~ ~

~·. """
.a t '¾
'
•
.I •
'
Urut-.(,
"
:'
1 :.,t

l,
• },

- ' CD
H~b'YJcl fu ~.:ct/ Co l'lhou ru.: / ya b<::il •
c.

~
( I
j 1' ~d}ti ortiJ 'YY1ili~d.s 1 \f t'U:l oiu t&hrJ r rn f o!. €_ cl hrl •
: Con:ho1 !if- l"nc:lw+,,la.1 "nlbots and .·. • ruJo rn oJ-f c. @ _'Ve.hid u,
(I w hkh Ci fualu h, vrul ou.t ~n v/''Ytln man1 ~ and • p~fmm st Mfla

I 'Yefe.tJ'+,'\t_ +cuRA +tio.t ...-r'Te end :ff:,dun. pa-d+I o nl~nd

• ~oti on alo½ f1'~d foltu) •• • /' - ~ .•
( HOt,Jeru @ •'.
I o~e.u:d1' oru 1ri u n1ti'udtud ern\-ni n l'YUJ'\ts 'Yer" '"re.

".YD\io~ +o purrm' • 'n'lo-n enrf)plo.l( ~ . Unde.i suU\ wnd\tlo,-,.,

r-,c. d .tocJ k £':.ti N"Y\)l • \t 1
a i-1 oJ:fu.tr)oJ-1-re fhal- con be suc(e~ ~1~
~~d o-n -re.o,I ,_-h'm1. l-Omp6x .. sf're.nu),

( 1ke O.JJ -thore..o sq+ c.o ropiJ4 ~ h~ I:,"') d fu U"j pru ctdJJ ~
~. a.ufu!Ylolw •\~ru'd ' 1;¥1 --roboH~ sr}emi.
,: , n J- , • • , ., >: ;... ""' J .. 1 :.,
•r...- - ,_,..., _u +-w•-~·
u) n7~, s c.t-e..n;e. _-e.onc.~ 'ft.:rYctre.J:·_· 0. °'.1 ~ '·. ,~~ ,oy ~ i'\J
9 J '1 n0 ~

"r)e.wo! 1 n ~ . {l..!Ns) ¾ ad°'ff • a fvi:~. ~k co~t

( ~ l C) IY\ l'Y1 QV\J pu1oltrv ~ntm1 ' ' f0-1 k 5 ' ''

de ye.lop~ a_<tmo-h.:~ . · h1\uruJJc_j • f"ut-ci

s+.uduu +av ~,dbU-. tt1anl~~ · . • .·
. '
.
Q) Thl...-d sckn,e. •• ernplot cp m •ev-:olve_'l'\llu ts: ftjz:·cJ_
be.havim;, _±i ~ .fe. We.d ln !Y> Ci b\ le. '"Y'Obot LO ntni0 •

( G) ~~:~2:~ s~.~ • • ·
Ne.waA 'Y1eluio<~ . hru 11--t ah{~ +o~ ,le.run . po±t-W'\!. ~
. S~c _ o-Y cl.a>''°' YYU C
1
so't-eXY\s. ~Y) ,-0,1, °:f p<1la.c.h J .14..
;erul"iJ"i and po..i+E.t n ;oy-e.ro<}nl-H oY) of- NtJ o.u ~ Lei \te-d
I'Yi • +wn ..s+, V. : .
 Hri ct1't,nol\+tu / C\ t))u.li1f-mte~d'o o_WJ+-e_dw.t
1'1. .ne_e_d.d · -IN~-~
CCll'\ l:,e_ don~ b'.l T~' \· .C 320 00 d.t81t,:J. s r ~(<?~~
0
-y) -
. . j_ . ... . . ·. .
. .

• I

1 • I

I
I1
l
I I .

I
I

. .
··, •
534 Applications of Soft

the prominent robot control issues, namely, control of direct dnve robot motors, control of flexibl .
. . ---¾
intelligent navigation of mobile robots. This in near future allows us to combine soft compu. e links and
ting Par d'
for more intelligent and robust control. a igrns

17.6 Soft Computing Based Rocket Engine Control··

Many of the rocket engine programs initiated by NASA's Marshall Space Flight Center (MSFC) 1• H
Alabama, have been successful as evident by success of the Space Shuttl~ Main Engine, ground ~ttntsvillc,
former X-33 engine and Fastrac X-34 engine for the reusable launch vehicle program. As a result dang of the
c h. h • . ' a tabasc of •
test cases and lessons learned has been created rrom w 1c improvements to engme control for~
. h d ruturc en •
programs can be made. Such cases include premature engines ut owns, propellant leaks' and numcrous ginc
.
of anomalous sensors and data. Such cases are not only costly to the American taxpayer, but als cases
ris~n social acceptance of current and future space programs. ..,,..---.- J'- ~ Y~~t'i
{Jhe Space Transportation Directorate at MSFC has continuall expressed n intereJtA tmpr~·:~
control and many efforts in various areas f?r control and ano aly det . tion and :miti~ation ha~c
undertaken. Some successful attempts have included n ~ _el me :nal sis and engme vibration anal .
t:
Other efforts, although su<-:cessful in theory and simulation, have been partially successful in actual cnr)n:•s.
firings)t is the harsh engine environment of cryogenics, vibrations, real-time control demands and diffcr:cst
engine configurations from test to test that continually encourage researchers to determine alternative solutio:
or improvements to approaches for engine control and anomaly detection and mitigation.
C~rrent co~trol technolog~es depend on pro~en, _sometimes archaic, hardware and logical ~rogramming
techniques which are costly to implement and maintain, and do not account for unforeseen conditions leading
to the kinds of problems referenced earlier. The principle goal is to provide another a':enue to address MSFC's
Space Transportation Directora~e's interest in improving overall engine control.~ approach for investigating
and demonstrating how the application of soft computing technologies can further address presented control
issues in rocket engine control is presented in this section as a case study. The testbed engine is shown in
Figure I 7-21.

,. '.t}
,. ' ~.. .
·r•:~'::1'•..,1•'!.l'..''"·t:M:;, ... '/1"~,,rr:,'.~·,:;'I~ ·1":· •.;· .,+
1/, .-~~-:,_,.,~ ..

'•
~,,t•"
f;.,

-~~·:.:, '

,.If" "'

\
tt ·: .u o i:o m m L (_;~
I 'I L~....,,..,_..__,,.__._.._ •• - ... •··~ _ •r ,. •· •••-·-""'_...,__,.._...,. . , . . . , _ . , ~ 4 ~ . : , ) . , I IC --

. Figure 17·21 Turbine technologies SR-30 turbojet engine.

 _____________________
._..,... ,,,..wm --=i::. ar,:;"i .u_.--- --------

fng Based
Rocket Engine Control
5Sot tCom p~u~ 1~:= :.:=- ----- ----- ----- ----- ----- ----- 535 .
----- -
~
. articular work, automation and control of a small-scale turbojet engine is described
and some
I~ ~his p data obtained using a PID controller has been provided. Turbine technologies
turbojet engine
prehminary
. d with instrumentation • cror momtor• mg• t he operatin
• g cond'mons
• of the engme. • Some pre1·1mmary •
is equ~~~~ned to demonstrate the _safety ?f the_ engine unde_r expecte~ hazardous operatin
g con~itions and
data onstrate the applicability of one-d1mens1onal propulsion equations to calculate the thrust
mduced by
to dem ine are shown. Additional data obtained to determine the system transfer function
to design a PID
tbe engll rare also shown. The PID control algorithm design has been outlined.
cootro e
The instrumentation includes several thermocouples a~d pressure transducers and a load
cell to m~ure
the thrust generated by the engine. A valve controls the fuel-flow rate. In the present work two
separate control
, approaches were used. First the difference between the desired thrust from the engine and the thrust measured
using the load cell was used as the feedback signal to control the fuel-flow rate to the engine.
In the second
a proach the temperature and pressure sensor data were used to calculate the thrust produce
d by the engine
~ing the aero-thermodynamic equations applying to turbojet engine operations, and the differen
u • ce berween
the calculated thrust and the desired thrust was used as the feedback signal. In the present approach, turbojet
engine's operation will be automated and several control logics will be trimmed to show their capabili
ties. In this
hardware-in-the-loop control demonstration effort, first a simple PID control algorithm is demons
trated. The
testbed development and some preliminary results obtained are presented using the experim
ental apparatus.
Simply stated, the term "soft computing" here refers to computational mechanisms that
can determine
suitable relationships (in a system data set) to assess and determine a quantitative opinion(s)
based on future
conditions. Within MSFC, such computational mechanisms are viewed as a collection of
algorithms that
,can achieve optimal or near-optimal results in the presence of imprecise data, uncertainty,
unknown physics
and probabilistic outcomes. Such algorithms include automated reasoning, nondeterministi
c or probabilistic
methods. Examples of the latter include Bayesian networks, statistical resampling techniques,
chaos theory and
pans oflearning theory: Other well-known soft computing t~chnologies include fuzzy logic, neural
networks
and genetic algorithms. The term soft computing is used metaphorically to contrast with hard
computing.
Hard computing systems are based on those traditional approaches used commonly in most
event-driven
systems. Such approac~es are often viewed as crisp or binary. For example, in a propulsion
start p~ep~rations, ifliquid oxygen system for engine
tank temperature A < x and liquid oxygen bottom tank pressure A> y, then
open_ liquid oxygen engine supply valve can be opened. For this example, soft computing would
accommodate
aregion of acceptable temperature and pressure valves as well as observe other conditions such
as liquid level
and so on. A mechanism (e.g. NNs) for determining when to open the liquid oxygen engine
would be used. supply valve
th
e approach would differ in tha~ it would be tolerant of any imprecision and uncertainty.
one could In essence,
h . . view soft compuu•ng as bemg . . ·1
simi ar to t he way the human brain. works. Humans tend •
to we
keeunsuc
fi (object·
. ive) and su b'Jecuve
• knowl
edge before making decisions based on current states ofevents. The •
an~ eaturt m soft computing stems from addressing any inherent imprecision, unct:rtainty,
partial truths
effo ove~a l srstem knowledge. The central goal in soft computing is to attain more robwt response
. For this
seq n, t e primary technologies to be used are Bayesian belief networks and fuzzy logic. For the
engine start-up
to uence,
. the Bayes·ian be1·ie f networks wi·11 be used to ascertain the state of the engme
• prior• to procll;CU __ .J!!ng ·
nd stage controI• For mam
de rnain • stage engine
• control approach, fuzzy logic • b l d d • •
will e emp oye an it IS MU6...aY ,........., •
pe ent on the complexity of the engine control requirements and functions. ·
17•6,1
1l.· . '
Bayesian Belief Networks !\\~ ~~k .\d -lo
For the . • .
BBN )engine st~rt hase, the primary soft computing technology to be utilized is Bay .,•
5 th lief netWorks.
• e O e i,nten ~the BBN is to qualify each of the states during engine start-up • reaching main
~~
• I

~
 Applications of Soft Comput·ing
536

! ffx( u)

~ Figure 17-22 Bayesian belief network.

sta~ ~ addition
health of the engine and proceeding into main
stage. This will further assure certainty in the n proven
iding adde d assu rance into prev entin g any premature engine shut4owns. BBNs haveEi(e~
to prov events in environments
to be good predictive and diagnostic
mechanisms for reasoning about the state of
rooted in classic
details, the genealogy of this SCT is strongly
where uncertainty is universal. Suppressing the shows a Bayesian
subjectivist viewpoint is taken. Figure 17-22
statistical Bayesian inference theory where a
belief network. belief in
pretation of probability where one's degree of
In short, Bayesian inference uses a different inter declarative (prior
are computational architectures that permit
some event is part of the reasoning. BBNs d (factual),
al prob abilistic valu es) and subj ective opin ions (posterior probabilistic values) about worl
condition n and a unique
sment through a visual network representatio
knowledge to be part of the reasoning and asses
syntactic message-passing feature.
ated to update its parameters for belief upda
ting, it first inspects all
Beli efupda tin~ Whe n node Xis activ
t, it updates
and its children nodes (A). Then using all inpu
messages transmitted to it by its parents (Jr)
its belief.
mit to parent
mitted by Y and Z, compute message to trans
Bottom-up propagation: Using messages trans
node U.
s Yand Z.
s new rpessages to be sent to its children node
Top-down propagation: Node X then compute

1'§ .,~ 17 .6. 2 F~ Logic Control

it accommodates
logic. The use of fuzzy logic is suitable in that
For main stage control, the plan is to use fuzzy of mathematics that deals with
the un~rtainties ass_ociated with control during power. Fuzzy logic is a branch
apaox1mace reasoning. , probability
eley combines the topics of multivalued logic
Zadeh of the University of California at Berk r softw are as a medium.
theory and artificial intelligence for simulation of human thought by using compute
o? littfiftsic
to make decision based on vaguenes! or imprecisi
The technology of fuzzy logic en ab ~t er
l\(u r~ -\ I
<: ;. _ ✓°'j~ c >j✓~
i (.

(f' ' '

 --· ....
. Based Rocket Engine Control 537
•~
computin9
17,6 tv--4~\. '<"-~
. l systems. Fuzzy logic comprises s s and subsets where a set represents an input linguistic
. bsets represent the 1·mgu1st1c
. (llost Phys1ca • • values
,o bl and ns su
~ria e k rocess is followed usmg
. fu •c. • d d c. •c. • d .
zz1I1cat1on an eruzz1r1cat1on to etermme a suitable response
Aco~~,bo:n ~ny given system to be controlled. The centroid method (or center of gravity method) will
to cond1::n~zzificacion. The intent in employing fuzzy logic with the SR-30 engine is to utilize all engine
!,e used fi d conventional PIO data to design and develop the fuzzy logic based controller. •
rest data ~e soft computing techniques of BBNs and fuzzy lc,gic have been developed and tested and verified
o~ce)t both will be integrated into the SR-30 engine tested control environment for final integration and
(off,hne , .
•fication
1
tesnng. • •
ven
_ .3 Software Engineering in Marshall's Flight Software Group
17 6
oftware ~evelopment organization a set of processes and standards for their base produa line are
fu~cy S .
ically adhered too. Such processes and standards generally adhe_re to a type of software dev~lopment life
~le. For the Flight So~are Group the popular Waterfal~ Model 1s used. Furthermore, the Flight Software
6roup's process is ISO 9001 •certifie1, And more importantly, has recently been certified as a CMM Level 3
organization, a first for any NASA organization. CMM is the Capability Maturity Model for software that was
developed by Carnegie Melon's Software Engineering Institute and has beco~e an internationally recognized
srandard for evaluating software development processes where a level 5 is the highest certification a softwar_e
de¥elopment organization can achieve. ~he principle function of the Flight Software Group is to devel~p flight
critical software for embedded systems,· hence requiring all software development processes to be stringent
with software quality assurance functions underlying all activities of software development.
The Flight Software Group traditionally views software engineering as the establishment and uses sound
engineering processes to develop reliable software, based on human processes and thinking, that works on real
machines. Furthermore, software engineering is also viewed as the design and implementation of a set of user .
requirements into software using sound engineering processes. The emphasis here is that the FlightSofuvare
Group uses sound software development processes based on empirically prove? and sound pr~ces. ~ ~\ ,
~ . .. . r()~ ~,,1.~ ~\~.~~~ ~
17.6.4 Experimental Apparatus and Fac1litt Turbine Te9tinclog1es SR-30 Engine
\ ~.~"
ft computing techr:iologr. he«lware-in-rhe--loo~periments wFre coi;iducted using Turbine-Technologies
model S~-30 turbojet ~ngin} sho~n i~ Figure _17~21. th\~emonsfrati~t'fengine consists ofthe turbojet engine
m~uf~ctured by Turbine Technologies Ltd. m i t s ~ . § ~ . he enclosure includes a control panel
for e~~me o, eration and monitoring and a PC-based data a,c quisition u~t fo~ measuring the engine operating
tmons. c J..0{ . ~ . "' J\.. h. ~\ • _ .
futThe SR-30 _engine has a single-stagt radial flow compressofi.ith a maximum pressure ratio of PR= 3.1
glc-stage axial-flow turbine, and reverse-flow annular combustion chamber and ~. operates obeying the
!~yton ~ermodynamic cycle in the s~e fashion as the large turbojet engines.Q'he cJngine as produced by
Turbine Technologies includes piany pressure and tem~rature sensors, a load-cell for thrust measurc-
men~s, a c ~motor winding for readin the e inc r nd a fuel flow-rate measurement system to
monitor!measure the operating parameters of the engine. e engi~e generates 20 lbs of thrust at 90,000
rpm while ingesting m = 1.1 lb s- 1 of air. The engine has a length of 10.75 in. and the exit exhaust diameter
=
of Dexit 2.25 in. •
The engine available is instrumented with pressure transducers in the compressor inlet and exit,. in the
combustor, in the turbine exit and in the thrust nozzle exit, and Kcype thermocouples in the compressor inlet
and exit, in the turbine inlet and exit and in the thrust nozzle exit. The engine available was also equipped
 538 Applications of Soft Com .
~~------ --------- --------- -------__ __;::..:.. .:...:~~P Uting

with a National Instruments (NI) PCI 4351, AID board with 24 bit resolution for l6 analog inputs wich a 60
samples s- 1 capability, and a NI Virtual Bench Logger data acquisition program for monitoring the measured
parameters on a PC. . . .
Starting the engine requires an external source of high-~ressure ~tr~~ minimum lOO ~si to spin up the
engine to approximately I 0,000 rpm. Subsequent fuel i~jectton and 1gn1t1on sta~ts the ~ngme. The fuel-flow
rate is controlled by the person operating the engine with the ~se of a lever, which basically controls a valve
constricting the fuel flow to the engine. Engine idles at approximately 5O,?OO rpm and the thrust generated
increases with the increased rpm. To obtain higher thrust values the engine operator steadily increases th
fuel-flow rate from the idle conditions. In order to stop the engine it is brought to the idle conditions an~
run until the exhaust temperature drops under 100 °C, to minimize engine damage.
~ ---,_,
'-..

.6.6 System Modifications \

--
Turbine Technologi'esc htta-a-cqffisi~ system as purchased and used for classroo demonstrations is not
sufficiently fast enough for use with the ~ware-in-the -I UR control al orith s Since ?ne of the main
scopes of the present work is to implement an emonstrate different control algorithms in controllina a
turbojet engine thrust, a new data acquisition system and software has been implemented into the existing
system to increase the data acquisition speed and to increase the control capability. Additionally, the available
system was designed and used to collect and present data and it did not have provisions to send signals via
computer for closed-loop control applicatio~s. Changes implemented include the replacement of the data
acquisition board, connection panels for the sensors, addition of a low flow-rate fuel-flow rate measurement
unit, a fast acting linear servo-controller.

17.6.6 FueJ:FIQw Rate Measurement System

, /.'

The~engi ne as produced by the Turbine ~hnologie~ ~s a pressure transducer together with a

calibration curve to determine the fuel-flow rate ff> the engine. The pressure values read on the fuel line are
plotted against the fuel flow spent and also against the rpm of the enginto generate pressure vs. the fuel rate
and the pressure vs. the rpm calibration curves for long-term operations. owever, for the present purposes,
since the fuel-flow rate is mainly the only control input to control the esired thrust of the engine a more
accurate and fast~r fuel-flow rate measurement device has to be implemented.

~ 17.6. 7 E'
xIt Cd''
on ItIons Monitonng
••

~\lei.s! erop_ulsion equations used in cal_;ulation of the engine thrust requires the measurement of the
exir,conditio~, namely the,.exhaust total pressur~ and the - tC,!JlRera~IJI..~ Although the existing system
available from Turbine Technologies incorporated a eressure transducer and a-tb~rmocoupl~, for this purpose,
the response time for the equipment was rather slow. In order to increase ch; c'ime' resolution of the data
~brained at the exit condit~~s a new ,Q_ressure trac~s£1M~er with a 0.2 ms response rime and a ~ response
nme has been incorporated;.)
As a result of this effort, new insight has been gained into the beha~ior and application of soft computing
technologies in a rocket engine control environment. The methodology created here will provide a new
approach to the area of employing soft computing technologies in rapid response engine control systems
for future vision vehicles. It will yield better insight into incorporating soft computing technologies with
proven and practical software engineering methods. It is expected that chis effort will demonstrate that by
 -~- ffl
~erc,se
11.
9
uting technologies, issuesd in dquality and reliability of the overall schemc of cngme .
ft cornp . h c
lo0ng so rnent can be further improve an t us sarety be further insured. _
I0 technologies is• expected to supplement efforts•m 1mprovm • •
, ef11P oiler deVe P use of these soft computing . g
canrr ore t 1e
1 e,
processo r execution fa
l ' t software development time, sortware maintena nce,
furtherJl'nagemen , l. l l .. II . , u1t to1erancc
~,.rare ma_ d nonlinear contro m powerd eve .transmo ns, a of which contribute a better engme
to •
• uon an fi al . Id c
s01L•
d111iuga 1 . proJ· ected that the m pro uct wi11 yie a roundation for a path .
to further dev I
. _ c opment
an seem, t 1s . h Id b bl .
c0nrrol sy . low cost engme controlle r t at wou e capa e of performi ng in unique vision spacecraft
d d • • h' r
of an aiJcernauve
. • low cost and a vance avionics arc uectures ror autonomous operations· from enuine r:r
. les requ1nng
vehtC engine shutdown.
pre-scare co

.1 summary
111
In this chaptet we have de~lt with the a~plications of soft computing techniques: Th~ applicatio? areas of
th esoft computing techniques are growing day by day. Neural networks and fuzzy logic are effectively used
. :arious control applications. Genetic algorithm plays a major role in providing solutions for optimizing a
;oblem. The combinations of all these techniques give an accurate solution to complex systems. There are
various researches going around the world in the field of soft computing_.

111.s Review Questions

1. State the various applications of neural networks. 7. Explain the application of fuzzy logic systems to
2. Mention the application areas of fuzzy logic.
image processing applications.:

3. In what areas does genetic algorithm gives a best 8. Describe in detail the application of genetic
optimized solution? algorithm to Civil Engineering area.
4. List few applications of hybrid fuzzy GA systems 9. With suitable block diagram, explain the prin-
and neurofuzzy systems. ciple involved in a liquid level controller using
neurofuzzy technique.
5. Soft computing techniques gives best solution to
complex problems. Justify. 10. With a case study example, describe in detail the
application of soft computing.
6. With suitable case study, explain how neural
network best performs its control action.