Upload
muhammad-zeeshan
View
222
Download
0
Embed Size (px)
Citation preview
8/18/2019 Intelligent Control Assignment
1/14
SEMESTER REPORT
TITLE: OVERVIEW OF INTELLIGENT CONTROL TECHNIQUES
Submitted To
Professor Jian Sun
Subject: Intelligent Control
Submitted B
Muhammad Zeeshan
Masters student (control science and technology)
School of Automation
Stude!t I": 2820!00"
BEI#ING INSTITUTE OF TECHNOLOG$ %BIT&' P(R( CHIN)
1
8/18/2019 Intelligent Control Assignment
2/14
I!te**i+e!t Co!t,o* -!d it. -//*ic-tio!:I!t,oductio!:
Intelligent control achie#es automation #ia the emulation of
$iological intelligence% It either see&s to re'lace a humanho 'erforms a control tas& (e%g% a chemical 'rocess
o'erator) or it $orros ideas from ho $iological systems
sol#e 'ro$lems and a''lies them to the solution of control 'ro$lems (e%g% the use of neural netor&s for control)% In
this assignment I am going to discuss se#eral techni*ues
used for intelligent control challenging industrial a''licationdomains here these methods may 'ro#ide 'articularly
useful solutions%
+raditionally intelligent control has em$raced classical control theory neural netor&s fu,,y
logic classical AI and a ide #ariety of search techni*ues (such as genetic algorithms andothers)% -e em'hasis more on the first three% .igure/ illustrates the #ie of the relation $eteen
control theory and neural netor&s% eurocontrol is a &ind of su$set $oth of neural netor&
research and of control theory% one of the $asic design 'rinci'les used in neurocontrol is totallyuni*ue to neural netor& design1 they can all $e understood/and im'ro#ed/more effecti#ely $y
#ieing them as a su$set and etension of ell/&non underlying 'rinci'les from control theory%
3y the same to&en the ne designs de#elo'ed in the neurocontrol contet can $e a''lied 4ust asell to classical nonlinear control% +he $ul& of the 'a'ers on neurocontrol mostly discuss
neurocontrol in the contet of control theory1 also they try to 'ro#ide designs% +he discussion of
$iology may $e limited here $ut e $elie#e that these &inds of designs/designs that dra on the 'oer of control theory/are li&ely to $e more 'oerful than some of the sim'Jer more nai#e
connectionist models of the 'ast1 therefore e sus'ect that they ill 'ro#e to $e more rele#ant to
actual $iological systems hich are also #ery 'oerful controllers% +hese $iological lin&s ha#e $een discussed etensi#ely in other sources hich are cited in this $oo&%.igure/2 illustrates more generally our #ie of the relations
$eteen control theory neurocontrol fu,,y logic and AI%
Just as neurocontrol is an inno#ati#e su$set of control theoryso too is fu,,y logic an inno#ati#e su$set of AI% (Some other
'arts of AI $elong in the u''er middle 'art of .igure/2 as
ell $ut they ha#e not yet achie#ed the same degree of 'rominence in engineering a''lications%)% .u,,y logic hel's
sol#e the 'ro$lem of human/machine communications (in
*uerying e'erts) and formal sym$olic reasoning (to a far
less etent in current engineering a''lications)% In the 'asthen control engineers mainly em'hasi,ed the linear case
and hen AI as 'rimarily 3oolean so/called intelligent
control as mainly a matter of cutting and 'asting5 AIsystems and control theory systems communicated ith each other in relati#ely ad hoc and
distant ays $ut the fit as not #ery good% o hoe#er fu,,y logic and neurocontrol $oth
$uild nonlinear systems $ased on continuous #aria$les $ounded at 0 and % .rom the controller e*uations alone it $ecomes more and more difficult to tell hich system is a neural system and
2
Figure-1
8/18/2019 Intelligent Control Assignment
3/14
hich is a fu,,y system1 the distinction $egins to $ecome meaningless in terms of the
mathematics% +his mo#es us toards a ne era here control theory and AI ill $ecome far
more com'ati$le ith each other% +his allos arrangements li&e hat is shon in .igure/6here neurocontrol and fu,,y logic can $e used as to com'lementary sets of tools for use on
one common controller%
In 'ractice there are many ays to com$ine fu,,y logic and other forms of AI ith neurocontroland other forms of control theory% .or eam'le see .igure/7% +his $oo& ill try to 'ro#ide the
$asic tools and eam'les to ma&e 'ossi$le a ide #ariety of com$inations and a''lications and
to stimulate more 'roducti#e future research%
I!te**i+e!t Co!t,o* Tec0!i1ue.:In this section i 'ro#ide $rief o#er#ies of the main areas of intelligent control% i $riefly see& to
'resent the $asic ideas to gi#e a fla#or of the a''roaches%
Fu22 co!t,o*:.u,,y control is a methodology to re'resent and im'lement a (smart) humans &noledge a$outho to control a system% A fu,,y controller is shon in .igure/!% +he fu,,y controller has
se#eral com'onents5
• +he rule/$ase is a set of rules a$out ho to control%
• .u,,ification is the 'rocess of transforming the numeric in'uts into a form that can $e
used $y the inference mechanism%
• +he inference mechanism uses information
a$out the current in'uts (formed $y
fu,,ification) decides hich rules a''ly in
the current situation and forms conclusionsa$out hat the 'lant in'ut should $e%
• 9efu,,ification con#erts the conclusions
reached $y the inference mechanism into a
numeric in'ut for the 'lant%
aturally there are many ays to com$ine these $asic designs in com'le real/orlda''lications% .or eam'le there are many com'le 'ro$lems here it is difficult to find a good
3
Figure-2
Figure-4 Figure-3
Figure-5
8/18/2019 Intelligent Control Assignment
4/14
controller $y ada'tation alone starting from random eights% In such 'ro$lems it is crucial to
use a strategy called :sha'ing%: In sha'ing one first ada'ts a sim'ler controller to a sim'lified
#ersion of the 'ro$lem 'erha's $y using a sim'ler neurocontrol a''roach or e#en $y tal&ing toan e'ert1 then one uses the eights of the resulting controller as the initial #alues of eights of
a controller to sol#e the more com'le 'ro$lem% +his a''roach can of course $e re'eated many
times if necessary% ;ne can also $uild systems that 'hase in gradually from a sim'ler a''roach toa more com'le a''roach%
.u,,y control design5
As an eam'le consider the tan&er shi' steering a''lication in .igure/< here the shi' istra#eling in the x direction at a heading ψ and is steered $y the rudder in'ut δ% =ere e see& to
de#elo' the control system in .igure/8 $y s'ecifying a fu,,y controller that ould emulate ho
a shi' ca'tain ould steer the shi'% =ere if ψr is the desired heading e > ψr − ψ and c > ?e%
+he
design of the fu,,y controller essentially amounts to
choosing a set of rules (@rule $ase) here each rule re'resents &noledge that the ca'tain hasa$out ho to steer% Consider the folloing set of rules5
% I3 e is neg -!d c is neg T0e! δ is 'oslarge
2% I3 e is neg -!d c is ,ero T0e! δ is 'ossmall6% I3 e is neg -!d c is 'os T0e! δ is ,ero
7% I3 e is ,ero -!d c is neg T0e! δ is 'ossmall
!% I3 e is ,ero -!d c is ,ero T0e! δ is ,ero
8/18/2019 Intelligent Control Assignment
5/14
indicate @degrees of certainty% .or instance for the e uni#erse of discourse the triangular
mem$ershi' function that 'ea&s at e > 0 re'resents the (fu,,y) set of #alues of e that can $e
referred to as @,ero% +his mem$ershi' functionhas a #alue of for e > 0 hich indicates that e
are a$solutely certain that for this #alue of e e
can descri$e it as $eing @,ero% As e increases or decreases from 0 e $ecome less certain that e
can $e descri$ed as @,ero and hen its
magnitude is greater than π e are a$solutelycertain that is it not ,ero so the #alue of the
mem$ershi' function is ,ero% +he meaning of the
other to mem$ershi' functions on the e
uni#erse of discourse (and the mem$ershi'functions on the change in/error uni#erse of
discourse) can $e descri$ed in a similar ay% +he
mem$ershi' functions on the δ uni#erse of
discourse are called @singletons% +hey re'resentthe case here e are only certain that a #alue of
δ is for eam'le @'ossmall if it ta&es on onlyone #alue in this case 70π/ 80 and for any other
#alue of δ e are certain that it is not @'ossmall%
.inally notice that .igure/8 shos therelationshi' $eteen the scaling gains in .igure/" and the scaling of the uni#erses of discourse
(notice that for the in'uts there is an in#erse relationshi' since an increase an in'ut scaling gain
corres'onds to ma&ing for instance the meaning of @,ero corres'ond to smaller #alues)%
It is im'ortant to em'hasi,e that other mem$ershi' function ty'es (sha'es) are 'ossi$le1 it is u'to the designer to 'ic& ones that accurately re'resent the $est ideas a$out ho to control the
'lant% @.u,,ification (in .igure/) is sim'ly the act of finding e%g% μpos(e) for a s'ecific #alue
of e% et e discuss the com'onents of the inference mechanism in .igure/ min{μneg (e) , μzero(c) }
-hy +hin& a$out the con4unction of to uncertain statements% +he certainty of the assertion of
to things is the certainty of the least certain statement% In general more than one μpremise(i)
ill $e non,ero at a time so more than one rule is @on (a''lica$le) at e#ery time% Gach rule thatis @on can contri$ute to ma&ing a recommendation a$out ho to control the 'lant and generally
ones that are more on (i%e% ha#e μpremise(i) closer to one) should contri$ute more to the
conclusion% +his com'letes the descri'tion of the inference mechanism%9efu,,ification in#ol#es com$ining the conclusions of all the rules% @Center/a#erage
defu,,ification Hses
5
Figure-8
8/18/2019 Intelligent Control Assignment
6/14
-here bi is the 'osition of the center of the
out'ut mem$ershi' function for the ith rule
(i%e% the 'osition of the singleton)% +his issim'ly a eighted a#erage of the conclusions%
+his com'letes the descri'tion of a sim'lefu,,y controller (and notice that e did not
use a mathematical model in its construction)%
+here are many etensions to the fu,,y
controller that e descri$e a$o#e% +here areother ays to *uantify the @and ith fu,,y
logic other inference a''roaches other
defu,,ification methods@+a&agi/Sugeno fu,,y systems and multi/
in'ut multi/out'ut fu,,y systems%Shi' eam'le5 Hsing a nonlinear model for a tan&er shi' e get the res'onse in .igure/B (tuned using ideas
from ho you tune a 'ro'ortional/deri#ati#e controller1 notice that the #alues of g > 2 /π g 2 >
2!0 and g 0 > 8π/ 8 are different than the first guess #alues shon in .igure/8 and the controller
surface in .igure/0% +he control surface shos that there is nothing mystical a$out the fu,,ycontroller
It is sim'ly a static (i%e% memoryless) nonlinear ma'% .or real/orld a''lications most often the
surface ill ha#e $een sha'ed $y the rules to ha#e interesting nonlinearities%
6
Figure-9
Figure-10
8/18/2019 Intelligent Control Assignment
7/14
Neu,-* !et4o,5.:
Artificial neural netor&s are circuits com'uter algorithms or mathematical re'resentationsloosely ins'ired $y the massi#ely connected set of neurons that form $iological neural netor&s%
Artificial neural netor&s are an alternati#e com'uting technology that ha#e 'ro#en useful in a
#ariety of 'attern recognition signal 'rocessing estimation and control 'ro$lems% =ere e illfocus on their use in estimation and control%
Mu*ti*-e, /e,ce/t,o!:
+he feedforard multilayer 'erce'tron is the most 'o'ular neural netor& in control system
a''lications and so e limit our discussion to it% +he second most 'o'ular one is 'ro$a$ly the
radial $asis function neural netor& (of hich one form is identical to one ty'e of fu,,y system)%
+he multilayer 'erce'tron is com'osed of an interconnected set of neurons each of hich has
the form shon in .igure/"% =ere
And the wi are the interconnection @eights and b is the @$ias for the neuron (these
'arameters model the interconnections $eteen the cell $odies in the neurons of a $iologicalneural netor&)% +he signal z re'resents a signal in the $iological neuron and the 'rocessing that
the neuron 'erforms on this signal is re'resented ith an @acti#ation function f here
+he neuron model re'resents the $iological neuron that @fires (turns on) hen its in'uts aresignificantly ecited (i%e% z is $ig enough)% @.iring is defined $y an @acti#ation function f
here to (of many) 'ossi$ilities for its definition are5
+he threshold function
7
Figure-11
8/18/2019 Intelligent Control Assignment
8/14
Sigmoid (logistic) function
+here are many other 'ossi$le choices for neurons including a @linear neuron that is sim'ly
gi#en $y f ( z ) > z %
G*uation () ith one of the a$o#e acti#ation functions re'resents the com'utations made $yone neuron% et e interconnect them% Fet circles re'resent the neurons (eights $ias and
acti#ation function) and lines re'resent the connections $eteen the in'uts and neurons and theneurons in one layer and the net layer% .igure 8 is a three @layer 'erce'tron since there are
three stages of neural 'rocessing $eteen the in'uts and out'uts%=ere e ha#e
• In'uts5 xi i > , 2 , . . ., n
• ;ut'uts5 yj j > , 2 , . . . ,m
• um$er of neurons in the first @hidden layer n in the second hidden layer n2 and in
the out'ut layer m
• In an N layer 'erce'tron there are ni neurons in the ith hidden layer i > , 2 , . . ., N − %
-e ha#e
8
8/18/2019 Intelligent Control Assignment
9/14
-ith j > , 2 , . . ., n% -e ha#e
-ith j > , 2 , . . ., n2% -e ha#e
-ith j > , 2 , . . .,m% =ere e ha#e
• wij (w2ij) are the eights of the first (second) hidden layer
• wij are the eights of the out'ut layer
• b j are the $iases of the first hidden layer%
• b2 j are the $iases of the second hidden layer
• bj are the $iases of the out'ut layer
• fj (for the out'ut layer) f 2
j (for the second hidden layer) and f
j (for the first hidden layer)are the acti#ation functions (all can $e different)%
E6/e,t co!t,o*:.or the sa&e of our discussion e ill sim'ly #ie the e'ert system that is used here as acontroller for a dynamic system as is shon in .igure/2% =ere e ha#e an e'ert system
ser#ing as feed$ac& controller ith reference in'ut r and feed$ac& #aria$le y% It uses the
information in its &noledge$ase and its inference mechanism to decide hat command in'ut to generate for the 'lant%
Conce'tually e see that the e'ert controller is closely related to the fu,,y controller% +here
are hoe#er se#eral differences% .irst the &noledge/$ase in the e'ert controller could $e arule/$ase $ut is not necessarily so% It could $e de#elo'ed using other &noledge/re'resentation
structures such as frames semantic nets causal diagrams and so on% Second the inferencemechanism in the e'ert controller is more general than that of the fu,,y controller% It can use
more so'histicated matching strategies to determine hich rules should $e alloed to fire% It canuse more ela$orate inference strategies such as @refraction @recency and #arious other 'riority
schemes% et e should note that .igure/2 shos a direct e'ert controller% It is also 'ossi$le
to use an e'ert system as a su'er#isor for con#entional or intelligent controllers%
9
Figure-12
8/18/2019 Intelligent Control Assignment
10/14
P*-!!i!+ S.tem. 3o, Co!t,o*:Artificially intelligent 'lanning systems (com'uter 'rograms that are often designed to emulate
the ay e'erts 'lan) ha#e $een used for se#eral 'ro$lems including 'ath 'lanning and high/le#el decisions a$out control tas&s for ro$ots
8/18/2019 Intelligent Control Assignment
11/14
that in the 'ast may ha#e $een 'erformed $y the co'ilot% In manufacturing systems efficiency
o'timi,ation and flo control are $eing automated and ro$ots are re'lacing humans in
'erforming relati#ely com'le tas&s% .rom a $road historical 'ers'ecti#e each of thesea''lications $egan at a lo le#el of automation and through the years each has e#ol#ed into a
more autonomous system% .or eam'le automoti#e cruise controllers are the ancestors of the
(research 'rototy'e) controllers that achie#e coordinated control of steering $ra&ing and throttlefor autonomous #ehicle dri#ing% And the terrain folloing and terrain a#oidance control systems
for lo/altitude flight are ancestors of an artificial 'ilots associate that can integrate mission and
tactical 'lanning acti#ities% +he general trend has $een for engineers to incrementally @add moreintelligence in res'onse to consumer industrial and go#ernment demands and there$y create
systems ith increased le#els of autonomy%
In this 'rocess of enhancing autonomy $y adding intelligence engineers often study ho humans
sol#e 'ro$lems then try to directly automate their &noledge and techni*ues to achie#e highle#els of automation% ;ther times engineers study ho intelligent $iological systems 'erform
com'le tas&s then see& to automate @natures a''roach in a com'uter algorithm or circuit
im'lementation to sol#e a 'ractical technological 'ro$lem (e%g% in certain #ision systems)% Such
a''roaches here e see& to emulate the functionality of an intelligent $iological system (e%g%the human) to sol#e a technological 'ro$lem can $e collecti#ely named @intelligent systems and
control techni*ues%It is $y using such techni*ues that some engineers are trying to create highly autonomous
systems such as those listed a$o#e% .igure 6 shos a functional architecture for an intelligent
autonomous controller ith an interface to the 'rocess in#ol#ing sensing (e%g% #ia con#entionalsensing technology #ision touch smell etc%) actuation (e%g% #ia hydraulics ro$otics motors
etc%) and an interface to humans (e%g% a dri#er 'ilot cre etc%) and other systems% +he
@eecution le#el has lo/le#el numeric signal 'rocessing and control algorithms (e%g% PI9
o'timal ada'ti#e or intelligent control1 'aram/eter estimators failure detection andidentification (.9I) algorithms)% +he @coordination le#el 'ro#ides for tuning scheduling
su'er#ision and redesign of the eecution/le#el algorithms crisis management 'lanning and
learning ca'a$ilities for the coordination of eecution/le#el tas&s and higher/le#el sym$olicdecision ma&ing for .9I and control algorithm management% +he @management le#el 'ro#ides
for the su'er#ision of loer/le#el functions and for managing the interface to the human(s) and
other systems% In 'articular the management le#el ill interact ith the users in generating goalsfor the controller and in assessing the ca'a$ilities of the system% +he management le#el also
monitors 'erformance of the loer/le#el systems 'lans acti#ities at the highest le#el (and in
coo'eration ith humans) and 'erforms high/le#el learning a$out the user and the loer/le#el
algorithms% Con#entional or intelligent systems methods can $e used at each le#el%
11
8/18/2019 Intelligent Control Assignment
12/14
)//*ic-tio!.:
In this section some of the main characteristics of the intelligent system methods that ha#e
'ro#en useful in industrial a''lications are outlined% +hen eam'les are gi#en for use of the
methods%
Heu,i.tic Co!.t,uctio! o3 No!*i!e-, Co!t,o**e,.:
+he first area e discuss here intelligent control has had a clear im'act in industry is the area
of heuristic construction of nonlinear controllers% +o areas in intelligent control ha#e made
most of the contri$utions to this area5 fu,,y control and e'ert systems for control (here e ill
focus on fu,,y control one ty'e of rule/$ased controller since the ideas etend directly to the
e'ert control case)% +he reason that the methods are @heuristic is that they normally do not rely
on the de#elo'ment and use of a mathematical model of the 'rocess to $e controlled%
"-t-7B-.ed No!*i!e-, E.tim-tio!:
+he second ma4or area here methods from intelligent control ha#e had an im'act in industry is
in the use of neural netor&s to construct ma''ings from data% In 'articular neural netor&
methods ha#e $een found to $e *uite useful in 'attern recognition and estimation% 3elo e
e'lain ho to construct neural netor& $ased estimators and gi#e an eam'le here such a
method as used%
E.tim-to, Co!.t,uctio! Met0odo*o+:
In con#entional system identification you gather 'lant in'ut/out'ut data and construct a model
(ma''ing) $eteen the in'uts and out'uts% In this case model construction is often done $y
tuning the 'arameters of a model (e%g% the 'arameters of a linear ma''ing can $e tuned using
linear least s*uares methods or gradient methods)% +o #alidate this model you gather no#el 'lant
12
Figure-14
8/18/2019 Intelligent Control Assignment
13/14
in'ut/out'ut data and 'ass the in'uts into your constructed model and com'are its out'uts to the
ones that ere generated $y the model% If some measure of the difference $eteen the 'lant and
model out'uts is small then e acce't that the model is a good re'resentation of the system%
eural netor&s or fu,,y systems are also tuna$le functions that could $e used for this system
identification tas&% .u,,y and neural systems are nonlinear and are 'arameteri,ed $ymem$ershi' function 'arameters or eights (and $iases) res'ecti#ely% Lradient methods can $e
used to tune them to match ma''ings that are characteri,ed ith data% Ealidation of the models
'roceeds along the same lines as ith con#entional system identification%
In certain situations you can also gather data that relates the in'uts and out'uts of the system to
'arameters ithin the system% +o do this you must $e a$le to #ary system 'arameters and gather
data for each #alue of the system 'arameter (the gathered data should change each time the
'arameter changes and it is either gathered #ia a so'histicated simulation model or #ia actual
e'eriments ith the 'lant)% +hen using a gradient method you can ad4ust the neural or fu,,y
system 'arameters to minimi,e the estimation error% +he resulting system can ser#e as a
'arameter estimator (i%e% after it is tunednormally it cannot $e tuned on/line $ecause actual#alues of the 'arameters are not &non on/line they are hat you are trying to estimate)%
E6-m/*e:
)utomoti8e E!+i!e F-i*u,e E.tim-tio!:
In recent years significant attention has $een gi#en to reducing ehaust gas emissions 'roduced
$y internal com$ustion engines% In addition to o#erall engine and emission system design
correct or fault/free engine o'eration is a ma4or factor determining the amount of ehaust gas
emissions 'roduced in internal com$ustion engines% =ence there has $een a recent focus on the
de#elo'ment of on/$oard diagnostic systems that monitor relati#e engine health% Although on/
$oard #ehicle diagnostics can often detect and isolate some ma4or engine faults due to idely
#arying dri#ing en#ironments they may $e una$le to detect minor faults hich may nonetheless
affect engine 'erformance% Minor engine faults arrant s'ecial attention $ecause they do not
noticea$ly hinder engine 'erformance $ut may increase ehaust gas emissions for a long 'eriod
of time ithout the 'ro$lem $eing corrected% +he minor faults e consider in this case study
include @cali$ration faults (here the occurrence of a cali$ration fault means that a sensed or
commanded signal is multi'lied $y a gain factor not e*ual to one hile in the no/fault case the
sensed or commanded signal is multi'lied $y one) in the throttle and mass fuel actuators and in
the engine s'eed and mass air sensors% +he relia$ility of these actuators and sensors is
'articularly im'ortant to the engine controller since their failure can affect the 'erformance of
the emissions control system%
=ere e sim'ly discuss ho to formulate the 'ro$lem so that it can $e sol#ed ith neural or
fu,,y estimation schemes% +he &ey to this is to understand ho data is generated for the training
of neural or fu,,y system estimators%
+he e'erimental setu' in the engine test cell consists of a .ord 6%0 F E/< engine cou'led to an
electric dynamometer through an automatic transmission% An air charge tem'erature sensor
13
8/18/2019 Intelligent Control Assignment
14/14
(AC+) a throttle 'osition sensor (+PS) and a mass airflo sensor (MA.) are installed in the
engine to measure the air charge tem'erature throttle 'osition and air mass flo rate% +o
heated ehaust gas oygen sensors (=GL;) are located in the ehaust 'i'es u'stream of the
catalytic con#erter% +he resultant airflo information and in'ut from the #arious engine sensors
are used to com'ute the re*uired fuel flo rate necessary to maintain a 'rescri$ed air/to/fuel
ratio for the gi#en engine o'eration% +he central 'rocessing unit (GGC/IE) determines the neededin4ector 'ulse idth and s'ar& timing and out'uts a command to the in4ector to meter the eact
*uantity of fuel%
An GCM (electronic control module) $rea&out $o is used to 'ro#ide eternal connections to the
GGC/IE controller and the data ac*uisition system% +he angular #elocity sensor system consists
of a digital magnetic ,ero/s'eed sensor and a s'ecially designed fre*uency/to/#oltage con#erter
hich con#erts fre*uency signals 'ro'ortional to the rotational s'eed into an analog #oltage%
9ata is sam'led e#ery engine re#olution% A #aria$le load is 'roduced through the dynamometer
hich is controlled $y a 9N/F;C IE s'eedOtor*ue controller in con4unction ith a 9+C/
throttle controller installed $y 9yneSystems Com'any% +he load tor*ue and dynamometer s'eedare o$tained through a load cell and a tachometer res'ecti#ely% +he throttle and the
dynamometer load reference in'uts are generated through a com'uter 'rogram and sent through
an DS/262 serial communication line to the controller% Physical *uantities of interest are digiti,ed
and ac*uired utili,ing a ational Instruments A+/MI;/