Manufacturing a Ceramic Turbine Rotor for a Compact Jet Engine

Commercial compact gas turbine engines operate in a wide variety of applications, producing thrust outputs that range from 22N (5 lb_f) in the JetCat P20-SX for small radio-controlled aircraft up to 1100N (247 lb_f) seen in the JetCat P1000 Pro to power a human jet pack [1]. Military-grade compact engines are costly for research and development purposes, which led to a program designed to increase the power and thrust output of a commercial compact engine. Commercial compact engines, such as the JetCat P400, offer a low-cost research and development test bed at $12,500 per engine [1] with a high power density. To gain the desired power and thrust increase from these small commercial engines and achieve an even higher power density would necessitate an increase in the turbine inlet temperature (T₄), which would exceed the maximum temperature limitations of the metallic turbine rotor currently in the JetCat P400. These compact engines are also too small to use traditional blade cooling methods in the turbine section.

Technical ceramics offer the ability to maintain similar strength parameters to metals while withstanding much higher temperatures. A drop-in replacement ceramic turbine rotor would allow for an increase in T₄, from 765 °C for the stock engine, to 1200 °C, improving the thrust and power output by a factor of 1.44, as outlined in Sec. 2, using the uninstalled thrust and Euler pump equations [2].

The present research determined that silicon nitride (Si₃N₄) could operate under elevated T₄ conditions utilizing finite element analysis (FEA). The simulation showed the Si₃N₄ turbine rotor could survive the elevated T₄ conditions at 1200 °C and centrifugal loads at 90,000 RPM representing an elevated temperature full power condition of the JetCat P400. Additionally, this research produced an alumina turbine rotor as a proof of concept sized as a testable replacement of the stock Inconel rotor in a JetCat P400. The alumina rotor was manufactured using digital light projection (DLP) additive manufacturing, which is cost-effective, timely, and scalable for production. An alumina rotor was made because Si₃N₄ was still in development for additive manufacturing at the time of publication.

Ceramic turbine rotors have been manufactured to allow such an increase in performance output. However, the current production of complex-shaped ceramic designs requires high tooling costs and lengthy timelines adding complexities to the process which make it undesirable for research and development activities. These issues have prevented ceramic turbines from being researched on a wider scale in engines. Small-scale engines are considered expendable, so any process that causes an increase in cost (such as tooling) is a major detracting factor. Traditional methods in ceramic manufacturing, such as tape casting or slip casting, can take anywhere from 9 to 14 weeks [3]. This is predominantly due to the time it takes to produce a metallic mold for casting purposes which also adds to the high cost of material used.

With the implementation of additive manufacturing, research in manufacturing complex designs has shown a potential to greatly reduce or eliminate tooling costs, shorten timelines, and increase the complexities of parts that are able to be produced which now includes producing ceramics. Recently, large-scale commercial additive manufacturing techniques, such as DLP or stereolithography (SLA), have become more readily available. These printing methods allow for the ability to directly print a complex ceramic shape without the use of any mold materials or complex tooling [4,5]. The DLP method was used in the present study.

Metal alloy turbines are easier to manufacture than ceramics for many technical uses, but in areas such as aerospace applications and engines these metal alloys tend to be a heavier weight option to meet operational requirements and are limited in operating temperature. Technical ceramics used in aerospace and engine applications have lower densities and similar strength properties to metal alloys while operating at much higher temperatures. To provide a comparison of different material classes and specific material properties, Dr. Michael F. Ashby created multiple plots comparing different material parameters known as Ashby plots [6]. The Ashby plot depicted in Fig. 1 compares various materials’ zero-time strength and density properties together with technical ceramics, grouped in yellow, and metals grouped together in red. When observing the plot, the higher strength technical ceramics such as silicon carbide (SiC), alumina (Al₂O₃), and Si₃N₄ show similar strength properties, between 500 and 800 MPa, to high strength metals such as tungsten, steels, nickel alloys, and titanium alloys [6]. However, the technical ceramics have lower densities than these metals, which make them beneficial in aerospace applications where weight reduction is important. As a comparison, Inconel 625 is a nickel-based representative turbine rotor material and has an average density of 8440 kg/m³ [7], compared to the Si₃N₄ and Al₂O₃ densities shown later in Table 1 at less than half that value.

Figure 2 depicts zero-time strength ranges of materials based on their maximum service temperature range with metals highlighted in red and technical ceramics in yellow [6]. Technical ceramics show a larger maximum temperature capability and much lower strength degradation compared to metals whose strength profile varies greatly at maximum service temperature levels [8]. The ability for technical ceramics to maintain similar strength characteristics to metal while at elevated temperatures allows for increased turbine operating temperatures beyond metallic capabilities with limited strength degradation, which allows for increased overall power and thrust output.

The ionic and covalent bonds in ceramics are much stronger than in metallic bonding, which makes ceramics stiffer and more brittle than metals while also achieving a lower density. Because of these properties, ceramics have a larger specific modulus and specific strength making them more efficient for use as a material. While ceramics are more mechanically efficient and maintain better strength properties at higher temperatures, the brittle nature of ceramics combined with complex geometries, such as those of turbine rotors, leads to many manufacturing complexities. In many cases, these complex geometries are not possible to fabricate using conventional processes and/or are prohibitively time and cost-intensive.

To try to make use of the thermomechanical properties of technical ceramics, research in industrialization and scalable manufacturing has increased in recent decades. However, this process is quite complex due to the high bonding energy of ceramic compounds. While this high bonding energy typically results in desired material properties (high melting temperatures, hardness, chemical stability), it necessitates the use of powder consolidation routes for most traditional bulk ceramic processing [9,10]. Powder processing can be challenging and costly when compared to liquid-based processes commonly used in metallic forming methods. Due to the brittle nature of ceramics, fine control of the raw materials and processing is imperative to minimize microstructural defects [9]. Additionally, powder processing routes are a multiple-step process that typically requires taking a raw or processed powder, forming it into a desired shape, removing a fugitive binder, and then applying a heat treatment to produce a dense sintered body. Multistep processes can be costly and time-consuming, especially when complex shapes are necessary. Therefore, any reduction in the production cost and time to manufacturing is highly desired.

Ceramic additive manufacturing offers an approach to create dense and near-net shapes with no (or minimal) post-machining required. It has been shown that suspension-based additive manufacturing, which includes DLP, has similar flexural strength, as tested through 3-point bend tests, to traditional manufacturing techniques [11]. Using additive methods, new rotor designs can be rapidly manufactured through direct printing of the geometry to avoid extra cost and timeline of creating specialized tooling or die for each geometry as would be needed in traditional ceramic casting techniques.

An ADMATEC Admaflex 300 DLP printer, along with Admatec alumina resin (AdmaPrint A130), was used for this study [12,13]. The resin composition is proprietary but contains a UV-curable polymer with high solids-loading of alumina particulate. Alumina was used as a proof-of-concept material as there was no commercially available Si₃N₄ resin. While both Si₃N₄ and Al₂O₃ have a similar density and maximum service temperature, Si₃N₄ outperforms in both strength and fracture toughness due to its unique interlocking grain structure providing a self-reinforcement [14]. The significant increase in fracture toughness of Si₃N₄ over alumina (K_Ic = 7 − 10 MPa m^1/2 for Si₃N₄ [14] and 2.7–5 MPa m^1/2 for Al₂O₃ [7]) provides greater resistance to catastrophic failure from thermal shock and/or mechanical loads.

Most of the lessons learned and design considerations from working through the production process with alumina should carry over to a Si₃N₄ resin when it becomes available for the printer. The DLP additive manufacturing process utilizes a photopolymer resin cured by an image flashed by a high-definition projector for each print layer. Each layer’s image was created by slicing an uploaded computer aided design (CAD) file of the geometry. As shown in Fig. 3, to produce each print layer, a thin layer of alumina resin is applied to the printer film and pulled into the print section, then the print head (including any previously printed layers) is lowered onto the new layer of print material on the film, the photopolymer is then selectively cured with light from the projection thus bonding selected portions of the print material to the previous layer stack. This process is repeated, building upon each layer, until the entire ceramic design is produced [12].

Once printed, the ceramic part was placed in a water tank for water debinding to partially remove the proprietary water-soluble fugitive polymer. The water tank had a circulation system and was kept at room temperature. After undergoing water debinding, the turbine rotor was able to be removed from the base plate and easily processed to remove any supports or material using a razor blade and/or sandpaper. After drying and processing, the turbine rotor then undergoes binder burnout up to 600 °C at a heating rate of 12 °C/h to burn off the majority of the polymer and any absorbed water. The dried part is then sintered at 1650 °C for 2 h using a heating rate of 600 °C/h to fully densify the ceramic. The entire process from print to sintering takes a total of two weeks as outlined in Fig. 4. As this is a final net-shape manufacturing process, the machining step in the figure is only referring to the removal of build supports. Additional information on the specifics of the ceramic printing, post-processing, and defect mitigation can be found in a companion paper from the research team [15].

The first half of this study evaluated if a Si₃N₄ turbine rotor could operate at the necessary elevated T₄ requirements to gain increased power and thrust from the engine. For the initial study, it was assumed that an intermediary device, such as a tolerance ring, was used to negate any stresses resulting from different coefficients of thermal expansion between the shaft and rotor. Unlike a traditional metallic rotor that has a thicker cross-section at the bore and blade attachment location with a narrower web in between, the analyzed Si₃N₄ turbine rotor design has a solid, 16.94 mm thick, uncontoured hub, as shown in the cross-section in Fig. 5. The bore had a diameter of 10.16 mm to support a 9.98 mm stainless-steel shaft. The blade geometry was specifically designed to operate in the JetCat P400 cycle with thicker blades (2.54 mm at the thickest part of the tip) than the metallic version (1.27 mm at the thickest part of the tip), which is more conducive to ceramic manufacturing. To compensate for the thicker blades and change in curvature, the blade count was reduced from 23 in the metallic version to 21 in the ceramic version.

Using the enhanced profile in Table 2, derived from a separate in-house study, an individual FEA simulation was run for the centrifugal case and thermal case to get an understanding of where the stress was acting in the turbine rotor from each contributing factor. The enhanced profile represents a cruise-type condition with an elevated temperature but reduced (compared to maximum) rotor speed. The idle, half power, and max power conditions represent stock JetCat P400 performance parameters. The bore temperature is a function of secondary flows and hot gas entrainment which remains approximately constant across the operating range. After the individual cases, a combined thermal and centrifugal load analysis was performed to determine the point of failure. These results were then analyzed to see where the highest stresses or potential failure occurred within the material. The Si₃N₄ strength properties were based on previous research initially using a cast method [16,17] since additive materials were not available. The thermal profiles were taken from a separate in-house computational study of steady-state operation of the JetCat at each specified condition.

Next, to evaluate the difference in coefficient of thermal expansion in the bore region, the ceramic rotor was fitted a metallic shaft, as depicted in Fig. 6. Material properties are listed in Table 1 for the Si₃N₄ turbine rotor and 2507 stainless-steel shaft. The simulation was a thermal static study based on the given enhanced profile from Table 2. The joint between the shaft and rotor was modeled as a fixed connection representing a bonded or mechanical hold with a nut. This simulation confirmed that an intermediary device was necessary to counteract the stresses in the bore due to dissimilar coefficients of thermal expansion while maintaining concentricity.

The next half of the analysis determined what operating regimes of the JetCat P400 engine a Si₃N₄ turbine rotor and an alumina turbine rotor could operate in. At the time of this study, Si₃N₄ was still in development for use with DLP additive manufacturing. To prove the capability of the DLP additive manufacturing process for single-piece turbine rotor manufacturing, a representative alumina turbine was made. The alumina material properties were based on data from the manufacturer of the photopolymer slurry [18]. The simulations consisted of combined centrifugal force and thermal stress based on three stages for the JetCat P400: idle, half power, and maximum power, as well as the previously mentioned enhanced thermal profile, all outlined in Table 2. These simulations were first run with Si₃N₄ to ensure that it would meet the same capabilities as the Inconel turbine rotor under maximum power operating conditions with the same design from Fig. 5. The enhanced thermal profile was then analyzed for the Si₃N₄ rotor. Simulations were also run using the alumina material properties [18] to determine if there was an engine operating regime the alumina turbine rotor could operate in for future experimental testing purposes.

The first simulation evaluated the centrifugal forces on the Si₃N₄ turbine rotor alone, where no failure was indicated. Since the hub was uncontoured from the bore to the blade root, the stresses resulting from centrifugal forces were concentrated near the center of the bore. This is shown by the peak stress of 300 MPa located in the bore region, as shown in the principal stress plot of Fig. 7. On a contoured hub, larger stresses are present at the blade roots. The uncontoured hub effectively reduced the centrifugal effects in the radially outboard portion of the hub, while concentrating the stresses in the bore, reducing the effects of the centrifugal forces.

The second FEA evaluated the thermal stresses and indicated a peak stress region at the bore. Figure 8 shows the thermal stress analysis using the enhanced profile with a resulting peak stress of 676 MPa, slightly above the expected modulus of rupture. The thermal FEA indicated a dominating effect due to thermal stresses compared to those from centrifugal stresses. Under the given temperature conditions, greater stress was observed in the bore in the thermal analysis, due to a large temperature difference between the cooler bore and the heated radially outboard portion of the hub. Since the hub is solid more energy could be absorbed and increase the effects of the temperature differential between the inner and the outer portion of the hub leading to the increased stress in the bore.

The combined FEA provided an overall determination on how the turbine rotor would perform under expected operating conditions. The analysis predicted failure with a maximum stress of 928 MPa in the bore region under combined loading conditions, shown in Fig. 9. In both individual load simulations, the peak loads were found in the bore. Under the combined load condition this led to a failure because the peak loads were superimposed and amplified. After observing the effects of each individual situation it became clear that the thermal stress was the dominating factor that would need to be mediated.

In an attempt to reduce the temperature differential, an additional study was run to try to decrease the stress at the bore by increasing the temperature in the shaft region. Maintaining the same centrifugal forces and temperatures at the blade roots, new temperature profiles were created for the hub that increased the bore temperature from 427 °C to 927 °C in 100 °C increments. A simulation was run at each of these temperatures and the results were then plotted showing the peak stresses that occurred in the bore region. This plot is shown in Fig. 10.

As shown in Fig. 10, a reduction in the temperature differential between the bore and blade roots presented a linear decrease, shown in the solid line, in overall peak stress located in the bore. At a bore temperature of 827 °C the maximum stress was equivalent to the modulus of rupture, indicated by the horizontal dashed line. By adjusting the bore temperature, therefore decreasing the temperature differential, the turbine rotor peak stress was decreased to below the modulus of rupture. The temperature increase in the bore could conceivably be increased closer to the property limits of the stainless-steel shaft or bearings to accommodate a higher temperature, further reducing the bore peak stresses. If taken to 1000 °C at the bore, based on the linear correlation, the stress in the bore would be reduced to only 441 MPa. The increase in temperature at the bore could be achieved by modifying the spacing between the stator and rotor thus allowing hot combustion gases to enter the gap. While this temperature increase at the bore is beneficial for the rotor, it may complicate the design of the metallic shaft and the bearings.

For the Si₃N₄ turbine rotor, the FEA showed it would be able to meet the stock maximum power condition of the JetCat P400, based on the initial bore temperature of 427 °C and design criteria shown in Table 2. The analysis in Fig. 11 showed peak stresses located in the bore at 515 MPa. The Si₃N₄ turbine rotor stress was greatly reduced in the bore when compared to the previous enhanced operating regime, due to the fact that the temperature gradient across the hub was much lower for the maximum power profile. This shows that when a Si₃N₄ photopolymer slurry is developed it could meet the requirements to operate in a stock JetCat P400 under maximum power conditions, provided the material still meets the strength requirement above 500 MPa.

For the printing of the alumina turbine rotor, the ceramic turbine rotor design had to be altered in the hub region to account for thickness restrictions of 10 mm in any two of the X, Y, or Z build directions, based on guidance from the manufacturer of the alumina photopolymer slurry. Adhering to these dimensional restrictions ensured water could fully infiltrate the printed parts and effectively remove the binding agent, and also so volatile species from binder burnout could be removed from the system with no defect generation. This requirement led to a design with a single row of nine extruded circular holes in the hub to ensure the material never reached the thickness limit. The resulting modified hub design was developed and is shown in Fig. 12. Circular cutouts were chosen to reduce material in the hub to ensure the dimension restriction was met while minimizing features that could cause stress concentrations. It was also found that these holes were required to be extruded all the way through the hub to avoid delamination that originated at the junction of the hole and the remaining hub that then extended throughout the rotor and into the blades as shown in Fig. 13. This delamination was not evident until the final step of the manufacturing process, sintering.

After the extruded hole design was created and analyzed, an alumina turbine rotor was successfully manufactured with no delamination. Initially, radial cracking was observed along with some small binder marks located on each side of the hub. It was found that the initial water debind time of two days was insufficient for complete binder removal and was the cause of the radial cracking. Additionally, small binder marks on the surface of the hub that was in contact with the build plate were due to water saturation being impeded from resting on the base plate throughout the debinding process. To correct this, the turbine rotor was removed from the base plate after one day of water debinding to allow for complete exposure to the water, while the debind time was also increased to a total of five days. Once these changes were incorporated into the process, a successful turbine rotor with no noticeable cracking was manufactured, as shown in Fig. 14. There was one turbine blade that was chipped off due to mishandling of the turbine rotor as it was moved between furnaces during processing. This error highlights the care that must be taken when handling the parts prior to densification.

The DLP additive manufacturing method showed a lot of promise in manufacturing. The photocuring of the slurry eliminates the introduction of macro air bubbles or other defects being introduced during the manufacturing process. Due to the slurry using a partially water-soluble photopolymer, the water-soluble binding agent was removed in water if the thickness limitations were not exceeded. The results from each print were consistent and issues were mitigated by slightly altering the process or changing the design quickly using cad software. The DLP additive manufacturing is a near-net shape manufacturing process that can produce parts with higher degrees of complexity when compared to casting. Prior to printing, the cad file needed to be enlarged by 22% in rotor diameter and 26% in thickness to account for shrinkage that occurs during the densification phase. The companion study [15] produced 40 samples of varying geometries and size-scale and found roughly a 1% standard deviation in shrinkage values, giving confidence for part accuracy. The DLP process described in this paper can be used to fabricate prototypes that consistently resemble each other in outward appearance, quickly and affordably. This is a valuable initial step along the path to testing DLP-manufactured silicon nitride turbine rotors for the first time, and then to develop a viable end-to-end manufacturing process that can produce structurally reliable turbine rotors with a predictable service life. Over time, DLP seems to have the potential to be scalable to manufacture complex parts that are not otherwise manufacturable by traditional means in greater-than-prototype quantities, as well.

The FEA results in Fig. 15 for the alumina turbine rotor design show it has the ability to operate in a JetCat P400 up to idle conditions with a peak stress of 295 MPa. For experimental testing purposes, running under idle conditions would allow for a 1.4 factor of safety. Unlike the Si₃N₄ which calculated failure in the bore, for the alumina design the middle of the turbine blades held the largest stress and would be the first location to fail.

When the final alumina turbine rotor design was analyzed at the half power operating condition it showed failure throughout the entire turbine rotor as shown in Fig. 16. Failure occurred throughout the middle part of the blades, middle of the bore region, and between the circular holes in the hub due to stress concentrations. The peak stress was found to be 665 MPa located in the blade region.

Analysis of Si₃N₄ showed that it was a suitable material to use to produce a drop-in replacement for an Inconel turbine rotor in the JetCat P400. Si₃N₄ offers a significant weight saving and a higher operating temperature limit over Inconel with a potential for production cost savings depending on the manufacturing method used for each. The Si₃N₄ turbine rotor is predicted to be able to operate under current maximum power conditions without modification and meet the desired turbine inlet temperature increase of 1200 °C to achieve a 44% increase in thrust with the enhanced profile if thermal stresses in the hub can be reduced.

Managing the thermal stresses induced from the increased temperature conditions of the enhanced profile will require an increase of the rotor temperature near the bore to at least 827 °C. This increase would reduce the thermal stress caused by a temperature differential of 873 °C between the blades and the bore, which was the dominating stress causing initial failure. The ability to manage this increase would be dependent on the shaft and bearing temperature limits, which will ultimately lead to new design criteria for a turbine rotor. All these results and analyses are dependent on accurate thermal gradients as the thermal stresses dominated the centrifugal stress. Additional rotor cross-sectional geometries should also be investigated to include variable thickness hubs; however, this change complicates the manufacturing process because there is no longer a flat surface in contact with the printer’s build plate.

Additional consideration was given to the rotor/shaft integration to account for the difference in coefficient of thermal expansion between a Si₃N₄ turbine rotor and stainless-steel shaft. Analysis confirmed material failure would result in the rotor bore if an integration device, such as a tolerance ring, was not present to alleviate the stress. A means of accounting for the differences in thermal expansion between the shaft and rotor while maintaining concentricity is required.

Increasing the temperature in the bore and incorporating a tolerance ring were all based on solely creating a drop-in replacement turbine rotor which did not change any of the design aspects of the JetCat P400. However, further consideration could be given to changing the bearings and shaft material to allow for an even greater temperature increase in the bore region. Additionally, other ceramic to metal integration techniques could be used such as designing the turbine rotor with a nub to be used with a metal shrink fitter. In this arrangement, a metal cup on the end of the shaft is heated and the ceramic nub is inserted. When cooled, the metal contracts and grips the ceramic, which would keep the ceramic in compression where it is significantly stronger than in tension. These techniques would be beneficial if a larger redesign of the engine is allowed.

The DLP additive manufacturing method produced an alumina turbine rotor that is predicted to run under idle conditions in a JetCat P400. Alumina does have poor thermal shock characteristics, and cracking may occur. The photocuring of the slurry eliminated the introduction of macro air bubbles or other visible defects during the printing process. Due to the slurry containing a partially water-soluble photopolymer, the water-soluble binding agent was easily removed in water if the thickness limitations were not exceeded. Issues due to stress cracking were due to thickness limitations and stress concentrations introduced by the design of the turbine rotor. Longer than recommended water debinding ensured the water-soluble binder was fully dissolved. The stress concentrations became predictable, and the design was quickly altered through CAD software and then reprinted. The results of the prints were consistent and repeatable, and after modifying the process to remove the induced stress concentrations and adjust the water debind timeline, the print had no issues producing the desired results. The DLP additive manufacturing process removed the need for any mold to hold shape greatly reducing tooling costs. By directly printing the ceramic turbine rotor, the timeline was expedient at two weeks with a low material cost of $158 per turbine rotor. At the time of publication, only alumina and zirconia were available for this additive process, but silicon nitride was in development. The predictability and consistency of results, ease of processing, and fast timeline make this process potentially scalable and beneficial for higher-volume manufacturing of complex ceramics in the future.

Once a Si₃N₄ photopolymer is developed it can then be used to create a Si₃N₄ turbine rotor for testing. A redesign of the original Si₃N₄ turbine may be required, similar to the modification for alumina, depending on the new strength properties of the material as well as any thickness restrictions. Once a suitable design is determined through FEA, it can then be manufactured and tested under full power conditions in the JetCat P400.

There are no conflicts of interest.

The authors attest that all data for this study are included in the paper.

c =

critical flaw size

$m˙$ =

mass flowrate

A =

cross-sectional area

P =

pressure

T =

thrust

V =

velocity

$W˙$ =

power

g_c =

gravitational constant

r_m =

mean turbine blade radius

v₂ =

tangential velocity from the stator vane

v₃ =

tangential velocity from the rotor blade

K_Ic =

fracture toughness

T₄ =

turbine inlet temperature

Al₂O₃ =

alumina

Si₃N₄ =

silicon nitride

CAD =

computer aided design

DLP =

digital light projection

FEA =

finite element analysis

SiC =

silicon carbide

SLA =

stereolithography

ρ =

density

σ_f =

modulus of rupture

τ =

torque

ω =

angular velocity

Ω =

rotational rate

9 =

nozzle exit

0 =

engine inlet

 c1 =

initial turbine inlet temperature condition

c2 =

increased turbine inlet temperature condition