TAP CHl KHOA HQC VA CONG NGHE - SO - TRLfONG DAI HQC PHAM VAN DpNG DIEU KHIEN VI TRI DQNG CO DC SC DUNG BO DIEU KHIEN MANG NO RON Nguyin Dinh Hoang'" TOM TAT Bdi bdo gidi thieu phuong phdp dieu khiin vi tri ddng co DC sic dung mang no ron Bp diiu khiin Id thich nghi vd hoat dpng dua ti-en thuat todn hii qui dich chuyin diu phi tityen (nonlinear autoregressive moving average) Kit qud mo phdng phin mem MATLAB vd md hinh thuc nghiem cho thiy hiiu qud vd lpi ich cua hi thing diiu khiin vi tri ddng ca DC bdng mqng no ron so vol bp diiu khiin PID thdng thuong l.B^TVANDi: Do dac tinh didu khidn tdc dp va vi tri cua ddng co DC rit tdt nen nd dupe six dung rat rpng rai cong nghiep mac dCi chi phi bao tri va gia cao hon ddng ca AC Bd didu khien thdng dung dimg dieu khidn vi tri va tdc ddng co DC la bd didu khien PID Tuy nhien, bd didu khidn PID cimg cd mdt sd han che vide didu khiin cac ddi tupng cd tham sd thay ddi hay cac ddi tupng phi tuydn De nang cao chit lupng didu khien, bai bao trinh bay mdt phuong phap didu khien thdng minh dua tren thuat toan hdi qui djch chuyen ddu ciia mang no ron nhan t^o de dieu khidn vi tri ddng ca DC Vdi dac diem nhan dang va duac huan luyen dap ling ngd ciia ddi tupng nen bd dieu khien sii dung mang no ron dimg dieu khidn vi tri dpng ca DC tdt hon so vdi bd dieu khien PID thdng thucmg DOI TUQNG VA PHUONG PHAP NGHIEN CUtT 2.1 D6i tuong nghidn cuu Thidt kd bp dieu khien vi tri ddng ca DC sii dung mang no ron dua tren thu$t toan hdi qui dich chuydn ddu phi tuydn (nonlinear autoregressive moving average) 2.2 Phuffng phap nghien cuu Nghien ciiu ly thuyet va th\rc hidn bd dieu khidn na ren tren chip vi didu khidn dsPIC30F4011 KET QUA NGHIEN CUtJ 3.1 MS hinh toin hpc ciia ddng co DC kich t i flde lap Dap ling ddng hpc cua ddng ca DC kich tu ddc lap dupe md ta bdi cac phuong trinh sau: V(t) = R i.(t) + L.[di,(t)/dt] + eb(t) Tm(t) =J[d^e/dt^] + B.de/dt (1) (2) T„ =KT.i eb(t) = Kb.[de/dt] (3) (4) • JOURNAL OF SCIENCE AND TECHNOLOGY • 69 TAP CHl KHOA HOC VA CONG NGHE - SO - TRU&NG OAI HQC PHAM VAN OONG Trong V,(t) : dien ap diu vao phan ling (V) B i,(t) ddng dien phin ling (A) Ra : dien trd phin ung (Q) J : Moment quan tinh (Nm ) he sd ma sat (Nms) La dien cam phan ling (H) T„(t): Moment ddng co (Nm) Cb : Slie phan didn ddng (V) KT : dich chuydn gdc (vi tri) cua true ddng ca [rad] he so moment (Nm.A"') Kb : he sd siic phan dien ddng(V/rad/s) Til cac phuong trinh (1), (2), (3), (4) ta cd thd viet ham truydn vi tri ciia ddng co DC nhu sau: e(s)/V(s) = Kb/[J.La.S'+(R J+BLa)S^+ (Kb^R,B)S] 0) Cac tham sd ciia ddng co : R.=2.45 ohm ; La=0.035H ; Kb=l,2v/(rad/s); J=0.022Kgm^/rad ; B=0.5.10''Nm/(rad/s) Thay vao (5) ta cd : e(s)A'(s) = 1.2/[0.00077s'+0.0539sV 1.441s] (6) 3.2 MS hinh FID dieu khien dong cor DC Dd didu khien chinh xac vi tri dpng ca DC bd didu khidn thudng dupe sii dung la PID U"u diem cua bd dieu khidn PID la dan gian, de thuc hidn Tuy nhien, neu thiet ke bo dieu khidn khdng chinh xac thi cd the dan den he thdng khong dn dinh Md hinh bd dieu khien nhu hinh PID mput DCMotor Measurement Hinh Md hinh diiu khiin ddng coDC sic dung bp dieu khiin PID 3.3 MS hinh dieu khien ddng cff DC diing m^ng noTon Vdi ddng CO DC hoac cac ddi tupng didu khidn phi tuydn, md hinh toan hpc dupe cc the bieu didn dudi dang khdng gian trang thai nhu sau: x' = f(x) + g(x).u y = h(x) 70 • JOURNAL OF SCIENCE AND TECHNOLOGY • (7) _TAP CHi KHOA HQC VA CpNG NGHE - S6 - TRLfQ-NG DAI HOC PHAM VAN D O N G Trong dd, f(x) va g(x) la hai ham cd tiie dugc xac dinh hoac khdng xac dinh Neu cac ham khdng xac dinh thi bd di6u Idiien PID truyen thdng khd cd tiie hieu chinh dung cac tham sd de dap ijcng dau ciia ddi tugng bam theo tin hieu dau vao ©i6u dugc giai thich la cac ham f va g cd chda cac yeu td khdng xac dinh, vi dy nhu su thay ddi cua tai trgng Tuy nhien, vdi hg dieu khien mang no ron, nd cd the udc lugng cac ham phi tuySn Bd dieu khiin s6 ludn cap nhat cac thdng sd ciia ddi tugng va vdi mot thuat toan hoc thich hgp thi su hdi tu cua sai lech ham myc tieu se tiin din v§ khdng Vi vay, dk thuc hien dugc cac y6u cau tren, bd dieu khien thich nghi gian tiSp dua tren mang ncr ron hdi tiep tuySn tinh hda (NNFBL: Neural Network Feedback Linearization) dugc lua chgn Hai budc ckn thuc hif n vi?c thi^t kd bd di^u khiSn thich nghi gian ti6p dua tren mang no ron hdi tiep tuyen tinh hda gdm: nhan dang md hinh he thdng va thiSt k§ bd diiu khiin 3.3.1 Nh^n d^ng h€ th6ng Nhan dang he thdng bao gdm viec thu thap mot tap dii lieu huan luyen thdng qua thuc nghiem Dieu dugc thuc hiSn bang each cho cac gia tri d ngo vao thay ddi va thu thap cac dii lieu dap ung d dau ciia he thdng Tap dii lidu tuong quan d ngo vao va sau dd se dugc su dung huan luyen cho mang no ron de udc lugng dgng hoc cua he thdng Qua trinh nhan dang he thdng dugc md ta d hinh 3; dd, u la dien ap dau vao, va ® la cac ngo cua ddi tugng va md hinh mang no ron tuong ung Tin hieu sai lech ciia ngd dugc su dyng de huan luyen mang no ron udc lugng he thdng Ddi tugng Md hinh mang no ron Sai lech Thuat hgc Hinh Mo hinh diiu khiin dpng caDC su dung bp diiu khien PID Toan bd qua trinh nhan d ^ g h? thdng gdm: Thuc nghiem trSn md hinh de thu thap dii li$u vao/ra, lua chpn ciu tnic md hinh, md hinh udc lugng va md hinh hgp le 3.3.2 Chu4a bi s6 li^u thy-c nghif m Trong giai doan thyc nghi?m, mdt tap cac sd lieu vao/ra dugc thu thap thdng qua qua trinh th\rc nghiem vdi md hinh didu khiin Dd dam bao tinh chinh xac, thdi gian lay mau nen chpn dii nhd T^p dii li$u cd the bieu dien theo phuang trinh Z'' = f{[u(k),y(k)];k=l,2, ,n} (8) • JOURNAL OF SCIENCE AND TECHNOLOGY • 71 TAP CHl KHOA HOC VA C N G NGHE - SO - TRLfONG DAI HOC PHAM VAN B O N G day, u(k) la tin hieu vao, y(k) la dap ling ngd ra, k la mau tuc thai va N la tdng so mau can lay 3.3.3 Lua chon cau triic mo hinh Dua vao md hinh toan hoc ciia he thdng nhu md ta d phuang trinh 7, cau tnic mo hinh mang no ron phai cd kha nang uac lupng dupe ham f(.) va ham g(.) Vi vay, md hini mang no ron hdi qui dich chuyen ddu (NAMAR) dupe sii dung dd ap ling ydu cau D^ng rdi rac ciia nd dupe bieu dien dudi dang : y(k+d) = F [ y(k), y(k-l), , y(k-na +1), u(k), u(k-l), ,u(k-nb +1)] (9) d day, F la ham phi tuydn, d la dd tri ciia he thong, na la sd ngd qua khii va nb la sd ngd vao qua khii Mang dupe huin luyen qua trinh nhan dang dS xip xi ham F Bing each khai tridn chuoi Taylor cua ham F, phuong trinh 3.4 trd th^mh : y(k+d) f [y(k), y(k-l), , y(k-na+l), u(k), u(k-l), ,u(k-nb+l)] + g[y(k),y(kl), ,y(k-na+l),u(k-l), ,u(k-nb+l)].u(k) (10) Phuang trinh (10) co dang gidng phuang trinh ddng hpc cua he thdng Mang no ron hdi tidp tuydn tinh hda hoat ddng gipng nhu cac md hinh h6i tiep tuydn tinh hda chuan vdi muc dich triet tieu eac phin phi tuyen Iam cho he thdng trd nen tuyen tinh Dd thiic hien didu nay, hai mang no ron ML? (Multi-Layered Perceptron) dupe huan luyen dd xap xi ham %.) va g(.) nhu sau : y"(k+d) = r [y(k), y{k-l), , y(k-na+l), u(k), u(k-l), ,u(k-nb+l)] + g^[y(k),y{k1), ,y(k-na+l),u(k-l), ,u(k-nb+l)].u(k), (11) Trong dd, f^ va g'^ dupe dimg de xap xi ham f va g Til phuong trinh 11, bp didu khidn duac thidt kd dd bam theo tin hieu tham chidu cho trudc dupe xac djnh nhu sau : „(j.^ ^ yrik + d)-fly(k),y(k-).), ,y(k-n + l),uik),u(k-l), ,u(k-n + l)] sbiklyik-l) y(k-n + l),u(,k-l), ,u(k-n-l-l)] ( jj) Tuy nhien, luat didu khidn theo 12 khdng kha thi vi phai xac dinh cd u(k) va y(k) tai cimg thdi diem Vi vay, thay cho mo hinh trdn ta sii dung md hinh: y(k+d) = f [ y(k), y(k-l), , y(k-n+l), u(k), u(k-l) n+l),u(k-l), ,u(k-n+l)].u(k+l) Vdi d>2 u(k-n+l)] + g[y(k),y(k-l), ,y(k- Luat dieu khidn trd : „f^,^,^^yril^ + d)-fbm.y(k-i).-.yik-n-l-\),u(k).u(k-l), ,u(k-n sbW, y(k -1), , y{k-n +1), a(t - \), ,u(k - ô +1)] 72 ã JOURNAL OF SCIENCE AND TECHNOLOGY • + l)] ^jj^ TAP CHl KHOA HQC VA C N G NGHE - S O - TRU'ONG OAI HOC PHAM VAN DpNG 3.3.4 Mjng no- ron ML? (Multi-Layered Perceptron) xap xi ham Dd xip xi ham phi tuydn f va g, md hinh sii dimg mang Perceptron nhidu lop (MLP) Mang MLP gdm ba lap vdi mpt lop vao, mpt lap an va mdt lop vdi ham sigmoid d lap an va ham tuydn tinh d lop cd the xap xi mdt ham bat ky D^c diem cua mang MLP la cd kha nang huin luyen nhanh Thuat toan truydn nguac (BP: Back Propagation) dupe su dung dd hieu chinh cac trpng sd W ciia MLP nham cue tieu hda ham muc tidu So dd khdi vdng kin ciia bd didu khien no ron hdi tiep tuydn tinh hda dupe trinh bay d hinh Hinh Sa dd bp dieu khien no ron hdi tiip tuyentinhhda 3.4 Ket qua mo phong 3.4.1 So Simulink MARMA-U Controller H Random Reference ^ Cloci K 12 • Pi - • DCMOTOR Hinh SadSmd phong ven bp dieu khiin NAMAR L2 • JOURNAL OF SCIENCE AND TECHNOLOGY • 73 TAP CHi KHOA HQC VA CONG NGHE - SO - TRLfQ'NG DAI HOC PHAM VAN P P N G 3.4.2 Tham so md phdng Sdnaronldpan Gia tn Idn nhat ngo Thdi gian lay mlu 0.01 s Gia tri nhd nhdt ngd Inf -Inf Sd ngd vao trS Sd the h? huan luyen 40 Sdngdratrl Sd mlu huin luyen 10000 Thuat todn huan luyen: Levenberg_Marquaidt Gia tri Idn nhat ngd vao Gia tri be nhat d ngd vao -1 3.4.3 Ket qua mo phdng K.k qua md phdng bd diiu khi§n mang na ron dugc thuc hien tren phan mem MATLAB Sai lech hudn luyen theo ham muc tieu la 1.6934e-010 Dap ling ngo ciia dpng CO so vdi tin hiSu mong mudn d ngd vao dugc bieu dien tren hinh I m ã MRW) Ơ Hinh Ddp ung ngo (6) cua DC Motor 3.5 Ket q u a t h y c nghiem Md hinh diSu khidn ddng co su dung chip dsPIC30F4011 cua hang MICROCHIP KSt qua thirc nghiem cho thay dap u n g vi tri ngo ciia ddng ca vdi vi tri dieu khien ngo vao thdng qua p h i n m i m TERMINAL IN/OUT la kha tdt, thdi gian dap ung nhanh, sai so khoang +/- d i n xung, vgt Id khoang % MOt sd ket qua diSu khien vi tri dugc cho bang Xung Xung ENCODER (hex) (degree) ENCODER (hex) (degree) 0x0000 0x0640 360 0x0320 180 OxOeSO 720 Bdng Cdc vi tri diiu khiin dpng ca vai cdc xung encoder tuong ung • JOURNAL OF SCIENCE AND TECHNOLOGY ' TAP CHi KHOA HOC VA C N G NGHg - SO - TRUONG DAI HOC PHAM VAN B O N G K E T LVA.ti Ket qua md phdng Matlab va md hinh thuc nghiem didu khidn dpng cp tren chip dsPIC30F4011 ehe thiy dap ling ngd ciia ddng ca DC thep vi tri dat cd dac tinh tdt hon so vdi bd didu khien PID vd dp vpt Id, thdi gian dap ling, dao ddng Mat khac, bd didu khidn duac huin luyen vdi tap da lieu cho trudc ndn dap ling ciia he thdng van bao dam dupe ydu cau eac thdng sd ciia ddi tupng didu khidn thay ddi TAI LIEU THAM KHAO [1] Pham Huu Diic Due, Nguydn Cdng Hidn, Nghien cicu img dung mang no ron md diiu khiin thich nghi robot hai khau, Tuyen tap cac bae cao khoa hpc tai hdi nghi toan qudc lin thii VI vd tu ddng hda, 107-112, (2005) [2] M.Norgaard, O Ravn, N.K Poulsen and L.K Hansen, Neural Network for Modelling and Control of Dynamic System, Springer, (2000) [3] Fukuda,T., and Shibata,T.,raeory and applications of Neural Networks for industrial control system, IEEE Transaction on Industrial Electronics, 39, 1992472-489 [4] Halici, Ugur Artificial neural netwwork, Ankara, (2005) [5] Neural Netwwork Toolbox User' Guide - Matlab Sunulmk, (2007) [6] dsPICSOFdata sheet Microchip Technology Inc, (2005) [7] MPLABIDE User' Guide Microchip Technology hic, (2005) POSITION CONTROL OF DC MOTOR USING NEURAL NETWORK CONTROLLER Nguyen Dinh Hoang ^ ABSTRACT This paper introduces a method for position control of BC motor using neuron network The controller is adaptive and based on a nonlinear autoregressive moving average (NARMA-L2) algorithm Simulation results in Matlab and the experimental model are presented to demonsti-ate the effectiveness and advantage of control system of DC motor with ANNs in comparison with conventional PID controller Keywords DC motor, neural networks, NAMAR_L2, conO-ol system C8SL) '"Khoa Ky thuat-Cdng nghe, Trudng Dai hpc Pham Van Ddng; email: nguyenhoang64@gmail.com • JOURNAL OF SCIENCE AND TECHNOLOGY • 76 ... khien nhu hinh PID mput DCMotor Measurement Hinh Md hinh diiu khiin ddng coDC sic dung bp dieu khiin PID 3.3 MS hinh dieu khien ddng cff DC diing m^ng noTon Vdi ddng CO DC hoac cac ddi tupng didu... ciia ngd dugc su dyng de huan luyen mang no ron udc lugng he thdng Ddi tugng Md hinh mang no ron Sai lech Thuat hgc Hinh Mo hinh diiu khiin dpng caDC su dung bp diiu khien PID Toan bd qua trinh... luyen cho mang no ron de udc lugng dgng hoc cua he thdng Qua trinh nhan dang he thdng dugc md ta d hinh 3; dd, u la dien ap dau vao, va ® la cac ngo cua ddi tugng va md hinh mang no ron tuong ung