Wrld Academy f Science, Engineering and Technlgy 5 007 Level Cntrl in Hrizntal Tank by Fuzzy-PID Cascade Cntrller Satean Tunyasrirut, and Santi Wangnipparnt Abstract The paper describes the Fuzzy PID cascade cntrller t cntrl the level f hrizntal tank that has diameter 300 mm and 480 mm lng. Interface card mdule PCI-604E in cmputer and Lab View sftware prgram is used fr built the cascade cntrller. Structure f its where the inner lp is a PID cntrller fr regulating flw rate f system and uter lp is a Fuzzy lgic cntrller fr cntrl the level. The respnse time, steady state errr, lad disturbance, and cntrl valve actin f cascade are tested and cmpared with the simple cntrller. The experimental shws that at the same water level 50% set-pint, the rising time t set-pint f cascade cntrller is less than the simple cntrller abut 1750 ms, and it has steady state errr less than simple cntrller abut ±1%. The lad disturbance n the plant has n affect when using the cascade cntrller. The cntrl valve actin f the cascade cntrller has peratin high frequency than the simple cntrller s the water level in hrizntal tank is smthly cnstant. Oversht affect n cascade and simple cntrller has nt ccurred. the cascade cntrller that cnsist f the PID and the fuzzy lgic cntrl t imprve the dynamic characteristic f the level cntrl in hrizntal tank. The perfrmance f it will cmparing with the simple lp r fuzzy lgic cntrller. Keywrds Fuzzy PID, level cntrl, hrizntal tank. I. INTRODUCTION OWADAYS, the varius parameters in the prcess f N industrial are cntrlled such as temperature, level, and etc. Sme prcess needs t keep the liquid level in the hrizntal tank such as il, chemical liquid in its. The level cntrl is a type f cntrl methd fr cmmn in prcess system. The level cntrl system must be cntrlled by the prper cntrller. The bjective f the cntrller in the level cntrl is t maintain a level set pint at a given value and be able t accept new set pint values dynamically. The cnventinal prprtinal-integral-derivative (PID) is cmmnly utilized in cntrlling the level, but the parameter f thse cntrllers must be turned by tuning methd either in time respnse f frequency respnse t meet their required Manuscript received Nvember 30, 006. The research has been supprted in part by Faculty f Engineering, Pathumwan Institute f Technlgy, Bangkk 10330, Thailand and Research Center fr Cmmunicatins and Infrmatin Technlgy, King Mngkut's Institute f Technlgy Ladkrabang (KMITL), Bangkk 1050, Thailand. Satean Tunyasrirut is with Asst. Prf at Pathumwan Institute f Technlgy, Thailand. His research interests include mdern cntrl, intelligent cntrl, Pwer Electrnics, Electrical Machine, and Drives (email: satean000@gmail.cm). Santi Wangnipprant is with Asst. Prf. at Department f Electrical Engineering, Faculty f Engineering, Pathumwan Institute f Technlgy, Bangkk 10330, Thailand. His research interests include sensr technlgy, cntrl system, enhance heat transfer, and energy cnversin. (e-mail:nipparnt@gmail.cm). perfrmances [1,]. On the ther hand, the fuzzy cntrller is als ppularly implemented in many practical industrial autmatin applicatins. II. THEORY A. Cylindrical Hrizntal Tank System The structure f the liquid vlume in hrizntal tank is shwn in Fig. 1. Fig. 1 The hrizntal tank mdel The cntinuity equatin is essentially the equatin fr the cnservatin f mass as fllws; dm = ρq i ρq (1) Where m = ρv = Mass f water, kg Q, Q i = Vlume flw rate, inlet and utlet respectively. ρ = Water density = 1000 kg/m 3 dv = Q i Q () When dv = (Dh-h )dh and Dh = Rh are substituted in Eq.() and given as; dh Qi Q ( Dh h ) = (3) Where Q = Cv h and Cv = value cnstant are substituted in Eq.(3) and given as; dh Qi Cv h = f( h, Q ) i ( Dh h ) ( Dh h ) = (4) H 78
Wrld Academy f Science, Engineering and Technlgy 5 007 Taylr-series expansin and Laplace transfrm are used, and we get: H( s) 1 K = = Q ( ) (( ) s i s A Dh h 1 + 1 + ωs A (5) fuzzificatin, cntrl rule and defuzzificatin stages. T create the membership functin, Membership functin editr is used fr expressing input and utput variables. The input variable is the level errr. The number f membership functins fr each input variable is designated in seven ranges. The setting level f the hrizntal tank is defined as the fuzzy cntrller utput, which has 49 membership functins. where K= -1/(A(Dh-h )) and ω = -1/A. Frm the transfer functin shws that K and ω as a functin f nnlinear h. III. PID, FUZZY, AND CASCADE CONTROLLER A. PID Cntrller The mst industrial prcess can be cntrlled with PID cntrl (Prprtinal-Integral-Derivative) prvided in equatin: t 1 det ( ) ut) = K e( t) + e( t) + Td Ti 0 ( (6) Where u is the cntrl variable and e is the cntrl errr (e=y sp -y). The parameters f its can be determined by ultimate sensitivity methd that using Kc=4.5 and received the data frm Fig.. All data is used fr calculated the parameters f PID cntrller as shwn in Table I. Pu = 5 ms Kc = 4.5 PBu =. Delay Time = 300 ms Fig. Data result frm the ultimate sensitivity methd TABLE I PARAMETERS OF PID CONTROLLER Type f cntrl Kc T i T D P.5 0 PI.045 0.1037 0 PID.647 0.065 0.0156 B. Fuzzy Lgic Cntrller Fuzzy cntrl uses a list f rules than cmplicated mathematical expressins. These rules are mdeled after decisins previusly made by humans thrugh the prcess cntrl system. The inputs f fuzzy lgic cntrller are the level errr between the actual level and target level f the level in hrizntal tank. The fuzzy lgic cntrller cnsists f (c) Fig. 3 Membership functin f input and utput variables fr the level cntrl; errr, CE, and (c) utput assumptin The frm f all input and utput membership functins is selected t be triangle-shaped functins fr simplicity. In fuzzificatin stages, the input variables, the level is cnverted int fuzzy variables by using the membership functin as shwn in Fig. 3 where e and ce f level is the fuzzy subsets in the universe. The linguistic rules are shwn in Table II. 79
Interface card PCI-604E Wrld Academy f Science, Engineering and Technlgy 5 007 TABLE II CONTROL RULES FOR INPUT AND OUTPUT VARIABLES ce e NB NM NS ZE PS PM PB NB DE0 DE0 DE0 DE0 DE0 DE0 DE0 NM DE1 DE1 DE1 DE1 DE1 DE1 DE1 NS DE DE DE DE DE DE DE ZE DE3 DE3 DE3 DE3 DE3 DE3 DE3 PS DE4 DE4 DE4 DE4 DE4 DE4 DE4 PM DE5 DE5 DE5 DE5 DE5 DE5 DE5 PB DE6 DE6 DE6 DE6 DE6 DE6 DE6 C. Cascade Cntrl Cascade cntrl can be used fr imprved disturbance rejectin when there are several measurement signals and ne cntrl variable. Cascade cntrl is built up by tw cntrl lp as shwn in Fig. 4. The inner lp is called the secndary lp that using the PID cntrller, the flw transmitter is used fr sending the feedback signal. The uter lp is called the primary lp that using the fuzzy lgic cntrller, the level transmitter is used fr sending the feedback signal. 0-10V Cntrller A/D D/A PI FT CV1 HV5 TANK Fig. 5 Plant level cntrl set up V. EXPERIMENTAL RESULTS In this paper shws level cntrl frm 10-90% f the hrizntal tank by cascade cntrller. There are three types f the experimental as the respnse time, lad disturbance, and cntrl valve actin are investigated. 0-10V Terminal blck CB-68LP HV6 HV4 HV3 HV HV1 I / V V / I SUPPLY TANK 1 4-0mA MOTOR PUMP 4-0mA 4-1mA H L LT A. The Respnse Time The prcess respnse r respnse time f the level cntrl system, when changing the step set pint frm 0 t 10, 5, 75 and 90 are tested. At the step set pint frm 0 t 50%, the respnse time f the simple lp r fuzzy cntrller and Cascade cntrller are cmpared and shwn in Fig. 6. Fig. 4 Plant mdel f cascade cntrl system IV. EXPERIMENTAL SETUP The simplified f the level cntrl system is shwn in Fig. 5. It cnsists f a micrcmputer, interface card, level transmitter, Flw transmitter, and linear cntrl valve. The micrcmputer is a cascade cntrller that used fr cntrls the level in hrizntal tank. The uter lp, the fuzzy cntrller is received the level signal, cmputatin and sending cntrl signal as the set pint f PID cntrller. The inner lp, PID cntrller is received the flw signal, cmputatin and sending cntrl signal t the cntrl valve in rder t keep the water level at the set pint in the hrizntal tank. Fig. 6 The prcess respnse result at the step set pint 0-50% when the fuzzy lgic cntrller nly, the cascade cntrller 80
Wrld Academy f Science, Engineering and Technlgy 5 007 Frm the Fig. 6 shws that the respnse time f the cascade cntrller is less than the fuzzy lgic cntrller abut 7.9% and it has steady state errr less than the fuzzy lgic cntrller t. B. Level Cntrl with Interrupt Lad At water level 50%, interrupting lad by decrease water t level 5% and ff valve quickly. The respnse time f the fuzzy cntrller nly and the cascade cntrller are cmpared and shwn in Fig. 7. It is seen frm the Fig. 7 that the fast respnses can als be btained, the respnse time f the cascade cntrller is less than the fuzzy cntrller nly abut 4.6%. Fig. 8 The actin f cntrl valve at level set pint 50% when the fuzzy lgic cntrller nly, the cascade cntrller Fig. 7 The prcess respnse result at interrupting lad when the fuzzy lgic cntrller nly, the cascade cntrller. C. Cntrl Valve Actin Cntrl valve actin f the fuzzy cntrller nly and the cascade cntrller are cmpared and shwn in Fig. 8. It is seen frm the Fig. 8 that the cntrl valve f the cascade cntrller perates at high frequency and keep level nearly the set pint. VI. CONCLUSION PID and fuzzy cntrller are used fr built the cascade cntrller in rder t cntrl the level in the hrizntal tank. It has been shwn that the speed f respnses f the level cntrl system with and withut lad interrupt in the tank are fast. Hence, it can be cnclusin that; 1. The respnse time f the cascade cntrller less than the single lp r fuzzy lgic cntrller abut 7.9% at the step level set pint 0 t 50%.. Bth f the fuzzy lgic cntrller and the cascade cntrller give the smallest state errr. 3. The interrupt lad has slightly effect t the fuzzy lgic cntrller and the cascade cntrller. 4. The cntrl valve actin f the cascade cntrller is perated at high frequency than the single lp s the water level in the tank is smthly ACKNOWLEDGMENT The authrs gratefully acknwledge the Mr. Kampl- Prmprnchai and Mr. Watee Kaewpet wh wrk hardly and testing this research. REFERENCES [1] J. G. Ziegler and N. B. Nichls, Optimum settings fr autmatic cntrller, ASME Trans. Vl64, 194, pp.759-768. 81
Wrld Academy f Science, Engineering and Technlgy 5 007 [] W. K. H, C. C. Hang, and J. H. Zhu, Perfrmance and gain and phase margins f well-knwn PI Tuning frmulas, accepted fr publicatin in IEEE Trans. Cntr. Syst. Technl., 1995. [3] Jacb, J.Michael., Industrial Cntrl Electrnics, New Jersey: Prentice-Hall, 1988. [4] Passin, Kevin M., and Yurkvich, Stephen., Fuzzy Cntrl, Sydney: Addisn-Wesley, 1998. 1995 he was awarded with the Japan Internatinal Cperatin Agency (JICA) schlarship fr training the Sensr Technlgy at Nara Natinal Cllege f Technlgy, Japan. Since 005, he has been a Asst. Prf at Department f Electrical Engineering, Pathumwan Institute f Technlgy, Thailand.. His research interests include Sensrs, Cntrl system, Energy cnversin, and Enhance heat transfer. Satean Tunyasrirut received in B.S.I.Ed. in Electrical Engineering and M.S.Tech.Ed in Electrical Technlgy frm King Mngkut s Institute f Technlgy Nrth Bangkk(KMITNB),Thailand in 1986 and 1994, respectively. In 1995 he was awarded with the Japan Internatinal Cperatin Agency (JICA) schlarship fr training the Industrial Rbtics at Kumamt Natinal Cllege f Technlgy, Japan. Since 1995, he has been a Asst. Prf at Department f Instrumentatin Engineering, Pathumwan Institute f Technlgy, Thailand. His research interests include adaptive cntrl, intelligent cntrl, electric drives. Santi Wangnipparnt received in B.S.I.Ed. in Electrical Engineering frm King Mngkut s Institute f Technlgy Nrth Bangkk(KMITNB),Thailand in 1988 and M.eng and D.eng in Energy Technlgy frm King Mngkut University f Technlgy Thnburi (KMUTT),Thailand in 1994 and 001, respectively. In 8