Industrial Applications of CAD-Packages for Control Systems in the USSR

Copyright © IFAC Computer Aided Design in Control and Engineering Systems, Lyngby, Denmark. 1<,)85 INDUSTRIAL APPLICATIONS OF CAD PACKAGES FOR CONTROL SYSTEMS IN THE PEOPLE'S REPUBLIC OF CHINA Chen Zhen-Yu Automatioll Research Institute uf Metalllllgical /VIil/istl)', Peuple's Republic of China Abstract. This paper introduces the development of CAD techniques for control systems in P.R. China. A general description has been made about the industrial applications of CAD packages for control systems. Some examples have been illustrated about the MIMO discrete-time systems, LQG systems, SISO control systems, self-tuning control systems and adaptive prediction systems. Some prospect ideas have been described in conclusion. Keywords. CAD package; industrial automation; control systems. 1. INTRODUCTION The techniques of CADCS are gOing to progress and also to carry forward the ap plications in P.R. China. Following, I will introduce the general situations about the applications of CADCS. The techniques of CAD for control systems had been developed gradually at the seventies in P.R. China. At the initial stage, a lot of calculating programs of single and multi variable control systems as well as simulation softwares had been constructed to aid for design and analysis of control projects (Chen, 1979; Han, 1979; Liu et al. 1979; Wang, 1972; Wang et al., 1979; Wu, 1978; Xu et al., 1976 ). Those softwares were special kinds of programs which must were used by experts and were alw 'lYs wri tten by themselves. Anyone who wanted to use those programs, he must read over the whole programs which always contructed as subroutines, and must wrote a main program as well as the another subroutines which can described the appointed jobs. These circumstance got some limitations of the possibility to spread the applications of control technology into the engineering areas, especialy the modern control theory contained a lot of complex algorithms, and the engineers got difficulty to be familiar with all the details of algorithms. When entering the eighties, through the domestic and foreign technical exchanges for which some activations were orgenaized by Chinese Association of Automation, the CAD techniques of control systems have developed more quickly. A lot of CAD packages have been constructed in the universities and institutes, which have the features including management systems to dispatch all the functions within packages (Bao et al., 1982; Chen et al., 1985; Gan, 1982; Gao et al., 1984; Han, 1980; Han, 1982; Han, 1983; He, 1982: Lu et al., 1984; rluan, 1982; Tu et al ., 1982; Wang et al., 1984; Xiong et al., 1982). There have interactive dialog system to simplify the operations of the users, who can get annotations from packages during every operating step. And also have command system to support the efficient running which looks like batch process, even the users want to compare the results of different choices of means. MIMO Discrete-time Systems The linear multivariable stochastic system can be described as (Ruan, 1982) y(k)=A l y(k-l)+., , +A p y(k-p )+B u(k-l-d) o +··,+B p u(k-p-d)+e(k) (1) where noise e(k)=V(k)+A 1V(k-l)- .. -ApV(k- P ) +B OW(k-d)+B 1 W(k-l-d)+ . , . +BpW(k-P-d) (2) The equivalent state space equation can be described as x(k+l)=Ax(k)+Bu(k-d)+BW(k-d) (3) and the measuring equation y(k)=Hx(k)+Bou(k-d)+BOW(k-d)+V(k) (4) The optimal filter of state x(k) which may satisfies the equation (3) and (4) can be described as x(k)=~/-l+KY (5) where Kk is the gain matrix of Kalman fil ter and AY(k)=y(k)-Hx(k/k-l)-B oU(k-d) x(k/k-l)=Ax(k-l)+Bu(k-l-d) The quadratic performance criteria s T T J(u)=E!: [x (k)Qox(k)+u (k-d)U(k-d)] (6) k=d and the steady-state optimal feedback control d-l u(k)=Ld~+ L M.u(k+i-d) (7) i=O l where Ld and Mi are the functions of steady state optimal feedback coefficient matrix L. 2. APPLICATIONS OF CADCS A MU10 discrete-time CAD package had been constructed, and the main features of the 69 70 Chen Zhen-Yu algorithm functions are desc ri bed as above . This CAD package had been used for project design of optimal control for cove r- type anneal i ng furnaces (Ruan et al ., 1981). These furnaces were used for annealing of alloy steel coi l s. The scheme of furnace i s illustrated as Fig . 1. ._-_. - - - -------, --...._---- -- - , ' -. outer c ov er ---.-- 0;"" -• inner cover ----------0- .--.-----, r - '- -- - c. ,---:2 ~ steel co lls ~ electri c -' heaters ...... or 0'0 ;:> te mperature sens ors o .,"o " .-~ .i ~ ~ -I () t i i I. Fig . 1. Scheme of annealing furnace The furnace was he ~ ted by electric power , and can descr ibed a s a 3 input 3 output system. The practical mea s uring had been made for collecting the step response data series of temperature outputs yl, y2, y3 . Then used the CAD package to carry out the identification and modelling, t~e calculating procedures i llustrates as Fig . 2 . L ~ - r:~ ' : ! I - ' - '- - d;t~ - ir;put- ;" - ~ ; :;~ ~ meth od method est 1ma te parameters -- - A, -- Ap, B() ,B, '''Bi' J __ _ 0[-- --1 I - -' d~t - ~ine t h: orciers L.model using Ale criteria ... __.. i _.... __._ ___._._ _L_.. .. _____ .. _ ~ 'L estimate noise covariance matri x Q,R -- -- .. - -. - - - - ----- -- - - ... ~- :;~'r[:::::: _ _____ J -: :--------::-.-=-cient matrix L,Ld,M .~ . I check t he control effect by s imula...ti on run n ing I I ~'ig ;" ~ '";O~ . i 2. " 0 0- r~u J C:_~) '['he cal cul a ting procedures of i d en d ificati on and ~ o deling When the process model haa been constructed , the CAD packa ge can carried on Simulation study ~ s well as real time control running , which is illustrated a s Yig . 3. This furna ces control systems had been implemented i n Shanghai Iron an d Steel Factory , and have got more fine performance to satisfy the production o bjects the n common instru mental systems . Fig . 3 . The bl ock diagram of real time control running This CAD package also had been used for project deSign of temperature control systems of creep - testing furneces (Ruan et al . 198 1). The control oriciple a nd sy s te m con struction were Similar a s the above exarnole and the oractical control performance we~ al s o satis f actory . The another ex a mple of using this r;AD pac kage was the project de si p n of electric power control syte~ of electroslag remelt ing process (:tu a n et a1., 198 4) . Ev'On that wa s a singl e variable system, the design as wel l as the practical control performance had got fine results . LQG Systems The Linear- Quadratic - Gaussian problems general l y can be described as state space epuation and me a suring equation (Gan et al . 19R2) ('n x ( t)=Ax(t)+Bu(t)+w(t) y(t)=Cx(t)+v(t) (g) The quadratic performance criteria J=E [xT(tf)?X f )+ J : ~ (XTt) ~ x(t) +u T ( t ) ;~ u ( t ) ) d t ( 10 ) For LQ p roblems the optimal feedback con trol law is giv en by (11 ) u(t)=- K(t)x(t) K(t)=:{- I BT.?(t) (12 ) where .?(t) can be obt q in e d by s olving rliccati differential matrix equation - P(t)=P(t)A+AT?(t) - .?(t)BH- 1BTp (t ) (1 ')) +) -"'or steady- st a te th e eq ll ation (13) beco mes algebraic equation fo r m .?A+AT.2_.r'13 ;{- I B 'l'.r'+ '-b O (iLt) And the corresponding control law by i~ given 71 Industrial ApplicatioJlS of C ·""!) I'<lck<lgcs u ( t)=- Kx(t) ( 1 '5 ) K= .-C 1ET p ( 16) the resuming time t~0 . 8 s. For Kal man- Bucy filter, the calculating e quat i on i s g iven by ~(t)=A2 ) + BU (t)+G y(t) -x (t) J ( 17) The block di ag ram of LQG CAD package is i llustrated as Fig . 4. I L__man~g':. •I .--·~-= :, .. . -L~ I 1 I _ ~t - - .-0] F I ------'--- _ ! I I optimal I optimal lIIJ_ter data t;an"formation ~gulato [-----._ .... __.. _L___ ~ . - l -l __ L I .f inite ~ime LQ I r - Ri~cat , : Idifferentiali ~.J , L~ qua~o 1_d__ T I y!,."gram -;i~lat I program .~ - in .. finite.. 1 time LQ I ' H~ '~ ~: - i -1 1 algebraic . : I~qu a tion 1. - . __ ._ ....... __ ..... .I I ..L __ '-- -'1 Fig . 4. Th , The block diagram of LQG CAD package i~ I ~AD package had been used for project design of wide range speed regulating ser vo system ( Su et al. , 1983) as Fig . '5 . This CAD package also had been used for design of some ca r riage control sy t e ~s of vehicles. S13 0 '.;ontrol Sys tems Ther e ha ve wide spread demands to design the S1S0 control system in the real engi neering problems , therefore the 8130 GAD uac kage had been constructed which included a lot of usefu l algorithms of time domain and frequency domain methods ( Tu et al ., 19A2) . The appl ication examples have be en in many cases . One of the user ' s departmen t introducing some examples are illustrated as Fig . 7 , Fig . 8 (Li et al., '1983 ). Fig . 7 . Fig . 5. Sche!'1e of a nti-disturbance speed regulating system J input and output of d at a. and graph 1--_.-. .... .. .•....._.. . _-- .. "._-_._----"_."- " Fig . 6 . Scheme of wide range speed regulating servo system The speed r egulating r ange wa s 0 .1 rpm to 1'500 rpm , and the speed up time was ~17 0 ms from 0 to 1'500 r pm , the brake down time was ~1 70 ms from 1'500 to 0 rpm, the reversi ng time was ~2'50 ms from +1500 to -1 '500 rpm, the regula ting ov ershoot" 2%. The another exa!'1ple of using the L')G C; ,! Jl package was the project design of effective anti - disturbance speed regul ating system ( Gao et al ., 1983 ). The block diagram of sDeed regulating system is illustrated as ."'ig. 5 . There had added a state feedback t r ansient comcensator into ~he sys tem to i mprove the ant i-di sturba nce performance . When the step load of 10~h rate had been app li ed into this system , the danymic speed drop An,fO. 67~ I f :·Lt;'2~"·:r L Blo ck diagram of high p recision servo system ,'-- -- - ilt e-lfS+S" (,.. . l _.. - - - I )(1""0) ------ - ._---- - - - - - - - - - Fig . 8 . Block diagram or servo system of on line robot Al l of the use r s had apnreciat pd that th ~ y had s a ved many times for using CAD package then their own calculating , and had advant agA to comDare their d ifferent plans of design from variOUS viewpoints at a short 72 Che n Zhen-Yu period of computer runn ing. Equatio n (1 8 ) also can be rewritten as CAHMA form Self-tuning Control Syst ems Before the self-tuning CAD packag e had been constuucted, some projects of self - tun in g sys tems had been designed and applicated in laboratories and factories (Chen et a l., 1983 ; Deng et al. , 1982 ; Shu et al., 1982 ; Yuan et a1., 1980 ; 'tIu et a1. , 1984), some of which are illu strated i n Fig. g, Fig.l0. c ry s tal diamater measur~ 18: 1 ' 1" ~ temperature measurement ~ r 11 I u y controller Scheme of self- tun ing control system of silicon crystal drawing process y ,_____ ---, r_1 i ~- ' E , all ~:Fn u -Iheat.:~ - controller - g I&-_ y i - vaciUm-+ ~ I ~i + U t ) (21 ) where z (t ) are p dimention vector~ , u(t) are deterministic inputs of r dimentions, yet ) are measurable random inputs of m di mentions . The self - tuni ng recur s ive mul tistep predictor i s described as z(t+k /t) =~A(t)z+k-i/ ) +l + ~ D(t)y y(t + k/t)= r~ ~ (t) B(t)u+k-i (t)y + k-i/t) (t+k-i /t) ( 22) (23) ! where : This CAD package had been applicated for predi ~ tions of the productivity of oilfield , the product i on, worth and market ing of light industry commercial product.s etc ( Deng et al., 1983; Guo et al ., 1984 : Han, 1983 ; Han et al ., 1983). One example w~s for the oilfield , the prediction was on the bases of data, which was collected from 1972 to 1979 . The productivity prediction equation was i n time-varying parmet~ form, so first used th i s CAD package to predicte time- varying para meters , then predicted the productivity of oilfield from 1980 to 1981 . The predictio n results are illustrated in tabl e 1. I -.- electri c Fi g . 10. ~.oY'CX)(lO testing sample , e b 2'f (k) [y( k)-'f(k)T (k-l») (20) 1I'f(k)1\ where j (k) are measu ra bl e vectors , b is certain posit ive value . The first step i s go ing to predicte time-varying parameters by using equation ( 20), if the one level prediction of 9 (k ) are not sat i sfactory, then can be go ing to predicte mul ti-levels as s i milarly as before . The second step is go ing to predicte outputs of systems from equation (19) . + Another i s the sel f - tuning recursive multistep pr edi cti on method (Deng et a l., 19 83 ), wh i ch considers th e f"lIMO systems as A( q- 1 ) z ( t ) =B (q -1 ) u ( t ) +D( q -1 ) Y ( t ) 1--------1 s e i f -t un ing 1+------00--,-----' Fig . 9 . Y ( k )=~( k)T e(k)+v k) (19) For s ingle variable time-varying system, the est imated values e( k)= e( k-l ) sensors Scheme of self-tuning control system of elect ri c vacuun testing equipment Gradually the ~AD package of self - tu ni ng contro l systems have been constructed in universiti es and in st itutes (Cao et al ., 1984 ; Chen et a l., 1985 ; Yin et al . , 1984) . These CAD packages contain the algorithms of 81SO and MIMO self - tuning control sys terns , which include self-tuning regulator ; se l f - tun i ng controller; pole ass i gnment self-tu ni ng controller : all parameters sel f - tuning controller , and the feedforward terms a re app li ed into the a lgorithms. The algori th ms of parameter s estimation include l east square; extended least square: maximum like l ihood etc, and can use variable or nonva ri able forgetting factors methods . Adapt i ve ?rediction Systems The adaptive prediction CAD package contains t wo kinds of algorithms. One is the multi l evel progressive order method (Han , 1982), which conside r s the dynamic system model as y ( k) =f [Yk _ l , uk ,e, k) +v (k ) ( 18) where y(k) are outputs of n dimensions, u (k ) are inputs of p dimensions , e are timevarying pa r ameters , v(k) are random noi ses . are est imated parameters . Table 1. The productivity prediction of an oilfield - - ---- - - - - - - - - ---- - ---------- ------_ --::-=I real predi c ted : rel at. : : time p roductivity productivity ' error 1_____ ___ ~ _ t~_n _ ~ / day k t~ _ ~ s~day __ ;_- ~ i 41 . 97 42 . 22 · - 0 . 59 ' ' 1980.3 , 1980 . 6 41 . 21 41 . 63 -1 . 02 : : 1980 . 9 40 . 51 41 . 07 - 1 . 38 ! 39 . 87 40 . SS - 1 .7 I : 1980 . 12 11981.3 39 . 28 40 . 05 - 1.q'il I 1 38 . 72 3Q . S7 - 2 . 10 I 38.22 39 . 13 ; - 2~3' 1981.12 1 37 . 75 38 . 71 ' - 2 . 'i4 _ . .. _---..1 __ _ _ _ _ ____ _ _ __ __ ... _ .. ___. "____ - ' - _ 11981 . _6 ~ 981 :I' .9 The average error was 1. 72% . Another example was the prediction of pla~ tic production of a province. The predic tion results are illu strated in t able 2 . ­ Ind ustria l App licatiolls or C\ 1) l'acbgTs ' ~ao Tabl~ . The prediction of plastic productio n production time tons 1972 - -.--. -g: - 40-. 197) i L .. \~hen 1974 197" 1976 1977 1973 1979 1930 1981 inve stment k yuan , -·i2~0 - -: 12 . 43 14 . )2 14 . )2 18 . 8 1 2'i . SS 31 . 29 3 1. 67 33 . 60 37 . 'i0 l S)O 2170 2700 49':;0 2030 4990 1)20 3700 33SO .... _ _.._-- - - this CAD package had oeen used for predi~tng the time- varying parameters and the plastic production, the average predic tion error of 1980 and 1981 was 4 . 2% , whi ch was obtained by using the real data from 1972 to 1979 . The above descript~n show a part of industrial aDP licati ons of ';ADCS i n rriC, hoping that wiil g ive out a gene r a l i mpression about this fields . ~vel the techniques of ~ADC S have been de veloDed and applicated in some mai~ engi neering areas , but there are sti ll not fam ili ar ized by many engineers , so it needs m~kig efforts to popularize th is techniques for engineering app licati ons , through organizing seminars , technique exchan~s and another possible ways. The enginee ring experiences helVe been shown that the '; ADCS techniques become the bridg es between the theory and prat i ce , and it also can be for ecasted that these kinds of bridges will g ive more pl ay to act. on the engineering fields. We also hope to hold much more t.echn i que exchanges in the international range , and use the advanced eauipments as far a s possible . BeSides to develope and enri ch the contai ns of ,;AD';S oackages , the cOr.Jbination of control techni~us and sp e cial aDD licated techniques wil l provides another wide work ing area . tor example the biomedi cal identification CAD package have been const.ructed for doctors , who want to u se the identification techniques but can not car ry on any part icu lar calculations , this CAD package provides as confortable instru ments and have been aDpreciated by doctors . of disciDline of C.AJ};S i s a l so ;1 The ext~d broad p ro blem , which is not l imited into the enginee ri ng area , especial ly the control science have permeated into the SOCiOlogy , economics nowadays. Such as syte~ Dlani~g , s ys te'C Dro"ramming as well as la~ge - s cale system , there h~ve a l ot of problems p~ov i ding to be developed concerni':e' with C.-\.D'--;S and its appl ic ations . The techni~us of data bqsa and expert. system are 'llso '1se 1',,1 for ';AD(;S . bao ¥Il an - Lu , et al. ( 1992) . ') "D or" the multi variaole feedoRck svte~ ba sed on mo dern fre4uency do~ain methods . Inform~ ­ tion and ';ontrol, ifol.ll . No . S . 24 - 33 . 73 Shu - Guang , et a l. ( 1934 ) . CAD package of Belf - tuning control systems . Treatise of Shan Xi Mechiner So lle e . Che n lhen- ¥u . 1979) . Simulation study of automation systeT.s of cold tandem mills . ~ [etalurgy Automation, No.l. 9 - 19 . Ch en Zhen - ¥u, et al . (193)). Self- tuning regulator for roll eccent r icity of cold tan dem mills . 4th IFAC Symp. on Automation in Wining , Mineral and Metal Pro ceSSing , Aug . 22 - 25 , 1983 . Helsinki, r'inla nd . Chen l hen - ¥il, et al. (19 3S ) . Self- tuning CAD package and its ~pDlicato n . )rd IF ft ,; 'iymp . on Conpu ter Aided Design in ~ontrl and Engineering Systems , July 3 - Aug . 2 , 19as . Lyng by , Denmark . Deng lheng- Duan , et al. (19 82 ) . Self - tuning control of silicon crystal drawing Dro cess . ~eDort of Shan Xi Machinery Col lege . Deng li - Li , et al . ( 19B)). ~ltivar iable selt - tuning recursive mul tistep predic tor s and their applications. Acta Auto matica Sinica, ifol . 9 . No . 4 . 241 - 247 . Gan Hua-Lei . (1982) . LQGSP- .'\. ,;AD software package for the LQG systens . Information and ';ontrol , \{ ol.l1. No . i) . 11 - 1'1 . Gao Long, et al . (1933) . The LQ deS i gn of effective anti - disturbance speed regu l ating system . rieport of Automation Dep ar t ment of TSin~ University . Guo Yi - Xin, et al . (1984 ) CAD package for adaotive prediction . ,(epo rt of Heilongiiang Institute of ADp lied Mathematics . Han Jing-Qin g . (197<.) . ·;anonical form of linear systems and calculation of feedback systems . Annual I·-reeting of Theory ;om. ot' ';hines" ASSOCiat i on of Automation, nay . 10-17, 1979 . Xia Hen, ';hina. lhn Jing- )ing . (1 0 80 ) . The theory of multi va r iabl e control systems -- polynomial matrix operations Rna frequency doma in methods . Treatise of Institute o~ = tern s Scien~ Academia Sioi ca . Han lhi - Gang . (1 982 ) . I~lti - lev progres sive order method . Treatise of Heilong~In s ti t u te of . ~ . plied i"athematics . Han Zhi - Gang . (1 9~) . A new met.hod of dvnamiC system prediction . Acta Au torna tica ~";inc' , ifoL Cl . No . 3 . 161 - 168 . Han lhi - Gang , et al . (108)) . An adaotive forecast about the production , worth and ma~keting of l ight industry commerc i a l products . In format i on and Cont rol, Vol.12 . No . l . 14- 18 . He Huan-Xi . (198 2 ). The computer aided deSign software package of the multivari able control system . Information Rnd Control , Vol . ll . No . 6 . SO - 'i8 . Li Jie- lin, et al . (19 9 )) . Apolications of S')IT ') .ill oackage . "eport of Shen - ¥an g In s ti tu te of Au:j;omation , AC'l.demia Sinica . Liu Zhi-Jun , et a l . (1979) . The me thod of digital continuous system Simulation . Acta !.utomatica Si ~ica , ifoLS . ;~0 . 4 . )11) - 324 . JJu GUi - !.hang , et a1. ( 1984) . ·jAD softwRre pack~e fo~ identification, si~ulat ion and control . Treatise of Nan ~ai Unive r sity . feed Jtu',n .c{onfl' - ¥ao , e t a L (19 '31) . Optim~l back control for cover- type ~n ne aling furnaces . The SY'!l:; . .2roce edin g of th " Si~o - US ~ilater fleeting on ') o ntro l ~st"m s , Sci9nce .r'ress of ':;hina . ,(U:ln ,tong- Yao, et '1.1. (1981) . The mathematical model of t ~ mperatu control for creep - testing furnaces . ibid . .{uan ,{ong- Yao . (19 82 ) . Software for 74 C h en Zhe n- Yu i dent if i cat i on a nd optimizat i on of mul t i-inpu t mU lt i - outpu t discrete- time sys tems . I nformation and Control, Vol . 11 . No . 6 . 17- 23 . Rua n Rong-Yao , et al. ( 19'34) . A ma thema ti cal model for computer control of the elect r osl ag remel ting process. Acta Automat i ca Si nica , Vol . l0 . No . 3~7 2')2 . Shu Di - Qian , et a l. (1 9820 . J1ultivariable sel f -tu ni ng contro l of el ectric furna ces . ~eport of Beijing Iron and Steel Co ll ege . Su Zhong- Fei , et a l. (1 983) . CAD of wide ra nge speed regu l ating system . rleport of Au t omat i on Departmen t of No r th - East Techno l ogy University . Tu Qi-Lie , et a l. (1982). An introduct i on t o SCTI, a CAD software package for co ntrol system des i gn . I nfo r mat i on and Co ntrol, Vol . l l. No . 6 . 43 - 49 . Wa ng En- 2ing. (1 972). Recursive methods of opt i mal l i near filter . J1athematical Pract i ce and Cognition , No . 5 . Wa ng Ki a - Sang , et al. (1 979 ). A mathemeti cal model of computer cont r oll ed creep tes t ing furnaces . Acta Automatica Sinica , Vol . 5 . No . 1. 6- 14 . Wang Zhi- Bao , et a l. (1984) . The design of a sof tware package for CAD of contro l sys t ems. Cont r ol Theory and Applicat i ons , Vol.1 . No . 4. 28- 34 . Wu J ~ 1978 ). An algorithm of inertial navigat i on control system. Journal of South- Chi na Institute of TechnQlQgy, Vol . S No . 2 . 75 - 7g . Wu Hong- Xin, et al . (198 4) . All parameters self- tuning control method . Annual Heeting of AP·;OI'l of~hines ASSCiCiation of Aut omation . Xi- An , P . lioChina. Xio ng Uuang-Leng , et al . ( 1982) . IDr.: J' T- A computer aided program for the identi fication of l i near systems . Information and Control , Vol . 11. No . ~ . 34- 42 . Xu Jian- }ian , et al . (1976) . Adaptive filtering . Journal of i"u D:·l.ll University , No , 2 . 22 - ) 8 . package of Yin Yi - Xin , et al . ( 1984). ~AD multivaria'ole self - tun i n~ control sy s tems. Treatise of Beijing I r on and "lteel College . Yuan Zhu - Zhi , et al. (1980) . Self- t lllling adaptive control of two instal lation s in chemical i ndustry . ~th LFAC/H'l.t' Dig i tal Computer Applicat i on to Process ~ont r ol , ?reprints , 161 - 166 .