In my work on gas turbine blade manufacturing, I treated rapid investment casting as an integrated full mold casting problem rather than as a sequence of isolated operations. The term full mold casting is used here to describe the complete sacrificial-pattern mold assembly: the printed blade pattern, the wax gating system, the ceramic shell, the backup layers, and the final poured metal. When the entire full mold casting assembly is heated, de-waxed, burned out, and filled, every component interacts thermally and mechanically. My objective was to replace the traditional trial-and-error method for determining die dimensions with a faster, lower-cost full mold casting route based on additive manufacturing, deformation coordination, and finite element analysis.
A conventional investment casting cycle usually follows a fixed but expensive sequence: preliminary process design, die manufacturing, wax pattern injection, pattern assembly, shell building, de-waxing, shell burning, pouring, inspection, and rework. If the first casting fails, the die must be corrected, and another complete full mold casting trial is required. In many cases, one to three months are needed before a stable process is reached. That long cycle is not only costly but also slows down design iteration for gas turbine blades. Because blades have complex airfoil surfaces, hollow internal passages, thin trailing edges, and thick root sections, the full mold casting mold must accommodate non-uniform expansion, contraction, and thermal gradients.
My rapid full mold casting route used a fused deposition modeling printer to create an ABS blade pattern. The ABS pattern was bonded to a conventional wax gating system to form the investment mold cluster. After five sand stucco layers, the cluster was de-waxed, the ABS pattern was burned out, and the ceramic shell was fired. Molten stainless steel was then poured into the fired shell. The first trials produced blade castings, but the shell cracked during processing, especially near the leading and trailing edges. The resulting blades showed surface flash and dimensional deviation. I therefore established a deformation coordination model and used finite element analysis to identify the stress concentration mechanism in the full mold casting shell.
The first observation I made was that the traditional full mold casting die-sizing problem is a shrinkage-compensation problem, but the rapid full mold casting problem is a coupled pattern-shell deformation problem. In conventional wax injection, the pattern is relatively compliant and can deform or be removed with less constraint. In my rapid route, the ABS pattern remained inside the shell until burn-out, and its thermal expansion imposed a local contact pressure on the green ceramic shell. The shell is brittle in the green and early bisque states, so tensile stress at high-curvature airfoil edges can easily exceed its strength. This is why the full mold casting shell cracked even though the layer count and slurry formulation were initially considered acceptable.
To describe the interaction, I simplified the blade cross-section as a circular ring problem. The ABS pattern was treated as an inner cylinder with inner radius \(a\) and outer radius \(b\). The ceramic shell was treated as an outer cylinder with thickness \(x\). Before heating, the two bodies were assumed to be in intimate contact. When the temperature increased from \(t_1\) to \(t_2\), the temperature change was
$$ \Delta t = t_2 – t_1 $$
The thermal expansion of the ABS pattern would normally increase its outer circumference, but the ceramic shell constrains that expansion. Therefore, the pattern is compressed inward, while the shell is pushed outward. The radial interference and contact pressure produce a deformation coordination condition. For the pattern, the net circumferential deformation is the free thermal expansion minus the elastic compression:
$$ X_1 = 2\pi b \alpha_1 \Delta t – \frac{2\pi b \sigma_1}{E_1} $$
For the shell, the net circumferential deformation is the free thermal expansion plus the elastic extension caused by the pattern:
$$ X_2 = 2\pi b \alpha_2 \Delta t + \frac{2\pi b \sigma_2}{E_2} $$
At the contact interface, the pattern and shell must remain compatible as long as no separation or slip occurs. Thus, the deformation coordination equation is
$$ X_1 = X_2 $$
Substituting the two deformation expressions gives the full mold casting compatibility equation:
$$ 2\pi b \alpha_1 \Delta t – \frac{2\pi b \sigma_1}{E_1} = 2\pi b \alpha_2 \Delta t + \frac{2\pi b \sigma_2}{E_2} $$
After canceling \(2\pi b\), the equation becomes
$$ \alpha_1 \Delta t – \frac{\sigma_1}{E_1} = \alpha_2 \Delta t + \frac{\sigma_2}{E_2} $$
This expression shows why the full mold casting shell stress depends strongly on the mismatch between the thermal expansion coefficients of the pattern and the shell. If \(\alpha_1\) is much larger than \(\alpha_2\), then the left side of the equation tends to be larger, and the shell must develop greater tensile stress \(\sigma_2\) to maintain compatibility. In my case, the ABS pattern had a thermal expansion coefficient far greater than that of the ceramic shell, so the shell experienced a large expansion-driven load during heating and burn-out.
The stress-strain relation for thermoelastic response can be written as
$$ \sigma = E \left( \epsilon – \alpha \Delta T \right) $$
For a fully constrained body, the mechanical strain is zero, and the thermal stress becomes
$$ \sigma = -E \alpha \Delta T $$
In the full mold casting shell, complete constraint does not occur everywhere, but the contact at the airfoil surface creates a constraint that approaches this condition locally. The resulting stress is especially severe at the leading edge and trailing edge because those regions have small curvature radii and thin cross-sections. The stress concentration factor can be approximated as
$$ K_t = 1 + 2 \frac{t_s}{r_c} $$
where \(t_s\) is the local shell thickness and \(r_c\) is the local radius of curvature. When \(r_c\) is small, \(K_t\) becomes large, and the full mold casting shell is more likely to crack.
I also used the transient heat conduction equation to represent the temperature field during burn-out and pouring:
$$ \rho c_p \frac{\partial T}{\partial t} = \nabla \cdot \left( k \nabla T \right) + q $$
Here, \(\rho\) is density, \(c_p\) is specific heat, \(k\) is thermal conductivity, \(T\) is temperature, \(t\) is time, and \(q\) is the internal heat source. In the full mold casting shell, the temperature gradient through the thickness is not uniform. The outer layers heat more slowly than the inner layers, and the ABS pattern burns away at a different rate than the wax gating system. This non-uniform thermal history adds to the mechanical mismatch and promotes cracking.
The equivalent von Mises stress used in my finite element post-processing was
$$ \sigma_{eq} = \sqrt{\frac{1}{2}\left[ \left(\sigma_1-\sigma_2\right)^2 + \left(\sigma_2-\sigma_3\right)^2 + \left(\sigma_3-\sigma_1\right)^2 \right]} $$
This scalar value allowed me to compare the severity of stress at different locations in the full mold casting shell, even when the stress state was multiaxial. The maximum equivalent stress appeared at the leading and trailing edges of the blade section, which matched the observed crack locations.
For the finite element model, I created two-dimensional cross-sections of the blade pattern and shell. The contact between the ABS pattern and the ceramic shell was defined as bonded contact, because in the early heating stage the shell remains in close contact with the pattern. I used four-node PLANE13 elements, which support thermal-structural coupling. The mesh was refined near the leading edge, trailing edge, and root fillet. I assumed the assembly was heated from room temperature \(22^\circ\mathrm{C}\) toward \(200^\circ\mathrm{C}\), and I evaluated the stress state at \(150^\circ\mathrm{C}\) because that temperature range is critical for ABS softening, expansion, and early shell stress development. The temperature history was represented as
$$ K(t) = 22 + at $$
where \(K(t)\) is the temperature at time \(t\), and \(a\) is a heating-rate constant. For the finite element solution, I used a static structural system with thermal loading rather than a fully coupled transient system, because the purpose was to compare stress states and identify the controlling geometric variables rather than to predict the exact failure time. This simplification was sufficient for ranking the full mold casting design alternatives.
The material properties I used in the finite element analysis are summarized in the table below. These values represent the effective behavior at the evaluation temperature and were used consistently across all cases.
| Material | Elastic modulus | Poisson ratio | Thermal expansion coefficient |
|---|---|---|---|
| ABS pattern | 1.0 MPa | 0.43 | \(92.0 \times 10^{-6} /^\circ\mathrm{C}\) |
| Ceramic shell | 630 MPa | 0.26 | \(4.0 \times 10^{-6} /^\circ\mathrm{C}\) |
The difference in thermal expansion coefficient is more than twenty times. This is the central reason why the full mold casting shell experiences high stress when the printed ABS pattern expands inside it. The shell has low thermal expansion and relatively high stiffness, so it resists the pattern expansion and accumulates tensile stress. When the local curvature is sharp, the stress concentration becomes large enough to initiate cracks.
I compared several full mold casting configurations. The first was a solid ABS pattern with a 6 mm shell. The second was a hollow ABS pattern with a 6 mm shell. The third was a hollow ABS pattern with a low-density lattice interior and a 6 mm shell. The fourth increased the shell thickness to 8 mm. The fifth combined a lattice interior, an 8.5 mm shell, and enlarged leading- and trailing-edge radii. The normalized maximum equivalent stress results are listed in the following table. The values are normalized to the original solid-pattern, 6 mm shell case.
| Case | Pattern core | Shell thickness | Edge radius treatment | Normalized maximum equivalent stress |
|---|---|---|---|---|
| A | Solid ABS | 6.0 mm | Original sharp edges | 1.00 |
| B | Hollow ABS | 6.0 mm | Original sharp edges | 0.86 |
| C | Hollow ABS with lattice | 6.0 mm | Original sharp edges | 0.71 |
| D | Hollow ABS with lattice | 8.0 mm | Original sharp edges | 0.58 |
| E | Hollow ABS with lattice | 8.5 mm | Enlarged radii | 0.39 |
The finite element results showed that increasing shell thickness alone reduced stress, but not enough to guarantee crack-free full mold casting. The largest improvement came from combining three actions: reducing the effective stiffness and mass of the ABS pattern, increasing the shell thickness, and increasing the edge radii. The edge radius change was particularly effective because it reduced the stress concentration factor. Instead of a sharp leading edge acting as a crack initiator, a larger radius distributed the contact pressure over a wider area.
The original shell-building schedule used five layers. The first layer was a fine face coat, and subsequent layers increased the sand particle size to build backup thickness. The slurry and stucco formulation I used is summarized in the following table. In the optimized full mold casting route, I changed the total from five layers to six and a half layers while keeping the same binder, powder, and sand chemistry.
| Layer | Binder | Powder | Stucco sand | Sand mesh |
|---|---|---|---|---|
| 1 | Silica sol | White corundum powder | Corundum | 70 |
| 2 | Silica sol | White corundum powder | Corundum | 30/60 |
| 3 | Silica sol | Calcined clay powder | Coal gangue sand | 30/60 |
| 4 | Silica sol | Calcined clay powder | Coal gangue sand | 16/30 |
| 5 | Silica sol | Calcined clay powder | Coal gangue sand | 16/30 |
| 6 and sealing half-layer | Silica sol | Calcined clay powder | Coal gangue sand | 16/30 |
The drying process was also important. After each slurry and stucco application, I placed the full mold casting cluster in a controlled temperature and humidity room. If the layer dried too quickly, shrinkage cracks formed before the next layer could reinforce the shell. If it dried too slowly, the green strength remained low for too long. I maintained a constant drying environment and verified that each layer reached a leather-hard condition before applying the next one. This was especially necessary for the face coat, because the face coat replicates the blade surface and any crack in it becomes a surface defect in the final full mold casting.
The de-waxing and burn-out stages were the most damaging to the shell. The wax gating system melted and drained first, while the ABS blade pattern remained until a higher temperature. As the ABS softened, expanded, and then decomposed, it pushed against the inner shell surface. At the same time, the shell was heating non-uniformly. The combination of mechanical loading and thermal gradient produced the cracks observed in the first trials. The burn-out temperature range was \(970^\circ\mathrm{C}\) to \(1030^\circ\mathrm{C}\). When the temperature was too low, the shell did not develop sufficient strength, and it cracked during pouring. When the temperature was too high, the shell deformed and the casting dimensions drifted. I therefore used a stepped burn-out schedule with a hold at intermediate temperature to allow the ABS to decompose gradually before the shell reached its maximum firing temperature.
The thermal schedule I adopted for the optimized full mold casting trials is shown below. The ramp rates were selected to reduce the instantaneous thermal gradient through the shell thickness.
| Stage | Temperature range | Purpose | Control variable |
|---|---|---|---|
| Heating stage 1 | 22 to 150°C | Controlled ABS softening and initial wax drainage | Low ramp rate |
| Heating stage 2 | 150 to 300°C | ABS decomposition and wax removal | Hold at 300°C |
| Heating stage 3 | 300 to 600°C | Carbon removal and shell strengthening | Moderate ramp |
| Heating stage 4 | 600 to 970°C | Shell sintering | Uniform furnace atmosphere |
| Burn-out hold | 970 to 1030°C | Full mold casting shell consolidation | Hold time and maximum temperature |
| Cooling | 1030°C to pouring temperature | Thermal equalization before pour | Controlled cooling |
After optimizing the full mold casting geometry and shell schedule, I repeated the experiment. The shell remained intact through de-waxing, burn-out, and pouring. The final blade castings had no through cracks, no major flash, and much smaller dimensional deviation. The table below compares the original and optimized full mold casting trials.
| Quality characteristic | Original full mold casting trial | Optimized full mold casting trial |
|---|---|---|
| Shell cracking | Multiple cracks at leading and trailing edges | No through cracks |
| Surface flash | Severe flash on airfoil surfaces | Minimal flash |
| Blade dimensional deviation | Large, requiring rework | Small, within trial tolerance |
| Shell integrity after burn-out | Broken and patched before pour | Intact and ready for pour |
| Casting result | Defective blade with surface and dimension problems | Sound blade casting |
The optimized full mold casting route also changed the ABS pattern design. Instead of printing a solid blade, I used a sparse internal lattice. The lattice reduced the apparent stiffness and the amount of material that had to expand and decompose. The outer skin of the ABS pattern still reproduced the blade shape, but the interior was not a solid mass. This reduced the contact pressure on the shell during heating. The lattice also shortened the burn-out time because the internal material could escape more easily. In the full mold casting context, the pattern is not merely a shape carrier; it is a thermal and mechanical load source until it disappears.
I also modified the blade model by increasing the leading-edge and trailing-edge radii. This was not a functional airfoil change in the final part; it was a pattern and mold-design adjustment to reduce the full mold casting stress concentration. A larger edge radius reduces the local curvature and distributes the contact pressure. The improvement can be estimated using the stress concentration expression
$$ K_t = 1 + 2 \frac{t_s}{r_c} $$
If \(r_c\) increases while \(t_s\) is held constant, \(K_t\) decreases. The finite element results confirmed that the maximum equivalent stress shifted away from the edge and became more uniformly distributed through the shell. This is one of the most important practical findings from my full mold casting study: small geometry changes at high-curvature features can have a larger effect than simply adding more shell layers.
The gating system also influenced the full mold casting result. I attached the printed blade pattern to a wax runner and sprue. The metal was introduced through the blade root, because the root is the thickest section and can feed the thin airfoil. A root-fed full mold casting reduces turbulence in the thin trailing edge and helps directional solidification toward the gating. The gating design variables included sprue diameter, runner length, gate cross-section, and the number of gates. I kept the gating system simple for the rapid trials, because the primary objective was to eliminate shell cracking and validate the full mold casting route, not to optimize yield. Once the shell remained intact, the gating system could be refined using the same finite element and deformation models.
Solidification shrinkage is another factor that must be considered in any full mold casting process. The volumetric shrinkage of the metal can be represented as
$$ \epsilon_V = \frac{\Delta V}{V_0} $$
and the linear shrinkage as
$$ \epsilon_L \approx \frac{1}{3} \epsilon_V $$
For a complex blade, the actual shrinkage is not isotropic because the root, shroud, airfoil, and internal passages cool at different rates. The full mold casting shell also constrains shrinkage. If the shell is too stiff and the alloy contracts strongly, hot tears can form. My optimization therefore balanced two opposite requirements: the shell had to be strong enough to resist pattern expansion and pouring pressure, but not so stiff that it prevented normal metal contraction. The six-and-a-half-layer shell provided this balance better than the original five-layer shell.
The solidification time can be approximated by Chvorinov’s rule:
$$ t_s = B \left( \frac{V}{A} \right)^n $$
where \(t_s\) is solidification time, \(V\) is casting volume, \(A\) is surface area, \(B\) is a mold constant, and \(n\) is typically close to 2. In the full mold casting blade, the root has a low surface-area-to-volume ratio and therefore solidifies slowly, while the trailing edge has a high surface-area-to-volume ratio and solidifies quickly. This difference creates thermal gradients and can produce shrinkage porosity if the gating does not feed the thin sections properly. The root-fed design helped maintain a favorable thermal gradient from the trailing edge toward the root.
I also evaluated the shell stress under the assumption of an effective contact pressure \(p\). The hoop stress in a thin cylindrical shell can be estimated as
$$ \sigma_{\theta} = \frac{p r}{t_s} $$
where \(r\) is the local radius and \(t_s\) is the shell thickness. This simplified formula shows why thin shells and small radii are dangerous: both increase hoop stress. In the full mold casting blade, the leading and trailing edges have small local radii, so they experience higher hoop stress than the mid-chord region. Increasing the edge radius or increasing the shell thickness reduces this stress. The finite element model captured the same trend.
For a more complete thermo-mechanical description, the total strain can be decomposed as
$$ \epsilon_{ij} = \epsilon_{ij}^{e} + \epsilon_{ij}^{th} + \epsilon_{ij}^{c} $$
where \(\epsilon_{ij}^{e}\) is elastic strain, \(\epsilon_{ij}^{th}\) is thermal strain, and \(\epsilon_{ij}^{c}\) is creep or viscoelastic strain. At the temperatures used in de-waxing and early burn-out, the ABS pattern can creep, while the green shell may have some viscoelastic behavior. A full transient viscoelastic model would be more accurate, but for process optimization the elastic-thermal model was sufficient to identify the critical locations and rank the design changes.
The general thermoelastic constitutive law can be written as
$$ \sigma_{ij} = C_{ijkl} \left( \epsilon_{kl} – \alpha \Delta T \delta_{kl} \right) $$
This equation is useful because it separates the mechanical strain from the thermal strain. In the full mold casting shell, the thermal strain is small because \(\alpha_2\) is small, but the mechanical strain imposed by the pattern is large. The shell therefore carries most of the compatibility strain as elastic stress. In the ABS pattern, the thermal strain is large, but the mechanical strain is compressive because the shell pushes back. The pattern may yield or creep, which relieves some stress, but not before the shell has already experienced the peak load.
The maximum principal stress criterion is often used for brittle ceramic shells:
$$ \sigma_1 \le \sigma_f $$
where \(\sigma_1\) is the maximum principal stress and \(\sigma_f\) is the flexural strength of the shell. If this condition is violated, a crack initiates. In my original full mold casting trial, the computed maximum principal stress exceeded the estimated green strength at the leading and trailing edges. After optimization, the maximum principal stress fell below the strength threshold. This explains why the optimized shell survived.
I also considered the Weibull probability of failure for the ceramic shell:
$$ P_f = 1 – \exp \left[ – \int_V \left( \frac{\sigma}{\sigma_0} \right)^m dV \right] $$
where \(P_f\) is the failure probability, \(\sigma_0\) is a scale parameter, and \(m\) is the Weibull modulus. This probabilistic view is useful because ceramic shells are brittle and contain flaws. Even if the average stress is low, a local high-stress region can initiate failure. The full mold casting optimization therefore focused on reducing peak stress, not only average stress. Enlarging the edge radii and using a compliant lattice pattern reduced the volume of highly stressed shell material.
The thermal expansion mismatch can be expressed as a mismatch strain:
$$ \epsilon_{mismatch} = \left( \alpha_1 – \alpha_2 \right) \Delta T $$
For my materials, \(\alpha_1 = 92.0 \times 10^{-6} /^\circ\mathrm{C}\) and \(\alpha_2 = 4.0 \times 10^{-6} /^\circ\mathrm{C}\). At \(\Delta T = 128^\circ\mathrm{C}\), which corresponds to heating from \(22^\circ\mathrm{C}\) to \(150^\circ\mathrm{C}\), the mismatch strain is
$$ \epsilon_{mismatch} = \left( 92.0 – 4.0 \right) \times 10^{-6} \times 128 \approx 0.0113 $$
This is a large strain for a brittle shell. If the shell were fully constrained, the elastic stress would be approximately
$$ \sigma = E_2 \epsilon_{mismatch} \approx 630 \times 0.0113 \approx 7.1\,\mathrm{MPa} $$
This value is consistent with the finite element results and with the observed cracking. It also shows why the original full mold casting shell failed. The mismatch strain was simply too high for the local shell strength at the sharp edges. Reducing the effective constraint by using a lattice pattern and increasing the edge radii lowered the local stress to a tolerable level.
The full mold casting process also depends on slurry viscosity and coating thickness. The viscosity of the slurry affects the amount of liquid that remains on the pattern and therefore the green thickness of each layer. I controlled viscosity by measuring the flow cup time before each coating. If the slurry was too viscous, the face coat was thick and prone to cracking during drying. If it was too thin, the face coat did not reproduce the blade surface and the shell lacked strength. The coating thickness per layer can be approximated as
$$ t_l = \frac{\eta Q}{\rho g A} $$
where \(t_l\) is layer thickness, \(\eta\) is effective viscosity, \(Q\) is drainage rate, \(\rho\) is density, \(g\) is gravitational acceleration, and \(A\) is coated area. This equation is qualitative, but it reminds me that slurry rheology is a first-order variable in full mold casting. The optimized route kept the same slurry chemistry but tightened the viscosity and drying controls.
The sand particle size also affects shell strength and permeability. Fine sand produces a smoother surface but lower permeability. Coarse sand increases permeability but can create a rougher surface and weaker interlaminar bonding. The layer sequence I used started with fine corundum and gradually changed to coarser coal gangue sand. This graded structure is standard in investment casting, but in rapid full mold casting it also helps accommodate the thermal mismatch. The fine face coat provides surface quality, while the coarse backup layers provide mechanical support and allow gases to escape during burn-out and pouring.
The permeability of the shell can be represented qualitatively by the Kozeny-Carman relation:
$$ k = \frac{d_p^2}{180} \frac{\epsilon^3}{(1-\epsilon)^2} $$
where \(k\) is permeability, \(d_p\) is particle diameter, and \(\epsilon\) is porosity. Increasing particle diameter or porosity increases permeability. This is beneficial for gas removal, especially when the ABS pattern decomposes and produces volatiles. However, higher porosity can reduce shell strength. The optimized full mold casting shell therefore used a graded structure rather than a uniform coarse or uniform fine shell.
Another critical variable was the metal pouring temperature. If the pouring temperature was too low, the metal solidified before filling the thin trailing edge. If it was too high, the shell was exposed to greater thermal shock and the metal could react with the face coat. I used a pouring temperature that provided adequate fluidity while minimizing shell thermal shock. The pouring temperature range for the stainless steel trial was selected based on superheat above the liquidus:
$$ T_p = T_l + \Delta T_{superheat} $$
where \(T_p\) is pouring temperature, \(T_l\) is liquidus temperature, and \(\Delta T_{superheat}\) is superheat. The full mold casting shell must withstand this thermal load immediately after burn-out. A preheated shell reduces the thermal gradient and lowers the risk of thermal shock. In the optimized trials, the shell was cooled to a controlled pouring temperature rather than being poured directly from the maximum burn-out temperature.
The filling behavior can be described by Bernoulli’s equation for the gating system:
$$ P_1 + \frac{1}{2}\rho v_1^2 + \rho g h_1 = P_2 + \frac{1}{2}\rho v_2^2 + \rho g h_2 + \Delta P_{loss} $$
Here, \(P\) is pressure, \(v\) is velocity, \(h\) is height, and \(\Delta P_{loss}\) represents friction and expansion losses. A well-designed full mold casting gating system minimizes turbulence and avoids air entrapment. In my root-fed design, the metal entered the thick root and filled the airfoil upward. This reduced the risk of cold shuts at the trailing edge and helped push gases out through the permeable shell.
The Reynolds number of the metal flow can be written as
$$ Re = \frac{\rho v D}{\mu} $$
where \(D\) is a characteristic channel diameter and \(\mu\) is dynamic viscosity. High Reynolds number indicates turbulent flow, which can entrain gas and oxide films. In the full mold casting of a thin blade, some turbulence is unavoidable, but a tapered sprue and a smooth runner system reduce it. I kept the gating cross-sections larger than the blade root gates to maintain a positive pressure gradient during filling.
The heat transfer during pouring and solidification can be modeled with the Stefan condition at the solidification front:
$$ k_s \frac{\partial T_s}{\partial n} – k_l \frac{\partial T_l}{\partial n} = \rho L \frac{\partial s}{\partial t} $$
where \(k_s\) and \(k_l\) are thermal conductivities of solid and liquid, \(L\) is latent heat, and \(s\) is the solidification front position. This equation shows that latent heat release slows the solidification front and can create mushy zones in the blade root. The full mold casting shell acts as a heat sink and influences the cooling rate. The thicker optimized shell changed the cooling curve slightly, but the main benefit was mechanical integrity rather than heat transfer.
I also recorded the casting dimensions after the optimized full mold casting trials. The table below summarizes the dimensional improvements observed in the trial blade. These values are relative to the original trial and are intended to show the process trend rather than a production tolerance.
| Feature | Original deviation | Optimized deviation | Improvement |
|---|---|---|---|
| Airfoil chord | Large positive deviation | Small deviation | Reduced by about 75% |
| Airfoil thickness | Local bulging | Uniform thickness | Surface flash eliminated |
| Leading-edge radius | Rounded by flash | Defined and smooth | Improved edge fidelity |
| Trailing-edge straightness | Warped | Straight within trial tolerance | Improved shell support |
| Root dimensions | Acceptable but requires rework | Acceptable | Reduced rework |
The full mold casting optimization actions are summarized in the following table. Each action addressed a specific cause identified by the deformation model and finite element results.
| Problem | Root cause | Optimization action | Effect on full mold casting |
|---|---|---|---|
| Shell cracks at leading edge | High stress concentration and pattern expansion | Enlarge leading-edge radius | Lower \(K_t\), reduced peak stress |
| Shell cracks at trailing edge | Small radius and thin shell | Enlarge trailing-edge radius and increase shell thickness | More uniform stress distribution |
| High contact pressure from ABS | Solid ABS expands as a rigid mass | Use sparse lattice internal structure | Lower effective stiffness and lower pressure |
| Low shell strength after burn-out | Five layers insufficient | Increase to six and a half layers | Higher green and fired strength |
| Thermal gradient during burn-out | Fast heating and decomposition | Stepped burn-out schedule | Lower transient thermal stress |
| Gas entrapment and flash | Poor permeability and sealing | Graded stucco and controlled drying | Better gas escape, less flash |
The finite element analysis also allowed me to compare shell thickness and edge radius quantitatively. The table below presents a process window derived from the simulations. The values are expressed as qualitative ranges because the exact shell strength depends on slurry formulation, drying, and firing. The purpose is to show the direction of optimization for full mold casting.
| Shell thickness | Edge radius | Pattern core | Predicted crack risk | Recommended full mold casting action |
|---|---|---|---|---|
| Thin | Small | Solid | Very high | Avoid |
| Thin | Large | Solid | High | Modify pattern core |
| Medium | Small | Lattice | Medium | Increase edge radius |
| Medium | Large | Lattice | Low | Acceptable |
| Thick | Large | Lattice | Very low | Preferred |
One important lesson from the full mold casting trials is that the pattern material and the shell material must be considered as a single thermo-mechanical system. It is not sufficient to select a shell thickness based only on pouring pressure or handling strength. The pattern must also be designed to reduce the load it imposes on the shell. In my case, the ABS pattern was the source of the largest thermal expansion mismatch. By changing the internal structure from solid to lattice, I reduced the effective expansion force without changing the external blade geometry. This is a design degree of freedom that is unique to additive manufacturing and is highly valuable in rapid full mold casting.
The full mold casting route also eliminated the need for a metal die. In the traditional process, the die determines the wax pattern dimensions, and any change in shrinkage requires a die correction. In the rapid route, the pattern is printed directly from a digital model. If the casting shrinks differently than expected, the digital model can be scaled and reprinted without cutting a new die. This is the main economic advantage of rapid full mold casting for gas turbine blade development. The table below compares the traditional and rapid routes.
| Aspect | Traditional investment casting | Rapid full mold casting |
|---|---|---|
| Pattern tooling | Metal die required | No metal die required |
| Pattern material | Wax | ABS or other printable polymer |
| Design change response | Die correction, long lead time | Digital model change, reprint |
| Typical iteration time | Weeks to months | Days |
| Cost for low volume | High | Low to moderate |
| Main process risk | Die dimensions and wax shrinkage | Pattern-shell thermal mismatch |
| Optimization method | Experience and trial | Finite element and deformation coordination |
I also compared the full mold casting process before and after optimization in terms of shell stress. The original design had a highly localized stress concentration at the edges. The optimized design reduced the peak stress and moved the maximum stress away from the shell surface. The following table lists the relative stress reduction by location.
| Location | Original normalized stress | Optimized normalized stress | Reduction |
|---|---|---|---|
| Leading edge | 1.00 | 0.38 | 62% |
| Trailing edge | 0.96 | 0.41 | 57% |
| Mid-chord suction side | 0.62 | 0.35 | 44% |
| Mid-chord pressure side | 0.58 | 0.33 | 43% |
| Root fillet | 0.71 | 0.42 | 41% |
The full mold casting shell cracking problem was therefore not solved by a single change. It was solved by a coordinated set of changes that reduced the driving force, reduced the stress concentration, and increased the shell resistance. The deformation coordination equation provided the physical explanation, and the finite element model provided the spatial distribution. Together, they allowed me to redesign the process before making another full mold casting trial.
The optimized full mold casting experiment produced a sound blade casting. The shell survived the complete cycle, including de-waxing, ABS burn-out, firing, and metal pouring. The blade surface was clean, the edges were well defined, and the dimensions were closer to the digital model. The successful result confirmed that rapid full mold casting can be used for gas turbine blade development when the pattern-shell interaction is properly controlled.
For future work, I would extend the model in several ways. First, I would add temperature-dependent material properties for both the ABS pattern and the ceramic shell. Second, I would include viscoelastic and creep effects for the polymer during softening. Third, I would use a fully coupled transient thermal-structural analysis rather than a static thermal load. Fourth, I would model the burn-out atmosphere and the decomposition kinetics of the ABS pattern. Fifth, I would optimize the gating system using casting simulation to reduce porosity and improve yield. Sixth, I would validate the full mold casting model with embedded thermocouples and strain gauges in a trial shell. These improvements would make the full mold casting route even more predictive and reduce the remaining trial iterations.
In summary, my rapid full mold casting study showed that the traditional die-sizing problem can be replaced by a digital pattern and a physics-based shell design. The deformation coordination equation revealed the importance of thermal expansion mismatch. The finite element analysis identified the leading and trailing edges as the critical crack locations. The optimization actions—enlarged edge radii, a lattice ABS core, a thicker shell, and a stepped burn-out schedule—produced a crack-free full mold casting shell and a sound gas turbine blade. The full mold casting route is therefore technically feasible and economically attractive for low-volume, high-complexity blade development. It shortens the iteration loop, reduces tooling cost, and provides a practical path from digital design to functional casting.
| Full mold casting parameter | Original setting | Optimized setting | Reason for change |
|---|---|---|---|
| Pattern material | ABS | ABS | Printable and burn-out capable |
| Pattern core | Solid | Sparse lattice | Reduce effective stiffness and expansion force |
| Edge radius | Original sharp geometry | Enlarged leading and trailing radii | Reduce stress concentration |
| Shell layers | 5 | 6.5 | Increase shell strength |
| Burn-out schedule | Direct heating | Stepped heating with holds | Reduce thermal gradient |
| Gating entry | Root entry | Root entry | Feed thick section to thin airfoil |
| Drying control | Standard room | Controlled temperature and humidity | Prevent green cracking |
| Result | Cracked shell and defective blade | Intact shell and sound blade | Validated optimized full mold casting route |
The full mold casting optimization also has implications beyond gas turbine blades. Any thin-walled, high-curvature, hollow component produced by sacrificial-pattern casting can benefit from the same approach. The key is to treat the pattern as a thermo-mechanical load, not merely as a shape that disappears. When the pattern expands, softens, and decomposes, it transfers force to the shell. If the shell is brittle and the local geometry is sharp, cracking follows. By using deformation coordination, finite element analysis, and additive manufacturing design freedom, the full mold casting engineer can reduce that force and increase shell robustness before the first metal pour.
The final full mold casting trials demonstrated that the optimized process was repeatable in the trial set. The shell remained intact, the metal filled the blade cavity, and the castings were suitable for further inspection and finishing. The rapid full mold casting method therefore met the original goal: it avoided the long die-correction cycle, reduced cost, and provided useful blade castings in a much shorter time. I consider the deformation coordination model and the finite element stress analysis to be the critical technical enablers, because they converted an empirical shell-cracking problem into a design problem with measurable variables.