In my development work on complex stationary blades, I have focused on a manufacturing route that can be described as low-allowance sand mould casting. The objective is not simply to produce a near-net shape, but to produce a blade whose as-cast surfaces, thicknesses, twist, and profile are already close enough to the final design that only grinding and polishing are required. In this article I summarize the reasoning, process design, trial stages, defect analysis, dimensional control, and economic results that I obtained while developing a stable sand mould casting process for a sabre-type blade. The blade is strongly curved and twisted, its wall thickness varies significantly from section to section, and its acceptance requirements are strict. Therefore, every decision in the sand mould casting route has to be made with deformation, surface quality, and dimensional repeatability in mind.
The general trend in foundry production is toward precision, meaning smaller dimensional tolerances, more accurate geometry, and lower machining allowance. Low-allowance sand mould casting is one practical expression of that trend. In this method, the allowance on surfaces that would normally be machined is kept very small. After casting, the part is not subjected to conventional mechanical cutting; instead, it is directly finished by abrasive grinding and polishing. This approach saves raw material, shortens the production cycle, and lowers total cost. For a sabre-type blade, however, the geometry is so complex that this approach is not optional. A conventional allowance would be difficult or impossible to remove by machining because the blade passages are curved, twisted, and narrow. Therefore, low-allowance sand mould casting became the only realistic route for producing this type of blade in our foundry environment.
The sabre-type blade is a relatively new blade form. Its shape is bent and twisted in a manner resembling a sabre, so it is also called a bow-shaped blade. Its practical value has attracted attention in turbine manufacturing because it can improve aerodynamic performance in the low-pressure final stages. The blade I worked on is used in retrofit units, and its geometry is defined by multiple profile sections. The blade has a complex profile, with several cross sections. Its maximum thickness is much larger than its minimum thickness, and its exit edge is very thin. The chord width also changes along the blade. Because the blade is long and slender, any inhomogeneous cooling or handling can produce distortion. For this reason, the sand mould casting process must control both solidification and subsequent heat treatment.
The acceptance conditions for the sabre-type blade include material requirements, dimensional requirements, and surface quality requirements. The chemical composition and mechanical properties must meet specified values. The maximum thickness has a tolerance, the exit edge thickness must not become knife-edged, and the profile must match the gauge within a specified leak gap. Surface quality is equally important: cracks, penetrating defects, and linear defects are not allowed. In important regions, certain porosity and cavity defects are also unacceptable. These requirements make low-allowance sand mould casting a challenging task, because the casting must be sound, dimensionally accurate, and clean enough that grinding and polishing do not expose unacceptable defects.
Before selecting the final process, I reviewed the conventional production method for stationary blades. In ordinary sand mould casting, the concave and convex surfaces are given machining allowance, usually between several millimeters and more than ten millimeters depending on blade size. Smaller blades take the lower limit, while larger blades take the upper limit. Because the allowance is large, casting distortion and heat treatment distortion are often absorbed by later machining. Surface defects can also be removed together with the machining allowance. As a result, conventional sand mould casting can usually meet design requirements for ordinary blades. For the sabre-type blade, however, the twisted geometry and complex profile make mechanical removal of such an allowance impractical. The machining tool cannot easily reach the curved passages, and the risk of damaging the thin exit edge is high. Therefore, low-allowance sand mould casting had to be developed.
Investment casting is another low-allowance method, but I found it less suitable for this blade. The blade is relatively large, its thickness varies greatly from point to point, and its rough weight is significant. If investment casting were used, wax pattern deformation, shell deformation, and solidification shrinkage would be difficult to control. The tooling would be large and complex, handling would be difficult, the production cycle would be long, and the cost would be high. For these reasons, I concentrated on low-allowance sand mould casting. The challenge was to obtain the dimensional accuracy and surface quality normally associated with investment casting, while keeping the economic and practical advantages of sand mould casting.
The process parameters I selected were based on the blade material and the required finishing allowance. A shrinkage scale was applied to the pattern. The concave and convex profile surfaces were both given a grinding and polishing allowance. The allowance had to be small enough to be removed by abrasive operations, but large enough to absorb casting distortion and heat treatment distortion. I expressed the basic allowance relationship as follows:
$$R_{\mathrm{total}} = R_{\mathrm{casting}} + R_{\mathrm{heat\ treatment}} + R_{\mathrm{handling}}$$
where \(R_{\mathrm{total}}\) is the total allowance remaining on the as-cast surface, \(R_{\mathrm{casting}}\) is the distortion introduced during solidification and cooling, \(R_{\mathrm{heat\ treatment}}\) is the distortion introduced during heat treatment, and \(R_{\mathrm{handling}}\) is the additional allowance required for cleaning, transport, and fixture effects. The working target was to keep \(R_{\mathrm{total}}\) within the grinding and polishing capability. In practice, this meant that the as-cast profile had to remain inside a narrow band around the nominal profile.
The production conditions were chosen to support this target. For the mould, I used self-hardening water-glass sand with an organic ester hardener. One advantage of water-glass sand is that the mould can harden before the pattern is removed, which helps to preserve mould cavity dimensions and reduces pattern-related distortion. This is important for low-allowance sand mould casting because any movement of the mould walls directly affects the final blade thickness and profile. The basic sand mixture contained quartz sand, water glass, hardener, and a breakdown agent. The exact proportions were adjusted to obtain adequate strength, good collapsibility, and acceptable surface finish.
For melting, the blade material is a martensitic stainless steel. I used a medium-frequency induction furnace. Because low-allowance sand mould casting demands good surface quality and complete filling of the thinnest sections, the metallurgical quality of the molten steel was strictly controlled. The pouring temperature was kept in a relatively high range, approximately \(1580\) to \(1620\,^{\circ}\mathrm{C}\), and slightly toward the upper limit. High pouring temperature helps gas escape and slag float, but it must not be so high that it causes excessive sand burn-on or coating erosion. The pouring temperature window can be written as:
$$T_{\mathrm{pour}} = T_{\mathrm{liquidus}} + \Delta T_{\mathrm{superheat}}$$
where \(T_{\mathrm{liquidus}}\) is the liquidus temperature of the stainless steel and \(\Delta T_{\mathrm{superheat}}\) is the selected superheat. For this material, the superheat was chosen to balance fluidity against mould reaction. I also controlled the pouring speed so that the metal front advanced smoothly and did not erode the coating.
The overall casting orientation was horizontal moulding with vertical pouring. Within that general scheme, I evaluated three gating options: side pouring, top pouring, and bottom pouring. Each option has a different influence on metal flow, temperature distribution, slag entrapment, and surface defect formation. The following table summarizes the main features of the three designs.
| Gating design | Metal entry position | Main advantage | Main disadvantage | Surface defect tendency | Yield |
|---|---|---|---|---|---|
| Side pouring | Along the exit edge through a fin-like ingate | Simple connection to the runner | Direct metal impact on the concave mould surface | High on concave surface, some on convex surface | Low |
| Top pouring | Through a shaped top riser at the large end | Simple, good feeding, good venting | Metal falls through the cavity | Scattered defects on both surfaces | High |
| Bottom pouring | Through a horn-shaped ingate at the small end | Smooth filling from the bottom | Low temperature at the thin exit edge | Concentrated at the exit edge | Medium |
In the side pouring design, the ingate was opened along the exit edge from the small end to the large end, almost like a casting fin. The sprue was connected directly to this ingate. A spherical blind riser was placed at the large end, with a vent to the atmosphere. The riser provided feeding and slag collection, while the vent allowed gas to escape. Although this design is simple, the metal stream enters the cavity directly and impinges on the concave mould surface. This impact can cause coating and sand to spall, which then become inclusions. In my trials, the concave surface showed many defects, and the convex surface also had defects, though fewer. The blade also required cutting of the fin-like ingate, which increased cleaning work. The yield was lower than that of top pouring and bottom pouring. Measurements of as-cast distortion showed that different sections deformed differently: the two ends deformed more, while the middle deformed less. The distortion amount was significant.
In the top pouring design, only a shaped top riser was placed at the large end. This riser acted as a feeder, a vent, and a pouring cup. Molten steel was poured directly into the mould cavity through this riser. The process is simple, the yield is high, and the riser provides good feeding. However, because the metal enters from the top, it falls through the cavity and can disturb the coating. In my trials, both the concave and convex surfaces had defects, but they were relatively scattered. The as-cast distortion was similar to that of side pouring. The top pouring design was therefore promising, but it still required improvement.
In the bottom pouring design, a shaped top riser was placed at the large end, and a horn-shaped ingate was placed at the small end. Metal entered the cavity smoothly from the bottom after passing through the sprue and ingate. This design reduces turbulence and impact, but the thin exit edge cools quickly, and the bottom pouring path causes a large temperature loss. As a result, slag and inclusion defects concentrated at the exit edge, which is an important region. In addition, a shrinkage depression appeared at the maximum thickness under the riser. Because the exit edge is critical for acceptance, bottom pouring was considered unsuitable for this blade. I therefore excluded it from further development.
The trial process was divided into three stages. In the first stage, I aimed to understand the distortion trend during casting and heat treatment and to identify the factors affecting surface quality. Each of the three gating designs was poured several times. The results confirmed that side pouring produced many defects on the concave surface and required extra cleaning; top pouring produced scattered defects but a higher yield; bottom pouring concentrated defects at the exit edge. After heat treatment, the distortion was measured again. I found that heat treatment distortion was larger than casting distortion. This was an important finding because it meant that low-allowance sand mould casting could not succeed unless heat treatment distortion was controlled. The first stage results are summarized in the following table.
| Trial stage | Objective | Gating designs tested | Main observation | Decision |
|---|---|---|---|---|
| Stage 1 | Understand distortion and surface defects | Side, top, bottom | Side: concave defects and low yield; top: scattered defects and high yield; bottom: exit-edge defects and shrinkage | Eliminate bottom pouring; continue side and top |
| Stage 2 | Improve dimensional control and surface quality | Side, top | Process correction and coating change reduced defects; top pouring better | Select top pouring for further work |
| Stage 3 | Verify stability and reliability | Top | Distortion within grinding allowance; surface quality acceptable | Adopt top pouring as production route |
In the second stage, I compared side pouring and top pouring after making two important changes. First, I adjusted the process correction amounts for different blade sections and corrected the pattern. Second, I replaced the original alcohol-based coating with an alcohol-based zircon coating. The zircon coating has high refractoriness, sinters more easily, and has low thermal expansion, so its high-temperature strength is better and it is less likely to spall. After these changes, each design was poured several times. The as-cast distortion was basically controlled within the grinding and polishing allowance. Surface defects on both the concave and convex surfaces decreased significantly. Comparing the two designs, top pouring still gave better surface quality. Therefore, under the existing production conditions, top pouring was considered the more reasonable choice.
In the same stage, I also improved the heat treatment stacking method. Originally, the blades were laid flat and mixed together on a plate. This caused uneven heating and cooling because the thick and thin sections responded differently. I changed the arrangement so that the blades were placed vertically on their sides, with a controlled gap between adjacent blades. This reduced heat treatment distortion. The principle can be expressed as a reduction in the thermal gradient:
$$\nabla T = \frac{\partial T}{\partial x}\hat{i} + \frac{\partial T}{\partial y}\hat{j} + \frac{\partial T}{\partial z}\hat{k}$$
When the blade is packed flat and mixed, the thermal gradient through the thickness is large and non-uniform. When the blade is placed vertically on its side with gaps, the thermal gradient becomes more uniform, so the differential expansion and contraction are reduced. The total distortion after heat treatment can be approximated by:
$$\delta_{\mathrm{HT}} = \int_{0}^{L} \alpha \left( T(x) – T_{\mathrm{ref}} \right) dx$$
where \(\alpha\) is the coefficient of thermal expansion, \(T(x)\) is the temperature distribution along the blade, and \(T_{\mathrm{ref}}\) is the reference temperature. A more uniform \(T(x)\) produces a smaller \(\delta_{\mathrm{HT}}\). This was confirmed by measurements after the stacking change.
In the third stage, I focused only on the top pouring design and kept the casting and heat treatment controls unchanged. The purpose was to verify process stability and reliability. Several heats were poured, and all blades were inspected. Apart from local areas that needed minor improvement, the as-cast distortion remained within the grinding and polishing allowance. Surface quality was satisfactory. These results showed that the process had a certain level of stability and reliability, and it was the best choice under the available production conditions.
Surface defects were the main problem during the trials. After grinding and polishing, some casting defects remained on the blade surface and seriously affected surface quality. Analysis showed that the defects were related both to the gating design and to the coating. They appeared as slag inclusions and coating inclusions on the concave and convex surfaces. The following table summarizes the defect sources and the corrective actions.
| Defect type | Location | Main cause | Corrective action | Result |
|---|---|---|---|---|
| Sand inclusion | Concave surface, convex surface | Metal impact and mould erosion | Change gating design; reduce turbulence | Fewer inclusions |
| Coating inclusion | Concave surface, convex surface | Coating spalling under high-temperature metal flow | Replace coating with zircon-based coating | Improved surface quality |
| Slag inclusion | Exit edge, scattered areas | Low pouring temperature, poor slag flotation | Raise pouring temperature; improve gating | Less slag entrapment |
| Shrinkage depression | Maximum thickness under riser | Insufficient feeding in bottom pouring | Use top riser with better feeding | Eliminated in top pouring |
The influence of the gating design on surface defects was clear. In side pouring, molten steel entered through the fin-like ingate and directly struck the mould cavity that forms the concave surface. This caused the coating and sand to spall. Many sand and coating inclusions were therefore found on the concave surface. As the metal flowed, some of the spalled material was carried to the convex surface, creating similar defects there. In bottom pouring, the exit edge was thin and cooled quickly, and the bottom pouring path caused a large temperature loss. Consequently, slag and inclusion defects concentrated at the exit edge. Top pouring avoided these unfavorable factors because the metal entered through a large top riser with a relatively smooth flow path and good feeding. Therefore, the blades poured by top pouring had better surface quality.
The coating also had a major influence. Samples taken from surface inclusions showed that the main components were those of the original corundum-based coating. This indicated that the defects were related to the coating. Further analysis showed that although corundum has high refractoriness, it does not sinter easily. Its high-temperature strength is not good enough, so the coating layer can spall under the washing action of high-temperature molten steel. This produced coating inclusion defects. In later trials and production, I replaced the corundum-based alcohol coating with a zircon-based alcohol coating. The zircon coating has high refractoriness, sinters easily, and has very low thermal expansion. Its high-temperature strength is excellent, so it does not spall easily and deforms little. The use of zircon coating not only suppressed surface defects but also helped control blade dimensions.
Deformation was the second major issue. The blade has non-uniform thickness, a long length, and a twisted profile. If a wooden pattern is used, it is easily deformed, has limited strength, and is easily damaged. To reduce pattern-related deformation, I used an epoxy plastic pattern. Epoxy plastic has high dimensional accuracy, low deformation, a smooth surface, and good wear resistance. These properties satisfy the requirements of low-allowance sand mould casting. The pattern correction can be expressed as:
$$P_{\mathrm{corrected}}(x) = P_{\mathrm{nominal}}(x) + \Delta P_{\mathrm{process}}(x)$$
where \(P_{\mathrm{corrected}}(x)\) is the corrected pattern profile at position \(x\), \(P_{\mathrm{nominal}}(x)\) is the nominal blade profile, and \(\Delta P_{\mathrm{process}}(x)\) is the process correction. Because the blade distorts differently at different sections, \(\Delta P_{\mathrm{process}}(x)\) cannot be a single constant. I therefore applied different correction amounts to different sections. After correction, the as-cast distortion was controlled.
The blade structure also affects deformation. Because the thickness varies greatly, the cooling rate during solidification and contraction is not uniform. The thick sections cool slowly, while the thin exit edge cools quickly. This difference produces thermal stress and distortion. In the early trials, I did not apply process correction based on the structural characteristics, and the as-cast distortion was large. Later, I applied different process correction amounts to different positions, and the as-cast distortion was reduced. The relationship between thickness difference and distortion can be written qualitatively as:
$$\delta_{\mathrm{casting}} \propto \frac{\Delta t}{t_{\mathrm{avg}}} \cdot L \cdot \Delta T$$
where \(\Delta t\) is the thickness difference, \(t_{\mathrm{avg}}\) is the average thickness, \(L\) is the blade length, and \(\Delta T\) is the temperature difference during cooling. Reducing \(\Delta T\) or compensating for it in the pattern reduces distortion.
Heat treatment also had a strong effect. In the early stage, the blades were laid flat and mixed on a plate before being charged into the furnace. Because the thickness difference was large, heating was not uniform. During cooling, the cooling rate also differed between thick and thin sections. As a result, heat treatment deformation was large. I concluded that uniform heating and uniform cooling of the whole blade are the key to controlling heat treatment deformation. To achieve this, I designed a special blade support plate for heat treatment. When loading, the blade was placed on the support plate with the exit edge downward and the inlet edge upward, that is, the blade was placed on its side. The large end was raised slightly above the small end, so the whole blade formed a small angle with the horizontal plane. A controlled gap was left between adjacent blades. With this arrangement, the blade was heated and cooled more uniformly, and heat treatment deformation was very small.
The heat treatment cycle itself was also controlled. After casting, the blades were heat treated by a sequence including holding, air cooling, holding, air cooling, and normalizing plus tempering. The exact temperatures and times were selected according to the material specification. The important point for low-allowance sand mould casting is that the heat treatment fixture and stacking must be repeatable. If the blade is allowed to sag or if adjacent blades touch, local cooling rates change and distortion increases. Therefore, the support plate, the blade angle, and the gap between blades are all part of the process specification.
Dimensional inspection was performed on finished blades. Special inspectors checked the maximum thickness, the exit edge thickness, and the leak gap between the blade profile and the gauge. The results showed that the maximum thickness tolerance was within the specified range, the exit edge thickness met the basic dimension, and no finished blade had a knife-edged exit edge. The following table lists representative leak gap measurements for several finished blades and several profile sections.
| Blade number | Profile section | Concave leak gap | Convex leak gap | Acceptance limit | Result |
|---|---|---|---|---|---|
| 1 | Section A | 0.18 | 0.22 | Concave 0.30, convex 0.50 | Pass |
| 1 | Section B | 0.21 | 0.25 | Concave 0.30, convex 0.50 | Pass |
| 2 | Section A | 0.16 | 0.20 | Concave 0.30, convex 0.50 | Pass |
| 2 | Section B | 0.24 | 0.28 | Concave 0.30, convex 0.50 | Pass |
| 3 | Section A | 0.19 | 0.23 | Concave 0.30, convex 0.50 | Pass |
| 3 | Section B | 0.22 | 0.27 | Concave 0.30, convex 0.50 | Pass |
| 4 | Section A | 0.17 | 0.21 | Concave 0.30, convex 0.50 | Pass |
| 4 | Section B | 0.20 | 0.26 | Concave 0.30, convex 0.50 | Pass |
The leak gap was evaluated using the following condition:
$$G_{\mathrm{leak}} = \max \left( d_{\mathrm{profile}} – d_{\mathrm{gauge}}, 0 \right)$$
where \(d_{\mathrm{profile}}\) is the measured profile deviation and \(d_{\mathrm{gauge}}\) is the allowable gap. If \(G_{\mathrm{leak}}\) is zero, the profile is within the gauge limit. In my measurements, all the checked sections satisfied this condition. The maximum thickness tolerance also met the requirement:
$$\left| t_{\max,\mathrm{actual}} – t_{\max,\mathrm{nominal}} \right| \leq \Delta t_{\max}$$
where \(\Delta t_{\max}\) is the specified tolerance for the maximum thickness. The exit edge thickness was checked against the basic dimension, and no knife edge was allowed. The condition was:
$$t_{\mathrm{exit}} \geq t_{\mathrm{exit,min}}$$
where \(t_{\mathrm{exit,min}}\) is the minimum allowable exit edge thickness. All finished blades satisfied this condition.
Surface quality was also inspected. No cracks, penetrating defects, or linear defects were found. Some localized surface defects were repaired by welding, and after repair the surface quality fully met the acceptance conditions. The repair was controlled so that the heat input did not introduce new distortion. The acceptance criteria for surface defects can be summarized as follows:
$$D_{\mathrm{linear}} = 0$$
$$D_{\mathrm{penetrating}} = 0$$
$$D_{\mathrm{crack}} = 0$$
$$P_{\mathrm{important}} \leq P_{\mathrm{allow}}$$
where \(D_{\mathrm{linear}}\), \(D_{\mathrm{penetrating}}\), and \(D_{\mathrm{crack}}\) are the presence of linear, penetrating, and crack defects, respectively, and \(P_{\mathrm{important}}\) is the porosity or cavity defect level in important regions. The low-allowance sand mould casting process was able to meet these requirements after the coating and gating improvements.
The material requirements were also verified. The chemical composition of the martensitic stainless steel was checked against the specification. The mechanical properties after heat treatment were tested. The following tables summarize the general requirements and the observed results in a generic form.
| Element | Specification range | Typical observed value | Compliance |
|---|---|---|---|
| C | Within specified range | Within range | Yes |
| Si | Within specified range | Within range | Yes |
| Mn | Within specified range | Within range | Yes |
| Cr | Within specified range | Within range | Yes |
| Ni | Within specified range | Within range | Yes |
| Mo | Within specified range | Within range | Yes |
| Property | Specified requirement | Typical observed result | Compliance |
|---|---|---|---|
| Tensile strength | Minimum specified value | Above minimum | Yes |
| Yield strength | Minimum specified value | Above minimum | Yes |
| Elongation | Minimum specified value | Above minimum | Yes |
| Impact toughness | Minimum specified value | Above minimum | Yes |
| Hardness | Specified range | Within range | Yes |
The economic benefit of low-allowance sand mould casting was significant. By reducing the machining allowance, raw material consumption was lowered. Because the blade no longer required extensive mechanical cutting, the machining cost and machining time were reduced. The production cycle was shorter, and the yield of finished blades was improved. In one retrofit project, two units successfully used low-allowance sand mould cast sabre-type blades to replace the original stationary blades. The direct economic benefit from saved raw material and reduced machining cost was close to several million monetary units. By the end of the reporting period, many turbine units had been produced, and as technology progressed, these units were expected to be retrofitted to improve economy. Therefore, the use of low-allowance sand mould cast sabre-type blades was expected to bring even greater economic benefits. If the same low-allowance sand mould casting process is extended to other stationary blades, the economic benefit will increase significantly.
The cost saving can be expressed in a simple model. Let \(W_{\mathrm{machining}}\) be the weight of material removed by machining in the conventional process, and let \(C_{\mathrm{machining}}\) be the cost per unit weight of machining. For \(N\) blades, the direct saving is:
$$S = N \left( C_{\mathrm{material}} W_{\mathrm{machining}} + C_{\mathrm{machining}} W_{\mathrm{machining}} + C_{\mathrm{cycle}} \right)$$
where \(C_{\mathrm{material}}\) is the material cost per unit weight and \(C_{\mathrm{cycle}}\) is the cycle-time cost reduction per blade. In the conventional process, \(W_{\mathrm{machining}}\) is large. In low-allowance sand mould casting, \(W_{\mathrm{machining}}\) is reduced to the small amount removed by grinding and polishing. The material utilization ratio can be written as:
$$\eta = \frac{W_{\mathrm{finished}}}{W_{\mathrm{rough}}} \times 100\%$$
where \(W_{\mathrm{finished}}\) is the finished blade weight and \(W_{\mathrm{rough}}\) is the rough casting weight. Low-allowance sand mould casting increases \(\eta\) because the rough casting is closer to the finished shape. The process also reduces the energy consumption associated with machining and the disposal of chips. Therefore, the total benefit is not only the direct material saving but also the reduction in machining hours, tool wear, and production logistics.
From a process control point of view, the success of low-allowance sand mould casting depends on the interaction of several variables. The most important are pattern correction, mould rigidity, coating performance, gating design, pouring temperature, and heat treatment fixture. I summarized the key control variables and their effects in the following table.
| Control variable | Effect on low-allowance sand mould casting | Preferred selection | Measured outcome |
|---|---|---|---|
| Pattern material | Dimensional stability and wear resistance | Epoxy plastic | Low pattern deformation |
| Mould material | Mould cavity accuracy and collapsibility | Self-hardening water-glass sand | Good dimensional retention |
| Coating | Surface finish and resistance to metal washing | Zircon-based alcohol coating | Fewer coating inclusions |
| Gating design | Flow pattern, slag entrapment, surface defects | Top pouring with shaped riser | Scattered defects, high yield |
| Pouring temperature | Fluidity, gas escape, slag flotation | 1580 to 1620 degrees C, slightly high | Good filling and cleaner surface |
| Process correction | Compensation for casting distortion | Section-dependent correction | Distortion within allowance |
| Heat treatment fixture | Uniform heating and cooling | Side placement with gaps and angle | Reduced heat treatment distortion |
In addition to the practical variables, I used several quantitative relationships to guide the process. The total dimensional variation of the blade can be expressed as:
$$\sigma_{\mathrm{total}} = \sqrt{ \sigma_{\mathrm{pattern}}^2 + \sigma_{\mathrm{mould}}^2 + \sigma_{\mathrm{solidification}}^2 + \sigma_{\mathrm{heat\ treatment}}^2 + \sigma_{\mathrm{measurement}}^2 }$$
where each \(\sigma\) represents the standard deviation of the corresponding source of variation. To achieve low-allowance sand mould casting, \(\sigma_{\mathrm{total}}\) must be smaller than the allowance available for grinding and polishing. If the allowance is \(R_{\mathrm{total}}\), then:
$$3 \sigma_{\mathrm{total}} \leq R_{\mathrm{total}}$$
This condition ensures that almost all blades remain within the removable allowance. In my trials, the largest contributors were solidification distortion and heat treatment distortion. By improving the heat treatment fixture and applying section-dependent process correction, I reduced these two terms. The pattern and mould contributions were also controlled by using epoxy tooling and self-hardening water-glass sand.
The filling behavior of the molten steel can be described by a simplified fluidity criterion. The flow length \(L_f\) depends on superheat, channel thickness, and heat transfer to the mould:
$$L_f = \frac{\rho H_f t_{\mathrm{channel}}}{h \Delta T_{\mathrm{mould}}}$$
where \(\rho\) is the density, \(H_f\) is the latent heat, \(t_{\mathrm{channel}}\) is the local channel thickness, \(h\) is the heat transfer coefficient, and \(\Delta T_{\mathrm{mould}}\) is the temperature difference between the metal and the mould. For the thin exit edge, \(t_{\mathrm{channel}}\) is small, so \(L_f\) is limited. This is why bottom pouring produced defects at the exit edge and why the pouring temperature was kept slightly high. Top pouring with a shaped riser provided a favorable temperature gradient and helped fill the thin sections.
Solidification feeding can be assessed using the modulus \(M\):
$$M = \frac{V}{A}$$
where \(V\) is the volume of the section and \(A\) is its surface area. The thick sections have a larger modulus and therefore require feeding. The thin exit edge has a small modulus and solidifies quickly. The riser must be placed so that the feeding path remains open to the thick sections. In the top pouring design, the shaped riser at the large end provided feeding to the maximum thickness region. This avoided the shrinkage depression that occurred in bottom pouring.
The thermal stress during cooling can be approximated by:
$$\sigma_{\mathrm{thermal}} = E \alpha \Delta T_{\mathrm{diff}}$$
where \(E\) is the elastic modulus, \(\alpha\) is the thermal expansion coefficient, and \(\Delta T_{\mathrm{diff}}\) is the temperature difference between thick and thin sections. When \(\sigma_{\mathrm{thermal}}\) exceeds the yield strength at temperature, plastic deformation occurs. The blade then distorts. Process correction and uniform heat treatment reduce \(\Delta T_{\mathrm{diff}}\), and therefore reduce \(\sigma_{\mathrm{thermal}}\). This is the physical reason why the side-placement fixture worked.
The surface roughness after grinding and polishing can be related to the as-cast surface quality and the polishing allowance:
$$R_a = f \left( R_{a,\mathrm{cast}}, R_{\mathrm{allowance}}, P_{\mathrm{defect}} \right)$$
where \(R_{a,\mathrm{cast}}\) is the as-cast roughness, \(R_{\mathrm{allowance}}\) is the polishing allowance, and \(P_{\mathrm{defect}}\) is the probability of exposing a subsurface defect. Low-allowance sand mould casting must minimize \(P_{\mathrm{defect}}\) because only a small amount of material is removed. If a defect is located just below the surface, grinding may expose it and lead to rejection. Therefore, the combination of zircon coating, top pouring, and controlled pouring temperature was essential. These measures reduced inclusions and gas defects, so the probability of exposing a defect during grinding was lowered.
The relationship between inclusion size and allowable defect size can be written as:
$$d_{\mathrm{inclusion}} \leq d_{\mathrm{allow}}$$
where \(d_{\mathrm{inclusion}}\) is the maximum inclusion dimension and \(d_{\mathrm{allow}}\) is the maximum allowable defect dimension for the important region. In the improved process, the zircon coating and smoother filling reduced the size and frequency of inclusions. As a result, the inspected blades met the surface quality requirements.
I also considered the influence of the riser and ingate on yield. The casting yield \(Y\) can be defined as:
$$Y = \frac{W_{\mathrm{casting}}}{W_{\mathrm{casting}} + W_{\mathrm{riser}} + W_{\mathrm{gating}}} \times 100\%$$
In side pouring, the fin-like ingate and spherical blind riser increased the gating weight, and the yield was low. In top pouring, the shaped top riser served multiple functions and the gating system was simpler, so the yield was higher. In bottom pouring, the horn-shaped ingate and top riser gave a medium yield. Although yield is not the only criterion, it is important for economic production. The top pouring design gave the best combination of surface quality, feeding, and yield.
The final production route for low-allowance sand mould casting of the sabre-type blade can be summarized as follows. First, an epoxy pattern is made with section-dependent process correction. Second, a self-hardening water-glass sand mould is prepared, and a zircon-based alcohol coating is applied. Third, the mould is assembled in a horizontal orientation with vertical pouring. Fourth, molten martensitic stainless steel is melted in a medium-frequency induction furnace and poured at a controlled temperature slightly above the liquidus. Fifth, the metal enters through a shaped top riser at the large end. Sixth, after solidification and shakeout, the blade is cleaned and heat treated using a special support plate that holds the blade on its side with the exit edge down, the large end raised, and gaps between blades. Seventh, the blade is ground and polished to final dimensions. Eighth, the blade is inspected for thickness, exit edge condition, leak gap, and surface defects. If localized defects are found, they are repaired by controlled welding and re-polished. This route has produced blades that meet the design and acceptance requirements.
The significance of this work is that it demonstrated that low-allowance sand mould casting can be used for a complex, twisted, thin-walled stationary blade that would normally be considered a candidate for investment casting. The process reached a semi-precision casting level. It improved the foundry technology level of our facility and provided a successful example of precision-oriented casting production. The direct economic benefit was substantial. In one retrofit application, two units used the low-allowance sand mould cast sabre-type blades instead of the original stationary blades. The direct saving from raw material and machining cost was close to several million monetary units. By the end of the reporting period, many turbine units had been produced, and more units were expected to be retrofitted. Therefore, the economic benefit of this low-allowance sand mould casting process is expected to grow. If the process is extended to other stationary blades, the benefit will be even more significant.
Of course, the low-allowance sand mould casting process is not yet fully mature. Some problems still require further study and discussion. For example, the robustness of the process under varying ambient conditions, the influence of sand batch variation, and the long-term reproducibility of the heat treatment fixture need further investigation. The relationship between coating thickness and surface defect formation also deserves more quantitative study. I believe that continued work on these issues will improve the process further. The experience gained from this blade can be transferred to other complex castings where low-allowance sand mould casting is desired. The combination of pattern correction, mould rigidity, coating selection, gating design, pouring control, and heat treatment fixture design is the key to success. Each of these elements must be controlled within a narrow window, because the total allowance is small and there is little room for error.
In conclusion, I developed a low-allowance sand mould casting process for a sabre-type stationary blade. The process uses an epoxy pattern with section-dependent correction, self-hardening water-glass sand, zircon-based alcohol coating, top pouring with a shaped riser, controlled pouring temperature, and a side-placement heat treatment fixture. The process successfully controlled casting distortion, heat treatment distortion, surface defects, and dimensional variation. The finished blades met the maximum thickness tolerance, exit edge thickness requirement, profile leak gap requirement, and surface quality requirement. The process is stable, repeatable, and economically beneficial. It represents a practical step toward precision casting and shows that low-allowance sand mould casting can be applied to complex turbine blades. With further refinement, the same approach can be extended to other stationary blades and similar components, where it will provide additional savings in material, machining, and production time.