Thin-walled turbine guides concentrate several conflicting requirements in one component. The part must be geometrically accurate, internally sound, and resistant to high-temperature oxidation, thermal fatigue, and vibration. At the same time, its trailing edges are so thin that they are difficult to fill before the metal freezes, while the surrounding ring and flange sections cool slowly enough to promote directional grain growth. Investment casting is attractive here because it can form the ring, vanes, and flanges as one near-net-shape piece, but the process is only successful when the ceramic shell and the solidification path are designed together.
A representative case is the second-stage power-turbine guide of a helicopter engine. The guide has a maximum diameter of 525.9 mm, a height of 95.3 mm, and 58 solid vanes. Its longest trailing edge is about 100 mm, with an edge radius of only 0.35 mm. The alloy is K418B, a cast nickel-base superalloy strengthened by Al, Ti, Nb, and Mo and containing small additions of B, C, and Zr. The combination of a large, stiff ring structure and thin vane edges makes the part a useful example of how investment casting shifts from a shaping problem to a thermal-engineering problem.
Component and alloy constraints
The guide operates behind the combustor, where the gas stream is accelerated and redirected before it enters the turbine rotor. Its function is not only to withstand temperature but also to set the flow angle and throat area within a narrow range. A distorted ring or a locally blocked trailing edge changes the aerodynamic matching of the stage. The alloy therefore has to provide creep and fatigue resistance at temperature, while the casting process has to preserve the vane profile and the thin exit edge.
K418B is a conventional cast nickel-base superalloy of the type used for integrally cast turbine components. Its nominal composition is summarized in Table 1. The relatively high Al and Ti contents support gamma-prime precipitation, while Cr, Mo, and Nb contribute to high-temperature strength and environmental resistance. These alloying elements also make the solidification range wide and the melt reactive, so the face coat of the shell must be chemically stable and the pouring temperature must be controlled within a practical window.
| Element | Ni | Cr | B | Mo | Nb | Al | Ti | C | Zr |
|---|---|---|---|---|---|---|---|---|---|
| Content | Balance | 11.8 | 0.01 | 4.5 | 2.1 | 6.1 | 0.8 | 0.05 | 0.09 |
Why investment casting fits the guide
Investment casting builds the mold around a disposable pattern rather than machining a cavity into a solid block. A wax pattern is assembled into a tree, coated with ceramic slurry and stucco, dewaxed, and sintered. Molten metal is then poured into the resulting shell. The route is well suited to the guide because it can produce the integral ring-and-vane geometry without a long sequence of welding, electro-discharge machining, and mechanical finishing operations. It also allows the wall thickness and trailing-edge profile to be established by the pattern rather than by cutting tools.
The trade-off is that the shell is not a passive container. It controls heat extraction, constrains shrinkage, and vents gases displaced by the metal. A shell that is too insulating keeps the thin trailing edge open but may leave shrinkage porosity in slow-cooling sections. A shell that extracts heat too quickly fills the edge poorly and steepens the thermal gradient along the vane. The practical task is therefore to tailor the shell locally, not merely to increase or decrease its overall thickness.
Ceramic shell: strength, permeability, and corner behavior
The shell system used for this class of guide separates the face coat from the backup layers. The face coat is based on zircon (ZrSiO4) powder and a silica sol binder, with a small latex addition to improve wet strength and slurry rheology. Zircon offers good refractory behavior and low chemical reactivity toward nickel-base melts; the face slurry contains about 81% zircon powder and 17.32% silica sol, with the balance made up by latex, water, wetting agent, and antifoam. The backup slurry is based on WDS3 powder, a fused-silica composite containing needle-like petroleum coke, with a silica sol binder. The backup stucco is 35-mesh fused mullite, while the face stucco is 80–120-mesh zircon sand.
Slurry control is part of the metallurgical control. The face slurry is maintained at a flow-cup viscosity of 26–28 s, a pH of 9.6–10.5, and a density near 3.2 g/cm3; the backup slurry is held at 10–12 s and about 1.5 g/cm3. These ranges keep the coating thickness, drainage, and sand pickup repeatable. The shell is then dried, dewaxed in steam, and sintered before pouring.
The modified shell was compared with a baseline shell in terms of green strength, wet strength after steam dewaxing, high-temperature strength, residual strength, and air permeability. The results are given in Table 2. The modified shell has substantially higher dry and wet strength, which helps it survive handling and dewaxing, and much higher permeability at elevated temperature, which aids gas escape. Its high-temperature strength, however, is lower than that of the baseline shell, especially above 1000 °C. That difference is a genuine trade-off: the modified shell is not universally stronger, but it is more permeable and more robust in the wet state.
| Property | Baseline shell | Modified shell |
|---|---|---|
| Dry unfired bending strength (MPa) | 5.59 | 9.32 |
| Wet strength after steam dewaxing (MPa) | 4.25 | 7.10 |
| High-temperature strength at 800 °C (MPa) | 7.81 | 3.07 |
| High-temperature strength at 1000 °C (MPa) | 10.32 | 3.27 |
| High-temperature strength at 1200 °C (MPa) | 7.40 | 3.03 |
| High-temperature strength at 1400 °C (MPa) | Softened; not measurable | 2.63 |
| Residual strength after 1000 °C, cooled to room temperature (MPa) | 3.35 | 2.35 |
| Residual strength after 1200 °C, cooled to 400 °C (MPa) | 5.53 | 3.01 |
| Residual strength after 1200 °C, cooled to room temperature (MPa) | 1.85 | 1.55 |
| Air permeability at 20 °C (×10-14 m2) | 0.09 | 0.15 |
| Air permeability at 600 °C (×10-14 m2) | 1.51 | 3.43 |
| Air permeability at 800 °C (×10-14 m2) | 1.65 | 4.37 |
| Air permeability at 1000 °C (×10-14 m2) | 2.40 | 11.13 |
The residual-strength behavior reflects the phase changes in the shell. After sintering at 1200 °C and cooling to 400 °C, both shells retain more strength than after cooling to room temperature, because the transformation from alpha- to beta-cristobalite has not yet produced the same level of cracking. The drop on cooling to room temperature is therefore a materials effect, not simply a test artifact. The wedge-strength tests tell a related story: corners and sharp transitions receive less stucco per unit area and are therefore weaker than flat panels. The needle-like petroleum coke in the backup system improves coating pickup at these corners, but the corner remains a critical location for shell cracking during dewaxing and pouring.
Simulation model and gravity-casting behavior
The casting process was analyzed with ProCAST, using thermophysical properties calculated in JMatPro. The model represented the guide, the gating system, and a 7.5 mm shell, with tetrahedral elements of about 6 mm along the runners and automatic refinement on the casting surface. The gravity route used top pouring. The poured volume was 11,678,649 mm3, of which 2,671,109 mm3 was the gating system, giving a poured mass of about 95.76 kg.
During filling, metal entered through the sprue, divided into six vertical runners, filled the inner ring, passed outward through the vanes, and then filled the outer ring and risers. The last region to fill was the bend of the outer-ring runner. The flow was generally smooth, but the vane inlet reached velocities above 1.2 m/s, whereas most of the casting filled below 0.3 m/s. Since shell erosion becomes a concern above roughly 0.35–0.5 m/s, the inlet region carries a local risk of chemical and physical sand burn-on even when the bulk filling is calm.
Solidification was directional. The vanes began to freeze at about 260 s and were substantially solid by about 620 s, while the gate remained liquid beyond 1000 s. That sequence is desirable for feeding, because the gate and riser can supply liquid while the vanes contract. The problem is the internal gradient within each vane. The trailing edge is thin, and the adjacent shell is also thin, so heat is removed rapidly from the edge. The thicker airfoil and ring sections cool more slowly. The resulting temperature gradient points from the trailing edge toward the vane body and promotes columnar growth normal to the edge.
The same gradient explains the filling defect. In the initial production condition, the trailing edge could freeze before the last metal arrived, producing a misrun. Shrinkage porosity also appeared in the upper thin-wall region of the vanes and at the mid-face of the outer ring, in addition to the porosity that is normally accepted in the gating system. Changing the runner layout alone altered the local flow velocity but did not correct the temperature field, so it did not solve the coupled filling and grain-structure problem.
Trailing-edge insulation: targeting the thermal gradient
The effective correction was local rather than global. A 10 mm self-curing insulation layer was applied to the shell at the trailing-edge region. The layer was made from fused-silica powder, needle-like petroleum coke, silica sol, and ammonium chloride as the curing agent. After application, the layer was allowed to cure and was pre-calcined before pouring. In effect, the thin edge was given a local thermal reservoir, so that it no longer acted as a fin dissipating heat into the surrounding shell.
The process window was adjusted at the same time. Shell preheat was raised to 1150 °C, about 100 °C above the earlier condition, and the pouring temperature was reduced to 1560 °C, about 40 °C lower. The higher mold temperature helps keep the trailing edge open; the lower melt superheat reduces the risk of shrinkage porosity in sections that now cool more slowly because of the insulation. Top pouring was retained because the runner design itself had not been the controlling factor.
The trial castings showed complete filling of the trailing edges. The columnar zone perpendicular to the vane surface was greatly reduced and largely replaced by an equiaxed region. The limitation is equally important: columnar grains remained near the inner and outer flanges, where the local thermal mass and constraint are different from those at mid-vane. The insulation therefore mitigates the problem rather than eliminating it everywhere. Its value lies in demonstrating that the defect is driven by a local thermal boundary condition that can be changed without redesigning the entire gating system.
| Route | Key process condition | Observed response |
|---|---|---|
| Gravity, initial | Top gating; uninsulated trailing-edge shell | Trailing-edge misrun and coarse columnar grains normal to the edge |
| Gravity, insulated | 10 mm self-curing insulation; shell preheat 1150 °C; pour 1560 °C | Complete filling; columnar zone much reduced, with residual columnar grains near the flanges |
| Centrifugal | Sand-backed shell; preheat 980 °C for 12 h; pour 1540 °C; 150 rpm | Complete filling; fine equiaxed structure; porosity mainly in the gating system and outer-ring edge |
Centrifugal casting and grain refinement
Thermal insulation addresses the gradient but does not mechanically disturb the growing dendrites. A second route used a vertical-axis centrifugal casting machine: the table was horizontal, the rotation axis was vertical, and the mold assembly rotated at 150 rpm. The shell was embedded in a sand box filled with crushed coal-gangue sand below 3 mm, preheated at 980 °C for 12 h, and transferred to the vacuum chamber for pouring at 1540 °C. After pouring, rotation continued for 1 min before the table was allowed to stop.
The centrifugal filling simulation used six inner runners and eight outer runners. Metal first entered the inner ring through the inner runners, then moved outward through the vanes while the outer runners supplied additional metal. Some trailing edges were temporarily unfilled during the early stage, but the final confluence completed the casting. The velocity field, however, was much more aggressive than in gravity casting. The outer ends exceeded 4 m/s, the mid-radius region was about 2 m/s, and the center was about 1.2 m/s. In this route, shell erosion rather than incomplete filling becomes the limiting filling constraint.
Solidification remained nonuniform. The outer ring developed an isolated liquid pool because of its thick bottom section, and the riser was the last region to freeze. The vanes solidified in roughly half the time required by the outer ring. At the vane scale, the trailing edge began to freeze at about 5 s, the middle region within about 40 s, and the two ends after more than 40 s. A steep gradient therefore still existed. The difference is that the rotating field was intended to fragment the dendrite network mechanically, distribute the fragments through the liquid, and hinder further growth. Under those conditions, the final structure is favored to be equiaxed rather than columnar.
Macro-etched sections of the trial casting supported that expectation. The columnar grains perpendicular to the vane surface were almost absent, and the surface structure was predominantly fine equiaxed. The result is consistent with dendrite fragmentation and restricted growth, although the centrifugal field also improves feeding and changes the local flow history. The practical significance is that the guide could be filled completely while avoiding the continuous columnar structure that dominated the gravity-cast condition.
Centrifugal casting is not a drop-in replacement for gravity investment casting. It requires a rotating vacuum machine, a sand-backed shell, careful balance, and a shell strong enough to resist the higher metal velocities. The process also changes the filling signature: a route that solves a misrun by increasing velocity can create a shell-erosion or inclusion problem if the shell is not designed for it. The insulation route and the centrifugal route therefore address different parts of the same problem. Insulation changes the local thermal gradient and protects the thin edge; centrifugal motion disrupts dendrite growth and promotes a fine equiaxed structure.
Engineering implications
The case shows that investment casting of thin-walled nickel-base guides should be treated as a coupled design problem. The ceramic shell is a thermal component, not merely a refractory container. Face-coat chemistry controls reaction with the melt; backup composition and stucco size control strength and permeability; local shell thickness controls the temperature gradient that determines whether the trailing edge fills and whether the vane solidifies as columnar or equiaxed. Simulation is useful because it exposes those couplings before a trial, but the decisive variables are still the local boundary conditions: shell mass, preheat, pouring temperature, and the presence or absence of a rotating field.
For production, the insulation approach is the simpler intervention. It can be added to an existing gravity-casting route, it directly targets the trailing-edge gradient, and it preserves the established gating design. Its weakness is that the flange regions remain difficult to protect, so some columnar structure can persist. The centrifugal approach is more powerful for grain refinement, but it raises the equipment and shell-performance requirements and shifts the limiting defect from misrun to shell erosion and process control. A robust process development sequence would therefore use gravity casting with local thermal control to establish filling and soundness, then apply centrifugal casting only where the grain structure or fatigue requirement justifies the additional complexity.