Document
. / NASA-TM-8686519850001762
NASA Technical Memorandum 86865
, I
Cornputational Thermo-Fluid Dynamics
Corttributions to Advanced Gas
, i
'Turbine Engine Design
Robert W. Graham, John J. Adamczyk,
and Harold E. Rohlik
Lewis Research Center
Cleveland, Ohio
Prepared for the
Twenty-third Aerospace Sciences Meeting
sponsored by the' American Institute of
Aeronautics and Astronautics
i
A Reno, Nevada, January 14-17,1985
,.
NI\SI\
COMPUTATIONAL THERMO-FLUID DYNAMICS CONTRIBUTIONS TO ADVANCED GAS TURBINE ENGINE DESIGN Robert W. Graham, John J. Adamczyk, and Harold E. Rohlik National Aeronautics and Space Administration Lewis Research Center Cleveland, Ohio 44135 Abstract heat transfer, thermodynamics, structural mechanics and materials. The engine companies, The early history of design practice for the universities and government laboratories became aircraft gas turbine engine can be characterized joint participants in this research endeavor. The advent of the electronic computer and the elec- as the application of empiricism founded on actual tronic acquisition and recording of data has had a test experience. Trial and error processes gave major influence on how fundamental research infor- the designer the needed information to modify or redesign the principal components for more strin- mation is utilized in the design of advanced gas turbines. There is a definite movement toward gent operating conditions but the changes between reliance on computational methods to establish the engine models were conservatively small. In many cases, this approach was both extremely costly and design of advanced engines. The mechanical calcu- time consuming. A sizeable fraction of the cost lator and slide rule have given way to the large storage computers; the manometer tube banks and of marketing a new engine went into equipment strip charts have been replaced with transducers modification and the iterative substitution of and CRT's.
design changes. Aircraft gas turbine engines being envisioned for the last decade of this cen- tury and the early decades of the next century It is still too early to predict what poten- will operate at higher pressures and temperatures. tial capability this new computational tool will Tolerances and clearances will be tighter than make possible. Hopefully, it will enhance innova- current practices. Design practice will rely tive design, improve the certainity of predictions heavily on computational codes to predict gas and minimize costly design changes during the buildup of a new engine.
flows and thermal conditions throughout the flow path of the engine so that iterative changes in the development of the engine will be minimized. In this paper the authors will trace the design practices for the gas turbine throughout The introduction of more computational methods into the design process will enable bolder changes its history with particular emphasis on the calcu- lational or analytical methods. With such a back- in the evolution of new engines. Improved cycle efficiency, quieter operation and more reliable ground and a picture of the current situation, we performance will be among the benefits derived will make an assessment of the future impact of from this design practice. computational fluid mechanics on future gas tur- bine design. Three principal components of the Introduction gas turbine engine will be considered: namely, the compressor, the combustor and the turbine.
Over the past 40 years the gas turbine power plant for aircraft propulsion has made remarkable Components of the Gas Turbine progress. However, this progress has not been made without considerable difficulty. The learn- The Compressor Early Design Approaches ing process in design has often been an arduous, frustrating experience marked with failure as well In the beginnings of the gas turbine power plant, the performance of the compressor was as achievement. During most of the evolutionary histoY'y of the aircraft gas turbine, the designer probably the most crucial design issue of all the found himself dealing with technical decisions that components. It was the inefficiency of the avail- were beyond the state of available technological able compressors that delayed the introduction of knowledge or experience. He was compelled to take the gas turbine as a viable power plant. Advances risks of innovation and extrapolated technical in the design and production of superchargers for practice to meet the operating demands of the aircraft reciprocating engines contributed signi- engine. In the early days there were few analyti- ficantly to the improvement of compressor perfor- mance. These initial improvements came about from cal tools to bridge the gap between established practice and the design demands of the gas turbine empirical information derived from development engine. Trial and error was one of the practical experience. Along with the improved compressor ways of overcoming the gap. This process eventu- machinery, more systemmatic design methods were ally become a development testing program in which also evolving which enabled subsequent improve- new concepts were tried and modified until they ments in later models. This "bootstrap" process advanced the design methodology dramatically in worked and proved themselves. While the process was reasonably successful, it became increasingly the first generations of the aircraft gas turbine.
apparent that the iteration process between fabri- cation and test was becoming prohibitively costly The earliest U.S. compressor used in an air- and time consuming. craft gas turbine was a centrifugal type. For the larger engines, the centrifugal was quickly re- In conjunction with experimental development, placed with the axial flow because of its smaller frontal profile and multistaging possibilities.
fundamental research activity expanded which pro- vided a foundation of information or data for The axial-flow has remained the preferred compres- sor configuration for the large engines used in improved design correlations or analysis. This research encompassed a broad spectrum of disci- civil transports and most military aircraft, plines including fluid mechanics, aerodynamics, except for helicopters. The airfoil shapes of the blades and vanes of the axial flow machine lend calculations are carried out on two orthogonal surfaces within a typical blade passage (see Fig.
themselves to airfoil wing theory in the design 3). The solutions are carried out for inviscid, practice. The "blade element" design approach steady flow. Viscous effects are included by resulted from this association. In this design boundary layer computations along the wall and approach, shown schematically in Fig. 1, each blade row was considered to be made up of a finite num- blade surfaces.
ber of elements which are radially stacked.
Velocity diagrams for the leading and trailing Future Design Approaches edges of each element were selected to give the proper flow turning and pass the required quantity Under development are numerical codes for of air. A blockage correction was applied to the three dimensional flows that apply to steady, flow area to account for the presence of the inviscid and viscid conditions. As these codes become verified, they will replace the codes now blades. An airfoil shape was selected from a family of known airfoils (generally the NACA in use which approximate the three dimensional types) which would accommodate the turning and geometry by gas path solutions on meridional and velocity diagram specifications. Demands for b1ade-to-b1ade surfaces. It is hoped the truly increased pressure rise per stage in more advanced three-dimensional analyses will enable 1ess- machines called for maximized loading or turning empirical loss estimates and also provide predic- in each blade row. This demand necessitated load- tion of the secondary flows. These three ing and loss information obtained in experiments dimensional codes are premised on the assumption with stationary cascades of blades which simulated of steady flow. The inclusion of the grossly the compressor stages. unsteady features of the real flow is an issue which still needs to be addressed. A proposed scheme for doing this will be described in the In addition to the cascade approach, the per- formance of single stage research compressors was next section of this paper.
used to ascertain loading limits and losses.
Analyses of these types of data from both cascades Comparing the new design procedure for and single stage research compressors were uti- compressors with the initial blade element lized in establishing empirical guides for loading approaches, it is apparent that more information limits and losses for individual blade rows. about the internal flow characteristics of blade rows is being incorporated. The early versions depended on information pertaining to the inlet The early multistage design procedure amounted to extensions of the single stage proce- and discharge stations of the blade row only. The dure. One method was called "stage stacking." new computational ability and also the availabil- This empirical method made use of dimensionless ity of nonintrusive flow measurements within the pressure and flow coefficient maps of each single blade rows have made this major change possible.
stage similar to those shown in Fig. 2. By match- This comment pertains to turbine design as well as ing these stage performance maps with the flow to the compressor.
angles of the velocity diagrams, an initial multi- stage design resulted •. Performance checks of An Analytical Approach for Multistage compressors designed in this fashion revealed Turbomachlnery severe limitations of the approach, particularly in integrating the secondary flow losses and the In the discussion on compressor design analy- three-dimensional reality of the flow through the sis it was mentioned that the current aerodynamic machine. The relative success of the method design of turbomachines is generally based on a depended heavily on how well the dimensionless combined through-flow and b1ade-to-b1ade analysis performance maps of each stage were known. to simulate three dimensional flow. Such an anal- ysis described the average passage flow field Comprehensive descriptions of the design which is considered to be steady and is periodic procedures and pertinent comments about the his- from passage to p~ssage. A more rigorous method has been proposed in which the equations are tory of compressor design are documented in refer- ence 1. Early on in the design history, the major truly three-dimensional and methods for averaging limitations of the practice were identified. the passage flows are suggested.
Analytical approaches as well as experimenta11y- founded empiricisms were initiated to overcome The derivation of the steady three- these limitations. Some of the analytical dimensional average-passage equation system begins approaches predate the general availability of the by ensemble averaging the three-dimensional electronic computer. For example, the three- Navier-Stokes equations. This is done to elimi- dimensional theory2 appeared in the early 50's. nate the need to resolve in detail the structure This reference was a major contributor to the of turbulent flows. The global effects of turbu- pseudo-three dimensional numerical techniques that lence on the deterministic flow field are accoun- followed in the 60's and 70's in which "flow ted for by means of a Reynolds stress and a scalar through" or "flow path" computational methods were flux tensor. The resulting system of equations, developed. The flow path approach yields mean1ine often referred to as the Reynolds-averaged equa- information along prescribed flow surfaces. The tions, are then averaged in time at every point in equation set includes the momentum equation, the space to remove unsteady time scales which are on continuity equation, the energy equation and the the order of the period of shaft rotation or equation of state. greater. The global effect of the organized (as opposed to random) b1ade-to-b1ade unsteady flow Current Design Practice structure manifests itself in the resulting equa- tion system as body forces, energy sources, momen- At this writing, the prevalent analytical tum and energy temporal mixing correlations. For methods used by the industry in design are the a single-stage machine subjected to an axisym- above pseudo-three dimensional types in which the metric inlet and exit condition this equation set support the computational effort. Certainly, the describes the three-dimensional average-passage verification of the closure approaches will flow field associated with either the rotor or stator blade row. require a major experimental test program as well.
In regard to multistage machines the ensemble The Combustor Early Design Practice time-averaged equations govern, in general, a flow field which varies from passage to passage around The combustor appears to be one of the most mechanically-simplistic components of the gas tur- a given blade row. To extract the average-passage equations from this equation system, they must be bine engine. Its outward simplicity is deceiving, averaged on a passage-to-passage basis. The math- however, because it houses a complex aerodynamic ematical operator which accomplishes this task and chemical reaction processer that challenge the averages out the details of the passage-to-passage ingenuity of the designer (see Fig. 4). Those who variation in the flow field. However the global designed the first combustors for aircraft gas effect on the average-passage flow field is not turbines had very little prior engineering experi- eliminated. Its existence is accounted for ence to build on. The principles of flame combus- through body forces, energy sources, momentum, and tion had been a scientific curiosity from the energy spatial mixing correlations.
early experiments of Bunsen but making an energy- intensive reactor work was a challenging new tech- A set of average-passage flow equations con- nology. This challenge incited a major effort in tain correlations which arise from turbulence, combustion research for the conditions encountered unsteady organized flow, and passage-to-passage in the gas turbine combustor.
flow variations. These correlations are generic to this equation system, just as the Reynolds The principal functions of a combustor are to stress tensor is generic to the Reynolds-averaged accept the high pressure air from the compressor Navier-Stokes equations. They serve as the mecha- discharge, reduce the airflow velocity to values nism by which the passage-average flow field is approximating flame speeds, distribute and mix mixed out in the multistage environment. The need fuel in that air stream and homogenize the combus- for accounting for this mixing process in turbo- tion gases into a uniform or prescribed tempera- machinery flow analyses was recently documented. ture profile at the combustor discharge. Research and development efforts were initiated that exam- ined elements of the process that occur within the In Ref. 3, two reduced forms of the average- passage flow equations are developed. In deriving combustor. Among the principal topics were the the first set of equations it is assumed that the following.
body forces and energy sources generated by the presence of neighboring blade rows can be evalua- 1. the diffusion process in the combustor ted using a through-flow blade-to-blade analysis. in 1 et.
It is further assumed that the mixing correlations can be modeled in both the through-flow blade-to- 2. the fuel spray patterns of the fuel injectors.
blade analysis and the average-passa~e system according to the analysis presented. The resulting set of equations describe a three- 3. the ignition of the fuel/air mixture.
dimensional flow field whose axisymmetric component is equal to that predicted by the 4. the mixing and diffusion of the combus- through-flow blade-to-blade model. In the second tion reactants.
approach, an average-passage equation system is developed for each blade row in the machine. The 5. dilution of the combustion products.
two sets of equations are coupled through common expressions for the mixing correlations, the body The major overall objective of these studies forces and the energy sources. Thus they must be was to aid the designer in providing satisfactory solved simultaneously. The closure problem asso- mixing, stable flow and burning while maintaining ciated with this formulation requires the develop- minimal pressure drop through the combustor.
ment of models which globally describe the mixing Much of this research was experimental and pro- produced by the coherent spatial and temporal vided data sets or correlations that could be nonuniform flow in addition to models for the applied to design practice. Most of these were Reynolds stress and associated turbulent empirical but some theoretical effort to under- correlations. stand flammability, flame propagation, diffusion and other phenomena were being pursued along with An alternative approach to the closure prob- the experiments.
lem is to consider separately the mixing stress, the body forces and energy sources not associated During the decades of the 50's, 60's and even with the flow turbulence. The contribution of the some of the 70's, combustor design was more of an coherent velocity field to the mixing correlations art than a science. Combustor hardware was built, can be decomposed into two components - an inci- tested and modified until it performed in conform- dent gust and a velocity field which is the source ity to the constraints of specifications. In the of the gust. .
course of developing a satisfactory combustor, some of the changes were easily performed such as The above-mentioned closure suggestions have adding new holes in the combustor liners or the not been evaluated. A considerable effort in introduction of new vane geometrics to channel the cooling air.7 Sometimes more major changes were numerical analysis will be necessary for this task. An analytical effort alone will not be needed that required complete revision of the combustor configuration. The designer was sent sufficient because of the complexity of the prob- lem. Detailed measurements of the fluctuating back to the drawing board. Modification of a air loads and the aerodynamic noise associated with combustor design, to any degree, is an expensive, mult i-stage blade rows will be necessary to time consuming process. As labor and material costs soared during the late seventies, it became Important beginnings in the modeling of tur- increasingly evident that the cut-and-try-type bulent flows show promise as methods for repre- process of development was becoming prohibitively senting the effects of energy release in the costly. Improved design techniques were needed reacting, turbulent flows of combustors. Within that were founded on fundamental knowledge of the next decade, we can expect to see a signifi- aerodynamics, mixing, combustion and dilution. cant impact of these modeling developments in the Analysis must become the principal component of design procedures for real combustors.
the design method. Design changes are rendered more economically on paper than in sheet metal. The Turbine Early Design Methods A New Combustor Design Philosophy The first aircraft sas turbines were designed blade-row by blade-row. This aerodynamic Over the past decade significant progress has design process was similar to the early blade been made in modeling the processes that go on in element methods used in the design of axial flow compressors. The main emphasis was selection of a combustor. The modeling has involved definitive experiments to elucidate the processes and compu- the velocity diagrams entering and leaving the tational methods which simulate the better- blade row and the use of empirical loss data and understood physical and thermodynamic processes. flow blockage factors. Generally the endwalls were assumed cylindrical, although conical config- For example, one of the areas has been spray production. Laser diagnostic systems have been urations were considered for some multistage employed to characterize the drop sizes, distribu- applications. Inter-blade-row velocity diagrams tion and velocities. Compared to the early visual were calculated assuming simple radial equilibrium and in most cases free vortex distribution of the and wax-solidification methods used to get Sauter mean diameters as characterization of the spray, whirl component. Blade-to-blade velocity distri- these new laser measurements are a major step butions were then calculated with a simple two- forward. dimensional velocity gradient (stream filament) approach with assumed variations in total pressure Another important development in combustion loss and the mass flow parameter from the blade research is the ability to predict the combustion inlet plane to the blade exit plane. These calcu- flow field. Reference 8 reports on a method that lations were used iteratively with assumed blade models combustion in a two-dimensional combustor profiles until aerodynamically satisfactory shapes geometry. The foundation of the analysis is a were evolved.
random vortex method developed at the University of California-Berkeley. The method models turbu- The gas temperatures and pressures were low enough in the early gas turbines that uncooled lence from first principles, tracking the vor- ticity and predicting the interaction of the blades were adequate. As the cycle pressure vorticity with the bulk flow. The random vortex ratios and turbine inlet temperatures increased, method has been incorporated in a computer program cooled turbine blading became a requirement. The called MIMOC (Modeling the Interface Motion of gas-side heat transfer estimates for the turbine Combustion,) which can be ·used to predict the were based on empirical turbulent flow correla- reacting flow field behind a rearward facing tions for the flat plate representation of the step. This computer program evolved from a blade row. Inside the blade or vane coolant pas- cooperative effort between University of Cali- sages one-dimensional channel or pipe flow corre- fornia Berkeley and the Lewis Research Center. lations were employed in estimating the heat The output of this program can be coupled to com- transfer coefficient. By iterating between the puter graphics to render images of the combustion gas-side and coolant-side calculations, it was flow field. Figure 5 shows the images of the com- possible to estimate a wall temperature at a spec- bustion field in three time steps downstream of a ified blade or vane location. This early heat step in a two-dimensional channel. These transfer approach failed to comprehend the com- computer-generated images compare favorably with plexity of the convection mechanisms and the intricate geometry of the blades or vanes. It was Schlieren images of real combustion in a similar geometry. apparent in this time period that more sophisti- cated design procedures were needed for the heat Reference 8 represents one effort to model a transfer and aerodynamics of the turbine.
particular turbulent flow field. Currently there Research programs were initiated to correct this is an extensive effort within the computational deficiency. A comprehensive picture of the devel- fluid mechanics community in the modeling of tur- opment of the turbine design process during the bulent flows. The general progress in this area 50's and 60'1 and into the early 1970 decade is will have important impact on the modeling of com- documented. This three-volume set was com- bustion flows. In this paper we cannot attempt to piled by the staff of the Turbine Branch at the review this extensive body of work on turbulence Lewis Research Center to serve as a turbine design modeling and interpret the progress being made as handbook.
it effects combustion modeling. Another single example of a continuing effort in this area is the Current Design Methods turbulence modeling effort being sponsored by the Naval Research Laboratory.9 In this reference Currently, quasi-three-dimensional inviscid flow analyses are standard design tools for flow computational studies of unstable two-dimensional shear flows transitioning to turbulence have been analysis and these are used with empirical and reported. The first versions of the model include calculated boundary layer and endwall effects.
only convective terms but continuing efforts will The more modern design approaches place great include the effects of buoyancy and diffusion - reliance on knowledge about the entire flow path important physical phenomena in combustion. in the turbine blade passage. Viscous secon- dary flows and such boundary layer phenomena as Future Design Codes transition are considered in this design method- ology. The blade aerodynamic geometry of advanced Current efforts in computational fluid turbines is becoming more complex to allow for mechanics, instrumentation, and computer technol- high loading distributions over the span of the ogy hold promise of another period of rapid blades. As a further consequence of this geomet- advancement in turbine design technology. We ric complexity and higher loadings, the secondary won't see dramatic increases in turbine inlet tem- flows are an increasingly important part of the perature such as the 550 K (1000 OF) increase in overall flow pattern.
the 50's and 60's, but we will see greatly increased computer involvement in design optimiza- In addition to the turbine aerodynamics, tion and the simulation of component and full current thermal design procedure involves more engine operation. For the aerodynamic and heat advanced heat transfer predictions to define more transfer codes, we will see significantly more precisely the metal temperatures of the compo- accurate definition of boundary conditions because nents. Hot gas-side heat transfer calculation of improved high temperature instrumentation and methods range from the use of flat plate heat greatly improved modeling of combustors. Three- transfer correlations to fairly sophisticated dimensional viscous computer codes will be stan- boundary layer codes with various turbulence dard design tools, and the uncertainty margins in models, transition criteria, and provisions for heat transfer coefficients will be reduced by a geometric shapes such as curvature. Internal flow factor of three. This, in combination with accur- and heat transfer calculations for coolant side ate knowledge of the environment, should permit heat transfer typically use one-dimensional con- metal temperature prediction with an accuracy of vection models with empirical correlations for 14 to 28 K (25 to 50 OF) and greatly reduce the impingement, pin fins, and turbulence promoters.
time and cost of engine development.
Metal temperature prediction systems use these heat transfer calculations with three-dimensional Improved aerodynamic and thermodynamic tur- conduction codes to estimate the gas-side and bine design tools are needed to decrease the life coolant side wall temperatures along the blade cycle costs of operating engines. In addition to profile at several spanwise positions. The calcu- reducing development costs, these design methods lation is performed iteratively until prescribed will lead to increased hot part life, and reduced wall temperatures are met through adjustment of fuel consumption. There has been an awareness of the coolant-side conditions.
these specific needs for many years as incremental improvements in design techniques were continually While currently used design methods provide developed. Future improvements will depend upon turbines with high aerodynamic efficiencies, sub- improved instrumentation, new and improved compu- stantial uncertainties continue in the prediction ter codes, and the continuing improvement in com- of local metal temperatures. An analysis of the puters. Two-dimensional codes for annular ducts ability to ~redict turbine metal temperatures is and turbomachinery blade-to-blade flow surfaces discussed. Figure 6 depicts a learning curve are already in wide spread use, and at least one in turbine design technology and also shows an three-dimensional code has been used with some assessment of the current ability to predict metal success by one of the engine companies. Many temperatures. Note that gas-side heat transfer individuals and organizations are working on coefficient uncertainty is 35 percent while the diverse approaches to the three-dimensional prob- coolant side uncertainty is 25 percent. These lem. These codes, with various levels of sophis- uncertainties, combined with only approximate tication will be integrated into the design knowledge of gas and coolant temperature, lead to system. An inviscid three dimensional code an uncertainty of about 100 K (180 OF) in local presented has been tested against experimental metal temperature for the operating conditions of blade paIsage flow measurements in an annular current engines. This in turn leads to uncer- cascade. Generally the inviscid three dimen- tainty in life prediction of a factor of ten. It sional code predicted the features of the flow should be noted that a significant contributor to very well and is definitely more accurate than the this limitation on metal temperature prediction two dimensional codes used in turbine design.
ability is the lack of precise knowledge of the Figure 8 outlines the features that will be incor- real engine environment. There is a need for porated into the advance design codes. In addi- accurate measurements of temperature, pressure, tion to those listed, unsteady flow effects must and turbulence of the gases entering the turbine.
be handled in some way. The other needs have The price paid for this uncertainty in metal tem- already been discussed to some extent, and cer- perature prediction can be quite high in terms of tainly m~st be included in this list. The para- development cost as well as subsequent maintenance meter AN is commonly used to identify blade costs. Figure 7 shows a pie chart on the cost of root centrifugal stress level. This product of development of a new engine from information sup- annular flow area and the square of rotational plied by several of the gas turbine engine compa- speed is directly proportional to blade centrifu- nies. This representation averages the response.
gal stress given a material density and a taper The total engine development costs in 1979 ranged factor. Increases in currently limiting AN2 from $500 million to $1.2 billion. In 1985 this values are desired by all aircraft turbine design- would scale up to at least $600 million to $1.6 ers for the improvement of aerodynamic perfor- billion. Of that total cost, 10 to 40 percent mance. Improved materials and cooling are the were incurred in the core turbine. And two thirds keys to improvement here.
of that was in fixes. That amounts to about $50 to $500 million in core turbine changes through Concluding Remarks flight certification and perhaps one year of operating experience. Recent conversations with, Judging from current trends, the design of and research proposals from, the engine companies the principal components of a gas turbine engine indicate that this picture has not significantly is transitioning from an empirical, trial and changed.
error process to a more computationally-based These benefit~ will accrue because of: process. Computational codes which can describe steady, three dimensional fluid flow and heat 1. improved predictability of fuel sprays, transfer are becoming operational. Codes which 2. models of combustion reactions in a gas will comprehend unsteadiness are being thought of, stream but in 1984 nothing is availarrle which includes three dimensions along with the complication of unsteadiness. The ultimate in the design codes 3. three dimensional models of mixing.
must include both of these. The type and charac- teristics of the codes that will be used widely in The turbines will be able to accept higher enthalpy gas. Despite the more severe thermal the future cannot be predicted. Just a few years ago it appeared that implicit type calculations conditions of the working gas, more efficient use (solving a set of differential equations at each will be made of the cooling air bled from the com- step) appeared highly promising because of reduced pressor. Tip clearance losses will be minimized.
Each stage of the turbine will operate at a higher computer time and inherent stability. However, the introduction of the vectorized computers has work factor and the parasitic losses will be mini- mal. The more accurate prediction of wall temper- made the explicit methods (solve the unknowns one at a time at each step) appear attractive. From atures from codes will result in more reliable conversations with computational experts at the life predictions. This advance should mitigate Lewis Research Center, it appears that the archi- the turbine maintenance uncertainty which troubles the airlines in current operations.
tecture of future computers will be the principal deterministic factor affecting the overall charac- ter of the codes. The most successful and power- The greater capability to predict the operat- ful codes will make optimum use of computer ing conditions throughout the engine will encour- architecture. Former debates about implicit and age designers to make bolder changes between explicit coding may not be too relevant. engine designs. Innovation rather than just extrapolation will be practiced in proposing new designs.
The output information from computational fluid mechanics and heat transfer codes will be part of the input to the structural codes that All in all, the computational design method- will predict the life and durability of the engine ology will introduce a new era into the aviation components. In current design practice, the pre- gas turbine industry. In the international compe- diction of structural lifetimes is highly uncer- tition for markets, the companies which introduce tain because of a combination of inaccuracies in reliable, high performar.ce gas turbine engines will thrive. They will be the companies that have the prediction of local parameters such as temper- ature and pressure. Future codes are expected to adopted and implemented sophisticated computa- bring about major improvements in the precision of tional methods into the design procedure.
estimating component life.
REFERENCES The new code predictions will have to undergo rigorous verifications before they can be accept- Compressor Section ed. Experiments will be devised to validate and certify the accuracy of the analyses and the 1. Johnsen, I.A., and Bullock, R.O., "Aerodynamic Design of Axial Flow appropriateness of the physical modeling. Judging from recent experience in verification of codes, Compressors," NASA SP-36, 1965.
this will be a major effort. However, the rewards from the availability of comprehensive design 2. Wu, Chung-Hua, "A General Theory of Three- Dimensional Flow in Subsonic and Supersonic codes for the compressor, combustor, turbine and other components not discussed make the overall Turbomachines of Axial-, Radial-, and Mixed- effort to develop these codes very worthwhile. Flow Types," NACA TN-2604, 1952.
The saving of time and expense in the development of future gas turbine engines will justify this Analytical Approach for Multistage effort. These engines should be more durable and Turbomachinery Section reliable than their predecessors.
3. Adamczyk, J.J., "A Model Equation for More specifically, we can expect to see some Simulating Flows in Multistage Turbo- of the following improvements in the major compo- machinery." Proposed TM.
nents of future engines. The compressors will exhibit higher pressure ratios per stage and will 4. Adkins, G.G., and Smith, L.H., "Spanwise Mixing in Axial-Flow Turbomachines," ASME be less susceptable to stall and surge over their range of operation. Structurally, the blades will Paper 81-GT-57, American Society of be less likely to encounter flutter and tuned Mechanical Engineers, 1981.
vibrations.
The Combustor Design Practice The combustors will be capable of operation at higher pressures and outlet gas temperatures. 5. Lefebvre, A.H., "Gas Turbine Combustion," More uniform temperature pattern factors will be McGraw-Hill, New York, 1983.
realized at the combustor exit. The future com- bustor will be more tolerant of the varying speci- 6. "Basic Considerations in the Combustion of fication of fuels including synthetic types. Hydrocarbon Fuels with Air," NACA Report 1300, 1959.
7. Norgren, C.T., and Childs, H.H., "Effect of 12. Boyle, R.J., Haas, J.E., and Katsanis, T., Liner Air-Entry Holes, Full State, and Com- "Comparison Between Measured Turbine Stage bustor Size on Performance of an Annular Performance and the Predicted Performance Turbojet Combustor at Low Pressures and High Using Quasi-3D Flow and Boundary Layer Air-Flow Rates," NACA RM E52J09, Jan. 1953. Analyses," NASA TM-83640, June 1984.
8. Ghoniem, A.F., Marek, C.J., and Oppenheim, 13. Stepka, F.S., "Uncertainties in Predicting A.K., "Modeling Interface Motion of Turbine Blade Metal Temperatures," ASME Paper Combustion (MIMOC) - A Computer Code for 80-HT-25, American Society of Mechanical Engineers, July 1980.
Two-Dimensional, Unsteady Turbulent Combustion," NASA TP-2132, August 1983.
14. Denton, J.D., and Singh, U.K., "Time Marching 9. Boris, J.P., Oran, E.S., Fritts, M.L. and Methods for Turbomachinery Flow Calculation," Oswald C., "Time Dependent, Compressible Application of Numerical Methods to Flow Simulations of Shear Flows: Tests of Outflow Calculatlons in Turbomachines. Von Karman Boundary Conditions," NRL-MR-5249, Naval Institute for Fluid Dynamics Lecture Series Research Laboratory, Dec. 1983. 1979-7, 1979.
Future Turbine Design Codes 15. Goldman, L.J., and Seasholtz, R.G., "Compari- son of Laser Anemometer Measurements and 10. Stewart, W.L., "Analytical Investigation of Theory in an Annular Turbine Cascade With Multistage-Turbine Efficiency Characteristics Experimental Accuracy Determined by Parameter Estimation," NASA TM-82860, Nov. 1982.
in Terms of Work and Speed Requirements," NACA RM-E-57K22b, Feb. 1958.
11. Glassman, A.J. ed., "Turbine Design and Appli- cation," NASA SP-290-Vol. 1, 1972; -Vol. 2, 1973; -Vol. 3, 1975.
~ BLADE ELEMENT ~"
""
""
""
" "
"
"
FLOW " SURFACE OF REVOLUTION ----.
"
FLOW HUB Figure 1. - Compressor blade element design approach.
INLET RELATIVE MACH NUMBER, Min
o 0.30
o .45
o .60
'.3- L'::, .80
.3
LU • 0::1- :::>2 Vl!::!
.2 VlU LU - o::t±: 0... LU U .1 1. 00 .80 .60 .4 .5 .6 .7 .8 .9 FLOW COEFFICIENT, rp Figure 2. - Typical compressor dimension- less performance map.
BLADE (MERIDL) FLOW MID CHANNEL 1- SURFACE S2
"
I L BLADE- TO-BLAD SURFACE SI (TSONIC) Figure ~. - Analysis surfaces in a turbomachine • • FULLY 3-DIMENSIONAL FLOW • CHEMICAL REACTION/HEAT RELEASE • HIGH TURBULENCE LEVELS • 2 PHASE WITH VAPORIZATION Figure 4. - Combustor flow phenomena.
Figure 5. - Computational graphical representation of combustion process downstream of a 2D step. Three ti me intervals. CD-83-13778 DATA BASE; LOSSES AND HEAT TRANSFER, QUASI 3D FLOW ANALYSIS >- <..:> o z :c u UNCERTAINTIES ~ • 35% GAS SIDE HEAT TRANSFER • 25% COOLANT SIDE HEAT TRANSFER • 180 F IN LOCAL T M • lOx IN PREDICTED LIFE 1950 1960 1970 1980 YEAR Figure 6. - High pressure turbine technology advance since 1950.
-CORE TURBINE COOLING DESIGN CHANGES 0600 m TO 1600 m $ • 30 TO 4(J'fo IN CORE TURBINE .2/30FTHIS IN "FIXES" Figure 7. - New engine development cost, - 3D VISCOUS FLOW CODES - LEADING EDGES - TRAILING EDGES - ROTATION - HEAT TRANSFER - COOLANT ADDITION - FLOW AND HEAT TRANSFER CODES FOR LOCAL AREAS - COOLING HOLES AND SLOTS - COOLANT PASSAGES WITHIN VANES AND BLADES -FLOW AND HEAT TRANSFER CODES FOR DISI< CAVITIES -WEll DEFINED TURBINE ENVIRONMENT - TEMPERATURE AND PRESSURE PROFilES - TURBULENCE & UNSTEADINESS - COOLANT FLOW CONDITIONS AND DISTRIBUTIONS -REDUCED TIP CLEARANCES AND SENSITIVITY -INCREASED AN2 (PRODUCT OF ANNULAR FLOW AREA & (SPEED)2) Figure 8. - Long range needs in turbine design technology.
r----------------------------r------·---------------------,-------------------------, 1. Report No. 2. Government Accl'ssion No. 3. Recipient's Catalog No.
NASA H1-86865
4. Title and Subtitle 5. Report Date
Computational .Thermo-Fluid Dynamics Contributions to
Advanced Gas Turbine Engine Design
6. Performing Organization Code
505-31-42
7. Author(s) 8. Performing Organization Report No.
E-2194
Robert W. Graham, John J. Adamczyk, and Harold E. Rohlik
~1~0.~w~0~rk~u~ni~t~No-.--------------~ 9. Performing Organization Name and Address 11. Contract or Grant No.
National Aeronautics and Space Administration
Lewis Research Center
Cleveland, Ohio 44135
13. Type of Report and Period Covered
1-----.-------------- ------------------ ... - ---.-----
12. Sponsoring Agency Name and Address
Technical Memorandum
National Aeronautics and Space Administration
14. Sponsoring Agency Code
Washington, D.C. 20546
15. Supplementary Notes
PreparE!d for the Twenty-third Aerospace Sciences Meeting sponsored by the American
Institute of Aeronautics and Astronautics, Reno, Nevada, January 14-17,1985.
16. Abstract
The early history of design practice for the aircraft gas turbine engine can be
characterized as the application of empiricism founded on actual test experience ..
Trial and error processes gave the designer the needed information to modify or
redesign the principal components for more stringent operating conditions but the
changes between engine models were conservatively small. In many cases, this
approach was both extremely costly and time consuming. A sizeable fraction of the
cost of marketing a new engine went into equipment modification and the iterative
substitution of design changes. Aircraft gas turbine engines being envisioned for
the last decade of this century and the early decades of the next century will
operate at higher pressures and temperatures. Tolerances and clearances will be
tighter than current practices. Design practice will rely heavily on computa-
tional codes to predict gas flows and thermal conditions throughout the flow path
of the engine so that iterative changes in the development of the engine will be
minimized. The introduction of more computational methods into the design process
will enable bolder changes in the evolution of new engines. Improved cycle effi-
ciency, quieter operation and more reliable performance will be among the benefits
derived from this design practice.
17. Key Words (Suggested by Author(s» 18. Distribution Statement
Computational fluid dynamics
Unclassified - unlimited
Gas turbine design
STAR Category 07
19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of pages 22. Price'
Unclassified Unclassified
* For sale by the National Technical Information Service, Springfield, Virginia 22161