Original Article

Proc IMechE Part G:

J Aerospace Engineering
Multi-objective regulating and protecting 227(2) 311–325
! IMechE 2012

control for ducted rocket using a Reprints and permissions:

sagepub.co.uk/journalsPermissions.nav

bumpless transfer scheme DOI: 10.1177/0954410011433237

uk.sagepub.com/jaero

Wen Bao, Yiwen Qi and Juntao Chang

Abstract
This article deals with the problem of multi-objective regulating and protecting control for a ducted rocket, in order to
get maximum thrust while avoiding extremely dangerous phenomenon like inlet buzz. First, the mathematical models
involving gas flow regulating system and ducted rocket are introduced. The description of inlet buzz margin is also given
and analyzed. Second, the multi-objective switching control problem of ducted rocket is proposed and discussed. Third, a
robust bumpless transfer scheme for solving the problem is presented. The designed bumpless transfer compensator
based on model reference adaptive sliding-mode control ensures that the switched system performs a smooth transition
at the transfer moment. The stability analysis is then given utilizing the Lyapunov functional approach. Finally, the
bumpless transfer strategy is applied to the ducted rocket control system, and the simulation results show its
effectiveness.

Keywords
Switching control, ducted rocket, inlet buzz, bumpless transfer, adaptive sliding mode control

Date received: 11 August 2011; accepted: 23 November 2011

method,5,6 since linear method cannot work well in the

Introduction whole flight envelope of highly nonlinear aeroengine
Ducted rocket is a supersonic flight propulsion system system. Nonlinear control approaches have been devel-
which takes the exhaust from a solid fuel gas genera- oped to overcome the shortcomings of linear methods, in
tor, mixes it with air, and burns it to produce thrust. particular, the switching control strategies. In switching
This type of engine has the advantages of traditional control, the basic idea consists of using simplified linear
solid fuel rocket engines such as longer range, higher controllers designed for specific operation points and
speed, and throttleability.1 In particular, the ducted some rules that allow to switch among them. Clearly,
rocket’s specific impulse is four to six times the the above-mentioned limitations on aeroengine result
normal solid rocket. Also, the use of a solid fuel in a family of subsystems which switch interactively
ducted rocket during the acceleration phase of a hyper- depending on various environmental factors. Thus, the
sonic cruise missile is shown to provide a significant aeroengine control can be constructed as a switching
increase in the missile average velocity when compared control problem.
to a conventional liquid fuel ramjet.2 Ducted rocket is In the literature, the stability issues have been a
first used in the Soviet surface-to-air missile in the early major focus in studying switched systems.7–10 Since
1960s. Recently, there has been some renewed interest rapid switching between two controllers may intrigue
in ramjet missile propulsion.3,4
Ducted rocket engine is a highly nonlinear, time-vary-
ing complex system. The property and parameters of School of Energy Science and Engineering, Harbin Institute of Technology,
engine widely and indefinitely vary when working con- People’s Republic of China
ditions change, which make the aeroengine inevitably
Corresponding author:
touch some of the safety limitations, such as tempera- Wen Bao, School of Energy Science and Engineering, Harbin Institute of
ture, air pressure, and material strength. It is difficult to Technology, Harbin 150001, People’s Republic of China.
build an efficient control system using traditional design Email: [email protected]

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

312 Proc IMechE Part G: J Aerospace Engineering 227(2)

dramatic amplitude oscillations and excite undesired The remainder of this article is organized as follows.
plant dynamics if no action is taken to avoid it, tradi- The model and control objectives are given in second sec-
tional control approaches have to focus on stability and tion, in which the multi-loop control system of ducted
safety at the expense of response speed. To solve this rocket is described and the control problems including
issue, bumpless transfer is required. Bumpless transfer the design contradictions in switching control are formu-
means that there is no controller-induced bump or lated in detail. In third section, the procedure to design the
undesirable transient to the process when the control bumpless transfer compensator along with the proof of
mode is changed. Over the past several decades, stability is developed. In fourth section the simulation
many contributions have been made to study the bump- results are presented. Finally, the article is concluded.
less transfer. Anti-windup bumpless transfer (AWBT)
schemes are concerned by Hanus et al.,11 Graebe and
Ahlen,12 and Edwards and Postlethwaite.13 Optimal Model and control objectives
linear quadratic control method is largely discussed
Mathematical model of the ducted rocket
by Turner and Walker14 and Zheng and Bentsman,15
and many applications are tentatively implemented.16,17 The system is described according to the dominant
L2 norm bounded structure is shown by Zaccarian and physical phenomena. The typical controlled plant for
Teel,18 moreover, interpolation,19 Min/Max strategy,20 thrust control includes an actuator (i.e. the gas regulat-
and observer-based technique.21 For more detailed dis- ing system) and the model of ducted rocket, which will
cussion on the advantages and conservativeness of be separately introduced in the following paragraphs.
these methods, see the study of Bao et al.22 In addition, for the inlet buzz protecting control, it
Particularly, a multi-objective switching control should also include the part of inlet.
scheme has been proposed and applied to control the
ducted rocket.20 In the study of Bao et al.,20 by decom- Gas regulating system. A pressure-balanced gas regulating
posing the two issues of regulation and protection, con- system, as given in the study of Shi et al.,23 is depicted
trollers can be individually designed and switched with in Figure 1. In the diagram, when the gas flow demand
each other, which tentatively deal with the conflict rises, the servo valve (2) can make the gas go into the
between rapid response and safety. Nevertheless, there valve head chamber (6) from a gas bottle (1). Then, the
is a limitation to this method, that is, the effectiveness champer pressure is increased and the valve head (5)’s
entirely depends on the designer’s experience.20 In other force balance is broken. Meanwhile, the valve head (5)
words, it depends on the choose of integral limiters in goes forward along the valve stem (7), which makes the
the subcontrol loops. Actually, a too small limiter will gas generator’s (3) pressure rise but its throat area (4)
lead the designed capability of corresponding controller decrease. The reader can refer to the study of Shi et al.23
to totally fail and the case that the controller is not able for further details. The mathematical model of gas flow
to switch to other loops, while a too large limiter will regulating system for ducted rocket can be described by
result in switch slowly, i.e. a switch should occur, but the following three equations.
did not happen. K1
Motivated by the above observations, this article p~ v ðsÞ ¼ ~
qðsÞ ð1Þ
T1 s þ 1
attempts to use a bumpless transfer method to solve
the problem of regulating and protecting control for K2
p~ r ðsÞ ¼ 2 2
p~ v ðsÞ ð2Þ
ducted rocket. First, the dynamic models, including T2 s þ 2T2 s þ 1
gas flow regulating system and ducted rocket engine,
are generally introduced. Since inlet buzz is a phenom- K3
w~ g ðsÞ ¼ ð1  T3 sÞp~ r ðsÞ ð3Þ
enon of ducted rocket that will extremely reduce sys- K4
tem’s performance and should be completely avoided,
inlet buzz margin description is introduced into ducted where q is the drive current of servo valve, pv, pr, and
rocket system. Then, the existing problem of switching wg, respectively, denote the pressure in the valve head
control is analyzed, leading to the construction of a chamber, the pressure in the gas generator chamber,
bumpless transfer problem. Moreover, an adaptive slid- and the gas flow rate. The overhead tilde sign indicates
ing-mode bumpless transfer design method is corre- Laplace transformation and the symbol  the deviation
spondingly presented. This enables the multi-objective from an equilibrium state which is a steady-state oper-
control problem addressed to be established in the form ating point of the gas regulating system.
of a compensator design problem. In addition to the
robustness to disturbances, the proposed method is Ducted rocket. Rocket control problem is actually its
comparatively simple and effective. At last, simulations thrust control problem. However, the online measure-
are carried out to validate the design method. ment of thrust is not easy, and the backpressure of inlet

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

Bao et al. 313

Figure 1. Schematic of pressure-balanced pintle-type gas regulating system.23

is instead. Recently, the mathematical model, parame-

ter analysis, and model reduction have been investi-
gated by Niu et al.24 giving a simplified mathematical
model of controllable engine. The model is given by

ð1 þ  1 sÞð1 þ  2 sÞ
p~ b ðsÞ ¼ K5 es w~ g ðsÞ ð4Þ
ð1 þ  3 sÞð1 þ  4 sÞ

where pb denotes the backpressure of inlet.

Nevertheless, the model is still complex for control
designs. Considering the physical meaning of parame-
ters, it is known that the time constants ( i, i ¼ 1,2,3,4)
of engine are made up of volume–time constants. The
volume of solid rocket is normally less than that of the
liquid rocket, so that the time constants of the solid
rocket are relatively smaller. Thus the quadric coeffi- Figure 2. Buzz boundary of a supersonic inlet.23
cient of dynamic model is very small. Moreover, the
small delay in equation (4) can be classified as a trans- instability caused by the shock wave rapidly being alter-
portation lag and it can be approximated by a first- nately swallowed and expelled at the inlet of the rocket.
order transfer function. Based on these analysis, the Also, inlet buzz can significantly deteriorate the propul-
dynamic model is reduced as follows sion system such as large pressure oscillation, combus-
tion instability, engine surge, and thrust loss. For not
1 þ ð1 þ 2 Þs encountering inlet buzz, it is necessary to acquire and
p~ b ðsÞ ¼ K5 w~ g ðsÞ ð5Þ
ð1 þ sÞ½1 þ ð3 þ 4 Þs manifest the inlet buzz boundary. The buzz boundary
of a supersonic inlet, given by Shi et al.23 and Chang
In addition to the above two sections, the descrip- et al.25 is adopted to investigate the protecting control
tion of inlet buzz boundary is a very important issue design problems, which is defined as the maximum pres-
which must be considered. Then, we give the following sure ratio of inlet pu at different free-stream conditions.
discussion. The inlet buzz boundary (Figure 2) at the point F (bF,
M0F, pbF) can be obtained by puF ¼ pu(bF, M0F). Then,
Inlet buzz boundary and margin. The inlet of ducted rocket the inlet buzz margin can be defined as Z ¼ (puF  pbF)/
is required to perform well over a wide range of alti- puF (where pbF is the current pressure ratio of supersonic
tudes, flight speeds, angles of attack, and at all engine inlet), which indicates the extent from the present work-
power levels. For a ducted rocket, inlet buzz is an airflow ing point to inlet buzz boundary. The closer the better of

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

314 Proc IMechE Part G: J Aerospace Engineering 227(2)

the supersonic inlet performance, the safety of inlet dete- authority will be transferred to the protecting control
riorates. Consider the practical application, a warning loop in order to avoid exceeding the safety limitations.
line lower than inlet buzz boundary is often chosen. Further, different subloops have different tasks.
Protecting control loops concern about the system’s
limitations, and protecting safety is the only target;
Control objectives
regardless of safety, regulation loop whose sole purpose
The function of an aeroengine is to provide necessary is to maximize the system’s performance.
power to aircraft for flight. For the purpose of achiev-
ing high flight performance, a superior engine control Bumpless transfer. The switching mechanism posed in
system is practically needed. This control system must above section is very simple, but it is not easy to achieve.
be capable of dealing with many unexpected limitations Actually, in switched systems, it is common that the
within a large working envelope. These limitations plant input is temporarily different from the controller
involve temperature, air pressure, fuel equivalence output, which is caused by a substitution due to the
ratio, material strength, etc. instantaneous switching between different control laws.
Moreover, switching among controllers implies control
Switching control. No matter what control method is discontinuities and undesired transient behaviors if no
adopted, the engine will encounter limitations on action is taken to avoid it. Although it is notable that
some occasions. For example, in the acceleration pro- a discontinuous control is not always harmful but useful
cess of ducted rocket, a large thrust command will lead to improve the performance of systems when it is well
to more fuel being burned to provide sufficient power. designed, jumps at the input of system are not desirable
Gas generated in gas generator will get into the com- in many systems, especially in ducted rocket control
bustion chamber and result in increased inlet backpres- system. Thus, finding a smooth way to switch between
sure. If the control system continues to pursue the two controllers comes up as a crucial problem. The sup-
thrust command, inlet backpressure will continue to pression of the jumps at the plant input and their asso-
rise, causing the system to pass through the inlet buzz ciated transient effects is referred to as bumpless transfer.
boundary. This high-risk phenomenon is not absolutely The following section will give a bumpless transfer
allowed. Traditional control design methods have to scheme for solving this issue.
leave enough margins of safety and only obtain conser-
vative design performances at the expense of the sys- Multi-objective adaptive sliding mode
tem’s rapid response. Hence, it is necessary to introduce
multi-objective control into ducted rocket to eliminate
bumpless transfer control
the contradiction (such as the case between thrust con- The controlled plant can be obtained by augmenting
trol and inlet buzz control). equations (1) to (3) and equation (5), the state-space
Considering the switching control scheme shown in form is given by
Figure 3, it contains several subcontrol loops, such as
_ ¼ AxðtÞ þ BuðtÞ þ d ðtÞ
xðtÞ
regulation and protection. When system is in ordinary
conditions, thrust control loop will work to provide a uðtÞ ¼ uðtÞ ðtÞ
quick and accurate response. When system operates yi ðtÞ ¼ Ci xðtÞ ð6Þ
near the safety boundary (or warning line), control

Figure 3. Diagram of a ducted rocket control system.

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

Bao et al. 315

where x(t) 2 Rn is the state vector, u(t) 2 Rm the control equations (6) and (9)
input, yi(t) 2 Rl, i 2 I N the ith output vector represent- " # " #
ing a regulating purpose. ui(t) 2 Rm is the ith subcon- A þ BDcon1 BCcon BDcon2
Adon ¼ , Bdon ¼
troller output and d(t) 2 Rn the external disturbance of Bcon1 Acon Bcon2
the system. s(t) : [0, þ1) ! I N ¼ {1, 2,. . ., N} is a  
I  
switching signal. A, B, and Ci are real constant matrices Bd ¼ , Cdon ¼ Dcon1 Ccon , Ddon ¼ Dcon2
0
of appropriate dimensions, and the triplet (A, B, Ci) is
controllable and observable. Hence, from the previous definitions, the on-line
closed-loop system is described as
Assumption 1. The disturbance d(t) is assumed to be 
x_ don ðtÞ ¼ Adon xdon ðtÞ þ Bdon ron þ Bd dðtÞ
bounded with a positive but unknown constant D, S don : ð11Þ
uon ðtÞ ¼ Cdon xdon ðtÞ þ Ddon ron
and it can be expressed as Id(t)I 4 D.
The selection and tuning of sub-controller is a very impor-  TSimilarly,T define an extended state vector xauoff ¼
tant part of the switching control design. Proportional– x xTcoff 2 Rnþl and the following matrices from
integral–derivative (PID) controllers are the most equations (6) and (10)
adopted controllers in industrial settings because of the " #
advantageous cost/benefit ratio that they are able to pro-  
A þ BDcon1 0 BDcon2
vide. Then, consider the PI controller described by Aauoff ¼ , Bauoff ¼
Bcoff1 Acoff 0
Rt   " #
Ci : ui ¼ KPi ðri  yi Þ þ KIi  yi Þd, i 2 I N ð7Þ 0 BCcon 0
0 ðri L¼ ; E¼
0 0 Bcoff2
where ri is a constant reference signal, KPi, KIi 2 Rml  
denote proportional and integral gain matrices, respec- Cauoff ¼ Dcoff1 Ccoff , Dauoff ¼ Dcoff 2

tively. In practice, these controllers (7) to be designed

may have different parameters for different purposes The augmented system for the off-line controller Coff
(e.g., thrust regulation, safe protection). At any moment, can be given by
only one controller is activated. Equivalently, the state- 8
space equation of equation (7) can be described as < x_ auoff ðtÞ ¼ Aauoff xauoff ðtÞ þ Bauoff ron þ Bd d ðtÞ
S auoff : þLxdon ðtÞ þ Eðroff þ vÞ
( :
uoff ðtÞ ¼ Cauoff xauoff ðtÞ þ Dauoff roff
x_ ci ¼ Aci xci þ Bci1 x þ Bci2 ri
Ci : , i 2 IN ð8Þ ð12Þ
ui ¼ Cci xc þ Dci1 x þ Dci2 ri
Let S don be the reference model of S auoff, and the
Rt model matching error is defined as
by defining the auxiliary state xci ¼ 0 ðri  yi Þd 2 Rl.
Where Aci ¼ 0, Bci1 ¼ Ci, Bci2 ¼ I, Cci ¼ KIi, Dci1 ¼ KPiCi, eðtÞ ¼ uoff ðtÞ  uon ðtÞ ð13Þ
and Dci2 ¼ KPi.
Without loss of generality, we consider the switched The objective of this section is to find a robust
system (6) with only two candidate controllers (8) (C1 (bumpless transfer) compensator, v, such that the
and C2). Suppose that C1 is the first to be used. output of equation (12) follows the output of model
Meanwhile, C2 is off-line. To avoid confusion, we (11) in the presence of external disturbances. To
denote in this section that Con ¼ C1 and Coff ¼ C2, that is ensure that the reference model S don is trackable, we
make the following assumption.

x_ con ¼ Acon xcon þ Bcon1 x þ Bcon2 ron
Con : ð9Þ Assumption 2. There exist matrices W1 , W2 , and W3
uon ¼ Ccon xcon þ Dcon1 x þ Dcon2 ron
such that the following equations are satisfied

x_ coff ¼ Acoff xcoff þ Bcoff1 x þ Bcoff2 ðroff þ vÞ Aauoff ¼ EW1
Coff : ð10Þ
uoff ¼ Ccoff xcoff þ Dcoff1 x þ Dcoff2 roff Cauoff L  Cdon Adon ¼ Cauoff EW2
Cauoff Bauoff  Cdon Bdon ¼ Cauoff EW3 ð14Þ
where v 2 Rl is the output of the robust bumpless trans-
fer compensator to be designed.
 TDefine Tnow an extended state vector xdon ¼ Taking the time derivative of the both sides of equa-
x xTcon 2 Rnþl and the following matrices from tion (13) and using the state equations of equations (11)

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

316 Proc IMechE Part G: J Aerospace Engineering 227(2)

and (12) lead to the error dynamic equation as shown veq ¼ W1 xauoff  W2 xdon  W3 ron þ Ke e  roff
below  ðCauoff EÞ1 ðCauoff  Cdon ÞBd d ð19Þ
e_ ¼ u_ off  u_ on
Substituting equations (14) and (19) into equation (15),
¼ ðCauoff Aauoff Þxauoff þ ðCauoff L  Cdon Adon Þxdon
then the error trajectories reaching the sliding surface,
þ ðCauoff  Cdon ÞBd d þ ðCauoff Bauoff  Cdon Bdon Þron the sliding motion on the sliding surface is
þ ðCauoff EÞroff þ ðCauoff EÞv ð15Þ
e_ ¼ ðCauoff EKe Þe þ ½I  ðCauoff EÞðCauoff EÞ1 
The PI sliding surface is chosen due to its ability to  ðCauoff  Cdon ÞBd d ð20Þ
reduce chattering and the ability to achieve zero steady-
state error.26 It is defined as It is clear that the error dynamics of equation (20)
would be influenced by disturbance d (d 6¼ 0). Then,
Rt
SðtÞ ¼ eðtÞ  0 ðCauoff EKe ÞeðÞd ¼0 ð16Þ what it should do next is to design an adaptive control-
ler, which not only drives the system trajectories onto
where  is a constant matrix and it is chosen such the sliding surface but also suppresses the effect of
that CauoffE is nonsingular, Ke is selected to satisfy disturbances.
the condition that CauoffEKe is Hurwitz. Taking the An adaptive control (i.e. the bumpless transfer com-
time derivative of the sliding variable in equation (16) pensator, Figure 4) is proposed as
results in
v ¼ W1 xauoff  W2 xdon  W3 ron þ Ke e  roff
S_ ¼ ½ðCauoff Aauoff Þxauoff þ ðCauoff L  Cdon Adon Þxdon S
 ðCauoff EÞ1 ð21Þ
þ ðCauoff  Cdon ÞBd d þ ðCauoff Bauoff  Cdon Bdon Þron kSk
þ ðCauoff EÞroff þ ðCauoff EÞv  ðCauoff EKe Þe ð17Þ where W1, W2, and W3 are the estimates of W1 , W2
and W3 , respectively, and  a constant. The last term of
From equations (14) and (17), the dynamics of S(t) compensator is to compensate the external
are disturbances.
Define the estimation errors as: W ~ 1 ðtÞ ¼ W1(t) 
S_ ¼ Cauoff E½W1 xauoff þW2 xdon þW3 ron Ke eþroff þv ~ ~
W1 , W2 ðtÞ ¼ W2(t)  W2 , and W3 ðtÞ ¼ W3(t)  W3 :
 

þðCauoff Cdon ÞBd d ð18Þ Then, from this and equation (21), equation (18) can be
further rewritten as
Consequently, the following equivalent control law can
S_ ¼ Cauoff EðW ~ 1 xauoff þ W
~ 2 xdon þ W ~ 3 ron Þ   S
_ ¼0
be obtained from SðtÞ kSk
þ ðCauoff  Cdon ÞBd d ð22Þ

Figure 4. The proposed multi-objective bumpless transfer scheme.

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

Bao et al. 317

For finding suitable updating laws for W1, W2, and _ ¼ 0, it indicates
the sliding mode occurs, S(t) ¼ SðtÞ
W3 to achieve stability of the closed-loop system, a that the error e(t) will be driven to zero according to
Lyapunov function candidate is chosen as equation (16).
" ! ! !# Consequently, we give the following theorem.
1 T ~ TW
W ~1 ~ TW
W ~2 ~ TW
W ~3
1 2 3
VðtÞ ¼ S S þ tr þ tr þ tr
2 1 2 3
Theorem 1. Consider the off-line augmented system (12)
where  1,  2 and  3 are three positive constants. Take with sliding surface (16) and compensator (21), satisfy-
the time derivative of V(t) along the trajectory of equa- ing assumptions 1 and 2. If the proposed control
tion (22) with some simple manipulations to have parameter matrices satisfy the adaptation laws (23),
then the following statements hold.
! ! !
~ T _~ ~ T _~ ~ T _~
_ ¼ ST S_ þ tr W1 W1 þ tr W2 W2 þ tr W3 W3
VðtÞ
1 2 3 1. The sliding variable S asymptotically converges to
~ 1 xauoff þ W ~ 2 xdon þ W
~ 3 ron Þ
zero.
¼ ST ½Cauoff EðW
! 2. The steady state of model matching error e(t) satis-
~ TW _~
S W 1 1 fies lim eðtÞ ! 0:
 þ ðCauoff  Cdon ÞBd d  þ tr t!1
kSk 1
! !
~ TW _~ ~ TW _~
W 2 2 W 3 3
þ tr þ tr Proof. This can be easily obtained from the above
2 3
analysis. #
¼ ST ðCauoff  Cdon ÞBd d  kSk
" #
W~T _ Remark 1. It can be seen that compensator equation
þ tr 1 ~ T T
ðW1  1 ðCauoff EÞ Sxauoff Þ
1 (21) involves a discontinuous term and the phenome-
" #
W~T _ non of chattering will appear. To eliminate the chatter-
þ tr 2 ~ T T
ðW2  2 ðCauoff EÞ Sxdon Þ
2 ing behavior, the compensator equation (21) is
" # modified to be
W~T _
þ tr 3 ~ T T
ðW3  3 ðCauoff EÞ Sron Þ
3 v ¼ W1 xauoff  W2 xdon  W3 ron þ Ke e  roff
S
 ðCauoff EÞ1 ð24Þ
_ 5 0, the updating laws are selected as
To make VðtÞ kSk þ
_ 1 ðtÞ ¼ 1 ðCau EÞT SxT ðtÞ where > 0 is a sufficiently small design constant.
W off auoff Therefore, the situation can be remedied by smoothing
_ T
W2 ðtÞ ¼ 2 ðCauoff EÞ SxT ðtÞ out the control discontinuities in a small boundary
don
_ 3 ðtÞ ¼ 3 ðCauoff EÞT SrT ðtÞ layer neighboring the sliding surface.
W on ð23Þ
Hence, the following can be concluded
Simulation and comparison results
_ ¼ ST ðCauoff  Cdon ÞBd d  kSk
VðtÞ In this section, to test the effectiveness of the multi-
4 kSkkðCauoff  Cdon ÞkkBd kD  kSk objective adaptive sliding-mode bumpless transfer
¼ kSk ½  kðCauoff  Cdon ÞkkBd kD method, a ducted rocket model with time-varying coef-
ficients is considered for numerical simulations. The
Define a constant  which always exists, satisfying actuator, i.e. the gas flow regulating system, is given by
 4 kðCauoff  Cdon ÞkkBd kD þ
0:87
p~ v ðsÞ ¼ ~
qðsÞ
where is a positive constant. Then, we can obtain 0:1s þ 1
_ 5  kSk 0:16
VðtÞ p~ r ðsÞ ¼ p~ ðsÞ
0:0000033s2 þ 0:013s þ 0:16 v
Using Lyapunov’s direct method, since V(t) is positive- w~ g ðsÞ ¼ ð1  0:15sÞp~ r ðsÞ
_ negative definite implies that S, W
definite, VðtÞ ~ 1, W
~ 2,
and W ~ 3 converge to zero. Then, W1, W2, and W3 will The engine simulation models used (Table 1) operate at a
converge to their true values. Moreover, lim SðtÞ ¼ 0 fixed height H ¼ 5 km and range in Mach number from 2
t!1
means that S(t) tends to zero as the time t tends to to 4 with a ¼ 0.00001. The linear model at the operating
infinity. Also, all trajectories starting off the sliding sur- point H ¼ 5 km, Ma ¼ 2 is selected as the nominal model
face S ¼ 0 must reach it asymptotically and then will for the design of controllers and compensators. Then, the
remain on this surface. The system’s behavior once on nominal augmented system (6) is obtained (the cubic coef-
the sliding surface is usually called sliding mode. When ficient of gas regulation system and the quadric coefficient

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

318 Proc IMechE Part G: J Aerospace Engineering 227(2)

of ducted rocket model are very small and are omitted Consider the regulating and protecting switching control
from the calculations) as follows between the subloops of thrust control and inlet buzz
2 3 protecting control (Figure 5). After the pre-adjustment
61:6900 16:0210 106:8100
6 7 for the single-loop control parameters, the parameters of
A¼4 0 22:2510 122:7700 5 two controllers can be obtained as: C1 (thrust controller):
0 1 0 Kp1 ¼ 0.01 and Ki1 ¼ 1.1, and C2 (inlet buzz protection
2 3 controller): Kp2 ¼ 0.01 and Ki2 ¼ 9, respectively.
0
6 7   Moreover, the bumpless transfer compensators are
B ¼ 4 1 5, C ¼ 14:3220 5:2524 35:0160
given in the form of equation (24). To avoid confusion,
0 subscripts 1 and 2 will be used below to distinguish
quantities that are C1 and C2, respectively. For example,
Table 1. Parameters of the ducted rocket model at different v1 is the compensator of C1 (C1 is off-line) and v2 the
operating points. compensator of C2 (C2 is off-line). For the sliding surfaces
(16) of vi (i ¼ 1,2), i and Kei are chosen as 1 ¼ 2 ¼ 1,
Ma K5 1 2 3 4
Ke1 ¼ Ke2 ¼ 100,
 and Cauoffi for each case are calculated

2 0.5600 0.0091 0.0039 0.0140 0.0022 as Cauoff1 ¼  0:1119 0:0500 0:3336 1:1000  and
3 0.4400 0.0055 0.0026 0.0100 0.0014 Cauoff2 ¼ 0:1119 0:0500 0:3336 9 , respec-
4 0.4000 0.0038 0.0019 0.0080 0.0009 tively. Besides, the sliding-mode gains of (24) for both
cases are chosen as 1 ¼ 2 ¼ 106, and the boundary

Figure 5. Scheme of switching controller of ramjet engine.

Figure 6. Comparison for the curve of inlet buzz margin with and without bumpless transfer.

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

Bao et al. 319

layers for both cases are set as 1 ¼ 2 ¼ 0.04. whose specifications are listed in Table 1. The warning
Furthermore, the adjustable parameters of adaptive line (  ) of inlet buzz margin is 0.18. Figures 6 and 7
rules (23) for each case are given by  11 ¼  12 ¼  13 ¼ 103 show the comparisons of inlet buzz margin and gas flow
and  21 ¼  22 ¼  23 ¼ 102, respectively. at both acceleration and deceleration processes of
The following simulations are conducted on the ducted rocket among three cases: single-loop thrust
ducted rocket model with time-varying coefficients control only, hard transfer, and bumpless transfer.

Figure 7. Comparison for the curve of gas flow with and without bumpless transfer.

Figure 8. Comparison of model matching error (u1  u2) with and without bumpless transfer.

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

320 Proc IMechE Part G: J Aerospace Engineering 227(2)

The acceleration process starts at 1 s, making thrust the control system goes on to satisfy the objective of
signal step arise and the deceleration process starts at thrust, the gas flow keeps increasing, which makes
5 s, making thrust signal step down. system cross and away from the warning line. For
For the case of thrust control only (the dashed lines double-loop cases, as thrust subcontrol loop gas flow
shown in Figures 6 and 7), in the acceleration process, demand increases, the trajectory of inlet buzz margin

Figure 9. Adaptation of the bumpless transfer compensator parameter W1.

Figure 10. Adaptation of the bumpless transfer compensator parameter W2.

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

Bao et al. 321

approaches the warning line, and once it crosses the and the protection controller is switched back to thrust
line, inlet buzz protection controller engages to control controller.
the engine. Furthermore, for verifying the bidirectional As illustrated in Figures 6 and 7, the solid lines are
switching performance, the authority is changed again continuous and moderately change due to use of the
as thrust subcontrol loop gas flow demand decreases, bumpless transfer. Also, the trajectory of inlet buzz

Figure 11. Adaptation of the bumpless transfer compensator parameter W3.

Figure 12. Comparison for the curve of inlet buzz margin with and without disturbance.

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

322 Proc IMechE Part G: J Aerospace Engineering 227(2)

Figure 13. Comparison for the curve of gas flow with and without disturbance.

Figure 14. Comparison of model matching error (u1  u2) with and without disturbance.

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

Bao et al. 323

margin has a small overshoot and a short settling time forcing the off-line controller output to track the on-
(Figure 6). On the other hand, the dotted lines are not line controller output. Also, the method is simple in
continuous and have many big jumps and oscillations terms of design and implementation and is robust to
without bumpless transfer. Since the gas flow flips disturbances. Finally, the switching strategy has been
from one end to the other, the inlet buzz margin is sig- applied to a ducted rocket system. Simulation results
nificantly deteriorated and far from the warning line. have been presented to verify the effectiveness of the
The comparison of output matching error between the proposed designs.
two subcontrol loops is shown in Figure 8. As can be
seen, the solid line by the proposed bumpless transfer Funding
gives a very small error magnitude (106 to 105) This study was supported by the National Natural Science
which guarantees ‘smooth’ transitions. Furthermore, as Foundation of China (nos. 90716012 and 90816028), the
one side of the bumpless transfer scheme with symmet- National Science Fund for Distinguished Young Scholars of
rical structure, we give the parameter adaptations of the China (no. 50925625), and the China Aerospace Support
bumpless transfer compensator v2, as shown in Figures 9 Foundation (no. 2008-HT-HGD-03).
to 11, which demonstrate that the estimates of compen-
sator parameters converge to their true values. Acknowledgment
Figures 12 and 13 show the comparisons about The authors gratefully acknowledge the helpful comments
curves of inlet buzz margin and gas flow with and and suggestions of the reviewers, which have improved the
without disturbance at both acceleration and deceler- presentation.
ation processes of ducted rocket. The time-varying
measurement disturbances of sinusoidal signals with References
different frequencies (i.e. 0.005sin(wdt), wd ¼ 5, 10) 1. Wilson R, Limage C and Hewitt P. The evolution of ramjet
and random noise signal (a random number belongs missile propulsion in the U.S. and where we are headed.
to [0.005, 0.005]) are, respectively, imposed. Each dis- In: Proceedings of the 32nd ASME/SAE/ASEE joint pro-
turbance signal is added to the backpressure of inlet pulsion conference and exhibit, Lake Buena Vista, FL, 1–3
at 1 s. Along with a increase of disturbance in the July 1996, AIAA paper no. 1996-3148.
2. Limage CR and Gainesville V. Solid fuel ducted
backpressure of inlet, settling time and peak value of
rockets for ramjet/scramjet missile applications. In: Pro-
inlet buzz control slightly increase, and there is ceedings of the 32nd ASME/SAE/ASEE joint propulsion
almost no change in the magnitude of output matching conference and exhibit, Lake Buena Vista, FL, 1–3 July
error (Figure 14). In summary, the results presented 1996, AIAA paper no. 1996-2916.
have shown that the robust multi-objective bumpless 3. Ronald SF. A century of ramjet propulsion technology
transfer design approach leads to satisfactory perfor- evolution. J Propul Power 2004; 20(1): 27–58.
mance for a ducted rocket engine which requires safe 4. Besser HL. History of ducted rocket development at
and fast switching during some special working Bayern-Chemie. In: Proceedings of the 44th AIAA/
conditions. ASME/ SAE/ ASEE joint propulsion conference and exhi-
bit, Hartford, CT, 21–23 July 2008, AIAA paper no.
2008-5261.
Conclusions 5. Watts SR and Garg S. A comparison of multivariable
control design techniques for a turbofan engine control.
The design of a switched control architecture for a In: Proceedings of the 40th gas turbine and aeroengine
class of ducted rocket has been described and simu- congress and exposition, Houston, Texas, 5–8 June 1995,
lated. The control design for the ducted rocket is ASME paper no. 95-GT-258.
highly challenging because of the complex model 6. Balas GJ. Linear, parameter-varying control and its
along with some limitations (e.g. temperature, air pres- application to a turbofan engine. Int J Robust Nonlinear
sure, physical acceleration). The mathematical models Control 2002; 12(9): 763–796.
of gas flow regulating system, ducted rocket, and inlet 7. Liberzon D, Hespanha HJP and Morse AS. Stability of
buzz margin have been introduced and analyzed, switched systems: a Lie-algebraic condition. Syst Control
respectively. In order to obtain optimal thrust control Lett 1999; 37(3): 117–122.
in the working process of ducted rocket while avoiding 8. Zhao J and Dimirovski GM. Quadratic stability of a class
of switched nonlinear systems. IEEE Trans Autom
phenomenon like inlet buzz, the problem of multi-
Control 2004; 49(4): 574–578.
objective switching control has been formulated. 9. Zhao J and Hill DJ. Passivity and stability of switched
Furthermore, regarding the bumpy phenomenon systems: a multiple storage function method. Syst Control
during controller switchings, an adaptive sliding- Lett 2008; 57(2): 158–164.
mode bumpless transfer compensator design has 10. Lin H and Antsaklis PJ. Stability and stabilizability of
been presented. It is shown that the designed compen- switched linear systems: a survey of recent results. IEEE
sator guarantees the ‘continuity’ of plant input by Trans Autom Control 2009; 54(2): 308–322.

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

324 Proc IMechE Part G: J Aerospace Engineering 227(2)

11. Hanus R, Kinnaert M and Henrotte JL. Conditioning

technique, a general anti-windup and bumpless transfer Appendix
method. Automatica 1987; 23(6): 729–739.
12. Graebe SF and Ahlen ALB. Dynamic transfer among
Notation
alternative controllers and its relation to antiwindup con-
A constant system matrix
troller design. IEEE Trans Control Syst Technol 1996;
4(1): 92–99.
Ar force area of gas
13. Edwards C and Postlethwaite I. Anti-windup and Av area of valve head chamber
bumpless-transfer schemes. Automatica 1998; 34(2): B constant control matrix
199–210. Ci controller label
14. Turner MC and Walker DJ. Linear quadratic bumpless Ci constant output matrix
transfer. Automatica 2000; 36(8): 1089–1101. d disturbance signal
15. Zheng K and Bentsman J. Decentralized compensation of D constant vector
controller uncertainty in the steady-state bumpless trans- e model matching error
fer under the state/output feedback. Int J Robust F parameter at point F(M0F, bF)
Nonlinear Control 2009; 19(10): 1083–1104. H altitude
16. Kowalski KK, Achterberg J, Van Lierop CMM, et al.
I identity matrix
Implementation of a linear quadratic bumpless transfer
Ke constant matrix
method to a magnetically levitated planar actuator
with moving magnets. In: Proceedings of the IEEE Ki gain of the transfer function
Int conference on control applications, Yokohama, KIi integral gain matrix
Japan, 8–10 September 2010, pp.2203–2208. KPi proportional gain matrix
17. Smirnov A, Tolsa K and Jastrzebski RP. Implementation M mass of valve head
of a bumpless switch in axial magnetic bearings. M0 Mach number of the free stream
In: Proceedings of the international symposium on Ma Mach number
industrial embedded systems, Trento, Italy, 7–9 July pb backpressure of inlet
2010, pp.63–68. pc pressure in the channel of the control valve
18. Zaccarian L and Teel AR. The L2 bumpless transfer pr pressure in the gas generator chamber
problem for linear plants: its definition and solution.
pv pressure in the valve head chamber
Automatica 2005; 41(7): 1718–1723.
q drive current of servo valve
19. Stoustrup J and Komareji M. A parameterization of
observer-based controllers: bumpless transfer by covari- ri reference signal
ance interpolation. In: Proceedings of the American con- S sliding mode variable
trol conference, St. Louis, MO, 10–12 June 2009, S auoff off-line augmented system
pp.1871–1875. S don online closed-loop system
20. Bao W, Li B, Chang JT, et al. Switching control of thrust Ti time constant of the transfer function
regulation and inlet buzz protection for ducted rocket. tr trace of a matrix
Acta Astronaut 2010; 67(7–8): 764–773. u, ui control input
21. Arehart AB and Wolovich WA. Bumpless switching con- uoff output of the off-line controller
trollers. In: Proceedings of the 35th IEEE conference on uon output of the online controller
decision and control, Kobe, Japan, 11–13 December 1996,
v output of the compensator
pp.1654–1655.
22. Bao W, Qi YW, Yu DR, et al. Bumpless switching
veq equivalent control
scheme design and its application to hypersonic vehicle V Lyapunov function candidate
model. Int J Innovative Comput Inf Control 2012; 8(1B): Vr free volume of the gas generator
677–689. Vv volume of the valve head chamber
23. Shi XX, Chang JT, Bao W, et al. Supersonic inlet buzz wg gas flow rate
margin control of ducted rocket. Proc IMechE, Part G: J Wi estimates of compensator parameters
Aerospace Engineering 2010; 224(10): 1131–1139. Wi true compensator parameters
24. Niu WY, Bao W, Cui T, et al. Dynamic modeling x system state
and model reduction order of controllable flow solid xauoff state of the off-line augmented system
ducted rockets. J. Solid Rocket Technol 2008; 31(4): xci controller state
325–330.
xdon state of the online closed-loop system
25. Chang JT, Yu DR, Bao W, et al. Dimensionless analysis
yi controlled output
of the unstart boundary for 2-D mixed hypersonic inlets.
Aeronaut J 2008; 112(1135): 547–555.
26. Sam YM, Osman JHS, et al. A class of proportional- a delay time
integral sliding mode control with application to active b angle of attack of the free stream
suspension system. Syst Control Lett 2004; 51(3–4): i gain of the updating law
217–223. positive constant

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014

Bao et al. 325

 deviation form an equilibrium state s switching signal

Z inlet buzz margin i time constant of the transfer function
 warning line of inlet buzz margin
 constant matrix
positive constant Subscripts
x damping factor  Laplace transformation
pb current pressure ratio of supersonic inlet i a natural number
pu maximum pressure ratio of supersonic inlet
 sliding mode gain

Downloaded from pig.sagepub.com at Tsinghua University on December 26, 2014