Printed in the United States of America. Available from National Technical Information Service U.S.

Department of Commerce 5285 Port Royal Road, Springfield. Virginia 22161 NTIS price codes-Printed Copy: A03: Microfiche A01

This report was prepared as an account of work sponsored by an agency of the United States Government. Neither theU nited StatesGovernment nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise, does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

ORNL/CON-135

Engineering Technology Division

A REVIEW OF STIRLING ENGINE MATHEMATICAL MODELS N. C. J. Chen F. P. Griffin

Date Published - August 1983

Prepared by the OAK RIDGE NATIONAL LABORATORY Oak Ridge, Tennessee 37830 operated by UNION CARBIDE CORPORATION for the U.S. DEPARTMENT OF ENERGY under Contract No. W-7405-eng-26

iii
CONTENTS Page ABSTRACT 1. 2. ......................................................... ................................................. ..................................... ............... ................ 1 1 3 3 3 4 4 5 5 7 8 8 8 9 9 10 10 11 12 13 13 13 14 15 16 16 17 18 18 19 20 20 22

INTRODUCTION

REVIEW OF DESIGN METHODS 2.1 2.2

First-Order or Approximate Design Methods Second-Order or Decoupled Design Methods 2.2.1 2.2.2 2.2.3

Isothermal analysis .............................. Adiabatic analysis ............................... Semi-adiabatic analysis .......................... .....................

2.3 2.4 3.

Third-Order or Nodal Design Methods Method of Characteristics

...............................

REVIEW OF MODELS 3.1

..... ........................................ .............................

Second-Order Design Methods 3.1.1 3.1.2 3.1.3 3.1.4 3.1.5 3.1.6 3.1.7

Model by Martini (1978) - isothermal analysis .... Model by Qvale (1967) - adiabatic analysis ....... Model by Rios (1969) - adiabatic analysis ........ Model by Lee et al. (1981) - adiabatic analysis . Models by Shoureshi (1982) - adiabatic and isothermal analyses .............................. Model by Heames (1982) - adiabatic analysis ...... Model by Feurer (1973) - semi-adiabatic analysis ........................................ ..............................

3.2

Third-Order Design Methods 3.2.1

Model by Finkelstein (1975) - less rigorous analysis ........................................ 3.2.2 Model by Tew et al. (1978) - common pressure analysis ....................................... 3.2.3 Model by Giansante (1980) - common pressure analysis ........................................ 3.2.4 Model by Chiu et al. (1979) - less rigorous analysis ........................................ 3.2.5 Model by Azetsu et al. (1982) - less rigorous analysis ...... .. ................... 3.2.6 Model by Vanderbrug (1977) - less rigorous analysis .................. ................... 3.2.7 Model by Urieli (1977) - rigorous analysis ....... 3.2.8 Model by Schock (1978) - rigorous analysis ....... 3.2.9 Model by Gedeon (1978) - rigorous analysis ....... 3.2.10 Model by Zacharias (1977) ....................... 3.3 Method of Characteristics 3.3.1 3.3.2 ............................... ............................ ...........................

Model by Organ (1981) Model by Larson (1981)

iv Pane 4. 5. SUMMARY ...................................................... ............................. ..... 23 26 29 30

CONCLUSIONS WITH RECOMMENDATIONS

ACKNOWLEDGMENTS REFERENCES

.................................

.......................................................

A REVIEW OF STIRLING ENGINE MATHEMATICAL MODELS N. C. J. Chen F. P. Griffin

ABSTRACT

As requested by the Department of Energy, a review of existing mathematical models for Stirling engine thermodynamic Twenty-five models were identianalysis has been performed. fied through extensive literature search; 19 of these were Each individual published in sufficient detail for review. model's assumptions, limitations, predictability, and applicability were assessed by using a two-part review format consisting of model description and validation. According to their design methods, models were grouped into four categoapproximate (first-order) ries by degree of sophistication: methods, decoupled (second-order) methods, nodal (third-order) The salient characanalyses, and method of characteristics. teristics of the models were summarized in two tables for cross-reference. In the course of this review, these points were established. 1. Utilization of a detailed design method does not enThere is no evidence that sure enhanced model performance. the existing third-order analyses are superior to the secondorder methods. 2. Model validation is largely limited to kinematic enWith increasgines with emphasis on thermodynamic analysis. ingly important applications for free-piston Stirling engines, it is highly recommended that dynamic analysis should be integrated into thermodynamic study in future modeling efforts. 3. To achieve an in-depth evaluation of the individual model's assumptions and validation would require model acquisition and more abundant experimental data than were available for this review. 4. The ranking of the various models is not possible by this review. Only when all models can be run with a common set of input data and compared with well-defined experimental data can a fair or valid comparison be made.

1.

INTRODUCTION

As requested by the Department of Energy (DOE),

screening of the ex-

isting computer programs for Stirling engine analysis has been performed. This report is intended to provide a user guide for quick reference to the

2 existing computer codes. analysis were reviewed. Only those programs related to thermodynamic Although dynamic analysis is it critical to free-

piston Stirling engine studies,

will not be discussed in this report. total), some 19

Among the existing computer programs identified (25 were published in ready evaluated, Urieli
s

sufficient detail

for reviewing; of these,

1 4

10 were al-

most extensively by Martini, -

and to a lesser extent by review through some

and Walker.6

We will conduct a state-of-the-art even though in

an independent assessment,

the process of reviewing, (ORNL),

degree of overlap among Oak Ridge National Laboratory Urieli, and Walker is inevitable.

Martini,

Those models reviewed were grouped by their engine design methods and basic assumptions for cycle analysis. fied: order), approximate (first-order), Four distinct methods were identi(second-order), nodal (third-

decoupled

and method of characteristics. evaluations.

First-order methods are good for

back-of-the-envelope

Second-order analyses are good for inThird-order methods are very detailed a way that would be diffiThe method of characThese design methods

teractive design and optimization.

and can be used to simulate engine operation in cult if

not impossible to measure experimentally. based on the theories of gas dynamics. Sect. 2.

teristics is

are defined and reviewed in In Sect. 3,

models will be reviewed one by one according to a premodel description and model

devised format containing two major parts: validation.

Model description will further discuss assumptions and limiComIn

tations; model validation includes predictability and application. ments on individual model performance Sect. 4,

are provided wherever possible.

two extensive tables that summarize the significant features of The tables include informamodel classification, Table 1 sum-

the models are presented for cross-reference. tion such as principal investigator,

affiliation,

code listing availability,

model validation,

and references.

marizes the features of seven second-order models, information about ten third-order models.

and Table 2 includes

Conclusions with recommendations are furnished in a bibliography essential to model review is

Sect.

5.

Finally,

attached for further studies.

2.

REVIEW OF DESIGN METHODS

The four identified engine design methods are first-order, order, third-order, and the method of characteristics.
3 7 8

second-

The definitions

given below are similar to those of Martini

and Organ. .

2.1

First-Order or Approximate Design Methods

First-order design methods are used for back-of-the-envelope engine performance predictions. an ideal loss-free analysis, Martini
3

Stirling

Calculation of power output starts with

9

such as the Schmidt equation published by A simple cor-

or the generalized Beale number derived by Senft.

rection factor is power output.

then used to find the brake power output from the ideal brake efficiency is usually computed from a cor-

Similarly,

rected Carnot efficiency.

The corrections for all of the various losses

in a Stirling engine are consolidated into generalized correction factors. These efficiency and power correction factors are determined from experience with real engines. For example, most well-designed Stirling engines First-

achieve a brake efficiency that is

50 to 70% of the Carnot value.

order analyses provide a quick way to estimate the relationship between the overall size of an engine and its power output, but they are not very useful as detailed design tools for Stirling engines.

2.2

Second-Order or Decoupled Design Methods

This design method begins with a simplified cycle analysis to determine a basic power output and heat input. subtracted from the basic power output, Various power losses are then

and heat losses are added to the T rj-or-i.provt-

heat input to arrive at a net performance prediction. [ment of the secon -orer methos relative
(-o isth-t-in- -~d'iVij--l~-s-m e1h-an-is
-

~te-fTrst-order designmejthi
quantifie-d

s-re t-ent-i'efi-a~xd-

Powe rLQss-e

can include fluid and mechanical losses in cylinders,

friction,

transient heat

transfer (hysteresis) seals.

and gas leakage past piston wall conduction, and

(e-atossesinclude

displacer shuttle losses,

4 imperfect heat transfer in regenerators. ods, it is

In all second-order design meth-

assumed that the energy losses are not dependent on each other;

that is,

they are decoupled.

Second-order design methods may be further subdivided into three categories according to the way the variable gas volumes are handled in the simplified cycle analysis: _sthE-iE adi, adiabat-fic, and ('f-'i=a'dTiba-t'ic .

These terms were derived according to-h't-t'ansf:errat-ee-taIs spaces-a.d.i-enggine._cyl-inde-s.tJ If On the other hand, is if the rate is the rate is zero, it is infinite, adiabatic. it is isothermal.

Semi-adiabatic

the process somewhere in between with a limited heat transfer rate. Isothermal analysis This analysis is based on the classical Schmidt isothermal is cycle,

2.2.1

which,

by allowing for sinusoidal volume variations,

a slightly more

realistic form of the ideal Stirling cycle. space

All gas in the expansion and all gas in the in-

is maintained at the heat source temperature,

compression space

is maintained at the heat sink temperature because

finite heat transfer coefficients are assumed. also assumed (i.e., temperature in the local gas temperature is and there is

Perfect regeneration is equal to the local wall

the regenerator,

no axial heat conduction). the expansion space, and

All heat input to the isothermal all heat output occurs in

cycle occurs in

the compression space.

A simple closed form

solution exists for the Schmidt cycle. 2.2.2 Adiabatic analysis The adiabatic cycle assumes that the compression and expansion spaces are perfectly insulated. All heat input to the cycle occurs in the heater, the cooler. Gases leave the heater at the

and all heat output occurs in

heat source temperature and are mixed perfectly as soon as they enter the expansion space. Similarly, gases leave the cooler at the heat sink tem-

perature and are mixed perfectly as soon as they enter the compression space. Again, perfect regeneration is assumed. The adiabatic cycle is a

more realistic simplification of a Stirling engine than the Schmidt cycle, especially for large engines operating at high frequencies. isothermal However, an

second-order analysis can be just as accurate as an adiabatic

5 second-order analysis as long as proper adiabatic loss terms are subtracted from the isothermal cycle predictions. cycle requires a simple numerical integration. 2.2.3 Semi-adiabatic analysis Semi-adiabatic cycles allow for nonzero, ficients. stein,'
1

Solution of the adiabatic

finite heat transfer coeffirst analyzed by Finkel-

The simplest semi-adiabatic cycle,

accounts for heat transfer in the expansion and compression The wall temperatures of these volumes are assumed to be constant to the heat source and heat sink temperacooler, and regenerator are assumed to effi-

spaces.

with respect to time and equal tures, respectively.

The heater,

behave perfectly.

This semi-adiabatic cycle can actually result in

ciency predictions that are lower than either the purely adiabatic cycle or the isothermal cycle. This is caused by irreversible heat transfer

losses across the temperature difference between the gas and the cylinder walls in the compression and expansion spaces. Solution of this semi-

adiabatic cycle requires a simple numerical

integration.

2.3

Third-Order or Nodal Design Methods also known as nodal analyses, consist of

Third-order design methods, three basic procedures: control volumes; mass, and (3) momentum, (2) (1)

divide the engine into a network of nodes or equations for conservation of

set up the differential

and energy,

plus equation of state for the working gas;

solve simultaneously the system of difference equations by some There are two subclasses under this method: less rigorous. The rigorous third-

adequate numerical method. one, most rigorous,

and the other,

order analyses solve all the equations except for the use of steady flow correlations for heat transfer and friction flow because no correlations of universal validity exist for unsteady flow in today's technology. less rigorous third-order models simplify the numerical The

computations by equations. It

omitting some of the terms from the governing differential is

assumed that certain losses can be decoupled from the main calculation There are three common simplificathe momentum equation, but flow

to improve the speed of computations. tions: (1)

inertial terms are ignored in

6 friction terms are retained; ignored, is that is, (2) both inertial and flow friction terms are not used and a uniform pressure kinetic energy terms are ignored

the momentum equation is the engine; and (3)

assumed throughout

in the energy equation. All of the nodal design methods use finite differencing of the spatial derivatives to convert the partial differential of ordinary differential equations equations to a system

(with only time derivatives remaining).

Each conservation equation is node.

represented by a difference equation at each

The numerical methods for solving this system of ordinary differenexplicit (forward-differIn the explicit

tial equations are divided into two categories: encing) and implicit integrations, ture) (backward-differencing)

techniques.

the thermodynamic information (such as pressure and temperacomputed from time derivatives that were evaluated The simplest explicit method is techniques, the Euler method, may be

at a new time is

at the previous time. although more accurate used.

such as the Runge-Kutta method,

Explicit techniques are sometimes plagued by numerical oscillations especially if time steps are too large. In contrast,

and instabilities,

an implicit integration is

always numerically stable.

The implicit method

solves the system of ordinary differential

equations by computing the therthat are evalu-

modynamic information at a new time from time derivatives ated at the new time.

A large matrix must be inverted at each time step. instabilities, implicit integrations can but it may

Because of the lack of numerical use larger time steps.

This reduces computer execution times,

also reduce the accuracy of the numerical approximation. Third-order design methods attempt to consider the many different complex processes coexisting in a Stirling engine. It is hypothesized

that the various processes assumed to be decoupled in sign methods do in tion is studies. reality significantly interact.

the second-order de-

Whether this assumpand experimental and by far that they

true remains to be seen after further theoretical

The third-order methods are the most sophisticated, computer time; but there is In fact, no evidence

the most expensive in give the best results.

the results from second-order codes are Furthermore, some

at least as good when compared with experimental data. workers have questioned the mathematical methods: it is

foundation of the third-order

believed that under certain circumstances the solutions

7 may converge to values that are mathematically and computationally but do not correspond to a real physical state. In any case, stable,

more experi-

mental data are required for a fair assessment.

2.4

Method of Characteristics solves systems of nonlinear partial

The method of characteristics differential

equations of hyperbolic type by determining the characterThe characteristic curves are used to equations into a system of ordinary curves.

istic curves for the equations.

transform the partial differential differential

equations that are valid only along the characteristic

This method has been used successfully in

the study of compressible gas unsteady

flow and has been applied to the analysis of one-dimensional, flow in Stirling engines. In one-dimensional, unsteady flow, the characteristic

curves are in

the position-time plane on which the partial derivatives with respect to position and time of the fluid properties temperature) continuities. are indeterminate and may, (such as density, velocity, and

therefore,

undergo arbitrary disthe con-

To establish the conditions for indeterminacies, (mass, momentum, energy)

servation equations

along with the total differ-

entials of the fluid properties are expressed in matrix notation with the partial derivatives of the fluid properties as the dependent variables. The characteristic curves are found by setting the determinant of the coFor more information on the theory and readers
12

efficient matrix equal to zero. excellent books by Shapiro l

l

applications of the method of characteristics,

should refer to

and Leipmann and Roshko.

The method of characteristics can be applied at different levels of complexity to Stirling engine analyses. In rigorous analyses, all three

conservation equations are solved simultaneously. ses, however,

In approximate analy-

some simplifying assumptions are used to solve one of the The two remaining conservation

conservation equations independently.

equations are then solved simultaneously by the method of characteristics.

8 3. REVIEW OF MODELS

Assessments of the 19 models reviewed are presented. methodology consists of model description and validation.

The reviewing The model

descriptions present further their basic assumptions and limitations; model validation discusses predictability and applicability. models fell naturally into the three major design methods: order, ten third-order,
13

Grouping of seven second-

and two methods of characteristics. but not

Six other Stirling engine computer models were identified, reviewed. Rauch has described a model that is
14
T

based on a second-order

design method.

Berggren

and Andersen s have developed models that utiSirettl6 has developed a model that

lize third-order design methods.

solves the complete set of conservation equations using the method of characteristics. Inc., Vincent et al.
1 7

of Energy Research and Generation, but insufficient

give a brief description of a thermodynamic model,

8

details are provided to allow a review. at Harwell,l

Models have also been developed

and some are thought to exist at Philips, but the authors

of this report have not acquired any documentation that describes these models.

3.1

Second-Order Design Methods this category: one isothermal, five adia-

There are seven models in batic, 3.1.1 and one semi-adiabatic. Model by Martini (1978) Martinil' model.
3

isothermal analysis

has published detailed documentation of his second-order Martini assumed that the time-dependent gas tem-

In his analysis,

peratures in

the expansion and compression spaces of an actual Stirling temperatures. The ef-

engine can be expressed as time-averaged effective

fective hot gas temperature will be less than the heater temperature and the effective cold gas temperature greater than the cooler temperature. These temperatures were derived from the computed heat transfer coefficients in both the gas heater and gas cooler as well as from the computed heat requirement. detail by Martini. An iterative procedure is needed as described in great

9 To validate the model, Martini 3 applied his code to two reference engines, GPU-3 and 4L23, both of General Motors (GM). When compared with

the experimental values for the GPU-3 and the values predicted by GM for the 4L23, Martini's calculated power and efficiency were found to be within 20% error bands, if no correction factor for flow resistance is used. Considerable improvement (reducing the error bands by half) can be

made if either (1) a correction factor of about 2.9 is applied to the flow resistance coefficients, or (2) the computed heat transfer coefficients are adjusted by a factor of 0.8. 3.1.2 Model by Qvale (1967) - adiabatic analysis
1 Qvale's' 9 20

second-order model is based on an idealized adiabatic The pressure changes, mass

cycle that has no friction or seal leakage.

variations, piston displacements, and volume changes are all assumed to be sinusoidal. The problem was formulated with pressure, temperature, and Thus, the piston displacements are comThis type of

mass as the independent variables.

puted values that depend on the three independent variables.

analysis is more suitable for engine synthesis than for performance predictions of a specific engine. Ovale
19

validated his model by comparing it with the Allison PD-67A His predictions for heat input, work out-

experimental Stirling engine.

put, and indicated efficiency compare favorably with the test data over a range of engine speed (1500 to 3000 rpm). 3.1.3 Model by Rios (1969) - adiabatic analysis Both Rios and Qvale did their graduate work for Professor J. L. Smith at Massachusetts Institute of Technology (MIT). work on their adiabatic second-order model. Rios 2 1 expanded Qvale's

Rios used the same basic as-

sumptions as Qvale but changed the formulation of the problem so that nonsinusoidal piston displacements (such as those resulting from crankshafts with short connecting rods) could be specified. Rios' graduate work was applied to Stirling refrigerators. Martini
3

However,

obtained the Rios computer code and modified it to suit a StirMartini then compared the modified Rios case The code overpredicted brake

ling engine application.

with 18 data points from the GM 4L23 engine.

power and efficiency by an average of 24 and 16%, respectively.

It"

10 3.1.4 Model by Lee et al. The Lee et al.

22

(1981)

- adiabatic analysis an application a unique

model of Foster Miller Associates is In this model,

of the Rios adiabatic second-order analysis. power loss mechanism was introduced. This is

the cyclic heat transfer

loss (or gas spring hysteresis loss) that results from periodic heating and cooling of the working gas near the gas-wall interfaces ders, manifold spaces, connecting tubes, engine, and reservoirs. inside cylin-

In their appli-

cations to the Viking-1

Lee et al. 2 3 were able to quantify the

cyclic heat transfer loss. (adiabatic, it

Among the four major power losses identified pressure drop, and heat exchanger AT), signifi-

cyclic heat transfer,

was shown that cyclic heat transfer loss was ranked second in contributing about one-third of the total power loss.
22

cance,

The Lee et al. Inc.,

model was refined and verified by the Sunpower, For model validation, the model compared

third-order analysis.

fairly well to the GPU-3 test data. power and brake efficiency by <15%, 2.5 was applied to flow resistance. 3.1.5 Models by Shoureshi (1982) isothermal analyses -

The model overpredicted the brake provided that a correction factor of

adiabatic and
24

With objectives in low temperature-ratio developed two Stirling engine mathematical

applications,

Shoureshi

models.

Roth models are

second-order design methods and are called the complete model and the simplified model. The complete model However, is based on Rios' adiabatic second-order analysis. mechanfric-

updated correlations for two important losses were used: For mechanical

ical friction and transient heat transfer losses. tion, Shoureshi developed a correlation that is

based on internal combusthe cylinders,

tion engine data.

For transient heat transfer losses in

Shoureshi provided an alternate approach that excluded heat transfer enhancement factors as originally derived by Lee et al. (see Sect. 3.1.4).

To achieve an efficient optimization design method and a closed-form solution, Shoureshi

2 5

developed a simplified model.

This method involves

the Schmidt isothermal analysis plus a two-step correction for the net power output and heat input. In the first step, an adiabatic correction

ftahl

11 (the correction from Schmidt isothermal engine analysis to adiabatic analysis with perfect components) was introduced. The second step involved These loss terms were

deductions of all identifiable decoupled losses.

similar to the ones used in the adiabatic analysis of the complete model. The procedure of the first step was to derive appropriate corrections for the Schmidt isothermal work output with three factors: corrections for

temperature ratio, phase-angle difference between the displacer and the piston, and dead volumes. These factors were obtained by comparing the

computed Schmidt work output with that computed from the adiabatic analysis in the complete model, which was assumed to be a reference model by Shoureshi. In addition, a correction for Carrot efficiency was determined In the process of the second step,

as a function of temperature ratio.

various losses, expressed in closed-form solutions, were further deducted from the basic work output and added to the heat input to obtain the net work output and heat input. To verify the complete model, Shoureshi 24 compared his predictions with measurements from the following high-temperature engines: Allison, and GPU-3. Philips,

It was shown that the complete model predicted engine Similar conclu-

performance within the range of experimental uncertainty. sions were claimed for the simplified model predictions. 3.1.6 Model by Heames (1982) Heames et al.
26

adiabatic analysis

of Argonne National Laboratory (ANL) developed a userThe computer program consists of

oriented Stirling engine analysis code. four-modules:

(1) input processor, (2) output processor, (3) standard The input and output modules

function module, and (4) analysis module.

use a flexible format that simplifies data entry and retrieval for many different Stirling engine configurations. The two modules also provide

the capability to specify multiple computer executions for parametric studies. The output module saves the parametric results on an external file The standard function module is a

for users with graphics capabilities.

library of subprograms that provides the user with numerous correlations and functions that are used commonly in Stirling engine analyses. In-

cluded in the function module are (1) temperature-dependent correlations

12 for the physical properties of many different fluids and metals; (2) fric-

tion factor and heat transfer correlations for many different heater, cooler, and regenerator configurations; (3) engine heat loss correlations and (4) sub-

such as shuttle heat transfer and cylinder wall conduction;

routines to compute cylinder volume variations for different types of crank drive mechanisms. The analysis module contains the Stirling engine A user will eventually be able to select from However, the only analybased on

thermodynamic computations.

a number of different thermodynamic algorithms. sis method in Rios'

the present edition of the computer program is

adiabatic second-order design method.

A tape copy and a user's man-

ual for ANL's Stirling engine design code will be available upon request through the National Energy Software Center. ANL's computer code has been validated against GPU-3 engine data. experimental

The computer predictions for indicated power output compare However, the computer program ap-

favorably with the experimental data.

pears to overestimate efficiency by as much as five percentage points, especially at low engine speeds. 3.1.7 Model by Feurer (1973) semi-adiabatic analysis

Philips has published very little ing activities. However,

about their Stirling engine model7

a paper by Feurer 2

of Entwicklungsgruppe Stirdisclosed their semian adiabatic cycle that in the cylinders,

lingmotor MAN-MWM (MAN/MWM),

a Philips licensee, This cycle is

adiabatic second-order analysis. allows for nonzero, regenerator, first

finite heat transfer coefficients

and heat exchangers.

The power output and efficiency are then are corrected for: (1) losses

calculated based on this cycle, crank motion, (2)

due to nonsinusoidal

residual adiabatic losses that the (3) flow fric-

simplified heat transfer coefficients do not account for, tion losses, losses. In his theoretical studies, (4) mechanical friction losses, and (5)

static conduction

Feurer showed that all the loss mechaOne major conclusion is that the maximum

nisms are phase-angle dependent.

efficiency and the maximum power output do not occur at the same phase angle. Model validation has not been found in the open literature because

of proprietary controls.

13 3.2 Third-Order Design Methods six use less rigorous meth-

There are ten models in this category:

ods, three use most rigorous methods, and the simplifications in the final one (model by Zacharias, Sect. 3.2.10) were not stated clearly. 3.2.1 Model by Finkelstein (1975) - less rigorous analysis As a pioneer in Stirling engine analysis, Finkelstein developed a third-order method in the early 1960s. recent version published in 1975. Finkelstein's 2 8 model is a third-order design method, but less rigorThis review is based on a more

ous than that of Urieli (see Sect. 3.2.7), which will be described later. Finkelstein made two major assumptions in his derivation of the governing differential equations. the energy equation. First, the gas kinetic energy term was ignored in

Second, the momentum equation was reduced to the

form of an equivalent orifice equation. Finkelstein divided both the engine components and gas spaces into nodal networks. and mass. The nodes were treated as regions of variable temperature

Energy and mass transfer between nodes resulted from the comAll paths for conduction, Finkelstein used a special

puted temperature and pressure differences. convection, and mass transport were included.

technique to reduce the convergence time for finding a cyclic, steady state, nodal temperature distribution. This was accomplished by manually

adjusting the temperatures of each metal node at the end of every piston

revolution based on net heat balances for the nodes.

For example, if a

nodal heat balance across one piston cycle shows that there is a net flow of heat into a metal node, then the temperature of the node is adjusted upward. Finkelstein reported that his model has been validated, but the results are not available for evaluation. However, his program is now com--

mercially available for general use on the CDC Cybernet computer system. 3.2.2 Model by Tew et ai. (1978) - common pressure analysis

The Tew et al. 2 9 '3 0 model of the National Aeronautics and Space Ad-

ministration (NASA) Lewis Research Center is a less rigorous third-order

14 design method. They used three basic equations (conservation of mass,

energy, and equation of state) to determine the thermodynamics of the gas (temperature and mass distributions and pressure level) at each of 13 nodes. Several simplifications were used to minimize the numerical inte(1) the momentum equation was totally ignored and a common

gration times:

pressure throughout the 13 gas nodes was assumed during each time step, (2) kinetic energy was neglected in the energy equation, and (3) the three processes that contribute to gas temperature changes (pressure changes, gas mixing, and heat transfer) were treated independently. The numerical

integration required 30 to 40 piston cycles to approach cyclic steady state distributions of gas mass and regenerator metal temperature. Pres-

sure drop, shuttle, and conduction losses were accounted for by including these calculations only during the last cycle of the numerical integration. In the last cycle, the common pressure at each time step was asThe compression and ex-

sumed to exist at the center of the regenerator.

pansion space pressures were then calculated by estimating the pressure drops in each control volume from steady state correlations and summing these pressure drops. The net area enclosed by the pressure-volume curves

of the compression and expansion spaces is equal to the indicated work output. Thus, the pressure drops are decoupled from the calculation of

mass and temperature distributions, but they do affect the work output predictions. Tew et al. 3 1 compared their theory with experimental data from the GM GPU-3 Stirling engine. When the regenerator friction factor was increased

by factors of 4.0 and 2.6 for hydrogen and helium, respectively, the model overpredicted both brake power and efficiency by 5 to 30%. Tomazic 32 also

compared the NASA-Lewis model with experimental power output data from the USS P-40 Stirling engine. For this comparison, measured flow resistances The predicted brake

were used in the model rather than computed ones.

power values were consistently high for all engine speeds (500 to 4000 rpm) and pressures (4 to 15 MPa). 3.2.3 Model by Giansante (1980) - common pressure analysis Giansante 3 3 of Mechanical Technology, Inc. (MTI), acquired NASALewis' code and applied it to the free-piston Stirling test engine built

15 for DOE. Because the NASA-Lewis code was developed specifically for kine-

matic type engines, (1)

the following modifications have been accommodated:

gas springs for the free pistons were modeled as separate control vol(2) centering ports for the free were

umes with heat transfer and seal leakage,

pistons (these eliminate free-piston migration due to seal leakage) simulated in a way similar to the NASA-Lewis method, to predict free-piston dynamics was implemented. and (3)

a subroutine

Two interesting types of

load were a linear alternator and a velocity-cubed dissipator. MTI compared their free-piston Stirling model with two experimental points (high and low powers) Engine. In the comparison, from their DOE 1-kW Technology Demonstrator the piston dynamics (positions, velocities, The

accelerations)

were specified in

the code from experimental data.

indicated power and indicated efficiency were overpredicted by an average of g and 55%, respectively. Unfortunately, the scope of MTI's validation

effort was very limited,

and they were not able to identify any specific assumptions. - less rigorous analysis developed a thermody-

improvements to the NASA-Lewis 3.2.4 Model by Chiu et al.

(1979)

Chiu et al. 34 of General Electric Company (GE) namic program (TDP)

that comprises thermodynamic and dynamic analysis for They (1) discretized the engine equations into difintegra-

free-piston Stirling engine applications. into a nodal network, ference equations, tion scheme. (2)

converted the differential

and (3)

applied a transient finite-difference

Some key assumptions were the use of steady state empirical plus the ideal gas

correlations for heat transfer and flow resistance, law.

The current code has not incorporated such important losses as shuttling losses, and transient heat transfer. This last as-

static heat conduction,

These losses were estimated manually and corrected ad hoc. pect has a close resemblance to second-order methods.

GE's model has been validated against a prototype machine with the results being claimed as relatively satisfactory. input, +15%. In fact, for the heat

predictions and test data agreed reasonably well, For the indicated power,

within about

the model overpredicted by an average of the error averaged about 37%.

about 41%; and for the indicated efficiency,

16 GE's predictions could have perhaps been made to fit the data better, the heat but

their policy was to minimize the use of correction factors in transfer and fluid friction correlations. 3.2.5 Model by Azetsu et al. The Azetsu et al. lished in English, it is less rigorous
35

(1982)

- less rigorous analysis

model,

the only Japanese mathematical model pubHowever,

may be classified as a third-order analysis. than those of Urieli (Sect. it is

3.2.7) or Schock (Sect. model, (2) because

3.2.8). (1)

In many aspects,

similar to the Tew et al.

a common pressure throughout all spaces was assumed, and (3)

gas kinetic

energy was ignored, other equations.

the momentum equation was decoupled from the

A steady state momentum equation was used for pressureSimilarity also has been found in the solution

drop calculations only.

method and numerical convergence procedure. adapt to various engine configurations, properties. The Azetsu et al. stration engine, Japan.

The model was designed to and thermal

operating conditions,

model was validated by a two-piston Stirling demon-

manufactured and constructed at the University of Tokyo, cyclic tempera-

Good agreement has been obtained for P-V diagrams, indicated work, and thermal efficiency.

ture variation,

They also con-

cluded that engine performance was influenced significantly by phase angle and dead volume. 3.2.6 Model by Vanderbrug Vanderbrug 36 (1977) less rigorous analysis (JPL) presented a general known as

of Jet Propulsion Laboratory

purpose program for Stirling engine analysis. Stirling Cycle Computer Model (SCCM),

The program,

was initially designed for Stirling This analysis used a simplified basic effects were ignored. Attributes

engines in underwater applications.

equation set in which the gas-inertial of the SCCM program include (1)

thermodynamic processes for each control (2) empirical or are

volume are quasi static during a small time interval, theoretical correlations for component performance

characteristics and (3)

readily modeled by a lumped parametric (nodal)

method,

usersystems

oriented subroutines can be assembled for any particular physical to be modeled.

17
37

Hoehn

validated Vanderbrug's model against the single JPL experiThe net indicated power predicted by even though expansion

mental data point published so far.

the SCCM program was only 1% higher than the measured value,

the predicted magnitudes of indicated power for the individual and compression pistons were underestimated by 15 and 32%,

respectively. The

Hoehn also applied a Schmidt analysis to the single data point. Schmidt predictions were nearly as good as the SCCO model. many more data points are needed for a meaningful 3.2.7 Model by Urieli Urieli's
38

Therefore,

evaluation.

(1977) is

- rigorous analysis He considered the

model

a rigorous nodal analysis.

full conservation equations by retaining the kinetic energy and the gas inertia effects. tion, that is, Salient features consist of (1) piecewise approxima-

discretizing the engine into control volumes of various (2) converting the partial differential equations into a

sizes and shapes;

system of ordinary differential

equations by transforming all differensolv-

tials to difference quotients except for the time variable; and (3) ing these ordinary differential Kutta method with a stationary

equations using the fourth-order Rungeinitial condition. Also, the model applied

a convergence scheme that adjusts the matrix temperatures at the end of each cycle in umes. accordance with the net heat transferred in
39

the control

vol-

Urieli stated that convergence The program,

usually occurs within ten cycles. was fully documented. number It

written in Fortran language,

is

efficient and very versatile.

Urieli has shown that a minimal

of cells (about 33) a particular machine, three-dimensional

can be used satisfactorily to find the performance of with a corresponding saving in computer time. Also, and

plots showing the behavior of the temperature,

flow,

pressure profiles through the cycle are possible,

thus helping to provide

further insight into the detailed behavior of Stirling cycle machines. Ilrieli's model has been validated at three engine operating frequencies of the University of Witwatersrand (South Africa) test engine.
40

When compared with the experimental data,

the model underpredicted the

heat transfer rates in the heater and cooler and the power output by averages of 7, 13, and 40%, respectively. However, it is important to note

18 that these predictions are for experimental net power. Thus, errors in cases that produce very little

the net power output predictions that are

small compared with the heat input may appear rather large relative to the small magnitude of the net power output. 3.2.8 Model by Schock (1978) - rigorous analysis

Schock of Fairchild Industries developed a Stirling Nodal Analysis Program (SNAP), but full documentation has not been released. extensive enough for review. Schock's model However,
41

published paper is

is

rigorous third-order design method.

He applied the same differential

equations as Urieli, but his method of computer modeling was different. Schock employed a finite-difference, nique. explicit-forward integration tech-

He further developed a special

scheme for enhanced mathematical How-

stability and for accelerated ever, no details were given in

convergence to a steady state cycle. the paper. Furthermore,

the option of an

inertialess gas was provided for a faster but less accurate calculation. The model has been applied to a free-piston Stirling engine built by Sunpower, Inc., for DOF. The model predicted (1) no significant differ-

ences in cyclic pressure variation between expansion and compression spaces; (2) nonuniform mass flow rate, which showed gas streaming out of at certain parts of the cycle; and (3) highly the

both ends of the regenerator

fluctuating gas temperature profiles in each space. model is

In addition,

capable of calculating cyclic heat flow profiles, mechanical and energy balances in tabulated forms or three-dimensional

power outputs, plots. 3.2.9

Model by Gedeon (1978) In a highly descriptive

- rigorous analysis and informative paper on the optimization of Inc., describes an optimievalua-

Stirling cycle machines,

Gedeon 4 2 of Sunpower,

zation computer program used for the development and performance tion of Sunpower's 1--kW engine (SPIKE). two parts: tion. It is

The computer program consists of simula-

an optimization scheme and a third--order thermodynamic

the sole interest of this review to isolate and concentrate Gedeon's third-order analysis differs from

on the thermodynamic analysis.

the third-order methods used by most other investigators in two important

19 aspects. First, the working gas is divided into only six nodes: cooler, two in and

the regenerator and one each in compression space.

the heater,

expansion space,

The computed results changed by only a few percent the analysis. Thus, it was

when additional control volumes were used in

decided that the model with six control volumes for the working gas was sufficiently accurate for their optimization study. In addition, a nodal

network was included to account for energy paths in the piston, displacer, and cylinder walls. hysteresis, Iosses caused by shuttle heat transfer, gas spring

and wall conduction could all be accounted for. thermodynamic simulation was

The other major difference in Gedeon's his integration technique. to integrate the continuity, integrations,

Gedeon used an implicit numerical method momentum, and energy equations. In implicit

the dependent nodal variables at a new time are determined This requires the inversion of a large matrix at each long time steps can be used because implicit integraGedeon's numerical method

simultaneously. time step.

However,

tion techniques are always numerically stable.

required about 200 time steps per piston revolution to maintain an integration accuracy of three significant figures. Losses due to gas leaks

past piston and displacer seals were properly accounted for in his mass and energy balances. Similar to other investigators, Gedeon included a

special subroutine to accelerate the convergence of the integration toward a cyclic steady state solution. At the end of each piston cycle, the

nodal temperatures were adjusted based on average temperatures and net energy accumulation during the cycle. about ten piston cycles. Hardware development at Sunpower, Inc., has complemented and augGedeon claims that the A steady solution was found in

mented the development of their computer models. agreement between their experimental, computer predictions is

free-piston Stirling engines and No

within +10% for all measurable parameters.

further details were provided by Gedeon. 3.2.10 Model by Zacharias (1977)

This review was made possible through an ORNL translation of a paper written in German by F. Zacharias
4 3

of MAN/MWM.

A full evaluation of the

20 model was not possible because few details about the model were given in the paper. method. The model described by Zacharias utilizes a third-order design it was not stated whether a complete set of conservation momentum, energy) was solved or if the equations. the gas inertia Iike other third-

However,

equations (continuity,

or kinetic energy terms were neglected in order models, umes.

the engine was partitioned into a network of control voleach control volume was defined by three and mass flow rate. After the conser-

The state of the gas in pressure,

variables:

temperature,

vation equations were converted from differential equations,

to finite-difference technique.

they were integrated by using an explicit numerical

A special algorithm was needed to accelerate the convergence of the regenerator nodal temperatures toward a cyclic steady state solution; but no details were provided. Zacharias applied the model ling engine. to a four-cylinder, double-acting Stir-

The predicted gas temperatures and pressures were presented plots as functions of position and time. A compari-

on three-dimensional

son between model predictions and experimental measurements was not included in the paper.

3.3

Method of Characteristics this category. Both models are based on

Two models were reviewed in

some simplifying approximations that decouple one of the conservation equations from the solution of the other two. The first model solves siThe second

multaneously the conservation equations of mass and momentum. model

is based on the simultaneous solution of the conservation equations

of mass and energy. 3.3.1 Model by OrRan (1981) Organ 7'

8

of Cambridge University modeled an alpha-configuration

Stirling engine as a series of ducts that can branch out into parallel paths and can have gradually changing flow areas. The gas velocity and The

pressure were assumed to be functions of axial position and time. temperature of the gas, however,

was assumed to depend only on position

21 (i.e., the gas at a particular location is isothermal). The gas tempera-

ture distribution was specified a priori and was set equal to the heater wall temperature in the expansion cylinder and heater, the cooler wall

temperature in the compression cylinder and cooler,

and a straight-line For

transition from hot-end to cold-end temperature in the regenerator. these assumptions, mass and momentum. acteristics, the flow is

defined by the conservation equations of

When these equations are solved by the method of charcurves are found to be the Mach lines. there are two families of Mach lines In

the characteristic

the physical plane (position-time),

that propagate either rightward or leftward at the local acoustic velocity relative to a moving fluid particle. The conservation equations are This technique ensures that

integrated numerically along the Mach lines.

7

pressure information propagates through the gas at the speed of sound. Organ applied his computer model to a Stirling machine that has He pre-

rather long heat exchangers and uses air as the working fluid.

sented a plot of the Mach line net for an instantaneous startup to 4000 rpm. Some features of the solution pertaining to the initial half revolu(1) a fan of rarefaction waves from (2) a triangular

tion of crankshaft motion include:

the impulsive withdrawal of the compression-space piston,

dead region where the gas remains undisturbed for a finite period of time after the instantaneous startup, and (3) a substantial change in the gra-

dient of the Macb lines across the regenerator caused by the temperature gradient over that component. For the conditions in Organ's example, it

took about 650 of crankshaft rotation for the pressure information to travel from one piston to the other. considerably smaller if was lower, However, this angle would have been the engine speed

the heat exchangers were shorter,

or the acoustic velocity of the working fluid was higher (using

helium or hydrogen rather than air). Organ also presented a plot of predicted work output per cycle vs engine speed. The net output was computed by integrating the pressures at

the surfaces of the compression and expansion pistons (after a cyclic steady state solution was reached) with respect to the piston positions. When Organ compared his predictions with the work output per cycle computed from the ideal Schmidt analysis, it was evident that the effects of

22 fluid friction and inertia are very important, especially at high engine speeds. 3.3.2 Model by Larson (1981) Larson 4 4 of Cleveland State University has developed a Characteristic Dynamic Energy Equations (CDEE) computer model based on the method of characteristics. ables: Larson formulated his analysis in terms of three variThe analysis was simpli-

gas density, velocity, and temperature.

fied by decoupling the momentum equation from the continuity and energy equations and neglecting kinetic energy in the energy equation. Approxi-

mate expressions for the gas velocity and its spatial derivative were derived by assuming a spatially uniform density and then correcting for the effects of pressure drop. The approximate velocity expression enabled

Larson to (1) separate the momentum equation from the system of simultaneous equations and (2) compute pressure directly from the momentum equation. The continuity and energy equations along with the total differen-

tials of density and temperature form a system of hyperbolic partial differential equations. The two characteristic curves for this system of

equations are the gas velocity and the gas velocity multiplied by the heat capacity ratio. The characteristic directions were used to transform the

partial differential equations into a set of ordinary differential equations that are valid along the characteristic curves. These equations

were solved numerically using a fourth-fifth order Runge-Kutta integration technique. Iarson 4 5 applied his model to the GPU-3 configuration. Gas tempera-

tures were computer plotted as a function of position and crank angle. Pressure-volume diagrams for the compression and expansion cylinders were also presented for a typical set of operating conditions. When compared

with GPU-3 experimental measurements, the CDEE model overpredicted power output by no more than 10% over the entire frequency range from 1000 to 3500 rpm. However, no details were provided about the heat transfer and

fluid friction correlations used in the model to achieve this good agreement.

23 4. SUMMARY

In

this state-of-the-art

review,

four distinct Stirling engine design approxiand

methods were identified based on the degree of sophistication: mate (first-order), decoupled (second-order), nodal

(third-order),

method of characteristics. system analysis. optimization. mass, pressure,

First-order methods are good for preliminary

Second-order methods are good for interactive design and

Third-order methods are good for detailed simulation of the and temperature distributions in a Stirling engine. determines the characteristic The

method of characteristics first

curves of

the conservation equations and then integrates the equations along the characteristic pressure waves. All of the engine design methods, were reviewed. except for first-order analysis, is approximately Among the 19 curves. This method can account for the finite velocity of

The number of models reviewed (19 total)

twice the number reviewed previously by others (10 total). models, 7 are second-order, characteristics. damental 10 are third-order,

and 2 use the method of fun-

Based on a carefully designed model review format, limitations,

assumptions,

and applicability of the individual

computer models were discussed. For quick cross-reference, the secondTables 1 and 2 summarize the attributes of The models are compared in terms of and model valiFive of the

and third-order models. simplification,

classification, dation.

code listing availability,

Classification of the second-order models reveals that adiabatic

analysis overwhelms the other second-order classifications. second-order models use adiabatic analysis, and one uses semi-adiabatic analysis. proximate analyses one margin.

one uses isothermal analysis, ap-

For the third-order models,

(six) outnumber rigorous analyses (three) by a two-to-

Table 1.

Summary of Stirling engine mathematical

models - second-order design methods

Principal investigator W. R. Martini

Present affiliation Martini Engineering

Second-order classification Isothermal

Code listing availability Yesa (Ref. 3)

Model validation Yes, GM GPU-3 and 4L23 engines (Ref. 3) Yes, Allison PD-67A engine (Ref. 19) 1, 2, 3, 4

E.

B.

Qvale

Laboratory for Energetics, Denmark; original work done at MIT General Electric Company; original work done at MIT

Adiabatic

Not published

19,

20

P.

A. Rios

Adiabatic

Yes, as modified by Martini (Ref. 3) Not published

Yes, GM 4L23 engine, by Martini (Ref. 3) Yes, GM GPU-3 engine (Ref. 22) Yes, GM GPU-3, Allison PD-67A, and Philips engines (Ref. 24) Yes, GM GPU-3 engine (Ref. 26)

21

K. Lee

Foster-Miller Associates

Adiabatic

22,

23

R.

Shoureshi

Wayne State University

Adiabatic

Yes (Ref.

24)

24,

25

T. J.

Heames

Argonne National Laboratory

Adiabatic

Yes, from National Energy Software Center Not published

26

B.

Feurer

Unknown; original work done at MAN/ MWM. West Germany (on a floppy disk)

Semi-adiabatic

Not published

27

Computer code

acquired by ORNL.

Computer code (on tape)

acquired by ORNL.

25

Table 2.

Summary

of Stirling

engine mathematical

models - third-order design methods

Principal investigator T. Finkelstein

Present affiliation Engine TCA Stirling R&D Company

Numerical simplifications No gas inertia or kinetic energy

Code listing availability Available for use through CDC Cybernet computer Yes (Refs. and 29)
a

Model validationferences Yes, but not published 28

R.

C. Tew

NASA-Lewis Center

Research

Common pressure, no kinetic energy

Yes, GM GPU-3 engine (Ref. 31), USS P-40 engine (Ref. 32) Limited, DOE 1-kW freepiston engine (Ref. 33) Yes, GE Proto I and 2 freepiston engines (Ref. 34) Yes, University of Tokyo test engine (Ref.

29, 32

30,

31,

J.

E.

Giansante

Mechanical Technology. Inc.

Common pressure, no kinetic energy

Yes

(Ref.

33)

33

W. S.

Chiu

General Electric Company

No gas inertia or kinetic energy

Not

published

34

A.

Azetzu

University of Tokyo, Japan

Common pressure, no kinetic energy

Not

published

35

35)
T. G. Vanderbrug Unknown, original work done at Jet Propulsion Laboratory Sunpower, Inc.; original work done at University of Witwatersrand, S. Africa Fairchild dustries InNo gas inertia Yes (Ref. 36) Limited, JPL Research Engine (Ref. 37) Yes, University of Witwatersrand test engine (Ref. 40) 36, 37

I.

Urieli

Rigorous

Yes (Refs. and 39)

38,

39,

40

A.

Schock

Rigorous

Not published

Yes, but not published

41

D. R. Gedeon

Sunpower,

Inc.

Rigorous

Not published

Yes, Sunpower free-piston engines, not published Not published

42

F. Zacharias

Unknown; original work done at MAN/ MWM,West Germany

Unknown

Not published

43

A copy of early NASA-Lewis computer program acquired by ORNL.

26

5.

CONCLUSIONS WITH RECOMMENDATIONS

This section provides some broad perspectives and recommendations about analyses of Stirling engines. lowing were established: thermodynamic models, (2) (1) In the course of our study, the fol-

a state-of-the-art

review of Stirling engine and (3) a qualitative

an information center,

comparison between the numerous models. A full and complete assessment was not attempted in this report be-

cause the limited time available did not allow us to obtain all of the documentation (especially some of the more obscure items such as theses, reports, and lecture notes) relating to some codes. In addition, Thus, it diffiis recdomes-

culties were encountered with incomplete draft reports. ommended that comprehensive tically and abroad as well. are numerous gine design,

literature surveys should be continued, For instance, on the domestic front,

there

companies that are or have been outstanding in Stirling enapplication, and manufacturing; yet their documentation Inc., is

not available. tries.

Typical companies are Sunpower,

and Fairchild IndusInformation

For those companies abroad,

the situation is worse.

from many active and leading companies development is

in Stirling engine research and

scarce or not even released because of proprietary or seExamples include Philips, Harwell, Extensive communication, MAN/MWN, the

curity restrictions. French,

and the Japanese.

information exchange,

and program cooperation may improve this problem, organizations see a commercial

but as long as these

or military future for Stirling machines,

some proprietary restrictions are likely to exist. Utilization of a detailed design method does not ensure enhanced model performance. According to our review, there is at present no evi-

dence to claim that the existing third-order and method of characteristics analyses are superior to the second-order methods. However, it should be

pointed out that many of the models required arbitrary corrections for the friction factor and/or heat transfer correlations to make the power output and efficiency predictions fit the validation data better. Whether these

correction factors represent weaknesses in nesses in

the models themselves or weaka question that

the friction and heat transfer correlations is

27 must be resolved. After this question is answered, the rigorous thermo-

dynamic models will be more accurate and more generally applicable. Contradictory opinions exist among the model developers about differences between the integration techniques used in the method of characteristics. the nodal analyses and the speed at etc.) propagates

The point in question is temperature,

which thermodynamic information (pressure, through the engine. As discussed earlier,

the nodal integrations use a

system of fixed grids and uniform time steps and can be classified into two major categories: (forward-differencing) explicit and implicit techniques. techniques, In explicit

thermodynamic information propagates The node

only from one node to an adjacent node during each time step. spacing and time step, therefore,

determine the speed at which information The explicit techniques are sometimes if the time steps are improperly technique is

propagates through the grid system. plagued by numerical instabilities, chosen. In contrast,

the implicit (backward-differencing)

always numerically stable,

regardless of grid spacings and time steps.

The implicit technique solves the finite difference equations describing

*r

the state of the working fluid by calculating the condition of the gas in each cell at a particular time from the condition of the gas in cells at that time. Consequently, all other

thermodynamic information can propagate

from one end of the engine to the other during one time step. Numerical integrations using the method of characteristics are nor-

mally based on fixed time steps and floating grids. spacing depends on the characteristic In Organ's model,

The choice of grid

curves of the governing equations.

in which the method of characteristics was applied to the nodes are spaced so

the conservation equations of mass and momentum,

that the pressure information propagates at the local acoustic velocity relative to the local gas velocity. It may be very important to account for the fact that pressure waves

propagate at the speed of sound when predicting the performance of certain engines, especially ones that have long heat exchangers and operate at However, this effect may not be very important in other

high frequencies.

~* engines,
curacy. gation.

and the nodal This is

integration techniques may provide sufficient ac-

a dilemma that will only be resolved by further investi-

28 We are not in a position now to rank with confidence the models reviewed. The only fair way to compare the models would be to run all of

the codes on the same computer and compare their predictions with data from well-defined experimental lines was begun by ANL. apparent reasons.
4
6

Stirling engines. this is

Some work along these

However,

a difficult task for three

First,

every model has its unique attributes and may

also be applicable only to a limited number of engine configurations. Second, codes. acquisition of codes would have to be limited to nonproprietary Finally, a lack of well-defined experimental data is a hindrance.

Numerous investigators have published experimental

data, but many of them parameter com-

have not provided enough information about engine dimensions, definitions,

and even the operating conditions to allow a meaningful

parison between simulation and experiment. It is obvious that a large variety of thermodynamic models for StirThey range in complexity from

ling engine analysis has been developed.

the simple first-order models up through the rigorous third-order and method of characteristics models. any new thermodynamic models. It does not seem necessary to develop

Time and effort would be better spent by One final observa-

increasing our understanding of the existing models. tion is

that validation of the thermodynamic models has been limited When experimental data from free-piston the experimentally and piston

mainly to kinematic engines.

Stirling engines have been used to validate the models, determined dynamic parameters amplitudes), (such as frequency,

phase angle,

rather than predicted values from a separate free-piston have usually been used as inputs to the thermodynamic

dynamics model, models.

Analyses of free-piston dynamics have been explored to a much but it is an

lesser extent than those of Stirling engine thermodynamics, area that is important and needs additional studies.

29 ACKNOWLEDGMENTS

The preparation of this report was supported by the Department of Energy as part of the Stirling Cycle Heat Engine Technology Program, aged by P. D. Fairchild of ORNL Energy Division. I,. Crowley of ORNL Enman-

The authors are indebted to C. D. West and J.

gineering Technology Division for reviewing the manuscript and offering useful comments. In addition, the authors of the models reviewed in (Note: Zacharias Sect. this

report were solicited for comments. this report too late to provide F.

3.2.10 was added to The con-

this opportunity.)

structive responses received from the ten model developers listed below were greatly appreciated. W. S. D. R. F. Chiu - General Electric Company Gedeon - Sunpower, Inc. (former colleague of

W. Hoehn - Rockwell International T. G. Vanderbrug)

V. H. Larson - Cleveland State University W. R. Martini - Martini Engineering

A. J. Organ - Cambridge University E. B. P. R. R. O.vale - The Technical University of Denmark

A. Rios - General Electric Company Shoureshi - Wayne State University C. Tew - NASA-Lewis Research Center

30 REFERENCES

1.

W. R. Martini, Stirling Engine Design ManuaZ, NASA-CR-1353 Aeronautics and Space Administration, April 1978.

82,

Natl.

2.

W. R. Martini, Thermodynamic Design of Stirling Engines by Computer, Martini Engineering, Richland, Washington, 1980. W. R. Martini, Stirling Engine Design Manual - Second Edition, to be published by NASA-Lewis. W. R. Martini, "Validation of Published Stirling Engine Design Methods Using Engine Characteristics from the Literature," pp. 2245-50 in Proceedings of the 15th IECEC, Paper No. 809449, 1980. 1. Urieli, "A Review of Stirling Cycle Machine Analysis," in Proceedings of the 14th IECEC, Paper No. 799236, 1979. G. Walker, Stirling Engines, Clarendon Press, Oxford, 1980. pp. 1086-90

3.

4.

5.

6. 7.

A. J. Organ, "Gas Dynamics in the Temperature-Determined Stirling Cycle," J. Mech. Eng. Sci. 23(4), 207-16 (August 1981). A. J. Organ, "Gas Dynamics of Stirling Cycle Machines, Stirling Fngines - Progress Toward Reality," pp. 131-40 in Mechanical Engineering Publications Limited, London (March 1982). J. R. Senft, "A Simple Derivation of the Generalized Reale Number," pp. 1652-55 in Proceedings of the 17th IECEC, Paper No. 829273, 1982. T. Finkelstein, Generalized Thermodynamic Analysis of Stirling Engines, SAE Paper No. 118B, Society of Automotive Engineers, January 1960. A. P. Shapiro, The Dynamics and Thermodynamics of Compressible Fluid Flow - Vols. I and II, Ronald Press, New York, 1953. H. W. I,iepmann and A. Roshko, Sons, New York, 1957. Elements of Gasdynamics, John Wiley and

8.

9.

10.

11.

12.

13.

J. S. Rauch, "Harmonic Analysis of Stirling Engine Thermodynamics," pp. 1696-1700 in Proceedings of the 15th IECEC, Paper No. 809335, 1980. R. Berggren, Analysis Documentation - STENSY, Interim No. 1, MTI Report No. 82 ASE286ER46, Mechanical Technology, Inc. (March 1983). N. E. Andersen, Stirling Engine Modular Analysis Program (SEMAP), Laboratory for Energetics, Technical University of Denmark, RE 79-9, 1979.

14.

15.

31 16. E. J. Sirett, "The Gas Dynamics of the Stirling Machine," first year report on postgraduate work, Cambridge University Engineering Dept., 1981. R. Vincent, W. Rifkin, and G. Benson, "Analysis and Design of FreePiston Stirling Engines - Thermodynamics and Dynamics," pp. 3686-95 in Proceedings of the 15th IECEC, Paper No. 809334, 1980.

17.

18.

R. Howlett, A Digital Computer Simulation of the Thermomechanical

Generators, AERE-M2294, Kingdom, 1970. Atomic Energy Research Establishment, United

19.

E. B. Ovale and J. I. Smith, Jr., "A Mathematical Model for Steady Operation of Stirling-Type Engines," J. Eng. Power, 45-50 (January 1968).

20. 21.

E. B. Ovale,
Thesis,

An Analytical Model of Stirling-Type Engines, Ph.D.

1967.

Massachusetts Institute of Technology,

P. A. Rios, An Analytical and Experimental Investigation of the

Stirling Cycle, Ph.D. Thesis, 1969. Massachusetts Institute of Technology,

22.

K. Iee, I. P. Krepchin, and W. M. Toscano, "Thermodynamic Description of an Adiabatic Second Order Analysis for Stirling Engines," pp. 1919-24 in Proceedings of the 16th IECEC, Paper No. 819794, 1981. K. Iee et al., "Performance Loss Due to Transient Heat Transfer in the Cylinders of Stirling Engines," pp. 1706-1909 in Proceedings of the 15th IECEC, Paper No. 809338, 1980.

23.

24.
25.

R. Shoureshi, Analysis and Design of Stirling Engines for Waste-Heat

Recovery, Ph.D. Thesis, Massachusetts Institute of Technology, 1981. R. Shoureshi, "Simple Models for Analysis and Design of Practical Stirling Engines," pp. 1647-51 in Proceedings of the 17th IECEC, Paper No. 829272, 1982. T. J. Heames et al., "A User Oriented Design System for Stirling Cycle Codes," pp. 1681-87 in Proceedings of the 17th IECEC, Paper No. 829278, 1982. B. Feurer, "Degrees of Freedom in the Layout of Stirling Engines," presented at Von Kaman Institute for Fluid Dynamics - Lecture Series 53, Feb. 12-16, 1973. T. Finkelstein, "Computer Analysis of Stirling Engines," pp. in Proceedings of the 10th IECEC, Paper No. 759140, 1975. 933-41

26.

27.

28.
a

32 29. R. Tew, K. Jefferies, and D. Miao, A Stirling Engine Computer Model For Performance Calculations, NASA TM-78884, Natl. Aeronautics and Space Administration, July 1978. R. C. Tew, Computer Program for Stirling Engine Performance Calculations, NASA/TM-82960, Natl. Aeronautics and Space Administration, January 1983. R. C. Tew, I,. G. Thieme, and D. Miao, Initial Comparison of Single Cylinder Stirling Engine Computer Model Predictions with Test Results, NASA 1M-79044, Natl. Aeronautics and Space Administration, March 1979. W. A. Tomazic, Supporting Research and Technology of Automotive Stirling Engine Development, NASA TM-81495, Natl. Aeronautics and Space Administration, April 1980. J. E. Giansante, A Free Piston Stirling Engine Performance Code 81TR17, prepared for NASA-Lewis by Mechanical Technology, Inc., November 1980. General Electric, The Thermodynamic Program (TDP) for Free-Piston Stirling Engine/Linear Compressor Simulation, GE No. GFHP-82-048, 1982. A. Azetsu, N. Nakajima, and M. Iirata, "Computer Simulation Model for Stirling Engine," pp. 57-63 in Stirling Engines -Progress Towards Reality, Mechanical Engineering Publications Limited, London, March 1982. J. G. Finegold and T. G. Vanderbrug, Stirling Engines for Undersea Vehicles - Final Report, JPL No. 5030-63, Jet Propulsion Laboratory, March 1977. F. W. Hoehn, B. D. Nguyen, and D. D. Schmit, "Preliminary Test Results with a Stirling Laboratory Research Engine," pp. 1075-81 in Proceedings of the 14th IECEC, Paper No. 799233, 1979. I. Urieli, C. J. Rallis, and D. M. Berchowitz, "Computer Simulation of Stirling Cycle Machines," pp. 1512-21 in Proceedings of the 12th IECEC, Paper No. 779252, 1977. 1. Urieli, A Computer Simulation of Stirling Cycle Machines, Ph.D. Thesis, University of Witwatersrand, South Africa, 1977. D. M. Rerchowitz and C. J. Rallis, "A Computer and Experimental Simulation of Stirling Cycle Machines," pp. 1730-38 in Proceedings of the 13th IECEC, Paper No. 789111, 1978. A. Schock, "Nodal Analysis of Stirling Cycle Devices," pp. in Proceedings of the 13th IECEC, Paper No. 789191, 1978. 1771-79

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

33

42.

D. R. Gedeon, "The Optimization of Stirling Cycle Machines," pp. 1784-90 in Proceedings of the 13th IECEC, Paper No. 789193, 1978. F. Zacharias, Motortech. Z. "Further Stirling Engine Development Work - Part 1," 38(9), 371-77 (1977).

43.

44.

V. H. Larson, "Characteristic Dynamic Energy Equations for Stirling Cycle Analysis," pp. 1942-47 in Proceedings of the 16th IECEC, Paper No. 819798, 1981. V. H. Iarson, "Computation Techniques and Computer Programs to Analyze Stirling Engines Using Characteristic Dynamic Energy Equations," pp. 1710-15 in Proceedings of the 17th IECEC, Paper No. 829283, 1982. J. E. Ash and T. J. Reames, "Comparative Analysis of Computer Codes for Stirling Engine Cycles," pp. 1936-41 in Proceedings of the 16th IECEC, Paper No. 819797, 1981.

45.

46.

35

ORNL/CON-135

Internal Distribution

1. 2-6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16-20. 21. 22. 23. 24. 25.

F. N. J. J. F. J. N. R. P. E. R. F. D. T. V. H. W.

Chen C. J. Chen T. Cockburn (Consultant) C. Conklin A. Creswick L. Crowley Domingo D. Ellison D. Fairchild C. Fox L. Graves P. Griffin S. Griffith J. Hanratty (Consultant) O. Haynes W. Hoffman L. Jackson

26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41-42. 43.

J. E. Jones R. E. MacPherson S. S. Mason (Consultant) J. W. Michel R. E. Minturn J. Petrykowski G. T. Privon S. D. Rose R. L. Rudman (Consultant) J. P. Sanders H. E. Trammell C. D. West ORNL Patent Office Central Research Library Document Reference Section Laboratory Records Department Laboratory Records (RC)

External Distribution

44.

45. 46. 47.

48. 49. 50. 51. 52. 53. 54. 55.

Thierry Alleau, Commissariat a l'Engergie Atomique, Centre d'Etudes Nuclearies de Grenoble, Service des Transferts Termiques, 85X, 38041 GRENOBLE CEDEX, France A. Azetsu, Department of Mechanical Engineering, University of Tokyo, Japan W. T. Beale, Sunpower, Inc., 6 Byard St., Athens, OH 45701 Donald G. Beremand, Stirling Engine Project Office, National Aeronautics and Space Administration, Lewis Research Center, Cleveland, OH 44135 Stig G. Carlqvist, Societe ECA, 17, avenue du Chateau, 92190 Meudon-Bellevue, France W. S. Chiu, Systems Engineering, General Electric Company, P.O. Box 527, King of Prussia, PA 19406 J. G. Daley, Argonne National Laboratory, 9700 South Cass Avenue, Argonne, IL 60439 B. Feurer, MAN-AG, Maschinenfabrik Augsburg-Nurnberg AG, Postfach 10 00 80, D-8900 Augsburg 1, West Germany T. Finkelstein, TCA, P.O. Box 643, Beverly Hills, CA 90213 D. R. Gedeon, Sunpower, Inc., 6 Byard St., Athens, OH 45701 J. E. Giansante, Mechanical Technology Inc., Stirling Engine Systems Division, 968 Albany-Shaker Road, Latham, NY 12110 L. Goldberg, University of Minnesota, The Underground Space Center, 11 Mines and Metallurgy, 221 Church Street, SE, Minneapolis, MN 55455

36

56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72.

73. 74. 75. 76. 77. 78. 79.

80. 81.

T. J. Heames, Argonne National Laboratory, 9700 South Cass Avenue, Argonne, IL 60439 F. W. Hoehn, Rockwell International Corp., 6633 Canoga Avenue, Canoga Park, CA 91304 R. E. Holtz, Argonne National Laboratory, 9700 South Cass Avenue, Argonne, IL 60439 Y. Ishizaki, Cryogenic Systems, 253-5 Yamanouchi, Kamakura 247, Japan Naotsugu Isshiki, 2-29-6 Kyodo Setagayaku, Tokyo 156, Japan Prof. Nobuhide Kasagi, University of Tokyo, Dept. of Mechanical Engineering, Bunkyo-Ku, Tokyo 113, Japan V. H. Larson, Cleveland State University, 1983 East 24th St., Cleveland, OH 44115 K. P. Lee, Manager, Facility and Manufacturing Automation, P.O. Box 809, Sudbery, MA 01776 W. Martini, Martini Engineering, 2303 Harris, Richland, WA 99352 G. McLennan, Argonne National Laboratory, 9700 South Cass Avenue, Argonne, IL 60439 Vincenzo Naso, Prof. Ing., Universita Degli Studi Di Roma, Instituto Di Macchine E Tecnologie Meccaniche, Rome, Italy A. J. Organ, University Engineering Department, Trumpington St., Cambridge, CB2 1PZ, England E. B. Qvale, Laboratory for Energetics, Technical University of Denmark, Bldg. 403 DK-2800, Lyngby, Denmark C. J. Rallis, School of Mechanical Engineering, University of the Witwatersrand, 1, Jan Smuts Avenue, Johannesburg, South Africa Lt. Cdr. G. T. Reader, Royal Naval Engineering College, Manadon Plymouth, Devon PL5 3AQ, England G. Rice, Department of Engineering, The University of Peading, Reading, Berkshire, England P. A. Rios, Electromechanics Branch, Electrical Systems and Technology Laboratory, General Electric R and D Center, P.O. Box 43, Schenectady, NY 12301 J. R. Senft, Dept. of Mathematics/Computer Systems, University of Wisconsin-River Falls, River Falls, WI 54022 B. Shaddis, Mueller Associates, 1401 South Edgewood St., Baltimore, MD 21227 A. Schock, Fairchild Industries, Germantown, MD 20874 R. Shoureshi, Wayne State University, Detroit , MI 48202 R. C. Tew, NASA-Lewis Research Center, 21000 Brockpark Rd., Cleveland, OH 44135 I. Urieli, Sunpower, T nc., 6 Byard St., Athens, OH 45701 Valerie J. Van Griethuysen, Energy Conversion Branch, Aerospace Power Division - Aero Propulsion Laboratory, Dept. of the Air Force, Wright-Patterson AFB, OI1 45433 G. Walker, University of Calgary, Dept. of Mechanical Engineering, 2920 24th Ave., NW, Calgary, Canada TZN 1N4 M. A. White, University of Washington, Joint Center for Graduate Study, 100 Sprout Road, Richland, WA 99352

37

82. 83-109.

Office of Assistant Manager for Energy Research and Development, Department of Energy, ORO, Oak Ridge, TN 37830 Technical Information Center, Department of Energy, Oak Ridge, TN 37830

Iz