Advances in Engineering Software 36 (2005) 127–134

www.elsevier.com/locate/advengsoft

Computational modular model library of gas turbine

Yuan Caoa,*, Xianlong Jina, Guang Menga, Jay Fletcherb
a
School of Mechanical and Power Engineering, Shanghai Jiaotong University, Shanghai 200030, China
b
United Technologies Research Center, East Hartford, CT-06108, USA
Available online 27 September 2004

Abstract
Designing and developing new aircraft engine is time-consuming and expensive. Computational simulation is a promising means for
alleviating this cost, but constructing flexible simulation software capable of evaluation of integrated aircraft engine system architectures is hard
work. This paper addresses the design of a tool—a generic modular-modeled library of air gas turbine, this library is based on object-oriented
technology and hierarchical decomposition and provides a flexible component-based representation for defining, modifying, and simulating the
aircraft gas turbine system, subsystem and components. It enables users to customize and extend the framework to add new functionality or
adapt the simulation behavior as required, and it allows new models to be composed programmatically or graphically to form more complex
models. The model library can be used in steady-state and transient analysis of the aero-engine. It is also a user-friendly, accurate, fast PCbased
and easily reusable simulating tool. The advanced object-oriented simulation language Modelica is used to construct this library. Modelica
provides a powerful tool to design the library. All of the work described in this paper is developed based upon Dymola/Modelica.
q 2004 Elsevier Ltd. All rights reserved.

Keywords: Gas turbine; Object-oriented; System-level simulation; Dymola/Modelica

1. Introduction and distributed to users and creators of engine models by

Applied Vehicle Technology Pane (AVT) in Research
As the aircraft industry enters 21st century, there is Technology Organization (RTO) of NATO. The responses
increasing pressure to reduce the time, cost and risk of varied in details, forms and sources, but were adequate to
aircraft engine development and maintenance. To compete lead to the following conclusions: ‘There is a trend towards
effectively in global marketplace, innovative approaches to object-oriented and graphical approaches; the general trend
reducing aircraft gas turbine engine design-circle span, is to move towards workstations and PC systems which offer
manufacture and maintenance cost are needed. An oppor- graphical user interfaces; flexible and modular systems are
tunity emerged to realize this point with computational required. Also, there is a requirement to model components
simulation. As modeling techniques have improved, and and systems at varying levels of detail’ [3].
computers have progressed, simulation has assumed an Based upon these, the objective of our work is to develop
essential role in planning, executing, and evaluating a set of extensible generic model libraries of components
operations. So too, in the design, manufacturing, and and subsystem of the aero engine system that may be used to
operating of aircraft turbine engines, accurate performance rapidly assemble a system level dynamic model and to
simulations have become essential. Therefore, throughout evaluate the dynamic performance of integrated power
the years, many computer-based models for the simulation system architectures during conceptual and preliminary
of the operational characteristics of the aircraft gas turbine design. As such, the intended use is not for particular
engine have been structured and evolved into a very wide component design but rather to understand the interaction of
range of applications [1–4]. components.
To make sure the requirement of commercial users Due to the wide range of interactions between different
of engine models, a detailed survey was prepared, engineering fields, especially in aero-engine, systems are
getting more complex and heterogeneous. For example, the
* Corresponding author. Tel.: C86-21-629-32256. gas turbine engine system includes several engineering
E-mail address: caoyuan@sjtu.edu.cn (Y. Cao). fields, mechanics, hydraulics, electrics, thermodynamics
0965-9978/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.advengsoft.2004.07.006
128 Y. Cao et al. / Advances in Engineering Software 36 (2005) 127–134

and so on. When trying to create models of these systems, 2.2. Object-oriented modeling and hierarchical structure
problems arise with the interaction of the different parts.
Fortunately, among the recent research in modeling and Traditionally, simulation software development has
simulation, there are two concepts that are closely related to followed the top–down structured design approach, which
these problems: object-oriented modeling language and applies the method of functional decomposition to establish
non-causal modeling. During the last few years, a new program structure. Although this method has been success-
advanced multidisciplinary object-oriented modeling ful in some applications, it fails to reflect the real world. As
language Modelica has been developed. It is designed to a result many attempts have been made to tackle this
allow convenient, component-oriented modeling of com- problem by applying Object-oriented technology (OO
plex physical systems, e.g. systems containing mechanical, technology).
electrical, electronic, hydraulic, thermal, control, electric So the whole aircraft gas turbine system is designed as an
power or process-oriented subcomponents [5]. So in this object-oriented framework, which is a set of classes that
paper the model libraries will be developed based no the embodies an abstract design for solution to a family of
Modelica language. related problems (Johnson and Foote, 1988) [6]. The set of
classes define ‘semi-complete’ applications that capture
domain-specific object structures and functionality. Specific
2. Basic design IDEA functionality in new applications is realized by inheriting
from, or composing with, framework components.
2.1. Library design In this paper, new object-oriented simulation software
Dymola/Modelica is used to develop the gas turbine model.
Following the top–down design philosophy and con- It is also structured to enable the assembly of object-
sidering the different engineering fields related to aero- oriented hierarchical libraries, which can be later used to
engine system structuring, the whole system can be divided assemble system-level dynamic models. And several other
into several sub-systems; each sub-system consists of Modelica libraries, such as mechanical, electrical and
various components. They are shown in Fig. 1. thermoflow library, can be obtained freely.
Based on the above decomposition, each subsystem has The layout configuration of the aircraft gas turbine is
its respective library of component models. Although the shown in Fig. 3. From a structural view point, a gas turbine
scope of this work is the development of the libraries of can be essentially as assembly of engine component—inlet,
several subsystems that can be assembled to capture the fan, compressors, combustor, duct turbine, shafts and
physical characteristics of the whole aero-engine system, nozzle. These components can be represented in the
our emphasis in this paper is to construct the aircraft gas computational domain as objects [6].
turbine system model library. So the library of gas turbine During the simulation, the engine’s mathematical model
system will be particularly illustrated in details in the next is mapped to collections of interacting objects, rather than
section. decomposed into segments of different functions that can
To build a complete aircraft gas turbine system-level implement certain algorithms. Each object mimics the
model, the strategy is determined and shown in Fig. 2. Both behavioral and structural characteristic of a physical or
the prospective model library and the system-level model conceptual entity shown in Fig. 3. And it represents an
will be designed according to the strategy. instance of a software class, while the classes are united into

Fig. 1. Sub-systems and components of aero engine system.

 Y. Cao et al. / Advances in Engineering Software 36 (2005) 127–134 129

a component of the gas turbine, and can be used to model the

corresponding component’s function. For example, the
variations of specific heat capacity and adiabatic exponent
can be acquired in the work fluid modular (medium) model.
Based on the modular model the hierarchical structure is
easily structured, in addition, ‘Message-based modeling’ [1]
is used for the interaction among models.

3. Common engineering model

In an aircraft gas-turbine library, physical engine

component structures and substructures are represented by
the different combinations of the volume model, flow
Fig. 2. Flow chart of aero-engine model design. model, medium model, and connector model objects. And
all these objects are designed based on the OO technology.
a hierarchy via inheritance relationships [6]. For example, The object composition provides a powerful mechanism for
the class (or partial model) ‘BaseCompressor’ is used to representing the physical topology of the gas turbine engine.
generate different kinds of compressors. However, it inherits
from a more generalized class ‘CompressorMap’. If a new 3.1. Volume model and flow model
compressor object need to be modeled, it can easily be
instantiated from the existing class ‘BaseCompressor’. In The difficulty in modular is keeping the independence of
this manner, a well-planned class hierarchy provides the every modular model, at the same time, considering the
‘slots’ into which the future codes are to be plugged. coupling among different models. So the concept of volume
and flow is developed. In recent years, control volume
2.3. Non-linear model models have been used to model the thermodynamics
systems that involve mass flow and energy in and out.
In the design of the mathematical model of the aircraft Besides the volume model, the flow model is also
gas turbine, the following factors are considered: indispensable in modeling the gas turbine components [8].
So another model named flow model is introduced for
The inertia of rotor and mechanical efficiency; calculating the mass flow and the convective energy
Mass and energy storage and release (mass, energy, and associated to the mass and energy flow. Flow models are
momentum balance); the result of modeling abstraction, where the volume is
Air of cooling system, or leaking; neglected. The flow models contain either an algebraic
Variation of specified heat capacity and adiabatic equation that relates pressure drop and mass flow, or a
exponent; mathematical expression for dynamic momentum balance.
Pressure loss in combustor, and combustion efficiency The storage of mass and energy are modeled in the control
etc are considered by setting relevant variables on volume model. Volume models and flow models always
customized windows of the models. have to alternate each other [8] as Fig. 4 shows.
Volume model and flow model are connected by flow
Non-linear mathematical model of the gas turbine has connectors (see Section 3.3). In the volume model the
been discussed in many books and papers. In this paper, the output variable quantities are pressure, specific enthalpy,
model is divided into several modular models with the density, temperature, specific entropy and ratio of specific
object-oriented technology. Each modular model represents heats, respectively. And the input variable quantities
are mass flow and convective heat. While the flow
model’s output variables are mass flow and convective
heat flow, the input information is pressure, specific
enthalpy and so on.
Suppose the gas (or air) in the volume model is uniform.
Any heat transfer and momentum are neglected. The energy
and mass balances of a volume can be expressed in the
follow balance equations (take the enthalpy and pressure as
the state variables):
 
vri dpi vri dhi vV
m2i K m1i C Vi C C ri i Z 0 (1)
Fig. 3. Aeroengine structure [7]. vpi dti vhi dti vti
130 Y. Cao et al. / Advances in Engineering Software 36 (2005) 127–134

Fig. 4. Volume and flow model.

    outlet should be equal to that at B’s inlet. If the value on one

vr vr side is known and that on the other side is unknown, the
m2i h2i Km1 h1i CVi ri Chi i CVi hi i K1
vhi vri fluid flow must assign the known value to the other side. The
transmitted data are of various types, ranging from
vVi temperature, pressure, to enthalpy, mass flow rate, and/or
Cðri hi KpÞ Z0 (2)
vti chemical composition [8]. Some unexpected types of data
Where m is the mass flow; positive and negative sign are may also need to be involved when the components are of
associated to flow into and out of the control volume, special type. Besides, although the working fluid flows are
respectively; V is the cubage of volume model; r is the unidirectional from upstream to downstream, the data
density of airflow; h is the enthalpy; u is the internal energy transmissions do not have to behave in the same way. The
and i is the subscript, representing the different element. For previous models tried to wrap all these complexities in one
simplicity, dissipative works has been neglected here. Based single object, whose mechanism of data transmission turned
on these balance equations, the basic energy equations in out to be too complicated for new parameter types to be
volume model can be acquired as following: inserted in and hence new component types and balancing
methods are extended and customized.
 
vri vri dpi
Vi ri C There are too many medium composition, it is not
vPi vhi dti
    possible to create all of them. So based on the OO
vr vr vr dVi technology, a new medium model can be easily added by
Z ri Chi i mi K i ui K r2i C i pi (3)
vhi vhi vhi dti users follow the old medium model structure in the library.
 
vr vr dhi
Vi r i i C i 3.3. Connector model
vpi vhi dti

vri

vri

vri

dVi Component interaction is defined by Connector models.
Z 1Khi mi C ui C pi Kri (4) The connector represents an exchange of information data
vpi vpi vpi dti
between successive models or components.
As Fig. 4 shows, between two volumes a flow model has A consequence of allowing multifidelity and multi-
to be connected with them. Here, the flow model is applied disciplinary models to be incorporated in an engine
in the compressor, turbine, nozzle, etc.; to calculate the component is that an engine model maybe composed of
corrected mass flow rate of air based on the pressure ratio, component models having differing fidelity and discipline.
temperature, angular speed, etc. In the flow model the There are different connector models available in the
mechanical power is transmitted to internal energy of the air library. Such as: mechanical connectors, thermo connectors,
by increasing the air pressure and vice versa. hydraulic connectors, etc. Structure of the Connector is tied
to the discipline level of the component model. That is to
3.2. Medium model say, the different connectors are used corresponding to the
variables transferred.
This paper wishes to provide the users with the In this paper several connector combinations are
capability to assemble an engine the same way as that in introduced and classified into the sub-package. For
thermodynamics, that is, by coupling the physical com- example, there are two output ports in the high-pressure
ponents together using thermodynamic linkages, such as compressor of a recuperative gas turbine, one for the main
the working fluid flows and the shafts. So in this modular
model library, it also includes a medium model for
calculating the thermo-physical properties. The medium
model is chosen according to the applications field.
Take the working fluid flow as an example. During the
gas path balancing, it acts as a means of data transmission
between two components. As depicted in Fig. 5, if
component A and B are coupled together by a working
fluid flow, the value of any thermodynamic parameter at A’s Fig. 5. Medium models.
 Y. Cao et al. / Advances in Engineering Software 36 (2005) 127–134 131

Fig. 6. Compressor flow model structure.

airflow and one for the coolant. So the three ports interface adiabatic efficiency his, is then computed by multiplying
model named ‘ThreePortsSingleDynamic’ is introduced. the normalized adiabatic efficiency by the designing
Although in some conditions, the thermo connector adiabatic efficiency value.
models used in the volume models are the same form as the  ðkK1=kÞ
p2
flow models’, the structures of them are different. In volume p1 K1
model pressure, specific enthalpy, density, temperature, his Z  ðkK1=h kÞ (6)
p2 p

specific entropy and ratio of specific heats of entrance p1 K 1

connector are respectively equal to these of export connector,
Based on shaft rotation speed, the pressure, temperature
however, in flow model the mass flow and connective heat
and enthalpy, etc., which are transferred from the previous
flow are the same but other variables are not.
and next volume models period. The power or rate of energy
transmitted from the shaft to the fluid in compressor flow
modes is:
4. Component example—compressor Pcomp Z mðhout K hin Þ (7)

In this section, a component of gas turbine—compressor Based on the OO technology, the compressor flow model
will be taken as an example to illustrate how to use the can be built through the inheritance and the objects’
model library explained above to build a component model. composition. Its structure is shown in Fig. 6.
The compressor model in the library can be divided into A complete compressor component model is constructed
two parts: one is an essential flow model. In this model, the through combining compressor flow model and volume
pressure potential and kinetic energy of the working fluid in model. It is shown in Fig. 7.
the compressor will be increased by transmitting the
mechanical energy of the shaft. And the other is the volume
model. The compressor model is also used to model the Fan 5. Gas turbine models library and an example
component in the engine model.
In the volume and flow models only thermodynamic 5.1. Gas turbine library
parameters are considered, i.e. pressures and temperatures,
as related to engine performance and performance analysis. The previous sections have described the architecture of
The engine shaft speed is another important engine Gas Turbine library, which is an object-oriented framework
parameter. Modeling of the thermodynamics depends on for modeling and simulation of aero gas turbine systems.
the shaft speeds and vice versa. In this case generalized This library can be classified into four parts: (1) Partial
semi-empirical engine component models (performance Components: it is the package of the aero-engine physical
maps) and mechanical model that are related to the components and volume components models. These
compressor may be used.
Compressor performance is represented by a set of
overall performance maps normalized to design point
values. Baseline performance maps provide normalized
corrected mass flow rate and normalized efficiency as a
function of normalized pressure ratio and normalized
corrected spool speed.
mcorrectedZ f ðu; p; TÞ (5)
The normalized adiabatic efficiency value is obtained
from the baseline compressor performance map. The Fig. 7. Compressor model.
132 Y. Cao et al. / Advances in Engineering Software 36 (2005) 127–134

the components are defined here; and (4) Demos. The

structure is shown in Fig. 8.
With the OO technology this library provides a flexible
and extensible environment that can be used to compose
new engine component models graphically or program-
matically, to inspect and edit the existing models,
and simulate the display execution results. It also provides
plug-compatible software components which users can
combine to form increasingly complex gas turbine engine
models according to application requirements.

5.2. An aero-engine model

In this section, we utilize the gas turbine library and

Fig. 8. Gas turbine library.
electrical library developed by our team to graphically build
and simulate the characteristics of gas turbine connecting
components are the basic blocks for building Aeroengine with a switch reluctance generator. An acceleration process,
models; (2) Base Classes: they are the bases of other models the shaft speed from 7000 to 14000 rpm, is simulated. The
in the library. They can be inherited or extended by other compressor performance map is compared with the GSP’s
models. The records sub-package is the class of the variables result. The NLR’s GSP (Gas turbine Simulation Program) is
and parameters, and the functions sub-package is the class of an off-line component-based modeling environment for gas
all functions. They are both put in the Base Classes package turbine. It is developed by the Netherlands National
as sub-packages. (3) Interfaces: the connectors among Aerospace Laboratory.

Fig. 9. Overview of gas turbine framework.

 Y. Cao et al. / Advances in Engineering Software 36 (2005) 127–134 133

Components are the basic blocks for building engine The Gas Turbine Engine sub-model represents a two-
models. In gas turbine library, a basic component model is spool engine. It includes the low-pressure compressor and
represented by an icon. Components are dragged-and- turbine, combustor, high-pressure compressor and turbine,
dropped from library windows onto the model window and low and high-speed shaft, fuel pump, nozzle, power turbine.
arranged to form the desired aero engine system configur- The volume models are placed between two flow models to
ation (see Fig. 9) according to the prototype. The main define the initial conditions, boundary conditions and handle
frame model (left figures of Fig. 9) is divided into three sub- inter-component fluid dynamics.
models: Atmospheres, Gas Turbine Engine, and The SR_Generator sub-model represents an aero gen-
SR_Generator. erator. It is driven by the gas turbine through a gearbox.
Atmosphere sub-model represents the environment. In In unsteady simulations, the independent variables are
this sub-model some variables, such as the atmosphere functions of time and may be implicitly integrated to predict
temperature, pressure, Mach number and so on are their value at a future time based on known past values. So
calculated. before proceeding to the transient analysis, the simulation

Fig. 10. Some results of the model.

134 Y. Cao et al. / Advances in Engineering Software 36 (2005) 127–134

first attempts to drive the engine to balanced (stable-state) The Gas Turbine Library described in this paper provides
conditions to make sure the values of initial operating framework components that, together, form an integrated
point. This ensures that the engine model is in a consistent, system for aircraft gas turbine engine. It can help one to
physically valid operating state before beginning the understand the engine performance behavior and to identify
transient operation. the causes of possible deficiencies in engine performance in
After the model being built, each of these components is a highly cost-effective manner. It has also become very
shown on a model window whose properties and charac- useful for efficient mission analysis, the preliminary design
teristics can be specified by user via data modification. studies of engines and their matching to airframes.
Compressor model has input fields named compressor The object-oriented technology is utilized to produce a
customizer for defining design point values and perform- reusable component-based architecture that can be extended
ance map. So do the turbines, nozzle, and duct. The initial and customized to meet future application requirements.
operating point values are set on volume model’s The modular individual components in this library can be
customizer. easily modified and extended, so new components models
The DASSL numerical algorithm method is used to can be added conveniently. As a result the library will be
resolve the problem of the modeling. Relative tolerances enlarged and changed by users, the library can be adjusted
have to be assigned to tell the integration methods how and updated by users with the aero engine developing.
accurately the solution should be computed. The accuracy The Dymola/Modelica is a new powerful simulation tool.
of the computed state variables is comparable to the relative In this paper it is applied in a new engineering field, and
error tolerances. The integration step-size is chosen in such proved to be successful. On the other hand its libraries group
a way, that the local error is smaller than the desired has been enriched too.
maximum local error, defined via the relative and absolute
tolerance. The relative tolerance is 1E-4.
Once the initial and boundary condition values are fixed, Acknowledgements
the model can be simulated. When the simulation finishes
successfully, then the results are available. All variables for The work described in this paper was made possible through
each component in the simulation model can be plotted. funds from the United Technology Research Center (UTRC).
Fig. 10(a) shows the compressor performance map, and We would like to thank all of the members of the team at
Fig. 10(b) shows the same map of GSP. Comparing Shanhai Jiaotong University and UTRC.
these two plots, we can conclude that our model is effective
and correct, Fig. 10(c) and (d) show the temperature before
turbine (T3*) and the compressor outlet’s pressure.
References
Especially, in Fig. 10(c) there is a decline of T3* during
the acceleration process. With the mechanical inertia and [1] Xie Z, Su M, Weng S. Extensible object model for gas turbine engine
cubage of volume model decreasing, the T3* decline will simulation [J]. Appl Thermal Eng 2001;21:111–8.
gradually reduce. That is to say, the inertia and the volume [2] Visser WPJ, Broomhead MJ. A generic object-oriented gas turbine
in gas turbine affect the acceleration remarkably. Fig. 10(e) simulation environment. ASME Turbo Expo 2000, Munich, Germany,
8–11 May 2000.
and (f) also show the plot of single-phase flux linkage circle
[3] RTO/NATO. Performance prediction and simulation gas turbine
and load current of the SR_Generator model. engine operation. Joseph Ottawa/Hull, Canada J8X 1C6.
[4] Bayt RL, Sahm MK, Lents CE. Collaborative engineering in integrated
aircraft power system design. ASME 2002 Design Engineering
6. Conclusion Technical Conferences and Computers and Information in Engineering
Conference, Montreal, Canada, September 29–October 2, 2002.
[5] Michael MT. Introduction to physical modeling with Modelica[M].
Designing and developing new aero engine is a time- Dordrecht: Kluwer Academic Publishers; 2001.
consuming and expensive process. Computational simu- [6] Johnson RE, Foote B. Designing reusable classes. J Object-Oriented
lation is a powerful means for alleviating this cost, thanks to Programming 1998;11(2):22–35.
the flexibility it provides for rapid and relatively inexpen- [7] Reed JA, Afjeh AA. Computational simulation of gas turbines.
sive evaluation of alternative designs in design phase. So a International Gas Turbine and Aeroengine Congress and Exhibition,
Indianapolis, Indiana, June 7–10 1999.
framework named Gas Turbine Library representing engine [8] Pérez A.G. Modeling of a gas turbine with Modelica[D]. Master Thesis.
components, subcomponents and subassemblies is devel- Lund, Sweden: Department of Automatic Control Lund Institute of
oped and illustrated in this paper. Technology; 2001.