You are on page 1of 5

dft

au-u
- -t l
SuPmkCdf*Aml TA4 - 10:15
H' Control of Gasturbine Engines for Helicopters
Ryo Watanabe * Masahir o Kur(-osaki t Takeshi Kusakawa t
Kenko Uchida § Masayuki Fujita I
February 22, 1993

Abstract
Engine tests were conducted in order to demonstrate efectiveness and pracability
of HI control theory to helkopter ingines. An engine control algorithm is desiged by
H control theory, whic meets robut stability requirements ad nominal performance
requirements described in frequency domain. Simulation ad engine test reslts show that
HI controllr with Leedioward element have better contro perlomasce than PI controler
witk ieedbward elment, especialy, less power turbine speed oversoots are oberved in rapid
torque decreasing tests.

1 Introduction bine. The compressor and the HP turbine are connected by a


shaft. Airflow is continuously compressed in the compressor
Gasturbine engines for helicopter are subject to significant and burned in the burner. Resulting high pressure and high
torque disturbance due to pilot's manupiration of collective temperature gas drives the HP turbine and then the power
pitch angle. Engine dynamics significantly vary with engine turbine which is connected to a helicopter rotor via a speed
operitiong conditions. In order to cope with the disturbance reduction gear. A pilot controls the rotor lift by manipulating
and the dynamics change, conventional controls use multi- the rotor blade pitch angle.
ple gain and gain-scheduling technique, respectively. A lot
of imulations, engine tests, and flight tests are necessary to Sn.!.~~~~ktc

validate the control laws. The theory which guarantee spec-


ified stability and performance under specified uncertainty Air~~~~~~~~u
and disturbance is expected to greatly reduce design and val-
idation efforts of the control laws. In this paper, control
laws are designed using H,0 control theory in order to evalu- TrbohatEngine
Fig1
ate effectiveness and practicability of the theory to helicopter 3.1 Control Ojci;ctson
engine control. IcRr es u t er
olbua e t atde
lLx-u to
This paper consists of the followings. At first, brief de-
scription of gasturbine engines for helicopter and control ob-
jectives are gven. Then requirements for closed loop transfer Fig.1 Trurbosha£t Enginle
functions and engine model are given. Engine model is iden-
tified uing least square method. H,,. controller is designed
using Glover-Doyle two Ricatti Algorism. Fianlly, engine test 3 Control of Gasturbine Engines for
results are presented.
The controller synthsis performed using Pro-Matlab and
Hel'icopters
the Matlab Robust-Control Tool Box software. These appli- 3.1 Control Objectives
ication packages were run on Vax 2000 workstation.
Rotor shaft torque changes significantly due to pilot mna-
nipulation of rotor blade pitch angle during flight. It is the
2 Gasturbine Engines for Helicopters most important objective of helicopter engine controls to keep
power turbine speed(NF) constant at a reference value under
Fig. 1 shows typical turboshaft engine, which consists of
a
the rotor shaft torque disturbance in order to maintain pro-
a compressor, a combustor, a HP turbine, and a power tur-
portional relationship between the rotor lift and the pitch an-
*Control Systems Etgineering Department, Aero-Engine and Space gle and also to prevent hazardous power turbine overspeed.
Operations, IkiklwajimfrHaima Heavy hdustries Co.,Ltd., 229 Tono- External signals available for NF control are power tur-
gap., M ko-Macki, Niskitamagun, Tokyo 190-12, Japan bine speed reference(NFREF), power turbine speed(NF) and
tControl Systems Engineering Department, Aero-Engine and Space
Operations, Iskihwajizna-Harima Heavy Iadustries Co.,Ltd., 229 Tono- pitch angle reference which is equivalent to shaft torque refer-
gays., Misiho-Machi, Nishitamagun, Tokyo 190-12, Japan ence(TQREF). Output from the NF control is fuel flow refer-
tEagieeriug Department, Aero-Eugine and Space Operations, ence signal(WFREF) to Fuel Metering Unit(Actuator 2). A
IshikawaJima-Harima Heavy Industris Co.,Ltd., 229 Tonogaya, block diagram of NF control system is shown in Fig.2.
Misuko-Macki, Nisitamagun, Tokyo 190-12, Japan
IDepartment of Electricai Engineering, Waseda Uniiversity, 3-4-1
Okubo, Skiajuku, Tokyo 169, Japan 3.2 Control Law
ISckool of Information Science, Japan Advanced Institute of Science
and Technology, Hokurink, Tatsunokuchi, IshLikawa 923-12, Japan Typical control law for helicopter engines consists of start
1123
the following anti-windup and bumpless transer tehnue
are used, which is a generalization of that by Haus. [i)
Torque
f
+~~~~~~~~~~
Nt e = e-G(ut-u')
Reference WFREF WFr
ACT2 Cvi .Nf d
NF + I~

-ax = Acx0+B,e'
Reterence Engine
Controller
U = Cze + Dce
Here, u means output of the controller and u' means input
Fig.2 NF Contrd Sysem to the plant selected finally through the nonlinearity. Block
control, acceleration and deceleration control, steady state diagram of anti-windup and bumples transfer technique is
shown in Fig.4.
speed control and selection logic of them The start con-
trol is to start engine and tablish idling state, preveting
from so caed hot start and hang start. The compressor
speed(NG) is controlled in this modte The accel/decel con-
trol is to accelerate and decelerate the compresor speed as
fast as possible. An upper limit(WFACC) of WFREF and
a lower limit(WFDEC) of WFREF are set as a function of
the compressor speed in order to prevent from compressor
stall/engine surge and flame out, respectively. The steady Fig.4 Anti-Windup and Bumpes Transfer Technqu
state speed control is to maintain power turbine speed con-
stant at NFREF. When the power turbine torque(NF) is
same as the rotor shaft torque, the power turbine speed is 4 Control Requirements
constant. Once the rotor shaft torque is varied by changing
the pitch angle, the power turbine torque must be adjusted by 4.1 Robust Stability Requirements
manipulating fuel flow to the combustor as quickly a posible Control system must be stable for not only nominal model
to prevent from power turbine speed fluctuation. Switching but also "nominal model" + "uncertainty model". Since a
of each control law is done by LVS(lower value selection) -and loop from the rotor shaft torque reference to NI is an open
HVS (higher value selection) to avoid sudden change of fuel loop, uncertainty in this loop does not have any efect on sys-
flow, which are nonlinear elements. Block diagram of control tem stability. Uncertainty in a feedback loop from NFREF
law is shown in Fig.3. to NF must be considered. Hence, requirement for robust
stability is given as follows. Control system hall be stable
for multiplicative uncertainty AW,1w1,pf. of tranfer func-
tion (G,,,aci2) from fuel reference(WFREF) to power turbine
speed(NF). this requirement can be expressed as follows.
< 1, T
G
ad2K.
1+GG,,,a42K,.
Where T is the complementary sensitivity function.

4.22 Nominal Performance Requirements


Main requirement for control objective, i.e., to maintain
Fig.3 Aircraft Engine Contrd Logc the power turbine speed -constant at reference value under
the significant rotor torque disturbance, is given as"lows.
Low frequency gain of the doed loop.transer function from
3.3 Anti-Windup and Bumpless Transfer TQREF to NF , Ggfptvqf, shall be small. This can be ea-
Technique pressed as follows.
The fuel flow referece to the fuel metering unit might be
different from the output of the steady state speed controller,
IIGn elqrefWnWtltwejoo < 1

since there are many noninear elements between them. Gen- Where W,fj.f is a frequency weighting function.
erally, when an output of a compesator with an integrator Requirement for sensor noise is given as lows. igh
is different from an input to a plant due to nonlinear el- frequency gain of the dosd loop transfer fuction from the
ements between them, so called integrator windup occurs, sensor noiseto NF shall be small. This can be expresed as
which often causs unexpected and unsatisfactory behavior follows.
of the control system. [1] [2]
The steady state speed controller must have an integrator IITW4I{ < l
to meet the steady state performance specification described
in Sec.4. For the folowing controler, Where W4 is a weighting fnction.
Requirement for offset is given as follo. Low frequency
d gain of the closed loop transfer function from NFREF to
dt
= A,T, + Bg, NFERR shall be small, which can be expressed as flows.
u = Cz. + De,
1124
I _
Augmented Plant

1-m

in
m.

AuT2Ii
_ ~ ~ w3 _2 w2
O

Control

Fig-5 Aupe Lte Plant

IISWerIIo< 1, S= i- ,iK < 1

Where S is a sensitivity function and W, is a weighUng An augmnented plant shown in Fig.5 is used in order to
function. design the controller which meets the above requirements. In
Requirement for fuel valve response is given as follows. the augmented plant P, zl/wIl,zl/w2 and zl/w3 represent
High frequency gains of the closed loop transfer functions the closed loop transfer function from the shaft torque to
from TQREF, NFREF and sensor noise to WF shall be small, the power turbine speed, G,qf'erf, the sensitivity function,
which can be expresed as follows. S, and the complementary sensitivity function, T, respec-
tively. Also z2/wl, z2/w2 and z2/w3 represent fuel valve
I G,flw,f W,1f tr,,fII,f < 1, rePOnse GwJ/tqef i Gwf/afro) and Of/j~j, respectively. Here
WI, W2, W3 and W6 are fr-equency weightings for the corre-
IlGlDf nf,ef Wwf nfref Iloo < 1, sponding transfer functions.
JjGw,f1WdWyjfIj1j
C
Where Gwf/tqef, Gwf/afref and Gwf/d are closed loop trans- 5 Desilgn Model
fer functions from TQREF, NFREF and sensor noise to NF,
respectively. W,,trq W,y,Wq1fq and W,1f4I are weighting The reference value of the power turbine speed(NFREF)
functions corresponding to the transfer functions, respectively. was selected to 49%. The desin model of the engine and
the actuators were derived from data obtained by system
4.3 Augmented Plant identification experiments at the folowing operation points:
1)NF=49%, NG=73%, TQ=134.2Nm and 2)NF=49%, NG=86%,
The requirements for robust stability and requirements for TQ=5O5.7Nme Here, 100% speed of NF and NG are 36,470
nominal performance given in the previous section are sum- and 46,499 rpm, respectively. The engine model at NG =
marised here. The requirements for transfer function from 73% was considered as a nominal model and a multiplicative
shaft torque reference to power turbine speed , Gf,t, is uncertainty was computed from the difference between both
models.
Wqe4jf
IIGn,j1tqrWnf C<
1
The requirement for the sensitivity function of the feedbck 5.1 Actuator Model
loop, S, is
Actuators of the system use for the engine tests are a shaft
ItSWr111oo < 1. torque generator(actI) and a fuel valve(act2). A dynamo was
The requiremnents for complimentary sensitivity function sed as the shaft torque generator in order to smulate the
of the feedback loop, T are helicopter rotor, which is a common practice in industries.
The identified models are as follows:
ltTAWnfiwf,jriIIo < 1, For NG=73%
IVTWAII. < 1. Actl =t 4.6382
+ 4.6382'
s
The requirements for fuel valve response are Actuator ~~7.2580
=t s +5.8484ex x 102
for t0 s+7.2580 x 102
ad 2
=

(Iwf tef WfwjjqjIr, c 1,


jGwf1nfrefWwflafrrefIIcc < 1,
1125
For NG=86% The uncertainty from the fuel flow reference(WFREF) to
3.2146 the power turbine speed(NF) and the frequency weighting for
actlI =
s + 3.2146' robust stability(W3) are shown in Fig.7.
7.5085 x 102 Muhtiplicive Uncrtainty and W3:Gain
S2 + 6.8013 x 101s + 7.5085 x 102 SA0_

5.2 Engine Model t20


Mlilciv -Ue\anW3
To design controllers, it is necessary to have models of i
engine dynamics from the shaft torque to the power turbine
speed(NF), Gth, and from the fuel flow(WF) to the power .201 2
MuliplcWe Ureuh
turbine speed(NF), G.0. The identified models are as folows, 10 lo" I0' 10I 102

of which transfer functions at NG= 73% and 86% are shown


in Fig.6. Fig.7 Gain Plots of Multiplicative Uncertainty and W3
For NG=73%,

Gt,.
= -6.4384 x 10-2 6.2 HOO Controller
s2 + 4.1393s + 6.1577xxO-"
5.1670 x 10-2 A Hr controller with a feedforward term was designed
n 2 + 9.2090 x 10-Is + 7.4000 x 10-2 by Glover-Doyle two Riccati algorism. [3] The order of the
controller was reduced from 12 to 8 by internal balance re-
For NG=86%, alization method in order to save computing time. Also, a
PI controller with a feedforward term were designed for per-
-3.0395+ x 10-2 x 10-1
G,, = 52 + 1.751f formance comparison which have the same transfer function
2.5334 as the HI controller and the HI feedforward term at low
1.1637 x 10-1 frequency range, respectively. The feedforwaxd term of the
92+ 1.9476s + 2.7561 x 10-1 H3 controller(Kt) and the PI controller are shown in Fig 8.
Om: Gain
The Hr controller (K,,) and the PI controller are shown in
Fig.9. A digital controller (FADE() was used on the engine
tests.
Kt: Gain

OwnG: Gn

i.: PMc.
501

.50 uI

Fig.6 Gain Plots of the Plant -im''.,o4 I3 to'


l 14-
10 ~ Iil O l Id
i
l )4

6 Controller Synthesis Fig.8 Bode Plots of Feedforward Controlers

6.1 Frequency Weightings 7 Engine Tests


The frequency weightings to meet the robust stability re-
quirements and the nominal performance requirements were The engine tests were conducted at IHI small engine test
defined as follows. cell. The power turbine speed(NF) response for small (-60
Nm) torque step at NG=73% and 86% are shown in Fig.10
WI =-
s+2.4024x102
s
and Fig.11, respectively. NF response for larger (-400 Nm)
s+2.4024x 10"3 torque step at NG=86% is shown in Fig.12. Those figures
show that overspeed of NF with the H" controller are less
= 1.0000 x 10-ls + 7.5008 x 10-1 than those with the PI controller. NF and NG responses at
s + 7.5008 x 1Ch5
switching from the start control(NG control) to the steady
W3 _ 1.0000 X 102s + 8.9312 x 10' state control(NF control) are shown in Fig.13. Smooth trans-
s + 1.7820-x 102
3.3333s' + 3.3008 x 10's + 7.9279 x fer of the control mode and no wind up of the integrator was
observed.
s2 + 9.9240 x 104s + 2.3784 x 107
1128
Kw Gain 10.
3z,

49
0

0
-

Fig.10 NF Respo
2
51 t~ ~ ~ ~-'J rP
4
NFP

6
le.
no

a
4. Tw Sep: NG073%

10

to -60Nm Torque Step: NG =73 %


12
PiControl

14 16 18 0

NPReqintn-C. Taue Ssep: ?6-S%


cn c.5--_, -__----,--

50 Ce
Pi Co"

10 1 JO-2 -t id
fhew=Ns)
--
~49.5
z
4g1 H-hiCHOlO
Fig.9 Bode Plots of Feedbak Controllers
A2 IK'
0 2 4 6 8 10 12 14 16 is 21
Ths)
8 Conclusions
The augmented plant was constructed to design the H' Fig.11 NF Respo to -60Nm Torque Step: NG =86 %
controller which meets practical control specifications required
for helicopter engine controls. The HW control logic with re- NF R_pmm m -40014 Tempe Swp: NG46% w 73%
duced order was installed in the digital controller(FADEC). 59,
The engine test results show that H' control has better per- 56
* ~~~~Pt
cofoai
formance than conventional P1 control in reducing the power
turbine speed (NF) deviation from the reference value under AK
the shaft torque disturbance. The anti-windup and bumpless t52
S0
transfer technique provides smooth transfer of the control
modes. The authors wish to thank Mr. Hiroyuld Miyag and '..
0 2 4 6 3 10 12 14 16 I1 20
Mr. Hiroaki Asahi of IHI for their support for the engine T_N
tests and Mr. Masahiro Egami of IST for his programming Fig.12 NF Responses to 40ONm Torque Step: NG =86 %
of control software.
NO - mud NO Rain=
References

I
9
[1] R.Hanus, M.Kinnaert, J.L.Henrotte : Conditioning
Technique, a General Anti-Windup and Bumpless asN
Transfer Method, Automatica, 23-6 729/739 (1987) 37 'I
Pedlmc

[2] P.J. Campo, M.Morari, C.N. Nett: Multivariable Anti-


Windup and Bumpless Transfer: A General Theory, In 0 tO 20 30
trams)
40 50 60 x
Proc. American Control Conference, pages 1706/1711 NF RespoI d NFRefd5m
(1990) 50--
^~~~~~~~~~~~~~~~ .1-4 VW ..
.. . ~_

[3] K.Glover, J.C.Doyle. State-Space Formulae for Stabiliz- '43


NF Rdetnaci
ing Controllers that Satisfy a Hw Norm Bounds and Re- NPPFeedlk
lations to Risk Sensitivity Systems and Control Letters,
11-2, 162/172 (1988) 44

A7
0
--
o10 20 30 40 50 60 70
twt($s)
Fig.13 Switching from NG Contrd to NF Control

1127

You might also like