supplied by this new riser was checked by the feeding volume criterion. The maximum casting volume that could be fed by the riser was greater than the required feeding volume, so the riser size was considered acceptable.
The second improvement was to add a ring of external chills on the lower surface of the inner ring. This region is surrounded by sand cores, and the cooling rate is very low. It is difficult to feed this region because the distance from the main risers is large. External chills increase the local cooling rate and create an artificial end zone that extends the feeding distance of the risers. I placed ten chills uniformly around the inner lower surface. Each chill had a size of 80 mm × 30 mm × 30 mm. The chills were again made of the same steel grade as the casting. The improved riser and chill arrangement is shown in the following photograph of the actual automated pouring line used in the production verification.
4.2 Simulation of the Improved Process
I repeated the numerical simulation with the improved riser and chill system. The filling results were unchanged; the filling was still smooth and the gas escape path was still efficient. The solidification sequence was more favourable. The new blind riser at the bottom flange solidified later than the flange itself, and it successfully fed the thick section. The external chills at the inner ring increased the cooling rate of the lower surface and eliminated the isolated liquid island. The shrinkage defects predicted by the software were no longer present in the flange and inner ring regions. The residual melt modulus analysis indicated that any remaining shrinkage was located inside the risers rather than inside the casting. This meant that the sand casting defects would be removed together with the riser when the casting was cut from the gating system.
4.3 Orthogonal Experiment for Process Parameters
Although the improved process eliminated the major shrinkage defects, the quality of the casting could still be influenced by the pouring temperature, the pouring time and the initial temperature of the sand mold. To find the optimal combination of these parameters, I designed a three-factor, three-level orthogonal experiment. The three factors and their levels are listed in Table 6.
| Level | A: Pouring temperature (°C) | B: Pouring time (s) | C: Mold initial temperature (°C) |
|---|---|---|---|
| 1 | 1550 | 28 | 20 |
| 2 | 1560 | 30 | 60 |
| 3 | 1570 | 32 | 100 |
The objective of the orthogonal experiment was to minimize the sand casting defects associated with shrinkage. I used the residual melt modulus as the evaluation index. The residual melt modulus is defined as the ratio of the residual liquid volume to its surface area at a certain time. A smaller value indicates a lower risk of isolated liquid regions and therefore a lower risk of shrinkage defects. The standard L9 orthogonal array was adopted. The results are shown in Table 7.
| No. | A | B | C | Error column | Residual melt modulus |
|---|---|---|---|---|---|
| 1 | 1 | 1 | 1 | 1 | 1.9200 |
| 2 | 1 | 2 | 2 | 2 | 1.9306 |
| 3 | 1 | 3 | 3 | 3 | 1.9404 |
| 4 | 2 | 1 | 2 | 3 | 1.9657 |
| 5 | 2 | 2 | 3 | 1 | 1.9759 |
| 6 | 2 | 3 | 1 | 2 | 1.9524 |
| 7 | 3 | 1 | 3 | 2 | 2.0147 |
| 8 | 3 | 2 | 1 | 3 | 1.9878 |
| 9 | 3 | 3 | 2 | 1 | 1.9986 |
The range analysis is summarized in Table 8. The range values show the importance of each factor. The order of influence was A (pouring temperature) > C (mold initial temperature) > B (pouring time). The optimal combination was A3 B1 C3, that is, a pouring temperature of 1570°C, a pouring time of 28 s, and a mold initial temperature of 100°C.
| Average response | A | B | C |
|---|---|---|---|
| t1 | 1.930 | 1.967 | 1.953 |
| t2 | 1.965 | 1.965 | 1.965 |
| t3 | 2.000 | 1.964 | 1.977 |
| Range R | 0.070 | 0.003 | 0.024 |
I also performed an analysis of variance. The results are shown in Table 9. The factor A had a significant effect at the 1% level, and factor C had a significant effect at the 1% level. Factor B had no significant effect. This confirms that the pouring temperature is the most important process parameter for the control of shrinkage defects in this steel.
| Source | Sum of squares | Degree of freedom | Mean square | F-ratio |
|---|---|---|---|---|
| A | 0.002710 | 2 | 0.001355 | 903.33 |
| B | 0.000014 | 2 | 0.000007 | 4.67 |
| C | 0.000836 | 2 | 0.000418 | 278.67 |
| Error | 0.000003 | 2 | 0.0000015 | — |
The results of the orthogonal experiment were applied to the actual production process. In the optimized process, the pouring temperature was controlled at 1570°C, the pouring time was set to 28 s, and the sand mold was preheated to 100°C before pouring. Under these conditions, the filling and solidification of the diffuser casting were stable, and the number of sand casting defects was minimized.
5. Dimensional Deviation Correction
5.1 Causes of Dimensional Deviations
The dimensional accuracy of a casting is influenced by the shrinkage of the alloy, the pattern and core box accuracy, the resistance to contraction, the cooling conditions and the mould properties. For a complex casting such as the diffuser, the shrinkage is not uniform in all directions because the wall thickness varies considerably. In the initial process, a single casting shrinkage value of 1.5% was used for all directions. This is a common simplification, but it is not accurate enough for a component with internal spiral passages. The non-uniform shrinkage, combined with the phase transformation expansion, led to dimensional errors in the internal flow passage of the diffuser.
After the first production run, I performed a three-dimensional optical scan of the casting. The scan showed that the outer surfaces were generally within the required tolerance. However, the internal flow passage had several deviations that exceeded the permissible tolerance band according to ISO 8062 CT11. These deviations were caused not only by the non-uniform shrinkage but also by the deformation of the sand core and by the difficulty of core setting. Because the internal flow passage is formed by cores, the dimensions of the core boxes needed to be corrected.
5.2 Calculation of the Actual Shrinkage
To correct the core box dimensions, I used the measured deviations to calculate the actual shrinkage of the casting at each critical location. The original core box dimension was calculated with the formula:
$$
L_M = (1 + \varepsilon_0) L_N
$$
where LN is the nominal casting dimension and ε0 is the original shrinkage of 1.5%. If the measured casting dimension is La, the actual shrinkage is:
$$
\varepsilon_a = \frac{L_M – L_a}{L_a} \times 100\%
$$
With this actual shrinkage, the new core box dimension can be selected to produce a casting within the tolerance band. For a total tolerance T, the acceptable casting dimension lies between LN – T/2 and LN + T/2. The corresponding core-box dimension range is:
$$
L_{M,min} = (1+\varepsilon_a)\left(L_N – \frac{T}{2}\right)
$$
$$
L_{M,max} = (1+\varepsilon_a)\left(L_N + \frac{T}{2}\right)
$$
For example, one point at the flow passage had a nominal dimension of 286 mm and a measured deviation of -4.62 mm. The measured dimension was therefore 281.38 mm. The original core box dimension was 290.29 mm. The actual shrinkage was:
$$
\varepsilon_a = \frac{290.29 – 281.38}{281.38} \times 100\% = 3.17\%
$$
The tolerance for this dimension was 6.2 mm, so the acceptable casting range was from 282.9 mm to 289.1 mm. The new core-box dimension to obtain the nominal casting size was:
$$
L_{M,new} = (1 + 0.0317) \times 286 = 295.07 \text{ mm}
$$
Several other measured points were treated in the same way. Some core-box dimensions needed to be increased, while a few needed to be reduced, depending on the sign and magnitude of the deviation. The selected correction values for the main locations are summarized in Table 10. The values were rounded to convenient machining increments.
| Nominal dimension (mm) | Measured deviation (mm) | Original core-box dim. (mm) | Actual shrinkage (%) | Corrected core-box dim. (mm) |
|---|---|---|---|---|
| 286 | -4.62 | 290.29 | 3.17 | 295.1 |
| 231.5 | -3.49 | 234.97 | 3.08 | 238.7 |
| 315 | -3.57 | 319.73 | 2.67 | 323.5 |
| 179 | -5.31 | 181.69 | 4.55 | 187.2 |
In practice, I did not rely solely on the calculated values. I also used standard gauge plates and sampling boards to verify the corrected core-box shape. When the core box was too small, I removed material by grinding. When it was too large, I applied epoxy resin to build up the surface. This approach allowed a quick and inexpensive correction of the core boxes.
6. Production Verification
After the process optimization and the core-box corrections, a second production batch was cast. The optimized parameters, namely a pouring temperature of 1570°C, a pouring time of 28 s and a mold temperature of 100°C, were used. The castings were again subjected to three-dimensional scanning. The comparison of the scan results with the previous measurements showed a substantial improvement. The shrinkage cavities and shrinkage porosity at the bottom flange and inner ring were no longer present. The sand casting defects that had appeared in the initial production run were completely eliminated.
The dimensional deviations at the internal flow passage were also reduced. Most of the measured points were inside the tolerance band. A few local deviations still existed, but they were minor. With the approval of the customer, these small deviations were corrected by grinding the areas where the casting was too large, and by welding where the casting was too small. After welding, a local heat treatment was performed to restore the required mechanical properties. The castings then passed the final inspection and were considered acceptable.
The results confirmed the validity of the numerical simulation and the usefulness of the design procedure that I followed. The combination of casting simulation, orthogonal experimental design and three-dimensional dimensional measurement made it possible to solve the sand casting defects in a relatively short time and with low cost. The production yield was significantly improved, and the time required for repair welding and rework was reduced.
7. Conclusions and Outlook
In this work, I investigated the solidification and shrinkage behaviour of a diffuser casting made of ZG0Cr13Ni4Mo low-carbon martensitic stainless steel. I proposed an optimized casting process that successfully eliminated the sand casting defects and improved the dimensional accuracy of the component. The main conclusions of my study are the following:
First, the numerical simulation of the filling and solidification process was able to predict the sand casting defects in the initial design. The simulation showed that isolated liquid regions would form in the bottom flange and at the lower surface of the inner ring. These predictions matched the actual defects observed in the first production batch. This indicates that simulation is a powerful tool for the diagnosis and prevention of sand casting defects.
Second, the shrinkage defects were eliminated by improving the feeding system. The addition of a blind riser at the bottom flange and the placement of ten external chills on the lower surface of the inner ring promoted directional solidification and increased the feeding distance of the existing risers. The simulation of the improved process showed no residual shrinkage defects at the critical locations.
Third, the orthogonal experiment showed that the pouring temperature had the greatest influence on the formation of sand casting defects, followed by the initial mold temperature, while the pouring time had only a minor effect. The optimal parameters were found to be a pouring temperature of 1570°C, a pouring time of 28 s and a mold initial temperature of 100°C.
Fourth, the dimensional deviations in the internal flow passage were caused by the non-uniform shrinkage of the casting and the insufficient accuracy of using a single shrinkage value for all directions. By calculating the actual shrinkage from three-dimensional measurements, I was able to correct the core-box dimensions and bring the casting within the required tolerance class. The use of gauge plates and gradual adjustment of the core boxes was an effective and economical method for small-series production.
Finally, the production verification demonstrated that the optimized process is stable and reliable. The improved casting process not only reduced the sand casting defects and the need for repair welding, but also shortened the production cycle and reduced the manufacturing cost. In the future, further improvement may be achieved by using a more accurate heat transfer coefficient in the numerical simulation and by applying computer-aided optimization methods to find an optimal riser shape and location. The same methodology can be extended to other large stainless steel castings with complex internal passages, where sand casting defects are a common problem.
In summary, the work presented in this paper provides a practical example of how modern casting simulation and statistical methods can be combined to solve complex foundry problems. The experience gained from the diffuser casting shows that a careful analysis of the solidification and shrinkage process is the key to producing sound castings and to minimizing sand casting defects in high-alloy steel castings.