Design of Neuro-Fuzzy System Controller for DC Servomotor- Based Satellite Tracking System

Transcription

1 IOSR Journl of Electrcl nd Electroncs Engneerng (IOSR-JEEE) e-issn: ,p-ISSN: 3-333, Volume, Issue 4 Ver. III (Jul. Aug. 6), PP Desgn of Neuro-Fuzzy System Controller for DC Servomotor- Bsed Stellte Trckng System Lnus A. Alwl *, Peter K. Khto nd Stnley I. Kmu 3,, 3 Deprtment of Electrcl nd Electronc Engneerng, Jomo Kenytt Unversty of Agrculture & Technology (JKUAT), P.O.BOX 6-, Nrob, Keny. Abstrct: A prbolc dsh ntenn postonng system s control unt whch drects the ntenn to desred trget utomtclly. In ll ts spects, utomton hs become wdespred nd promses to represent the future for stellte trckng ntenn systems. The term trckng s used to men tht the control system should utlze sutble lgorthms to utomte the process of pontng the ntenn to selected stellte thereby estblshng the desred lne of sght (LOS). Ths pper ms to present nd dscuss the results obtned from the development of DC servomotor-bsed Neuro-Fuzzy System Controller (NFSC) lgorthm whch cn be ppled n the postonng of the prbolc reflector ntenns. The dvntge of usng NFSC method s tht t hs hgh degree of nonlnerty tolernce, lernng blty nd solves problems tht re dffcult to ddress wth the conventonl technques such s Proportonl Integrl Dervtve (PID) strtegy. The rchtecture of the proposed ntenn control system employs Adptve Neuro-Fuzzy Inference System (ANFIS) desgn envronment of MATLAB/SIMULINK pckge. The results obtned ndcted tht the proposed NFSC ws ble to cheve desred output DC motor poston wth reduced rse tme nd overshoot. Future work s to pply the controller n rel tme to cheve utomtc stellte trckng wth prbolc ntenn usng dt derved from the sgnl strength. Keywords: ANFIS, Antenn-Postonng System, DC servomotor, NFSC I. Introducton Prbolc ntenns mounted t erth sttons whch re commonly used n stellte trckng pplctons, re prone to suffer from envronmentl dsturbnces []. For severl yers, DC servomotor-bsed controllers hve been ppled n closed-loop control systems to poston or stblze the stellte dshes. However, DC servomotors re known to hve nonlner prmeters nd dynmc fctors, such s sturton, bcklsh nd frcton whch cnnot be overlooked. Severl controller models hve been developed over tme to solve the problem of ntenn pontng n stellte nd movble trgets trckng usng servomechnsm [], [3], [4], [5], [6], nd [7]. These nclude conventonl controllers such s Proportonl-Integrl (PI), Lner Qudrtc Gussn (LQG) nd Proportonl-Integrl-Dervtve (PID) controller on one hnd nd new ntellgent technques such s Fuzzy Logc Controller (FLC), Neurl Network, Fuzzy-neurl networks nd Fuzzy-genetc lgorthm on the other hnd [8], [9] nd []. Conventonl controllers such s PI nd PID controllers re senstve to vrton n the motor prmeters nd lod. Also, tunng PI or PID gns to reduce the overshoot due to lod dsturbnce s dffcult. Moreover, n ccurte non-lner model of ctul DC motor s dffcult to fnd nd prmeter vlues obtned from system dentfcton my be only pproxmte vlues. For ths reson, mny reserchers tody re nterested n pplyng ntellgent dptve control technques to cheve fst speed response nd tolernce to prmeter vrtons. Fuzzy Logc Controller cn be n lterntve control technque to the PID controller [], [], [8] nd [9]. But the mjor drwbck of the FLC s nsuffcent nlytcl desgn technque wth respect to the selecton of the rules, the membershp functons nd the sclng fctors. On the other hnd, rtfcl neurl network (ANN) hs the blty to lern nd dpt but cnnot expln wht t hs lernt []. As result, ntegrtng FLC nd ANN to generte hybrd model cn tke dvntge of strong ponts of both [], strength whch s explored here... Problem Formulton The objectve s to desgn n ANFIS controller for DC servomotor-bsed ntenn pontng system to meet the followng tme domn step response trckng specfctons: rse tme ( t r ) 3s, settlng tme ( t s ) 3s nd mxmum overshoot ( M % ).These lmts hve been selected bsed on prctcl ndustrl stndrds. p.. System Descrpton Fg. s the control block dgrm of the DC servomotor ntenn pontng system. The frst nput to the summer s set poston r (t), the desred poston of the zmuth or elevton motor. The second nput s the DOI:.979/ Pge

2 Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System feedbck sgnl, the current poston of the respectve motor, cptured by some feedbck sensor lke the potentometer nd chnged to summer redble formt. The dfference between these two nputs, clled poston error sgnl e (t), s gven to the controller tht reds the sgnl nd produces pproprte output sgnl, controller output u (t).the controller output reches the motor drver, whch produces proportonl output to rotte the respectve motor n ether drecton ccordng to the sgn of the error sgnl. As the desred poston s pproched, the error sgnl reduces to zero nd the motor stops [6]. Fg..Block dgrm of ntenn control mechnsm II. Mthemtcl Modelng of DC (Servo) motor System Fg.. represents the DC (servo) motor model wth prmeters defned n Tble. For n rmturecontrolled seprtely-excted DC motor, the voltge ppled to the rmture of the motor s vred wthout chngng the voltge ppled to the feld. Fg.. DC motor Crcut Dgrm Usng Krchhoff s Voltge Lw, the output rmture voltge V (t) nd motor torque s relted to () whle motor torque T m (t) s relted to the rmture current I (t) by constnt fctor K T gven n (): di ( t) V ( t) RI ( t) L dt Eb ( t) () Tm ( t) KT I ( t) () The bck electromotve force (e.m.f) E b (t) s relted to the ngulr velocty by (3). Applyng Newton's Lw, Krchhoff's Lw nd Lplce trnsform [7] genertes (4) nd (5), from whch (6) results by elmntng current. d Eb ( t) KTm( t) KT dt (3) J s B s( s) K I (4) DOI:.979/ Pge T L si R I V K s( s) (5) V KT s J s Bs K (6) T R Ls The trnsfer functon from the nput voltge, V (s) to output ngle drectly follows (7): KT G (7) V s[( R L s)( J s B ) K K ] m T m T B

3 Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System In fxed feld motor t s ssumed tht KT K nd B R >> L whch smplfes (7) to (8): KT m( s) JR (8) E( s) BmR KT KB s( s ) JR Fg. 3 s block dgrm of DC (servo) motor system showng elements of the trnsfer functon. Fg.3. Block-Dgrm of Seprtely Excted DC (servo) motor III. Antenn Postonng System Trnsfer Functon nd PID Controller The system prmeters re gven n Tble [4, 6]. Detled dervton of the system trnsfer functons wthout ny compenstor s well s the ssumptons mde s presented n [6, 7]. The open-loop trnsfer functon for output ngulr velocty o (t) wthout feedbck s relzed s gven by (9): o.83 G (9) V s.7s 7 p Tble Prmeters of Model wth DC Servomotors Prmeter Defnton Azmuth/Elevton Power Amplfer Pole m Motor nd Lod Pole.7 B Motor Dmpenng Constnt[Nms/rd]. B L Lod Dmpenng Constnt[Nms/rd] B m Equvlent vscous frcton coeff. [Nms/rd]. J Motor Inertl Constnt[Kgm]. J L Lod Inertl Constnt[Kgm] J m Equvlent moment of nert[kgm].3 K Premplfer Gn _ K Power Amplfer Gn K B Bck emf Constnt[Vs/rd].5 K g Ger Rto. K m Motor nd Lod Gn.83 K pot Potentometer Gn.38 K T Motor Torque Constnt[Nm/A].5 L Motor Armture Inductnce[H].45 N Turns on Potentometer N, N N 3 Ger Teeth(Respectvely) 5,5,5 R Motor Armture Resstnce[Ω] 8 V Voltge cross Potentometer[V] The closed-loop trnsfer functon wthout the PID compenstor s obtned n () by usng (9) nd the block dgrm reducton. Accordng to Routh-Herwtz crteron, the system wll gve stble response wth the vlue of "K" n the rnge < K < 63, hence s tken [6], [7], gvng (): o s o s K.7s 7s 6.63K 663.7s 7s 663 () () DOI:.979/ Pge

4 Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System Fg.4. Block dgrm of Antenn Azmuth PID controller The structure of PID controller ppled to ntenn postonng system s shown n Fg.4 whle ts prllel form, [4], s gven n (): de( t) U( t) K pe( t) KI e( t) K () D dt where K, K nd K re proportonl, ntegrl nd dervtve gns nd e (t) = error. It s cler tht p I D the control sgnl s mde up of the sum of three stted gn components. The vlues of K P, K I nd K D re obtned usng Zegler- Nchols tunng lgorthm n [6, 7]. The three PID controller gn prmeters tht gve the most stble response were obtned n [7], nd quoted s follows: K 6, K 5 nd K [7]. IV. Fuzzy Logc Controller (FLC) Desgn As shown n Fg.5, the FLC system hs four mn components: Fuzzfcton Interfce (converts crsp pont nto fuzzy set), Knowledge Bse (conssts of dtbse hostng membershp functons of the fuzzy sets nd rule-bse contnng number of fuzzy IF-THEN rules), Inference Engne (derves concluson from the fcts nd rules contned n the knowledge bse), nd the Defuzzfcton Interfce (mps fuzzy set to crsp set) [8], [] nd []. p I D Fg.5. Block Dgrm of Fuzzy Logc Control System The FLC creted hd nputs: poston error e(t) desgnted s E nd chnge n poston error e(t) represented by DE nd sngle output gven s control nput to the DC servo motor drver denoted s CI. The lngustc vrbles hve been defned s: {NB, NM, NS, ZR, PS, PM, PB}, where the ntls correspond to negtve bg, negtve medum, negtve smll, zero, postve smll, postve medum nd postve bg respectvely. A totl of 7x7=49 Mmdn type rules were used to represent chnges n the two nputs. Trngulr fuzzy sets, beng smple nd esy to clculte, hve been selected for both nput (fuzzfcton) nd output (defuzzfcton). The spn of non-normlzed E s [-7 7], DE s [-7 7] nd tht of CI s [-7 7] s shown n Fg.6 together wth the constnt Sugeno output fuzzy set. The defuzzfcton method used s the centrod method becuse t cn be esly mplemented n dgtl control systems nd requres less computton tme. For gven set of nputs to the FLC, pproprte rule(s) must be fred from the rule bse nd defuzzfed to gve the control nput sgnl. Tble summrzes the control rules whch mp the fuzzy nputs to fuzzy output. DOI:.979/ Pge

5 Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System Fg.6.The membershp functons of E, CE nd CI n FIS Fle Tble 7x7 Rules Fuzzy Assoctve Memory (FAM) V. Neuro-Fuzzy System Controller Desgn 5. Adptve Neuro-Fuzzy Prncple Adptve Neuro-Fuzzy Inference System (ANFIS) technque s used s techng method for Sugenotype fuzzy systems nd ws proposed by Jng []. In pplyng ANFIS, the number nd type of fuzzy system membershp functons (MFs) hs to be specfed by user. The method s more effcent n the sense tht t combnes the dvntges of FLC nd NN pproch n order to construct nonlner self-tunng controller. In ddton, snce the rules re n lngustc formt, ntermedte results cn be nlyzed nd nterpreted esly. ANFIS method s lso vewed by mny reserchers, s hybrd method, whch conssts of two prts: grdent method tht s ppled to clculton of nput membershp functon prmeters nd lest squre method whch s ppled to clculton of output functon prmeters [3]. There re three constrnts of usng MATLAB ANFIS method s follows: only Sugeno-type decson method s vlble, there cn be only one output nd lstly, defuzzfcton method s weghted men vlue. A typcl rchtecture of ANFIS control structure s shown n Fg.7, n whch crcle ndctes fxed node, wheres squre ndctes n dptve node [4]. A A x y x A B A w N w 4, w z z y B B w N w 4, x y w z B Lyer Lyer Lyer 3 Lyer 4 Lyer 5 Fg.7. Correspondng ANFIS rchtecture For smplcty, two nputs x, y nd one output z re consdered. For frst order Sugeno fuzzy model wth two fuzzy f then rules, common rule set cn be expressed s n (3) nd (4): Rule : f x s A nd y s B, then DOI:.979/ Pge

6 Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System z px qy r Rule : f x s A nd y s B, then z px q y r (3) (4) where A nd B re the fuzzy sets n the ntecedent, nd p, q nd r re the desgn prmeters tht re determned durng the trnng process. The ANFIS network structure, s mde up of set of unts (nd connectons) orgnzed nto fve connected network lyers, to 5 s shown n Fg.7.The detled functons performed by ech lyer s explned n [] nd [3] nd re summrzed n the next secton under ANFIS desgn. The ANFIS structure cn be tuned utomtclly by the hybrd lernng lgorthm usng lest-squre estmton (for output membershp functons) nd bck propgton lgorthm (for output nd nput membershp functons) [4], [5] nd [6]. 5. Adptve Neuro-Fuzzy System Controller Desgn The ANFIS controller genertes chnge n reference drve voltge bsed on poston error E nd dervtve n poston error (speed error) DE defned by (5) nd (6): Error (E) = (Desred poston - Actul poston) (5) Error Chnge (DE) = (Current Error - Prevous Error) (6) In ths study, frst order Sugeno-type fuzzy nference s used for ANFIS nd the typcl fuzzy rule tkes the form of (7): If E s A nd DE s B then, z = f (E, DE) (7) where A nd B re fuzzy sets n the ntecedent nd z =f (E, DE) s crsp functon n the consequent. The sgnfcnces of ech lyer nd operton of the nput - output ANFIS structure [], [3] consdered re: Lyer : Ths lyer (the fuzzfcton lyer) enbles the entry of rw dt or crsp nputs from the trget system nto ANFIS. It s composed of number of computng nodes whose ctvton functons re fuzzy logc membershp functons, tken here s trngulr. Ech dptve node genertes the membershp grdes clled fuzzy spces for the nput vectors A,,, n nd B,,, n where n s the number of membershp functons of the nputs (E nd DE) chosen s n 7, defned by (8). The degree to whch the nputs le wthn the fuzzy spce s gven vlue normlzed between -7 nd 7. O A, = (E), O B, = (DE),,...,n (8) A Lyer : Is the rule lyer where ech node s fxed. Once the loctons of nputs n the fuzzy spces re dentfed, the product of the degrees to whch the nputs stsfy the membershp functons s found. Ths product s clled the frng strength of rule whose output s gven by (9). In other words, t selects the mnmum vlue of the nputs. In ths lyer, the totl number of Tkg-Sugeno rules used s 49. O = W mn( A(E). B (DE)) (9) Lyer 3: Ths s the normlzton lyer n whch the rto of ech rule s frng strength s clculted wth th respect to the sum of the frng strengths of ll the rules. Ech node n ths lyer s fxed. The node output s th the nput ctvton level dvded by the sum of ll the ctvton levels of the other nputs, s gven n (): 3 W () O = W n W Lyer 4: In lyer 4, the defuzzfcton lyer, the output of ech node s the weghted consequent vlue. th Adptve node n ths lyer clcultes the contrbuton of rule towrds the overll output, wth the followng node functon n (): 4 O = W Z W ( PE qde r ) () Lyer 5: Lyer 5 s the summton lyer nd ts output, whch s the sum of ll the outputs of lyer 4, gves the overll output for the respectve nputs wthn the fuzzy spce. The sngle fxed node n ths lyer computes the overll output s the sum of ech rule s contrbuton gven n (): B DOI:.979/ Pge

7 Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System O 5 = W Z W Z W Z W W () Before the ANFIS system cn be used for predcton, the prmeters of the rules re determned by frst genertng n ntl FIS where frst rndom vlues re ssgned to the prmeters. Next, n optmzton scheme s ppled to determne the best vlues of the prmeters tht would supply rules to delstclly model the trget system. After trnng, the rules remn so tht when new nput dt s presented to the model, the rules provde correspondng resonble output [3]. The optmzton technque used s hybrd lernng lgorthm tht mnmzes the error between the ANFIS model nd the rel system usng trnng dt from the trget system to generte sgnls tht propgte bckwrds nd forwrds nd updte the prmeters [4]. The prmeters to be trned re A, nd B of the premse prmeters nd p, q nd r of the consequent prmeters. The ANFIS Edtor GUI wndow ncludes four dstnct res to support typcl workflow. It enbles relzton of the followng tsks: Lodng, Plottng, nd Clerng the Dt Genertng or Lodng the Intl FIS Structure Trnng the FIS Vldtng the Trned FIS [5] For genertng FIS structure, the trngulr MF s used for the two nput vrbles nd output type s lner. The number of MFs for the nput vrbles E nd DE s 7 ech hence the number of rules s 7*7 = 49. Fg.8 shows the membershp functons for E nd DE before trnng. Fg.9 shows the generted ANFIS structure used for the DC servo motor ntenn postonng controller desgn. It hs been used 8% of the generted dtsets for trnng the ANFIS system model nd % ech s testng nd checkng dt. Hybrd lernng lgorthm ws used for trnng the generted FIS wth the number of epochs s nd tolernce of.. It s cler from [6] tht the trngulr MF s specfed by two prmeters. Therefore, the ANFIS used here contns totl of 37 fttng prmeters, of whch 8 (*7 +*7 = 8) re the premse prmeters nd 343 (7*49 = 343) re the consequent prmeters. Fg.8. Input nd Output Membershp functons before trnng Fg.9. 5-Lyer ANFIS model structure wth nputs & output DOI:.979/ Pge

8 Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System Fg.. Membershp functons for E nd DE nd Secton of ANFIS Rule vewer fter trnng Fg. shows optmzed membershp functon for E nd DE nd the rule vewer fter trnng. Thus n dptve network tht hs exctly the sme functon s Sugeno fuzzy model hs been constructed. In Fg., the trned ANFIS controller wth testng nd checkng dt plots re shown. Surfce plots showng reltonshps between nput nd output prmeters before nd fter trnng re gven n Fg. to help vew the control surfce. Fg.. Trnng, Testng nd Checkng ANFIS controller Fg. 3D-Surfce plot before trnng (left) nd fter trnng (rght) The steps for MATLAB smulton re summrzed s:. The SIMULINK model lyout creted (Fg.3) s opened nd fuzzy edtor nvoked by typng fuzzy.. The Mmdn FLC nd Tkg-Sugeno (TS) ANFIS bsed.fs fles (nmed mmdn, nfs3) re loded to the fuzzy edtor nd then exported to workspce. 3. Ech of the 49-rule bsed Mmdn FLC nd TS ANFIS models re run nd the smulton stopped fter seconds. 4. The creted MATLAB m-fle s executed to generte trnng, testng nd checkng dt used n rto 8::. 5. The vrbles x=poston error, y=error rte, z=output nd three other vrbles [,, 3 ] re selected to hold dt from step (4) n excel felds ech mde up of the correspondng three columns. 6. The nfs edtor s opened by typng commnd nfsedt nd the number nd type of Membershp functons re specfed. 7. Vrble [, ] s loded from workspce, grd prttonng s selected s FIS generton method, number of epochs s set (between 4 nd ) nd tolernce.. Trnng s done usng hybrd lgorthm method. 8. The.fs ANFIS fle s exported to workspce nd the SIMULINK control model (ANFIS model) run. Ths trned the Adptve Neurl-Network. 9. The testng nd checkng dt sets re loded, n turns, to the ANFIS.fs fle usng [, 3, ] from workspce nd performnce s tested nd vldted respectvely. DOI:.979/ Pge

9 Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System. The performnce chrcterstcs re observed nd nlyzed nd the desgn process s stopped once desgn constrnts/objectves hve been fulflled. VI. Smulnk Model The SIMULINK model for the desgned NFSC ncludng PID nd FLC controllers creted n MATLAB softwre for conductng the smultons s shown n Fg. 3. The FLC controller nd MATLAB code were used to generte the requred trnng, testng nd checkng dt. The nputs of the controller re tken s reference poston represented by step nput sgnl nd the ctul poston obtned from the ctul output sgnl feedbck. The output s the drvng voltge to the motor drver. The prototype crcut dgrm constructed wthn Proteus 8 Professonl Smultor envronment for softwre mplementton s s shown n Fg.4. Fg.3. PID, FLC nd NFSC (ANFIS) Smulnk Model Lyout Fg.4.Crcut Dgrm for ntenn poston control wth NFSC (ANFIS) VII. Results nd Dscusson The proposed NFSC system hs been studed by smulton. Further, t hs been compred wth the conventonl PID controller nd FLC n order to evlute ts performnce n presence of sturton nonlnerty nd to vldte the ccurcy of the desgn. In Fg. 5, Fg. 6 nd Fg. 7, the outputs of ech controller.e. PID, FLC nd ANFIS (before trnng) re respectvely shown whle Fg.8 shows the NFSC output fter trnng hs been conducted. From ths, t ws observed tht the ANFIS controller output ws the best n terms of fster rse tme, settlng tme nd mpltude stblzton. Fg.9 shows system response wth PID controller n whch t s seen tht the response s not tht good owng to hgh overshoot nd ncresed settlng tme. DOI:.979/ Pge

10 Ampltude (Volts) Ampltude (Volts) Ampltude (Volts) Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System 4 PID Controller Output Tme (Seconds) Fg.5. PID Controller Output.7 FLC Controller Output Tme (Seconds) Fg. 6. FLC Controller Output.4 Output of Untrned NFSC Tme (Seconds) Fg. 7. NFSC (ANFIS) Controller Output before trnng DOI:.979/ Pge

11 Ampltude (Volts) Ampltude (Volts) Ampltude (Volts) Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System.4 Output of NFSC After Trnng Tme (Seconds) Fg. 8. NFSC (ANFIS) Controller Output fter trnng.4 Antenn Control System Response wth PID Control Tme (Seconds) Fg.9 System step response wth PID controller Fg. shows the Mmdn type FLC response to step nput whch shows no overshoot but the rse tme s comprtvely longer. Fg. gves system response by usng the NFSC pror to trnng whle Fg. shows NFSC response fter trnng. Fg. 3 shows combned plot to compre system responses usng PID, FLC nd NFSC (before trnng) nd n Fg. 4, fter trnng the NFSC..4 Antenn Control System Response wth FLC Tme (Seconds) Fg.. Mmdn Type FLC step nput response. DOI:.979/ Pge

12 Ampltude (Volts) Ampltude (Volts) Ampltude (Volts) Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System.4 Antenn Control System Response wth Untrned NFSC Tme (Seconds) Fg. System response usng NFSC before trnng..4 Antenn Control System Response wth trned NFSC Tme (Seconds) Fg. System response usng NFSC fter trnng..4 Antenn Control System Response wthpid, FLC nd Untrned NFSC PID. Untrned NFSC FLC Tme (Seconds) Fg. 3. Response of PID, FLC nd NFSC (before trnng) DOI:.979/ Pge

13 Ampltude (Volts) Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System.4. Antenn Control System Response wth PID,FLC nd trned NFSC PID Trned NFSC(ANFIS) FLC Tme (Seconds) Fg. 4. PID, FLC nd NFSC response fter trnng NFSC Tble 3: Smulton Results wth Vrous Step Inputs PID FLC NFSC-Before Trnng Step t r t s M p t r t s M p t r t s M p Input (V) (s) (s) (%) (s) (s) (%) (s) (s) (%) Av NFSC Response After Trnng Step Input t r t s M p (V) (s) (s) (%) Av Key: t r- rse tme, t s- settlng tme nd M p- percentge overshoot, Av=Averge In ddton, the system response ws nlyzed usng the NFSC, FLC nd PID controllers when dfferent vlues of reference step nput sgnl correspondng to the desred postons were ppled. The correspondng results were studed nd the estmted tme response chrcterstcs before trnng the NFSC nd fter trnng hve been summrzed n Tble 3. By exmnng the results, t s seen tht NFSC provdes response wth fster settlng tmes nd mnmzed rse tme lbet the presence of sturton nonlnerty. From the verge vlues n Tble 3, NFSC recorded the best verge performnce. The rse tme s 3.sec for Mmdn FLC model,.sec for the Tkg- Sugeno NFSC model both pror to nd fter trnng nd.8sec for the PID control. The mxmum overshoot s.% wth PID, none wth the FLC nd.% wth NFSC (before trnng) nd only 4.% fter trnng due ts lernng blty. The settlng tme s 3. sec. for FLC,.6 sec nd.7sec for the NFSC before nd fter trnng respectvely nd 4.6 sec for the PID control. It cn be understood from ths study tht whle usng NFSC the system tends to pproch nd settle t the desred poston n the fstest tmes (.sec nd.7sec respectvely) s compred to both PID nd FLC controller. Ths mens tht the rte t whch the error between desred ntenn poston nd ts ctul poston cn be performed by the NFSC lgorthm wthn the lest mount of tme nd thus llowng for contnued estblshment of drect lne of sght for qulty communcton between the prbolc dsh nd the selected stellte. The PID controller prt from ts hgh overshoot tkes longer to rech DOI:.979/ Pge

14 Desgn of Neuro-Fuzzy System Controller for DC Servomotor-Bsed Stellte Trckng System stedy stte. Although no overshoot ws regstered wth FLC, t lcks the lernng blty offered by the NFSC whch hs been cqured t the expense of 4% overshoot ntroduced n stedy stte vlue but whch s stll wthn the desred lmt of less thn %. VIII. Concluson The objectve of ths pper to desgn NFSC Controller for DC servomotor hs been cheved. Ths ws verfed through smulted output responses of the system to step nput sgnl whch stsfed the desgn crter. The FLC lone ncresed both the rse nd settlng tmes by few seconds but decresed the overshoot sgnfcntly to zero degrees. Usng PID, FLC nd ANFIS pproch t s seen tht ANFIS provdes the best performnce n terms of stblzng the system response n much shorter tme wth mnmum rse tme nd overshoot. Therefore, the objectves were met n both desgn nd softwre smultons. Future Work Further work my focus on the hrdwre pplcton of the developed NFSC lgorthm to cheve utomtc nd rel tme stellte trckng wth prbolc ntenn. References [] J.K. Km, K.R. Cho nd C. S. Jng, Fuzzy control of dt lnk ntenn control system for movng vehcles," ICCAS, June 5. [] T.V. Ho, N. T. Xun nd B. G. Duong, Stellte trckng control system usng Fuzzy PID controller," VNU Journl of Scence: Mthemtcs nd Physcs, vol. 3, no., pp , 5. [3] M. N. Soltn, R.Zmnbd nd R. Wsnewsk, Relble control of shp-mounted stellte trckng ntenn," IEEE Trnsctons on Control Systems Technology, p. 99,. [4] L. Xun, J. Estrd nd J. D Gcomndre, Antenn zmuth poston control system nlyss nd controller mplementton," Desgn Problem, July 9. [5] M. Ahmed, S.B. Mohd Noor, M. K. Hssn nd A.B.Che Soh, A Revew of Strteges for Prbolc Antenn Control," Austrln Journl of Bsc nd Appled Scences, vol. 8(7), pp , My 4. [6] N.S. Nse, Control System Engneerng," John Wley & Sons, 6th Edton,. [7] L.A. Aloo, P.K.Khto nd S.I. Kmu, "DC Servomotor-bsed Antenn Postonng Control System Desgn usng Hybrd PID- LQR Controller", Europen Interntonl Journl of Scence nd Technology, Vol. 5 No., Mrch, 6. [8] A.K. Pndey, Speed control of DC servomotor by fuzzy controller," Interntonl Journl of Scentfc nd Technology Reserch, vol.,. [9] A. Mehmet nd T. Isml, Moton controller desgn for the speed control of DC servo motor," Interntonl Journl of Appled Mthemtcs nd Informtcs, vol., no. 4, pp. 3, 7. [] P.K. Khto nd J.N. Nderu, Drect Neuro-Fuzzy Approch to Moton Tunng of Servomotors, Keny Socety of Electrcl nd Electronc Engneers (KSEEE) nd Jpn Socety of Appled Electromgnetc nd Mechncs (JSAEM),. [] R.S.Burns, Advnced Control Engneerng, Butterworth-Henemnn, OXFORD, pp ,. [] J.S.R.Jng, ANFIS: Adptve Network Bsed Fuzzy Inference System, IEEE, Trnsctons on system, Mn & cybernetcs, vol. 3, 993. [3] B.A.A. Omr, A.Y.M. Hkl nd F.F.G. Areed, Desgn dptve neuro-fuzzy speed controller for n electro-mechncl system, An Shms Engneerng Journl (ASEJ) vol. pp. 99-7,. [4] J.P.Llj, P.M. Krstn nd M.H.C Alfonso, NEURO-FUZZY CONTROL, Soft Computng, Unversty of Icelnd, 5. [5] S.Ghdr, S.Jvd nd F.Frokh, Decresng Strtng Current for Seprte Excted DC Motor usng ANFIS Controller Interntonl Journl of Smrt Electrcl Engneerng, Vol., No., Sprng 3. [6] H.T. Nguyen, N. R. Prsd, C. L. Wlker nd E. A. Wlker, A Frst Course n Fuzzy nd Neurl control, by Chpmn & Hll/CRC, 3. DOI:.979/ Pge