Vous êtes sur la page 1sur 16

Research Findings

The basic notion of this research has been confirmed both experimentally
and on a simulation basis. The basic notion is that a linear control systems
design of the system where the closed-loop poles are placed on the right half
plane resulting in a power producing engine. Initial work regarding the
correspondence between the linearized model being closed-loop linearly
unstable and the full nonlinear model inducing a limit cycle has also been
observed both in simulation and experimentally.

Figure 1 Photograph and schematic of the


engine demonstrator used for model
validation.
Figure 2 Schematic of the engine demonstrator
with measured variables and parameters.

Our investigation has focused on measuring the displacements, velocities,


pressures and temperatures in a commercially available free-piston engine
demonstrator as an experimental platform for modeling and model
validation. Delays resulted in instrumenting this engine adequately and has
led to a 6 month no-cost extension of the project. We have solved our
instrumentation problems and are now collecting good data. Figures 1 and 2
show the engine demonstrator and a schematic of the engine defining
relevant variables.

Model Development

Our investigation has focused on casting physics-based models of the


demonstrator engine with varying degrees of model reduction. To date we
have formulated a 4th, 5th, 6th, 7th, and 8th order nonlinear models and
then validated each with experimental data. We have also linearized each of
these models to investigate the linear control design hypotheses as
described above in this document.

The models developed are described below in brief and then elaborated
upon further along in the report:

4th order nonlinear model: This model utilizes the Schmidt model treated
in the literature to reduce heat transfer and enthalpy energy flows to position
based internal energy approximations. Internal energies and therefore
temperatures of different sections of the engine as thus position dependent
leading to the proposed feedback perspective.

5th order nonlinear model: A single control volume is used to model the
interior of the engine. This multi-port results in pressure dynamics that effect
the inertial elements of the system. Heat transfer rates into and out of the
CV are position dependent. Temperature dynamics therefore result, whereas
the 4th order model assumes that the temperature dynamics are
instantaneous.

6th order nonlinear model: The interior of the engine is represented as


two CVs. Each CV is modeled in a manner similar to that of the 5th order
model except that now interaction enthalpy flow rates are present between
the CVs.

7th and 8th order nonlinear models: The 7th order model has three
interaction CVs and the 8th order model has 4. As the number of CVs is
increased, the temperature gradient of the regenerator is represented as n
discrete sections.

For all of the models, the inertial elements of the engine, namely the
displacer and power pistons, are subject to the following equations of
motion.

(1)

(2)

For model reductions (4th and 5th order models) in which the pressures of the
two sides of the displacer section are assumed equalized at all points in time,
, and the inertial equations of motion reduce to the following:

(3)

(4)

Also common to all models are the equations calculating the volume of the
hot space , the volume of the cold space , and the volume along the
sides of the displacers piston (the regenerator volume) . The derivative of
the volume is also convenient to define:

(5)
(6)
(7)
(8)
(9)

The various models also utilize the piston wall areas revealed by the piston
displacements. The variable exposed wall areas alter the heat transfer into
and out of the system:

(10)
(11)

4th order model (the Schmidt model)

The Schmidt model assumes rapid heat transfer into and out of the engine
such that the hot and cold volumes in the engine quickly reach the hot and
cold temperatures. The Schmidt model also assumes that the pressures in
the hot and cold spaces are equal, implying that the mass flow rates
between the two are large.

This commonly accepted method for describing the temperature dependent


pressure in the engine, as described by Schmidt [2], permits the following
equation:

−1
V V 
P = mR  c + e  (12)
 Tk Th 

This model simplification of the heat transfer and its effects, results in a state
dependent pressure more easily seen in the following expression:

P = ( C1 A − C 2 A + C 2 Ar ) x + C 2 ( A − Ar ) y (13)
Where the constants are defined as:

−2
V V   1 
C1 = mR  h 0 + k 0    (14)
 Th Tk   Th 
−2
V V   1 
C 2 = mR  h 0 + k 0    (15)
 Th Tk   Tk 

The 4th order model of the engine is thus given by Equations (3), (4) and (13).
It is important to note that Equation (13) can be interpreted as an input term
to the coupled system dynamics of Equations (3) and (4). It indicates that the
pressure developed above or below the equilibrium pressure is a function of
the positions of the piston and displacer. Further, this “input term” is a
function of two of the states of the system, x and y, and are effectively
feedback gains which can be altered by manipulating the parameters of the
system. This interpretation serves to cast the problem as a standard
feedback control design problem, thereby offering a constructive control-
based design methodology for the closed-loop dynamics of the system.

5th order model

The 5th order model of the system removes the assumed fast heat transfer
assumption of the 4th order Schmidt model. We view this as an important
characteristic to include given that the power density of Stirling engines is
limited by heat transfer rates. The 5th order preserves the assumption of a
single pressure inside the engine and therefore continues the assumption of
the 4th order model that mass flows between the hot and cold sides are fast
compared to the other dynamics in the engine.

An energetic control volume approach is taken with regard to the rate of


internal energy in the control volume equally the net enthalpy rates (zero in
this model), the net heat transfer rates and the work rate of the CV:

(16)

With zero enthalpy rates, this produces the following pressure dynamics:

(17)

(18)

Where the heat transfer rates are given by:


(19)
(20)
(21)
(22)

and the temperature in the CV is given by the ideal gas law:

(23)

The core of the 5th order model is thus formed by Equations (3), (4), (18-22)
with the variable wall areas given by Equations (10) and (11). In terms of the
feedback perspective, the 5th order model has the following state vector:
(24)

6th order model

The 6th order model of the system subdivides the control volume into two
portions: the hot side plus the upper half of the regenerator volume, and the
cold side plus the lower half of the regenerator volume. These two control
volumes are similar to the single control volume of the 5th order model with
the important distinction that they now each contain an enthalpy term for
communication between the two. The mass flow rate is given by the
simplified linear mass flow:

(25)

The resulting enthalpy rate communicating with each CV is then given as:

(26)

where the temperature is that inside the CV.

In terms of the feedback perspective, the 6th order model has the following
state vector:

(27)

7th and 8th order model

Each of these models further subdivides the control volumes. The 7th order
model has three CV’s; hot side, cold side, and regenerator space. The 8th
order model has four CV’s where the regenerator space CV of the 7th order
model is split into two. Each addition of a CV adds another state to the
system associated with the pressure within that CV.

Model Validation

The above models were validated by utilizing measured data from the setup
shown in Figures 1 and 2. Since many of the parameters of the model are
difficult to directly measure, system identification was performed on the
parameters of the highest order model. These same parameters were then
used in the lower order models that called for them. The figure below shows
the measured pressure (on the low temperature side), and measured
displacements of the displacer and power pistons. This representative piece
of data shows the engine during “start-up” after it was perturbed from its
equilibrium position. From the figure it is clear that the equilibrium position is
linearly unstable. The period of time for the collected data is rather short in
order to avoid internal collisions that occur within the engine after it reaches
sufficient amplitudes. Since we are trying to capture mainly the linear
aspects of the model and given that modeling collisions is difficult, short
segments of data such as the one shown below have been used.

5
x 10
1.1

1.05
Pressure (Pa)

0.95

0.9
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

0.14

0.135
x (m)

0.13

0.125

0.12
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

0.05

0.048
y (m)

0.046

0.044

0.042
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Time (sec)

Figure 3: Measured pressure and displacement data from the


experimental setup.

Two levels of system identification were performed. The first step was to
utilize the measured pressure as an input to a model of the inertial elements
of the system given by Equations (3) and (4). The parameters of Equation (3)
and (4) were then identified such that the modeled displacement evolution
was as close to the measured displacement evolution as possible.
The second step was to then identify the parameters of the portions of each
model besides Equations (3) and (4) [or (1) and (2)] for some models. This
step was difficult given the fact that we have no pressure data for the hot
side, or for the various subdivided CV’s, and these had to be estimated. Work
on this is still on-going, but the figure below shows the modeled and
measured pressures after the model parameters were identified. It should be
noted that the measured displacements were used as inputs to the models,
and only the initial conditions of each model’s cold side pressure were
matched to the measured cold side pressure.

5
x 10
1.15
Measured Data
Schmidt Model
5th Order
1.1
6th Order
7th Order
8th Order
1.05
Pressure (Pa)

0.95

0.9

0.85
0 0.05 0.1 0.15 0.2 0.25 0.3
Time (seconds)

Figure 4: Parameters of the various order models were identified

Finally, once all parameters for each system were identified each model was
checked by allowing it to evolve freely from the measured initial conditions
(meaning all pressures and displacements are generated by the model only
after the initial starting conditions). The figure below shows the model
responses. Although results from this system identification process are still
being finalized and will be the topic of a journal paper currently in
preparation, it appears that the different models capture different aspects of
the system. It should be noted that all models are subject to drift and
although the results at first do not look very good, these models are allowed
to evolve only from the measured starting condition. General phase matches
and magnitude growths are seen most clearly in the power piston evolution,
y.
5
x 10
1.1
Measured Data
Pressure (Pa) 1.05 Schmidt Model
5th Order Model
1 6th Order Model
7th Order Model
0.95 8th Order Model

0.9

0.01

0.005
x (m)

-0.005

-0.01

0.01

0.005
y (m)

-0.005

-0.01
0 0.05 0.1 0.15 0.2 0.25 0.3
Time (seconds)

Figure 5: The response of each model utilizing only the initial


measured states.

Linearized Models

Once parameters of each nonlinear model is found, a linearization of the


model is performed in order to study the closed-loop pole locations. Results
for the 4th order Schmidt model are below. Results for other models are still
being finalized.

With regard to the 4th order Schmidt model, the linearized system dynamics
of equations (1), (2), and (6) can be represented in state-space as:

x = Ax + Bu
 x   0 1 0 0  x  0 
 x   − k d − ( bd + b )
0 b     Ar 
  =  md md md   x  +  md u (28)
 y   0 0 0 1  y  0 
  0 b −kd −(b p +b )     A− Ar 
 y   mp mp mp   y   m p 

with an input term given by the state-dependent feedback equation:


u = −Kx
u = −( P − P0 )
x 
 x  (29)
= −[C1 A − C 2 ( A − Ar ) 0 C 2 ( A − Ar ) 0]  
y
 
 y 

The location of the closed loop poles under these conditions reveals valuable
insight into the operation of a Stirling engine viewed as a dynamic system
subject to a state-dependent physical feedback law. As shown below in
Figure 6, two of the closed loop poles are in the right-half plane, and their
location corresponds to the operating frequency of the engine. This
interpretation partially collaborates the hypothesis that a free-piston Stirling
engine produces power and maintains self-sustained oscillation if the system
has unstable closed-loop poles. This linearized interpretation is valid about a
small operating region of the equilibrium values, whereas the real nonlinear
dynamics will limit the amplitude of oscillation as the states depart far
enough from the equilibrium.

100

80

60

40

20
Imag

-20

-40

-60

-80

-100
-30 -25 -20 -15 -10 -5 0 5 10
Real

Figure 6: Closed-loop pole locations of the linearized 4th order


Schmidt model.

The model can also be used to investigate open-loop properties of the


system. The Figure below illustrates the open-loop transfer function. A
frequency response plot of the open-loop transfer function of the 4th order
Schmidt model is shown below in Figure 8. Notice that it has negative gain
and phase margins, indicating instability robustness – or in other words, the
engine will initiate self oscillations and produce power with a degree of
model variation, inaccuracy, or unmodeled disturbances.
( A − BK )
Closed-loop poles:
Inertial dynamics Open-loop
Eigenvalues of transfer
+ + .
R =0 u x I x
-
B
+ s function:
A
R =0 + u1 x

Pressurefeedback u2
K K

Parameter dependent
feedbackgain K

U 2 ( s)
= K ( sI − A ) −1 B
U1 ( s)

Figure 7: Linearized dynamics of the B-10B free-piston Stirling engine reinterpreted


as a negative feedback loop. The pressure forces generated in the displacer chamber
act as negative feedback on the inertial dynamics of the system and effectively form
a feedback control law dependent on physical parameters of the system. The open
loop transfer function of the system can then be used to investigate robustness
terms from the frequency response plot and an assessment of the stability margins.

Bode Diagram
Gm = -6.63 dB (at 81.6 rad/sec) , Pm = -79.5 deg (at 91.3 rad/sec)
20

0
Magnitude (dB)

-20

-40

-60
360

180
Phase (deg)

-180
0 1 2 3
10 10 10 10
Frequency (rad/sec)

Figure 8: Open-loop frequency response plot of the 4th order model showing
instability margins.

Findings common to all models: The unified view that is beginning to emerge
from all of the models is that the flow rates of heat and enthalpy are always
state dependent quantities, where the states are those given by the coupled
dual mass-spring-damper systems of the engine. This is significant because
it allows a generalizable energetic interpretation of the feedback
mechanisms in the physical system.

This unified view (referred to in the previous section on research and


education activities as “the generality of the feedback perspective”) is
important because it allows highly coupled systems such as free-piston
Stirling engines to be viewed as a feedback system as shown in Figure 7. The
linearized version of any of the models therefore offers the open-loop
dynamics of this physically closed-loop system. The general objective of
classical control to provide a means to shape the closed-loop response by
understanding and shaping the open-loop system can be applied to the
highly-coupled and dynamic nature of the free-piston Stirling engine.

Regenerator Modeling

We have also investigated more detailed models of a generic regenerator


section. This is fully documented in CMMI Grantees Conference report of
2009.

Experimental Prototype

We have done some developmental work on the displacer section of a


custom designed Stirling engine. Shown below are designs of a displacer
section where the displacer can be arbitrarily control independent of the
power piston or pressures in the engine. This engine section will allow us to
investigate arbitrary dynamic couplings of the displacer and power pistons.
This new free-piston Stirling engine design utilized rack and pinion gears to
shift the displacer and power pistons back and forth in the engine chamber.
Hence, the engine chamber has to be designed in such a way that the inside
chamber can be interfaced with a spur gear that is connected to a motor. It
is important to have a chamber with an airtight seal to prevent the working
fluid from leaking out. So, the challenge was to design the engine’s cold part
properly that can contain the motor and gear assembly with an airtight seal.
Two designs are shown below.

Design 1:

Initially, the Stirling engine’s cold side chamber design consists of three parts
– an engine chamber, a motor case top part and a motor case bottom part
(see Figure 9). The motor case follows the shape of the motor, the bearing,
and the gear. The three parts are supposed to be connected together with
RTV adhesive and bolts, with an assumption that this assemble is strong
enough and thus no leakage.
Motor case
top part

Motor case
bottom part
Back
Cold Side Engine plug
chamber
Figure 9: Cold side engine block design #1

However, there are a couple problems with this design. Firstly, the parts are
not simple and easy to machine. As shown in the cross sectional view (Figure
10), the motor case diameter varies according to the diameter of the motor,
the bearing, the shaft, and the gear. The motor case was designed this way
in order to hold the mechanical parts (bearing, shaft and gear) in place and
to prevent them from sliding. While it is not impossible to machine the parts,
the machining cost can be cut down if a much simpler design is introduced.
The critical problem with this design is leakage. The pressure inside the
engine chamber can get very high, thus the RTV might not be able to
withstand the pressure and cause leakage. This critical failure will result in a
nonfunctioning Stirling engine.

bearing Motor case top part


Spur gear

Shaft Motor and Gearhead

Motor case bottom part

Back
plug
Cold Side Engine
chamber

Figure 10: Cross sectional area of the engine’s cold part assembly
design #1
Design 2:

To avoid leakage problem, the cold part engine block is designed as a


one piece. The engine chamber and the motor case are made of the same
rectangular box. Figure 11 shows the picture of the cold chamber design.
The two plugs at both ends hold the mechanical parts in place and provide a
complete airtight seal for the engine.

Figure 11: Cold side engine block design #2

Figure 12: Transparent view. The motor and


other mechanical parts are contained inside
the block.
Front plug

Motor, Gearhead, bearing, shaft, and gear


assembly
Figure 13: Motor assembly and the front plug that secures it in
place

The front plug is specially designed in such a way that it holds the bearing,
the motor and the shaft in place and prevents them from sliding to the left
and right. The plug also functions to eliminate excessive dead volume.

A set screw prevents the bearing


Bearing from sliding to the right
Spur gear

Shaft Motor and Gearhead

A screw holds the motor in place

Figure 14: Cross sectional view of cold side


engine assembly design #2, showing the front
plug that holds the motor assembly in place
Figure 15: rapid prototyped engine part and the actual motor
assembly

This design will be sent to the university’s machine shop for


fabrication. When all the parts have been machined, the power piston
section of the engine will be designed. After the power piston section has
been fabricated, the engine will be assembled and experimental work will
continue beyond the life of this grant.

Vous aimerez peut-être aussi