IMPLEMENTATION OF FUZZY-NEURO CONTROLLER FOR DC-DC CONVERTER FED DC SERIES MOTOR USING EMBEDDED MICROCONTROLLER

IMPLEMENTATION OF FUZZY-NEURO CONTROLLER FOR DC-DC CONVERTER FED DC SERIES MOTOR USING EMBEDDED MICROCONTROLLER I. Thngrju 1 M. Murugnndm 2 nd M. Mdheswrn 3 1 Deprtment of Electricl nd Electronics Engineering, Government College of Engineering, Brgur, Tmilndu, Indi 2 Deprtment of Electricl nd Electronics Engineering, Muthymml Engineering College, Rsipurm, Tmilndu, Indi 3 Deprtment of Electronics nd Communiction Engineering, Mhendr Engineering College, Mllsmudrm West, Tmilndu, Indi E-Mil: murugnm1@gmil.com ABSTRACT The speed control of DC series motor with Hybrid -Neuro is pented in this pper. The motor is connected with DC Chopper. This configurtion hs been designed with current control nd speed control loops. The current controller blocks the PWM signl when the current inces beyond its limit. The speed controller is bsed on -Neuro type. The performnces of FNC re nlyzed in pect of lod vrition nd speed vrition using MATLAB/Simulink. This system is implemented in NXP 80C51 Microcontroller bsed Embedded System. From the ult of simultion nd hrdwre implementtion it is found tht the FNC cn hve better control compred with Logic (FLC). Keywords: DC Series Motor, - Neuro, DC Chopper, MATLAB Simulink, Embedded System. 1. INTRODUCTION DC series motor is normlly utilized for high torque pplictions. In such wy, tht the DC series motor drives re engged with mny pplictions to hndle hevy lod, s they offer high strting torque. Most of their pplictions re of n industril nture such s lifts, crnes, hoist nd electric trction. DC series motors re n idel choice for bttery-operted equipment over AC motors, s they do not require the use of expensive inverter circuitry to convert the DC voltge to n AC voltge required by the motor. Here the DC series motor is controlled by DC chopper. The duty cycle is vried to get vrible output voltge in the chopper [1]. The DC series motor in industril environment hs inced due to high performnce nd high strting torque for prticulr drive system. Such high performnce pplictions requi the motor drive with lest stedy stte error, over shoot, under shoot nd the settling time in its speed chrcteristic. It is noted tht the intelligent control bsed Artificil Neurl Network (ANN) cn give the better performnce for nonliner prmeter vrying system with lod torque disturbnce. The conventionl controllers like PI, PID nd Logic s were widely used formerly for chopper control nd motor control pplictions [2]. Although the performnce of such controller is deprived nd it filed to give suitble ults when control prmeters, loding conditions nd the motor itself re chnged. Thus the tuning nd optimiztion of these controllers re chllenging nd difficult tsk, prticulrly under vrying lod conditions, prmeter chnges nd bnorml modes of opertion. The min disdvntge with the conventionl controller is the high computtion time. It hs been found tht the computtion burden of Logic cn be reduced by hybrid -Neuro. Intelligent control techniques involving ANN is found to be simpler for implementtion nd powerful in control pplictions [3]. The ANN controller is simulted using MATLAB GUI nntool [4]. The DC series motor drive fed by single phse controlled rectifier (AC to DC converter) nd controlled by fuzzy logic is pented. It ws concluded tht the fuzzy logic controller provides better control over the clssicl PI controller which hs improved the performnce. It ws lso reported tht the settling time nd mximum overshoot cn be reduced [5]. A low-cost fuzzy controller for the control of DC drive fed by four-qudrnt chopper ws designed nd utilised. The fuzzy controller ws implemented in low-cost 8051 microcontroller. The simulted closed loop performnce of the fuzzy controller in pect of lod vrition nd reference speed chnge hs been reported [6]. A fuzzy controller ws designed for closed loop control of DC series motor drive fed by DC-DC converter. The performnce in pect of lod vrition nd speed chnges hs been reported. The performnce of the proposed controller ws compred with the reported ults nd found tht the fuzzy bsed DC-DC drive cn hve better control. The fuzzy logic controller ws utilized for different types of DC drives [8]. The ANN controller is developed nd utilized for speed control of DC motor. It is found tht PI trined ANN controller hs better ponse [9]. A comprtive study of PI, fuzzy, nd ANN controllers for chopper-fed DC drive with embedded systems pproch ws done. Here the trining ptterns for the neuron controller were 5083

obtined from the conventionl PI controller lso from fuzzy logic controller. The designed controller ws implemented in low cost 8051 Microcontroller [10]. The Neuro- control ws pplied to DC series motor. logic suffers from complex processing. This problem ws olved by implementing fuzzy logic controller on neurl network [11]. The Logic is implemented in Neurl Network nd utilized for seprtely excited DC motor. The experimentl ponses demonstrte the effectiveness of the proposed system [12]. A feed forwrd ANN ws developed for the speed control of DC motor. The network ws rndomly trined by online trining methods. The proposed controller simplifies the lerning lgorithm, reduces the computtion time nd ccelertes the trining speed [13]. An dptive Neuro- controller for the control of DC motor speed ws designed nd simulted. The proposed system ult demonstrtes tht the deigned ANFIS Swrm speed controller pprecite good dynmic behvior of the DC motor speed control with no overshoot, better performnce nd high robustness [14]. A new simultor model for winding process using nonliner identifiction bsed on recurrent locl liner neuro-fuzzy network trined by locl liner model tree, which ws n incrementl tree-bsed lerning lgorithm. The experimentl ults illustrte the effectiveness of the proposed Neuro- modeling pproch [15]. The PID-ANN controlled DC-DC converter fed PMDC motor is pented. The PID-ANN controller gives the proper speed regultion from 10% to 100% lod disturbnce thn the conventionl PID controller [16]. Artificil neurl network-bsed trcking controller for high-performnce stepper motor drives is implemented. The experimentl ults hve shown tht the proposed control structure trcks the trjectories successfully, even under externl disturbnces nd noisy environment [17]. A hybrid neuro-fuzzy controller (NFC) is pented for the speed control of Brushless DC motors to improve the control performnce of the drive under trnsient nd stedy stte conditions [18]. -Neuro control rchitecture is pplied to Brushless DC motor nd implemented in rel time using digitl signl processor [19]. An dptive neuro-fuzzy inference system (ANFIS) hs been used to control the speed of the switched reluctnce motor. The ANFIS lgorithm hs been implemented on digitl signl processor (TMS320F240) llowing gret flexibility for vrious rel time pplictions [20]. The Neuro- (NF) hybrid system is design nd implemented for Sumo Robot (SR) control. The ults showed tht intelligent control nd soft computing techniques cn be esily pplied to vrious robotic competitions [21]. Neuro- bsed nvigtion techniques for severl mobile robots re investigted in totlly unknown environment. By using Neuro- technique s mny s 1,000 mobile robots cn nvigte successfully neither collides with ech other nor colliding with the obstcles pent in the environment. It is concluded tht the developed Neuro- technique is most efficient [22-23]. In the proposed system -Neuro controller is designed. The motor drive system minly utilizes the FNC nd DC chopper (DC-DC converter). Such drive system hs the chrcteristics of precise, fst, effective speed reference trcking with minimum overshoot/undershoot nd miniml stedy stte error. The FNC bsed speed controller work effectively under input voltge vrition nd lod torque disturbnces lso it will work excellently even the motor prmeters chnges. Hence the system is mde for series motor with different lod nd different speed. In order to mintining the current t limiting vlue n ON/OFF type current controller is lso designed, it will limit the current in sfe vlue. 2. PROPOSED SYSTEM The block digrm of the proposed system is shown in Figure-1. The system minly consists of DC chopper to drive the DC motor. A pulse type tcho genertor is used s speed sensing element nd which is used for speed feedbck to the controller. A microcontroller or digitl signl processor cn be used to generte the PWM signl to switch the DC chopper, during the implementtion of experimentl setup. In this work NXP 80C51 fmily Microcontroller (P89V51RD2BN) bsed Embedded System is utilized. Figure-1. Block digrm of the proposed system. There re two loops used in the proposed system, which consist of nmely outer FNC bsed speed control loop nd inner ON/OFF type current control loop. The current control loop is used to blocks the PWM signl whenever the motor current exceeds the reference current (I Lref). In outer loop, the ctul speed (k) is sensed by 5084

pulse type tcho genertor nd given to the ADC of microcontroller, the set speed r(k) lso given to the ADC nd the error signl e(k) is obtined by compring the set speed r(k) with the ctul speed (k). The chnge in error e(k) is obtined from the pent error e(k) nd pervious error e previous(k). The error nd chnge in error re given s inputs to the FNC. The output of the FNC is denoted s duty cycle. The chnge in duty cycle dc(k) for the DC chopper is clculted from the new duty cycle dc(k) nd previous duty cycle dc(k-1). The input nd output gin of the FNC cn be estimted by simultion. The FNC cn reduce the error to zero by chnging the duty cycle of the switching signl to the DC chopper [7-8]. 3. MATHEMATICAL MODELING OF DC SERIES MOTOR AND DC CHOPPER 3.1 DC series motor model From the generl equivlent circuit of DC series motor the voltge nd torque equtions re obtined, which is given in eqution (1) nd (2) pectively. Consider, R R R ; L L L +2M V di rm se o i R L eb e (1) rm se K = Residul mg. Voltge const. e nd i ( i. e BeforeSturion) b e b i ; e K i d Angulr Speed Similrly, e ; e e K d b f K By rerrnging the eqution (1) by replcing e b nd e V o di d d Ri L K f i K (3) di 1 Vo Ri L K f i d K d Similrly the torque eqution lso derived s follows, T i nd i ( Before Sturtion) (4) T d J B T L (2) 2 2 d T i ; T K f i ; AngulrSpeed where, i = i se - Motor current V 0 = Motor terminl voltge R rm = Armture istnce R se = Series field istnce R = Totl istnce L rm = Armture inductnce L se = Series field inductnce L = Totl inductnce M = Mutul inductnce e b = Bck emf e = emf due to idul mgnetic flex T = Deflecting torque J = Moment of inerti B = Friction coefficient d = = Angulr speed = Angulr displcement = Series field flux T L = Lod torque K f = Armture voltge constnt nd By rerrnging the eqution (2) by replcing T K i 2 f 2 d d J B T 2 1 J 2 K f i d B T L The DC series motor is modeled using the modeling equtions (4) nd (6). Such n eqution modeling is more effective thn the trnsfer function model. In trnsfer function model, it is mnory to develop different model for every input nd output prmeter chnges. In this modeled eqution modeling the voltge nd lod torque re the input prmeters, the output prmeters re speed, current nd deflecting torque etc. 3.2 DC chopper The DC chopper switch cn be Power Trnsistor, SCR, GTO, IGBT, Power MOSFET or similr L (5) (6) 5085

switching device. In order to get high switching frequency (upto 100 KHz) the Power MOSFET my be tken s switching device lso the on stte voltge drop in the switch is smll nd it is neglected. Hence the power MOSFET is designted s switching device for the DC chopper. When the gte pulse is pplied the device is turned on during the period the input supply connects with the lod. When the gte pulse is removed the device is turned off nd the lod disconnected from the input supply. 4. SIMULATION OF THE SYSTEM USING MATLAB / SIMULINK 4.1 Simultion of fuzzy logic controller Initilly fuzzy logic controller is designed nd simulted with DC series motor model. The performnce of the fuzzy logic controller ws exmined with MATLAB/Simulink in terms of speed vrition nd lod vrition. The designed FLC is explined in this section. The fuzzy logic hs emerged s tool to del with uncertinty, imprecise or qulittive decision mking problems [5]. The FLC involves three stges nmely Fuzzifiction, Rule-Bse nd Defuzzifiction. There re two types of FLC nmed Mmdni type nd Sugeno type. In this work the Sugeno type controller is utilized becuse it hs singleton membership function in the output vrible. Moreover it cn be esily implemented nd lso the number of clcultions cn be reduced [6]. 4.1.1 Fuzzifiction In logic system the linguistic vribles re used insted of numericl vribles. The process of converting numericl vrible (rel number or crisp vribles) in to linguistic vrible (fuzzy number or fuzzy vrible) is clled fuzzifiction. In this work, the motor speed is controlled by FLC. The error e(k) nd chnge in error e(k) is given to the FLC s input vrible. As explined in section 2 the error is clculted by compring the ctul speed (k) with reference speed r(k). From the error e(k) nd pervious error epervious (k) the chnge in error e(k) is clculted. Then the error nd chnge in error re fuzzified. The eqution for error nd chnge in error re specified in eqution (7) nd (9). ek ( ) ( k) ( k) (7) r ek ( ) ek ( ) e ( k) (8) Pervious Five linguistic vribles were used for the ech input vrible error e(k) nd chnge in error e(k). Those re Negtive Big (NB), Negtive Smll (NS), Zero (Z), Positive Smll (PS) nd Positive Big (PB). In FLC there re mny types of membership functions vilble, those re tringulr-shped, Gussin, sigmoidl, pi-shped, trpezoidl-shped, bell-shped etc. In this work the tringulr membership function is used for simplicity nd lso to reduce the clcultions [8]. Normlly seven membership functions re preferred for ccurte ult [5]. In this pent work only five membership functions re used for the input vribles error nd chnge in error. In order to reduce the number of membership function nd mintin the sme performnce chrcteristics the wih of the membership functions re kept different. Here the wih of center membership function is considered nrrow nd it is wide towrds outer. The input nd output fuzzy membership functions re shown in Figure-2. Figure-2. memberships used for simultion. 4.1.2 Rule tble nd inference engine According to generl knowledge of the system behvior, the perception nd experience, the control rules re derived between the fuzzy outputs to the fuzzy inputs. However, some of the control rules re developed using tril nd error method. Tble-1. Rules. NB NS Z PS PB NB NB NB NB NS Z NS NB NB NS Z PS Z NB NS Z PS PB PS NS Z PS PB PB PB Z PS PB PB PB 5086

In generl the rules cn be written s if e(k) is X nd e (k) is Y, then dc(k) is Z, where X, Y nd Z re the fuzzy vrible for e(k), e(k) nd dc(k) pectively. The rule tble for the designed fuzzy controller is specified in the Tble-1 [8]. The element in the first row nd first column mens tht if error is NB, nd chnge in error is NB then output is NB. 4.1.3 Defuzzifiction The inverse process of fuzzifiction is clled defuzzifiction. In this process the linguistic or fuzzy vribles re converted in to numericl or crisp vrible [6]. Here the best well-known weighted sum method is considered for defuzzifiction method. The defuzzified output is the duty cycle dc(k). The chnge in duty cycle dc (k) cn be obtined by dding the pervious duty cycle dc previous (k) with the duty cycle dc(k) which is specified in eqution (9). The ANN controller uses complex network structure in mny works. In this work simple ANN controller is designed with few neurons nd one hidden lyer. The feed forwrd neurl network is formed with two neurons in the input lyer, three in the hidden lyer nd one neuron in the output lyer. The error e(k) nd chnge in error e(k) re the two inputs of the designed network nd the neurons re bised properly. The pure liner ctivtion function is used for input nd hidden neurons, the tngent sigmoidl ctivtion function for output neuron. The designed network is then trined with the set of inputs nd desired outputs from the FLC [6]. A supervised feed forwrd bck propgtion neurl network-trining lgorithm is used nd it is trined with minimum error gol. The output of the network is chnge in the duty cycle dc(k). dc( k) dc( k) pdc( k) (9) This type of controllers cn be esily implemented in ny embedded system. The Sugeno type of controller is selected minly for implementtion in reltime embedded bsed processors [6]. The designed FLS is simulted with DC series motor through DC chopper. The performnce chrcteristics of speed vrition nd lod chnges with FLC fed DC series motor is given in section 5 s ult nd discussion. But the performnce of FLC fed DC series motor is not stisfctory. The min disdvntge with the FLC is the high computtion time. It is found tht the computtion burden of FLC cn be reduced by hybrid -Neuro. The ANN required trining to trin the neurons in the ANN controller. The designed FLC is simulted with the drive system for extrcting the trining. The Logic controlled DC series motor ws simulted for 5 seconds with the smpling time of 0.0001seconds. Totlly 50001 is obtined from the system with FLC. Out of 50001 only 6000 re tken for trining the ANN controller by removing the sme vlue of. 4.2 Simultion of -Neuro The performnce of FLC bsed control of DC series motor is depicted in reference. But the ults re needed to be improved further. In order to improve the performnce of DC series motor, the -Neuro bsed controller is designed. The Artificil Neurl Network control lgorithm is designed with the outcome of Logic controller. The Hybrid -Neuro controller is working properly becuse of its trined network nd lso it reduces the computtionl time. In this section, the implementtion of FLC in ANN is pented. Figure-3. Performnce plot of ANN during trining. The designed ANN is trined with the error gol of 0.0067113 t 11 epochs. The performnce plot of ANN during supervised bck propgtion trining is grphiclly shown in Figure-3. The complete configurtion of the trined network with the weights nd bis is shown in Figure-4. Figure-4. Configurtion of trined neurl network. 5087

The simultion of DC-DC converter fed DC series motor is simulted bsed on eqution modeling technique, using MATLAB/Simulink toolbox. The complete Simulink model of DC series motor, chopper, current controller, PWM genertor with developed FNC is given in Figure-5. The Figure-6 shows the simulink model of FNC. The set speed nd ctul speed is tken s input to this block. Then the normlized error nd chnge in error re clculted nd then given s input to the FNC. The output of this controller block is duty cycle. The duty cycle is given s input to the PWM genertor. () (b) (c) Figure-7. ) Simulink model of PWM genertion, b) Simulink Model of current controller, c) Simulink model of DC chopper. Figure-8. Simulink model of DC series motor. Figure-5. Simulink model of the developed system. 5. RESULTS AND DISCUSSIONS The proposed model hs been simulted using MATLAB Simulink toolbox. The designed -Neuro controller ws tested with DC-DC converter nd DC series motor. The specifiction of DC series motor used for simultion is given in Tble-2. Tble-2. DC Series motor specifictions. DC motor prmeters Motor rting Dc supply voltge Motor rted current Vlue 5HP 220 V 18 A Figure-6. Simulink model of -Neuro. The Simulink model of PWM genertor block is shown in Figure-7. Here the repeting sequence is compred with the duty cycle to get PWM output. The PWM unit produces the pulse t 1 KHz of switching frequency. Then the PWM is given to the current controller. The Simulink model of current controller is given in Figure-7b. The current controller llows the PWM when the ctul motor current with in the limit of reference current. Now the PWM signl is given to the chopper unit to vry the output voltge of the chopper from fixed DC voltge in order to control the speed of the DC motor. The Simulink model of the chopper is shown in Figure-7c. The chopper controlled vrible DC output voltge is given to the DC series motor. The simulink model of DC series motor is shown in Figure-8. Inerti constnt J 0.0465 Kg-m 2 Dmping constnt B Armture résistnce R Armture inductnce L Motor speed Armture voltge constnt K f Residul mgnetism voltge const. K 0.005 Nm Sec./rd 1 0.032 H 1800 rpm 0.027 H 0.027 V.Sec./rd Tble-3. Performnce comprison of -Neuro system with the reference [5] for the speed r=1800 rpm. Rise time(sec) Clssicl PI [5] Not mentioned system - Neuro 0.73 0.68 Mx. over shoot (%) 6.72 0.36 Nil 5088

Settling time (Sec) 2.67 1 0.97 Stedy stte error (rpm) Not mentioned ±5 Nil Figure-10. Performnce for speed vrition from 500rpm to 1000rpm t 4sec nd 1000rpm to 1800rpm t 8sec with 20% lod torque. Figure-9. Speed vrition with pect to time ponse for r=1800 rpm for fuzzy controller nd FNC with mgnified view of settling prt. The speed ponse of Logic nd - Neuro controlled DC series motor is shown in Figure-9 for the set speed of 1800rpm. The trnsient to settling portion of the ponse is expnded nd shown in the sme Figure- 9. The trnsient nd stedy stte performnces of clssicl PI, nd -Neuro controller is given in Tble-3. It is seen from Tble-3 the fuzzy controlled DC series motor gives better performnce compring with clssicl PI controller further the developed -Neuro controller gives more ccurte performnce thn the fuzzy controller. In FNC the Mximum overshoot nd stedy stte error is zero nd the rise time nd settling time is lso less thn the fuzzy controller. The PI controller shows reltively high mximum overshoot nd long settling time but no stedy-stte error. Whe in the controller hve the Cpcity to repent inherent uncertinties of the humn knowledge with linguistic vribles nd it is flexible. The system cn be creted to mtch ny set of input-output. systems don't necessrily replce conventionl control methods. logic is bsed on nturl lnguge. It is more robust thn other non-liner controllers. It is not cpble to generlize, the fuzzy system only nswers to wht is written in its rule bse. It is not robust in reltion the topologicl chnges of the system, such chnges would demnd ltertions in the rule bse. The speed ponse for different set speed chnges of DC series motor with 20% lod torque is shown in Figure-10 for controller nd FNC. The corponding current vrition is shown in Figure-11. From this Figure-11 the set speed chnge from 0 to 500rpm, the current vrition with is between 0A to 20A, whe the current vrition with FNC is between 5A to 8A only. Figure-11. Current vrition for the speed chnges from 500rpm to 1000rpm t 4sec nd 1000rpm to 1800rpm t 8sec with 20% lod torque. From this, we conclude tht the current vrition is more for FLC due to more speed oscilltions nd it is less in FNC. The pek current never exceed beyond 20A during norml opertion of the drive system, becuse the reference current is 20A. Whe the set speed chnge from 1000rpm to 1800rpm the vrition of current is minimized in both the cses it is becuse of the mchine is running nerer to rted speed. The time domin specifiction of controller nd -Neuro controller for different set speed chnge with 20% lod is depicted in Tble-4. 5089

Tble-4. Time domin specifiction of controller nd -Neuro controller for different set speed chnge with 20% lod. Time domin specifictions Mximum over shoot in % Settling time in seconds Set speed chnge from 0 to 500rpm -Neuro Set speed chnge from 500 to 1000rpm -Neuro Set speed Chnge from 1000 to 1800rpm -Neuro 0.25 Nil 0.3 Nil 0.36 Nil 0.35 0.3 0.35 0.3 0.5 0.45 Tble-5. Time domin specifiction of fuzzy controller nd FNC for different lod chnge with rted speed. Time domin specifictions Mximum speed drop in % Recovery time in seconds Stedy stte error in rpm Lod chnge from 10% to 25% -Neuro Lod chnge from 25% to 50% -Neuro Lod chnge from 50% to 75% -Neuro 0.2-0.3 0.1 0.4 0.15 11-13 0.005 23 0.015 ±4 ±0.5 ±3.5 ±0.75 ±3 ±1 Figure-12 shows tht the speed ponse nd the corponding current vrition for vrious lod with -Nuero controller. During ech lod chnge there is negligible mount of distortion in speed ponse. At higher lod the oscilltion in the current is less compred to light lods. Tble-5 reports the trnsient nd stedy stte performnce comprison of fuzzy controller nd FNC for different lod chnge with rted speed of DC series motor. It cn be seen from the Tble-5 the stedy stte error inces when the lod inces in the cse of FNC but vice-vers for fuzzy controller. However the recovery time nd mximum speed drop of FNC is lmost negligible compred with the fuzzy controller. Figure-12. performnce for lod vrition 10% to 25% t 2 sec., 25% to 50% t 4sec nd 50% to 75% t 6sec for r=1800 with FNC. 6. EXPERIMENTAL IMPLEMENTATION The designed -Neuro controller ws implemented by using NXP 80C51 bsed microcontroller (P89V51RD2BN). A DC-DC buck converter ws built with the MOSFET using IRFP450, nd the controllers were tested with DC series motor. The microcontroller (P89V51RD2BN) hs n 80C51 comptible core with the following fetu: 5090

80C51 Centrl Processing Unit, 5 V Operting voltges from 0 to 40 MHz, 64 kb of on-chip Flsh progrm memory. PCA (Progrmmble Counter Arry) with PWM nd Cpture/Compre functions. The PWM is generted t frequency of 10 KHz. The PWM from the microcontroller ws then mplified for level through the open collector optocoupler CYN 17-1 nd fed to the DC DC power converter through n isoltor nd driver chip IR2110. The DC-DC buck converter output ws given to the DC series motor whose speed is to be controlled. The speed of the motor ws sensed by pulse type digitl speed sensor (Photo interrupter: GP1L53V). It hs n IR LED nd photo-trnsistor. Then the pulse signl is given to frequency to voltge converter (F to V converter: LM2907) nd feedbck the signl to the microcontroller through n ADC chip (8bit ADC: ADC0808CCN). The Figure-13 shows the experimentl setup of the proposed system with DC series motor. Figure-15. Experimentl grph of speed vrition for the step chnge in reference speed r=1800 rpm using FNC. Figure-14 shows the speed ponse with set speed of 1800rpm for fuzzy controller, it is tking lmost 6 seconds of time to settle the set speed lso seen tht it hs more oscilltions pent in the ponse due to fuzzy controller nture. The Figure-15 shows the speed ponse with the set speed of 1800rpm for -Neuro controller. The corponding experimentl current vrition is shown in Figure-16. Figure-13. Hrdwre setup of the proposed system with DC series motor. Figure-16. Experimentl grph of lod torque chnges nd speed vritions with motor current for the reference speed r=1800 rpm nd lod chnges up to 100% lod using FNC. Figure 14. Experimentl grph of speed vrition for the step chnge in reference speed r=1800 rpm using fuzzy controller From the Figurer-15 it is observed tht there is no overshoot, no stedy stte error nd the settling time lso 4.5 seconds it is less thn the logic controller. From the experimentl setup the were tken for current vrition with pect to lod torque chnges. Figure-16 shows the experimentl grph of lod torque chnges nd speed vritions with motor current for the reference speed 5091

r=1800 rpm nd lod vritions up to 100% lod using FNC. From the Figurer-16 it is observed tht the motor speed is regulted. The FNC mintin the speed s rted vlue even if the lod torques chnges. The motor current is lso within the limit of the reference vlue. Moreover the speed-current nd torque-current curves lmost follow the DC series motor chrcteristics. The Tble-6 exposes the performnce comprison of hrdwre of proposed system with logic controller. Tble-6. Hrdwre Performnce Comprison of proposed system with conventionl PID controller for the speed r=1800 rpm nd T L=10%. controller FNC Simultion Hrdwre Simultion Hrdwre Settling time in sec. 1 6 0.94 4.5 Mx. over Shoot in % 0.36 Nil Nil Nil Stedy Stte Error ±5 ±15 Nil ±2 7. CONCLUSIONS In this work the performnce of Hybrid - Neuro controlled DC-DC converter fed DC series motor ws pented. The dynmic speed ponse of DC series motor with FLC nd FNC ws estimted for different lod torque nd different set speed chnge. It ws found tht the speed cn be controlled effectively with hybrid FNC. The hybrid FNC reduced the pek overshoot, settling time nd stedy stte error of the DC series motors. The number of neurons used in ech lyer nd the number of lyers re reduced in the designed hybrid FNC. There by the trined hybrid FNC requi less computtion time nd reduced the complexity of the controller design thn the clssicl PI nd FLC. It ws implemented with simple low cost NXP 80C51 microcontroller (P89V51RD2BN) bsed embedded system, thus the cost of the system lso reduced. The nlysis provides the vrious useful prmeters nd the informtion for effective use of developed system. REFERENCES [1] Philip. T.Krein. 1998. Elements of power Electronics, Oxford University Ps. [2] Dimiter Drinkov, Hns Hellendoorn nd Michel Reinfrnk. 1996. An Introduction to Control, Nros Publishing House. [3] Zurd J. M. 1992. Introduction to Artificil Neurl Systems, Mumbi: Jico Publishing House. [4] MATLAB, Neurl Network Tool Box User s Guide, Version 3, Msschusetts: The Mthworks Inc. [5] H. A. Yousef, H. M. Khlil. 1995. A fuzzy logicbsed control of series DC motor drives. Proceedings of the IEEE Interntionl Symposium on. 2(10-14): 517-522. [6] M Mdheswrn nd M Murugnndm. 2012. Simultion nd Implementtion of PID-ANN for Chopper Fed Embedded PMDC Motor. Journl on Soft Computing. 2(3): 319-324. M.Murugnndm nd M. Mdheswrn. 2009. Performnce Anlysis of Logic Bsed DC-DC Converter fed DC Series Motor. IEEE interntionl conference, Chinese Control nd Decision Conference (CCDC 2009). pp. 1635-1640. [8] M.Murugnndm nd M.Mdheswrn, 2009. Modeling nd Simultion of Modified Logic for Vrious types of DC motor Drives. IEEE interntionl conference on Control, Automtion, Communiction nd Energy Conservtion -2009, 4 th -6 th. [9] M Murugnndm nd M Mdheswrn. 2013. Stbility nlysis nd implementtion of chopper fed DC series motor with hybrid PID-ANN controller. Interntionl Journl of Control, Automtion nd Systems. 1(5): 966-975. [10] M. Murugnndm nd M. Mdheswrn. 2012. Experimentl verifiction of chopper fed DC series motor with ANN controller. Frontiers of Electricl nd Electronic Engineering. 7(4): 477-489. [11] Young Im Cho. 2004. Development of new neurofuzzy hybrid system. Industril Electronics Society, IECON 2004. 30th Annul Conference of IEEE 5092

Volume 3, 2-6 November. 3: 3184-3189. Digitl Object Identifiers 10.1109/IECON.2004.1432322. [12] Buj G. S. nd Todesco F. 1994. Neurl network implementtion of fuzzy logic controller. IEEE Trns. Indust. Electron. 41(6): 663-665. [13] Ahmed Rubi, Rj Kotru. 2000. Online Identifiction nd Control of DC Motor Using Lerning Adpttion of Neurl Networks. IEEE Trnsction on Industry pplictions. 36(3): 935-942. [14] Boumediene Allou, Abdellh Loufi, Brhim Gsboui nd Abdesslm Abderrhmni. 2009. Neuro- DC Motor Speed Control Using Prticle Swrm Optimiztion. Leonrdo Electronic Journl of Prctices nd Technologies. (15): 1-18. [15] Hsn Abbsi Nozri, Hmed Dehghn Bndki, Mohmmd Mokhtre nd Somyeh Hekmti Vhed. 2012. Intelligent non-liner modelling of n industril winding process using recurrent locl liner neurofuzzy networks. Journl of Zhejing University- SCIENCE C (Computers nd Electronics) Springer- Verlg Berlin Heidelberg. pp. 403-412. [16] M. Mdheswrn nd M. Murugnndm. 2012. Simultion nd Implementtion of PID-ANN for Chopper Fed Embedded PMDC Motor. ICTACT Interntionl Journl on Soft Computing. 02(03): 319-324. [18] Mummer Gokbulut, Besir Dndil nd Cfer Bl. 2006. A Hybrid Neuro- for Brushless DC Motors. Springer-Verlg Berlin Heidelberg. pp. 125-132. [19] Cetin Gencer, Ali Sygin, Ismil Coskun. 2006. DSP Bsed -Neurl Speed Trcking Control of Brushless DC Motor. Springer-Verlg Berlin Heidelberg. pp. 107-116. [20] M. Ali Akcyol. 2004. Appliction of dptive neurofuzzy controller for SRM. An Interntionl Journl of Advnces in Engineering Softwre, Elsevier. pp. 129-137. [21] Hmit Erdem. 2011. Appliction of Neuro- for Sumo Robot control. An Interntionl Journl of Expert Systems with Applictions, Elsevier. pp. 9752-9760. [22] S. K. Prdhn, D. R. Prhi nd A. K. Pnd. 2006. Neuro-fuzzy technique for nvigtion of multiple mobile robots. Springer Science + Business Medi, Optim Decis Mking. pp. 255-288. [23] C. Ngrjn, M. Murugnndm nd D. Rmsubrmnin. 2012. Anlysis nd Design of CLL Resonnt Converter for Solr Pnel-bttery Systems. Interntionl Journl of Intelligent Systems nd Applictions (IJISA). 5(1): 52-58. [17] Ahmed Rubi, Mrcel J. Cstro-Sitiriche, Moses Grub nd Legnd Burge. 2007. Implementtion of Artificil Neurl Network-Bsed Trcking for High-Performnce Stepper Motor Drives. IEEE Trnsctions on Industril Electronics. 54(1): 218-227. 5093