Ceramics for turbine engines
Ceramic materials for gas-turbine engines have been in development for decades, but only now are those efforts showing signs of success.
||Since the mid-1940s,
researchers have envisioned ceramic materials as a way to improve the
performance of gas-turbine engines, lengthen their life span, and reduce
their fuel consumption substantially. Yet ceramics are just now
approaching their first commercial use in turbines.
The challenges involved have been considerable. The material in modern turbines must survive temperatures of more than 1,100°C for thousands of hours; high thermal stresses caused by rapid temperature changes and large temperature gradients; high mechanical stresses; isolated impact and contact stresses; low- and high-frequency vibrational loading; chemical reactions with adjacent components; oxidation; corrosion; and time- and stress-dependent effects such as creep, stress rupture, and cyclic fatigue. Early ceramic materials were not able to withstand these conditions, and early turbine-component designs were not compatible with brittle materials. Technological evolution has to be made over a broad front, and progress has been slow.
A variety of oxides, borides, carbides, and cermets were evaluated in the 1940s and 1950s for potential use as turbine components. Some ceramics had favorable strength and oxidation resistance, but none survived the thermal shock conditions imposed by an engine. Some cermets could survive thermal shock and impact conditions but did not have adequate oxidation resistance and stress rupture life.
Interest was renewed in ceramics for turbines when new materials in the silicon nitride and silicon carbide families of ceramics were developed during the 1960s. These materials had better thermal shock resistance, largely due to a combination of low thermal expansion, high strength, and moderate thermal conductivity.
The first promising silicon nitride and silicon carbide materials were fabricated in Great Britain by reaction sintering. The silicon nitride was prepared by a reaction of a powder compact of silicon with nitrogen to form silicon nitride. This resulted in a reaction-bonded silicon nitride (RBSN) material, typically with room-temperature strength approaching 140 megapascals. The strength was retained to at least 1,400°C, but the material weakened over time when exposed at high temperature to an oxidizing atmosphere. The silicon carbide was prepared by reacting a mixed powder compact of silicon carbide plus carbon with molten silicon to form an SiC-bonded silicon carbide, with the pores filled with silicon. Early reaction-sintered silicon carbide materials had strength that was similar to RBSN and superior oxidation resistance.
The Allison AGT-5 engine is one components test bed that was used to develop ceramics for automotive applications
A higher-strength pore-free silicon nitride was achieved in Britain in the late 1960s by hot pressing. This involved heating a mixture of silicon nitride powder plus a low percentage of oxides, such as calcium oxide or magnesium oxide, in a graphite die to about 1,700°C while applying a uniaxial pressure of about 13.8 to 34.5 megapascals. Room-temperature strength greater than 700 megapascals in three-point bending was demonstrated. However, additives to allow densification concentrated at the grain boundaries as a glass composition. This composition softened when the ceramic was heated to temperatures approaching 1,000°C, resulting in grain-boundary sliding when a stress was applied. This decreased strength.
Major efforts have been conducted worldwide since the early 1970s to improve the high-temperature properties of silicon nitride. Some have focused on finding a composition with a higher-temperature intergranular glass phase; others have focused on compositions that can be heat-treated to crystallize the grain-boundary phase and avoid the glass phase. Only recently have the properties been adequate to consider long-life applications.
Ceramic turbine components are fabricated starting with powders of the raw materials. The quality of the final part depends on the quality of the starting powder and on each step in the fabrication process. Early powders were coarse and contained impurities, and they were not widely available until the mid-1970s. Around that time, researchers demonstrated that silicon nitride and silicon carbide could be densified by pressureless sintering if the starting powder was of very small particle size. Powder synthesis techniques were refined during the 1980s, making powders with a smaller particle size (submicrons) and relatively high purity available.
Techniques to measure properties were critical to the evolution of ceramic materials for turbines. They provided both a means of comparing materials and the database needed for probabilistic design, and were an important part of quality control.
Initially, strength testing was conducted using a three-point bending procedure, with each company using a different test-bar size and fixture design. The data were impossible to compare, and a very small volume of material was exposed to the tensile load. Sometime around 1972, attempts were initiated to select a standard test-sample size and shape as well as conduct tests with four-point bending. More sample volume is exposed to tensile stress under four-point bending, so a reasonable tensile distribution can be established with 20 or 30 test bars. Four-point bend testing proved adequate to compare materials and to design small turbine components with localized regions of high stress, such as a turbine rotor blade inserted into a metal hub.
When components were designed with a larger volume under tensile stress, such as one-piece radial ceramic rotors, emphasis began to shift to tensile testing to simulate better the larger volume under stress and the increase in the proportion of internal flaws to surface flaws. Tensile testing is also important to gain an understanding of creep, stress-rupture life (slow crack growth), and cyclic fatigue, as well as to establish models for life prediction.
Earlier efforts at material characterization and database generation have continued under a project funded by the U.S. Department of Energy (DOE) through Oak Ridge National Laboratory in Oak Ridge, Tenn. The project addresses test standardization, a database for fast fracture and time-dependent fracture measured in uniaxial tension, correlations with nondestructive-evaluation (NDE) techniques and life-prediction methods, and iterative increases in material and component reliability. Statistical tensile databases now exist for all key silicon nitride and silicon carbide turbine materials from room temperature to at least 1,370°C, including some stress-rupture and creep tests lasting more than 10,000 hours. A parallel program at Garrett Turbine Engine Co. in Phoenix—now part of AlliedSignal—has established a flexural-strength database on the same materials exposed for up to 3,500 hours in a dynamic test rig cycling between a diesel-fired burner and an air-blast quench.
One of the biggest challenges has been fabricating to achieve reliable properties in the required complex shape at acceptable cost. The primary focus has been on developing near-net-shape fabrication processes that can produce complex turbine-component shapes with a minimum of machining, and on optimizing and controlling each step in the fabrication process to minimize the size of microstructural flaws to within design limits.
Turbine programs in the 1970s had to rely on reaction-bonded and hot-pressed materials. Ford Motor Co. in Dearborn, Mich., conducted extensive development of injection molding. The automaker succeeded in injection-molding one-piece stator-vane rings and rotor-blade rings.
Sintered silicon nitride and silicon carbide materials were developed in the late 1970s and became the primary candidates for turbine programs throughout the 1980s. These materials had intermediate strengths, between reaction-bonded and hot-pressed materials, but they had the potential to be fabricated to near-net shape at costs competitive with metal turbine components. Most of the effort focused on injection-molding and slip-casting radial turbine rotors, scrolls, and other turbine components. Given the properties of materials at that time, a microstructural flaw such as a pore, crack, or inclusion roughly 150 microns across in the high-stress region of the rotor was a critical flaw and would cause immediate failure. Early sintered materials had flaws substantially larger than this. Extensive development was required (and is still continuing) to fabricate turbine components free of these critical flaws.
The emphasis then shifted to hot isostatic pressing (HIP), either to achieve additional densification of a sintered part or to directly densify a powder compact encapsulated in a glass envelope. HIP achieved properties similar to uniaxial hot pressing but allowed complex shape fabrication. Turbine rotors and other components densified by HIP successfully operated at design conditions in experimental turbines. HIP is expensive, however, and results in less strength near the surface than in the interior of the part.
Research during the 1990s has been directed toward improving the properties of sintered materials to minimize flaw size and refining the microstructure to increase fracture toughness. Higher fracture toughness means a larger critical flaw size for a given stress. Whereas the early materials had a critical flaw size around 150 microns for a 200-megapascal stress, the improved materials can withstand flaws several times larger.
Improvement in the Weibull modulus of the materials is a good gauge of progress. In 1980, the typical Weibull modulus of candidate turbine ceramics ranged from about 5 up to 8, with occasional values of 8 to 12. Now most of the materials have a Weibull modulus consistently above 20.
The Garrett/Ford AGT 101 engine included ceramic components such as the radial rotor, the rotary regenerator, and even the attachment bolts
Fracture of a ceramic part in a turbine is likely to lead to fracture of adjacent ceramic parts and complete engine failure. Quality assurance must be rigorous enough to eliminate any ceramic parts with critical defects. This was a very difficult challenge when the critical defect size was about 75 to 150 microns. Materials with increased fracture toughness and improvements in processing have eased the challenge. In addition, NDE and proof-test procedures have improved, but these procedures are expensive.
Methods and tools for designing ceramic components have advanced dramatically since the 1960s. The field of probabilistic design for ceramics has been developed and validated in engine testing. The turbine-engine companies have established refined models and codes for preliminary design, detailed design, and life prediction. The first probabilistic life-prediction codes were based only on fast-fracture criteria, but improved databases have allowed time-dependent models to be incorporated in life-prediction codes.
The first key program in the United States was the Brittle Material DesignÐHigh Temperature Turbine Program, initiated in 1971 by the Defense Advanced Research Projects Agency (DARPA) at Ford and Westinghouse Corp. in Pittsburgh. The program focused on demonstrating that ceramic components could be successfully designed using finite-element-analysis codes and could survive under severe engine-operating conditions. Ford's portion of the program culminated in a successful 25-hour test at 50,000 rpm and 1,200°C to 1,250°C of a single-shaft engine with all hot-section components (rotors, stators, combustor, casing, and two honeycomb rotary regenerators) fabricated from ceramics.
The Ford program was a pioneering effort, demonstrating that brittle ceramics could be designed with a probabilistic FEA approach and could survive the conditions of engine operation. However, the materials available at the time had marginal properties, and other issues such as contact stress, durability, reliability and cost-effective fabrication needed further attention.
The program at Ford did not focus on demonstrating increased performance. DARPA initiated a program in 1976 with AiResearch Manufacturing Co. in Phoenix—now AlliedSignal Engines—to retrofit ceramics into an existing turboprop engine, with the goal of demonstrating a 40-percent increase in power output and a 10-percent decrease in fuel consumption.
To achieve the goals, the turbine-inlet temperature (TIT) had to be increased from about 1,000°C to 1,200°C. This was achieved by replacing the first two turbine stages with ceramics. The resulting design consisted of 104 ceramic parts. RBSN was selected for the stator vanes, shrouds, transition liners, and support structures. The rotors were a hybrid design that comprised Norton NC-132 hot-pressed silicon nitride blades with a single tang dovetail-inserted into a metallic disk. The blades and disk were separated by a thin metallic compliant layer. Performance improvement of 30 percent and fuel-consumption reduction of 7 percent were demonstrated in comparison with the baseline metallic engine.
An original goal of the program was to achieve 50 hours of successful engine operation. The team believed that this required a complete redesign of the static structure to avoid the source of biaxial contact stress, which had caused failures during engine testing. A U.S. Air Force program allowed redesign of the engine to avoid the contact-stress problems in the static structure. The structure was tested successfully in rigs and the rotor blades were validated by spin testing.
All engine testing in the Ford and AiResearch programs was conducted on test stands. The results were positive, but the question remained whether the ceramic components would be durable in an engine in an actual vehicle, especially a land-based vehicle subject to severe shock loading. DOE initiated a program in the mid-1970s at the Detroit Diesel Allison Division of General Motors Corp. in Detroit to design ceramic components into the Allison GT 404-4 truck engine. Testing included powering a truck on highways, city roads, and the GM proving grounds, where the engine was exposed to extreme vibrational and shock loading on the Belgian-block and truck-durability road courses. Testing clearly demonstrated that properly designed ceramic components could survive under the most severe conditions for a typical vehicle.
To take advantage of the higher temperature capability of ceramics, DOE funded programs from 1979 to 1987 to demonstrate proof of concept for a ceramic-based automotive gas-turbine engine that could power a midsize automobile over a standard federal combined driving cycle (city and open road) at 42.8 miles per gallon. The Advanced Gas Turbine (AGT) Program included one program at Detroit Diesel and another at Garrett.
The Allison engine (AGT 100) was relatively conservative in design. It had a shaft for the gas-generator turbine and a shaft for the power turbine. This split the work and minimized the TIT (1,285°C maximum) and engine speed (85,000 rpm) required to achieve the mileage goal. However, the design required two radial ceramic turbine rotors and a complex shaped scroll (to transition the hot gases from the combustor to the nozzle guide vanes).
The Solar Centaur engine (shown with a schematic of its hot section) is retrofitted with ceramic components
The Garrett engine (AGT 101) was an extension of the Ford 820 ceramic engine design. It had a single shaft, a radial rotor, and air bearings. It required a design speed of 100,000 rpm and a TIT of 1,370°C to meet the program goals. The AGT 101 design had the advantages of being symmetrical and simpler (fewer ceramic parts) than the AGT 100, but it posed a greater challenge to ceramic-materials technology because of the higher speed and temperature. For both projects, the TIT and the resulting engine performance remained short of program goals. Many ceramic components fractured during testing, some due to design and assembly problems but most from material limitations. The desire was to use net-shape fabrication processes, but the state of technology in 1985 did not result in reliable ceramic turbine components.
The AGT program focus was changed to improving the materials processing of components, emphasizing use of the AGT rigs and engines as test beds to guide ceramic-component development. Substantial improvements in shape forming, properties, and reliability were achieved between about 1985 and 1993.
DOE automotive programs began to change in the early 1990s. Cost, producibility, and durability became key issues. Gas-mileage goals were increased to 80 miles per gallon; hybrid propulsion-system concepts became popular. Allison, which had replaced its AGT 100 configuration with an axial turbine configuration designated AGT-5, modified its program to explore ceramic readiness for a hybrid turbine-electric concept. Several ceramic-component and engine failures have led to design or material modifications. The program recently achieved a 300-hour engine test with no sign of distress to any ceramic components.
Garrett discontinued the AGT 101 engine and switched to the AlliedSignal 331-200 auxiliary power engine, which was a better test bed for accumulating engine test time and field testing. More than 1,400 hours of engine testing have been performed with silicon nitride inlet guide nozzles, including more than 300 hours at design speed with a ceramic-bladed rotor.
AlliedSignal Engines, under a DARPA insertion program, is already conducting field tests of silicon nitride ceramic nozzles in its Model 85 auxiliary power unit. By April, more than 46,000 hours of engine testing had been successfully completed, including more than 7,500 hours on one engine. The primary concern has been cost, but AlliedSignal Ceramic Components has demonstrated 76-percent cost reduction over roughly the past year and predicts additional cost reduction.
A program at Solar Turbines Inc. in San Diego has reached the field-testing stage with ceramic components. DOE initiated this program in 1992 with the objective of retrofitting a Solar Centaur 50 industrial turbine engine with a ceramic combustor liner plus first-stage nozzles and blades. A major challenge of this program was to achieve design stresses that were low enough to enable the ceramic to survive for 30,000 hours.
A series of full-scale engine tests have been conducted at Solar. Silicon nitride rotor blades from AlliedSignal Ceramic Components and silicon carbide composite combustor liners from DuPont Lanxide Composites in Newark, Del., have been qualified in these tests, and are currently in a field test at an ARCO Western Energy oil field. As of this June, the engine had operated for more than 700 hours with no problems.
The recent field tests and other engine tests are encouraging. They indicate that the candidate ceramics, the component and engine designs, the manufacturing processes, and the life-prediction methods have simultaneously reached a level of maturity consistent with turbine-application needs. Concerns still exist, however, such as with contact stress as well as with the damage and cost due to foreign objects.
Some significant successes already have been demonstrated in spin-off applications. Silicon nitride has been in high-volume production for high-speed cutting-tool inserts for machining cast iron and superalloys since the late 1970s. Silicon nitride turbocharger rotors have been used in Japan since the late 1980s, with no reported incidence of failure. Silicon nitride cam-follower rollers are in production at Detroit Diesel, and silicon nitride bearings are in production at St. Gobain Norton Industrial Ceramics. Silicon nitride seal runners have recently entered production for AlliedSignal propulsion engines for business aircraft.
Silicon nitride is also used in the paper industry, in sandblast nozzles, and in many other commercial applications. Many applications also exist for silicon carbide materials developed primarily for turbine-engine programs. The substantial investments in advanced structural ceramics are clearly starting to pay off.
David W. Richerson is president of Richerson and Associates in Salt Lake City
© 1997 by The American Society of Mechanical Engineers