资源目录
压缩包内文档预览：
编号：128857697
类型：共享资源
大小：4.67MB
格式：ZIP
上传时间：2021-05-19
上传人：机****料
认证信息
个人认证
高**（实名认证）
河南
IP属地：河南
150
积分
- 关 键 词：
-
移动式
皮带
系统
设计
- 资源描述：
-
喜欢这个资料需要的话就充值下载吧。。。资源目录里展示的全都有预览可以查看的噢，，下载就有,,请放心下载，原稿可自行编辑修改=【QQ：11970985 可咨询交流】====================喜欢就充值下载吧。。。资源目录里展示的全都有，，下载后全都有,,请放心下载，原稿可自行编辑修改=【QQ：197216396 可咨询交流】====================
- 内容简介：
-
Journal of Process Control 62 (2018) 5565Contents lists available at ScienceDirectJournal of Process Controljournal homepage: /locate/jprocontDesign and implementation of an advanced controller in plantdistributed control system for improving control of non-linear beltweigh feederNayana P. Mahajana,1, Sadanand B. Deshpandeb,2, Sumant G. Kadwanea,3aDepartment of Electrical Engineering, Yeshwantrao Chavan College of Engineering, RTM University, Hingna Road, Wanadongri, Nagpur, Maharashtra,441110, IndiabDean R&D, Priyadarshini Institute of Engineering and Technology, RTM University, Hingna Road, Priyadarshini Campus, Digdoh Hills, Nagpur,Maharashtra, 440019, Indiaa r t i c l e i n f oArticle history:Received 28 May 2017Received in revised form21 December 2017Accepted 29 December 2017Keywords:Belt weigh feederDCSFuzzy logicModelling and simulationNon-linearitya b s t r a c tA Belt Weigh Feeder (BWF) is a flat belt conveyor designed for feeding bulk material into a chemicalprocess in a controlled fashion for better process control. The dynamic weight of bulk material is measuredwith a belt weigh bridge and the belt speed is controlled to compensate for any variation in the weightso that the mass feed rate is maintained as per the set feed rate. The problem of belt speed controlis challenging as the dynamic response of system is non-linear and there are frequent changes in beltload due to variation in bulk material characteristics. The control accuracy of belt weigh feeder is fullydependent on the controllers performance in providing precise control of speed of motor/belt. Any delayin achieving the set feed rate or frequent deviation between set and actual feed rate affects the quality andefficiency of downstream process. Conventional PI controller is unable to provide optimum control due tosystem non-linearity. To overcome this problem, in this paper first the operating data of the BWF systemis analysed and the nature and cause for the nonlinearity is investigated. The system is then modelledusing the design parameters of plant belt weigh feeder, which is then simulated to have a better insightinto its non-linear response. Subsequently, based on simulation results, a PI Fuzzy Logic (PI-FL) controlleris designed to improve the control accuracy of the system. Further, to ensure the stability of the system, anadaptive controller is introduced in cascade to fine tune the gains of PI-FL controller as per the operatingspeed of the BWF. Finally, an advanced PI-FL with cascade adaptive controller is implemented in theplant DCS (microprocessor based process control system). The actual test results indicate reduction inthe Integral of Absolute Error (IAE) of the system by about 34% using this controller. 2018 Elsevier Ltd. All rights reserved.1. IntroductionBelt Weigh Feeders (BWF) are used in many mineral/ore,cement, chemical, fertiliser and food industries to measure and con-trol the quantity of the bulk material going across the belt. The beltweigh feeder construction includes a small flat conveyor belt sup-ported on a steel frame which provides the mounting arrangementCorresponding author.E-mail addresses: nayana.mahajan.in (N.P. Mahajan),ep.sdeshpande (S.B. Deshpande), sgkadwane(S.G. Kadwane).1.2.in.3.for feed and discharge end pulleys, weigh bridge rollers and sup-porting idling rollers. An endless rubber belt moves over the pulleysand idling rollers. A DC motor is coupled to discharge pulley via agear box as can be seen from Fig. 1(b). The material load is sensed byload cell based weigh-bridge (in kg/m) and accordingly speed of thebelt (in m/sec) is varied so that the material feed rate (in kg/min orMT/hr) remains constant at the discharge of the belt weigh feederas per (1).Mass flow rate?MTh?= Belt load?kgm? Belt speed?ms? 3.6(Numerical constant)(1)/10.1016/j.jprocont.2017.12.0100959-1524/ 2018 Elsevier Ltd. All rights reserved.56 N.P. Mahajan et al. / Journal of Process Control 62 (2018) 5565Fig. 1. (a) Picture of Belt Weigh Feeder (BWF) System (b) Schematic Diagram of BWF construction.In order to sense the material weight accurately and allow forload cell time constant, weigh feeder belt is driven at a very lowspeed (less than 0.3 m/s) because of which gear box is required tocouple the drive pulley with DC motor.1.1. Problem statement and literature reviewThe belt weigh feeder system under study is designed for feed-ing Rock Phosphate (Tri-Calcium Phosphate) mineral, a key input inthe manufacture of fertilizers, into the reactor for reaction with 60%Nitric acid. Rock Phosphate, before being fed on the BWF, is groundinto a fine powder in Ball Mill to increase its surface area to achievebetter reaction rate. Rock Phosphate, being a mined mineral, hassome amount of impurities like silica and metal oxides in additionto moisture which leads to frequent changes in its bulk density. Thekey problem of BWF process is that feeding of powdered materialon the belt is non-uniform as it is fed with the help feeding hop-per above BWF. Depending upon the bulk density of bulk material,scaling inside the hopper, moisture and lump content of the bulkmaterial, the feeding rate of material on BWF belt varies contin-uously. The feeding is coarsely controlled manually by adjustingthe gate opening of the feeding hopper and adjusting the vibratingmotors attached to feeding hopper to facilitate uniform feeding ofmaterial on the belt to the extent possible as shown in Fig. 2. This inturn leads to frequent changes in load cell signal that causes BWFmotor set-point to change in order to maintain the mass flow rateconstant as per (1). Further, finely ground Rock Phosphate some-times has a flushing tendency which often leads to abrupt changesin belt load. All these factors in addition to changes in operator set-point requires the BWF controller to continuously adjust the motorand belt speed to maintain the desired feed rate. For example, if theset feed rate is 10 MT/hr and the material load on the belt changesfrom 25 kg/m to 85 kg/m then the belt speed has to be reduced from0.111 m/sec to 0.0327 m/sec to maintain the same feed rate.The accuracy, stability and control performance of Belt WeighFeeder has a direct bearing on quality of the resulting slurry in thereactor. The slurry viscosity and its moisture content are depen-dent on the precise control of Rock Phosphate feed rate and directlyaffect the efficiency of downstream crystallization and filtrationprocess.Hence, from the point of view of process optimization, it is crit-ical that the plant BWF maintains feed rate as per the set feed ratewith excellent control accuracy.The features of the Belt Weigh feeder system are inertia, time-variant, non-linearity and frequent load disturbance. There are twounstable phenomena which affect the performance and precisionof the system, the first one is non-uniform material character suchas coarse size, moisture content, and so forth, and the second oneis variation of the feed rate set-point 1,2. The belt feeder exhibitsnonlinear behaviour because of motor friction, motor saturation,and sensor noise in the measurement system. The dynamics of theweigh belt feeder is dominated by the motor 3,4. The separatelyexcited DC motor employed in belt weigh feeders has dead bandand friction which affects dynamic response of the system. Theexperimental study on DC motor friction shows that Coulomb andStribeck friction causes non-linearity in DC motor model in com-parison with viscous friction 5. The usual practice, however, is tomodel DC motor with only viscous friction while ignoring Coulomband Stribeck friction. The constant belt load disturbance, frequentchanges in bulk material properties and system non-linearity posea great challenge on the PI controller which is provided for main-taining motor/belt speed. These factors affect the system inertia,gain and dynamic response at different operating conditions. Pro-portional plus integral (PI) controller when applicable to DC motorFig. 2. Schematic diagram of an industrial BWF system with feeding arrangement.N.P. Mahajan et al. / Journal of Process Control 62 (2018) 5565 57based system has superior performance as compared to the fuzzycontroller under steady state conditions, but being a linear con-troller it needs mathematical model of the system 6,7. Due toits fixed proportional gain (Kp) and integral time (Ti) settings, theperformance of the PI controller is affected by parameter varia-tions such as system gain, inertia and friction. Any steady stateerror or delay in achieving the set speed (i.e. poor control accuracy)shall lead to off-spec product or loss in process efficiency. Henceautomated tuning of PI controller is desired 8 to improve the con-trol performance of the system at all operating points. In recenttimes, several authors have suggested adaptive control techniquesfor weigh feeder application because of its non-linearity 913.It has been established through several research that Fuzzy LogicControl (FLC) is suitable for controller design when the plant modelis unknown or difficult to develop. It does not need an exact processmodel and has been shown to be robust with respect to distur-bances, large uncertainty and variations in the process behaviour3. Reference 3 has developed three types of Fuzzy logic based PIcontrollers for belt weigh feeder and demonstrated that the self-tuning PI-like FLC performed better than the gain scheduled PI-likeFLC (as per nomenclature in reference 21), however the PI FLC per-formed significantly better than the two kinds of PI-like FLCs. Theauthors have neither developed the model of the system nor inves-tigated the cause and nature of nonlinearity present in the BWFsystem. While designing the PI-FL controller, authors have devel-oped a relatively smaller 9 rule base for auto-tuning of Kpand Kiparameters of PI controller 3.In the present research, first the operating data of the BWF sys-tem is analysed and the nature and cause for the nonlinearity isinvestigated and established. The system is then modelled usingthe design parameters of plant belt weigh feeder, which is thensimulated to have a better insight into its non-linear response.Subsequently, based on simulation results, a PI Fuzzy Logic Con-troller (PI-FLC) is designed incorporating 25 rule bases to improvethe control accuracy of the system. Further, to ensure the stabilityof the system, an adaptive controller is introduced in cascade to finetune the gains of PI-FL controller as per the operating speed of theBWF. Finally, this PI-FL with cascade adaptive controller (AdvancedBWF controller) is implemented in the plant DCS (microprocessorbased process control system) and tested with the plant BWF andthe test results indicate substantial improvement in BWF controlperformance with the Advanced BWF controller.This paper is organised as follows. Section 2 describes the math-ematical modelling of the system to better understand its dynamicbehaviour and the challenges posed by it for a conventional PIcontroller. Section 3 describes the design of PI-FLC with cascadedadaptive controller for the BWF system. Section 4 presents the com-parative performance of both types of controllers using simulation.Section 5 presents the actual implementation of the Advanced BWFcontroller for an industrial belt weigh feeder using the DistributedControl System (DCS) platform. Section 6 presents some discussionon the results and finally Section 7 provides the conclusion.2. Modelling of belt weigh feeder systemThe design parameters of the plant BWF system are as depictedin Table 1.2.1. Mathematical model of the weigh feeder systemThe Belt weigh feeder system can be modelled using followingmathematical equations.Total equivalent inertia of the system at the motor armature, Jeq, is given by(2)Table 1Parameters of Belt Weigh Feeder System.Sr. No. Design parameter Symbol Specifications1. Tag No/Service WIC-101 Rock Phosphate mineral2. Capacity/Max feed rate020 MT/h3. Pulley C-C length 3000 mm4. Pulley diameter D 200 mm5. Active belt length L 2500 mm of material bed6. Endless beltdimensionsLB6400 mm (L) X 1000 mm (W)7. Friction coefficientfactor? 0.38. Mechanical efficiency ? 0.99. Gear box ratio i.e.gear1/gear2I 1: 48 (N1/N2)10. Gear box transmissionefficiency?G0.6611. Rated motor speed 150 rad/s12 Maximum field voltage Ve210 V DC13. Field current Ie0.23 A DC14. Armature voltageVa240 V DC Max15. Armature current Ia15 A DC at Full load16. Armature resistance Ra5.98 ohm17. Armature inductance La0.276H18. Motor torque constant Kb/Kt1.219. Motor inertia Jm0.0221 kgm220. Motor viscous frictioncoefficientBm0.011 Nm sJeq= Jm+ Jgear1+?N1N2?2?Jgear2+ 2Jpulley+ (Mbelt+ Mmaterial) D24?(2)where Mbeltis the mass of belt (including idlers) and Mmaterialisthe mass of bulk material, which cause changes in system inertia toa large extent. Jgear1, Jgear2and Jpulleyare inertias of gear1, gear2andpulley respectively. Gear1and gear2are the internal gears of thegear box meshing with each other and having teeth ratio of 1:48.As the pulley employed is cylindrical hollow object with massMpulleywith inner Diameter Diits mass moment of inertia about itsshaft is given by (3)Jpulley=18Mpulley?D4 Di4?(3)The Load torque of the mechanical system before gear box andat motor shaft is given by (4) and (5) respectively.TL=?M(Belt+Material)gD2?(4)TLeq=TLI?G(5)where TLis the belt load torque before gear box and TLeqis theequivalent load torque seen by the DC motor and g is accelerationdue to gravity.The motor exerts an electromagnetic torque TMdue to the backemf induced in the armature and due to voltages supplied on thestator and armature. Eqs. (6) and (7) define back emf Eband TM.TM= KbIaandKb= KIe(6)where K is motor torque constant.The back emf Ebinduced in armature is proportional to fieldcurrent (or flux) and angular velocity.Eb= Kb i.e. Eb=TMIa?(7)The armature side can be modelled as (8).Va= Ladiadt+ Raia+ Eb(8)58 N.P. Mahajan et al. / Journal of Process Control 62 (2018) 5565Fig. 3. Actual BWF gains at different operating speeds.Electromagnetic torque TMis developed by DC Motor to over-come the equivalent load torque, TLeq, equivalent inertia, Jeq, andfriction as shown in (9).TM= TLeq+ Jeqddt+ Feq (9)where Feqis equivalent co-efficient of friction of BWF system.2.2. Identification of equivalent co-efficient of friction of BWFsystemThe BWF system is a multi-rotational system working at avery slow speed. The BWF motor typically operates at 2530% ofthe rated RPM at full load. Therefore, the effect of initial frictionbetween bearings of rotating parts that reduces as speed increasesand lubrication improves is expected to be prominent and nonlin-ear in the BWF system.The nonlinearity of BWF system can be established by record-ing and analysing its operating data armature voltage, armaturecurrent, belt load, motor RPM which are available on panel understeady state conditions.If we calculate the BWF system gain represented by /Volts andplot the same on Y axis versus corresponding motor speeds onX axis, we obtain the X-Y plot in MS Excel as shown in Fig. 3.The scattered plot clearly indicates the nonlinear response whilethe solid line depicts the best fit trend-line with nonlinear cor-relation equation relating X and Y variables. It can be seen thatwhen the machine operates at lower speeds (1320 rad/s) the sys-tem exhibits lower gain of 0.40.5 rad/s/volts while it improves toaround 0.60.7 rad/s/volts at higher speeds (3040 rad/s). This phe-nomenon can be attributable to changing frictional forces which arehigher at lower speeds and decreases as speed increases.Next, let us try to identify the friction torque seen by the BWFmotor based on the operating data using steady state torque equa-tion given by (10).TM= KbIa= TLeq+ Tf(10)where Tfis the torque seen by the motor due to friction is given byTf= Feq (11)and TLeqis given by (5).The X-Y plot in Fig. 4 in MS Excel depicts the scatter plot of calcu-lated friction torque,Tf, at different BWF speeds. The plot highlightsthat the BWF system operates in the region where Stribeck frictionis prominent due to its low speed. The solid black line representsthe best fit curve showing negative exponential relation 14 whichexplains the nonlinear response obtained in Fig. 2. Stribeck frictionis primarily responsible for the increase in BWF gains at higherspeeds.Fig. 4. BWF friction torque at different operating speeds.It is projected that viscous friction shall become active if the sys-tem speed were to increase beyond the operating range. However,for the purpose of modelling, we can approximate the best fit expo-nential line by a straight line. The average coefficient of friction canthen be deduced by calculating the slope of this line, which gives,Feq= 0.026N-m-s/radCombining system Eqs. (2)(11), we can build the system modelas depicted in Fig. 5 below. The system is provided with conven-tional PI controller for speed control of belt/motor. The belt load,expressed in Kg/m, when multiplied with the belt length signifiesthe material load on the belt which in turn provides the load torqueto the model. Jeq, which is the total equivalent inertia of the systemat the motor armature, in the model is given by (2).2.3. Response with PI controllerThe model should be first tested without PI controller in lineby changing Vastep wise in order to obtain the open loop reactioncurve. In order to obtain the parameters of PI controller i.e. optimumvalue of Kp(Proportion gain) and Ki(Integral gain), Zeigler NicholsReaction curve method can be applied if we have the open loopS type response of the system 15. Based on open loop responsecaptured earlier for the step change of Vafrom 0 to 100 V DC, opti-mum PI controller parameters are calculated (Kp= 1.49 and Ki= 30)10,16. The dynamic response of the system for a step change inspeed reference from 0 to 22 rad/sec is captured as shown in Fig. 6for above PI controller parameters.From the step response it is seen that the system with PI con-troller shows two overshoots and settles at set point after 1.4 s. Theresponse indicates optimal controller tuning with quarter ampli-tude damping for the given plant (belt weigh feeder) gain. However,as observed in Fig. 3, the belt weigh feeder gain keeps on chang-ing with changes in operating speed. Based on this, we can predictthat PI controller tuned for lower set point/load shall cause higherovershoots for higher set-points/load and vice versa due to non-linear gain exhibited by belt weigh feeder system. Reference 17highlights that Ziegler-Nichols tuned PI controllers fail to providesatisfactory performance for higher order and/or nonlinear sys-tems.3. Development of PI fuzzy logic with cascade adaptivecontroller for belt weigh feederIn view of limitations of PI controller for BWF application, itis proposed to modify the conventional PI controller such that itsgains namely the Proportional gain Kpand Integral gain Kigetscontinuously adjusted as per fuzzy rules based on conditions ofspeed error, e, and change in speed error signal, ?e. Traditionally,N.P. Mahajan et al. / Journal of Process Control 62 (2018) 5565 59Fig. 5. Block Diagram representation of Belt Weigh Feeder System Model.Fig. 6. Step response of system with PI controller.the Fuzzy Logic Controller (FLC) works well under transient condi-tions, while proportional plus integral (PI) controller is more suitedto eliminate offset near the steady-state condition. Also, as the sys-tem exhibits variation in its gain, it is proposed to further adaptthe PI-FLC gains as per the operating points for stability. It is pro-posed to take the combined advantages of these three controllersby developing PI- FL with cascaded adaptive controller (AdvancedBWF controller) which can be fast and yet stable as depicted inFig. 7.Fuzzy logic is applied to inputs error (e) and change in error?e along with scaling factors Ge and G?e to develop triangularmembership functions NL, NS, ZE, PS, PL as shown in Table 2A. Tri-angular membership functions are used in the controller design asthese are relatively easy to develop and require less computationaleffort. They are also quite popular in the literature.In this model G(e) and G(?e) are gains for normalizing the inputsignals e and ?e, Kpan Kiare PI controller gains, S(?) is adaptivegain factor based on operating point (speed), TLis load torque andu is controller output.Similarly triangular membership functions are developed forPI controller gains namely Kpand Kifor implementation of fuzzyrules. Based on the past operating data and simulations results ofsystem in open loop, the value of Kpand Kiare calculated usingZeigler Nichols Reaction Curve method across the operating rangeto develop universe of discourse for Kp(in the range 0.23) andKi(in the range 2.540) within which their triangular membershipfunctions PVL, PL, PML, PS, PVS are designed as shown in Table 2B. Using input and output membership functions, 25 fuzzy rules forPI FLC are developed as shown in Table 3. The crisp value of Kpand Kigains are derived as per the degree of membership using theCentroid method 18.The values of Kpand Kiare further adapted/fine-tuned as perthe system operating point (typically between 10 and 60 rad/s) byapplying the gain, S, which in turn is dynamically calculated as perthe operating speed, ?, of the machine as per (12). Eq. (12) is basedTable 2ADesign of Input Triangular Membership Functions.Sr. No. Fuzzy Set Membership Function of Speed error (e) Numerical Range Membership Function of ? error (de) Numerical Range1Negative Large (NL) Large difference inspeed in ve direction40 to 40Large error differencein ve direction-400 to 40040 to 10 400 to 2002Negative Small (NS) Small difference inspeed in ve direction25 to 10Small error differencein ve direction250 to 100105 100503Zero (ZE) Speed difference is zero105Error difference is zero10050520 501504Positive Small (PS) Small difference inspeed in +ve direction520Small error differencein +ve direction501502040 1502505Positive Large (PS)Large difference inspeed in +ve direction3060Large error differencein +ve direction2004006060 40040060 N.P. Mahajan et al. / Journal of Process Control 62 (2018) 5565Table 2BDesign of Output Triangular Membership Functions.Sr. No. Fuzzy Sets for KpNumerical Range Fuzzy Sets for KiNumerical Range1Positive Very Large (PVL)33Positive Very Large (PVL1)40403240202Positive Large (PL)32Positive Large (PL1)302021 20103Positive Medium Large (PML)21Positive Medium Large (PML1)201010.5 1054Positive Small (PS)10.5Positive Small (PS1)1050.50.2 52.55Positive Very Small (PVS)0.50.2Positive Very Small (PVS1) 2.52.5Table 3Rule base for PI-FL with cascade adaptive (Advanced BWF) controller.(a) Rule base for Proportional Gain Kpde/e NL NS ZE PS PLNL PVL PVL PVL PVL PVLNS PML PML PML PL PLZE PVS PVS PS PMS PMSPS PML PML PML PL PLPL PVL PVL PVL PVL PVL(b) Rule base for Integral Gain Kide/e NL NS ZE PS PLNL PL1 PL1 PL1 PL1 PL1NS PML1 PML1 PML1 PML1 PML1ZE PS1 PS1 PVS1 PS1 PS1PS PL1 PL1 PL1 PL1 PL1PL PML1 PML1 PML1 PML1 PML1on first approximating the nonlinear gain of the system shown inFig. 3 by a straight line and secondly reducing it from 0.9 so thatproduct of controller and system gain is always maintained lessthan 1.S = 0.9 (0.0034 + 0.3183)(12)For example, if the crisp values of Kpand Kigains as per PI FuzzyLogic is calculated as 1.5 and 15 respectively based on conditions ofe and ?e, then they are further adapted to a value of 0.745 and 7.45respectively if BWF motor speed is 25 rad/sec. However, if the BWFmotor is operating close to 50 rad/sec then final adapted values ofKpand Kishall be 0.617 and 6.17 respectively which shall be usedfor auto tuning of Advanced BWF controller. This cascade adaptivecontrol ensures stability of the system at all operating points asBWF system exhibits higher gain at higher speeds.Table 4Analysis of improvement in control performance of the BWF System with AdvancedBWF controller.Sr. No. Control Parameters PI Controller PI Fuzzy Logic with cascadeAdaptive (Advanced BWF)Controller1 Overshoot 22.72% 2.27%2 Rise Time 0.3 s 0.16 s3 Settling Time 0.9 s 0.45 s4 Oscillations 2 Nil5 IAE 28 18.954. Comparison of performance of advanced BWF controllerwith PI controllerThe model developed in Fig. 7 is simulated with same stepchange in set-point as in case of PI controller. Comparing the con-ventional PI controller response (Fig. 6) with that of PI-FL withcascaded adaptive (Advanced BWF) controller response as capturedin Fig. 8(b), we find a marked improvement in set point trackingand system stability when the Advanced controller is applied tothe system. The analysis of improvements achieved in the controlperformance of the system in terms of better rise time, lower over-shoot, lesser oscillations and smaller settling time is tabulated inTable 4.Based on IAE (Integral of Absolute Error) figures, the belt weighfeeder system control accuracy improves by 32.32% by employ-ing the Advanced BWF controller as compared to conventional PIcontroller alone.Further, in order to test the robustness of both the controllersin terms of set point tracking, disturbance rejection and stabilityof the system, the system is tested with multiple step change inbelt load and in mass flow set-point after the initial step change inset-point. The response of the system under both types of controlactions is depicted in Fig. 9.Not only in set-point tracking, but the system response improvessubstantially with Advanced BWF controller when tested with stepFig. 7. Model of PI Fuzzy Logic with cascade Adaptive Controller for BWF (Advanced BWF controller).N.P. Mahajan et al. / Journal of Process Control 62 (2018) 5565 61Fig. 8. (a) Rule Viewer for model simulation in Simulink (b) Step response of system with Advanced controller.Fig. 9. Comparison of system response for change in set-point and belt load with PI and Advanced BWF controller.change of 30 kg/m in belt load as depicted by Fig. 9(a) and (b). WhilePI controller takes more than 0.9 s to bring back the speed to newset-point after step change of 30 kg/m in belt load, it takes 0.4 s forAdvanced BWF controller to do the same. Every change in belt loadcauses change in BWF motor speed reference.5. Implementation of advanced BWF controller for anindustrial belt weigh feederIn the actual plant the belt weigh feeder is controlled by a ded-icated belt weigh feeder controller (Model INTECONT PLUS) of M/sSCHENCK Process Germany. This microprocessor based controller62 N.P. Mahajan et al. / Journal of Process Control 62 (2018) 5565Fig. 10. Existing Hardware Set-up for control of an Industrial Belt Weigh feeder in the plant.is designed to accept raw load cell signal (belt load) and pulse signalfrom speed sensor. It does the signal conditioning of the load cellsignal to calculate belt load in kg/m and calculates the actual mate-rial mass flow rate in MT/hr based on speed pulse counts. It hasinbuilt PI controller whose gains can be tuned from front keypadand it produces control output in form of 420 mA DC signal. Thissignal is given as reference signal to Motor full wave SCR convertorwhich in turns changes armature voltage and speed of motor/belt.Fig. 10 shows the existing system for control of belt weigh feederin the plant.This controller also transmits the mass flow rate value through420 mA DC signal which is hooked to M/s EMERSON make DELTA VDistributed (Digital) Control System (DCS) for indication at CentralControl Room (CCR).5.1. Implementation of plant BWF PI controller in DCSINTECONT PLUS controller program cannot be altered; itsparameters can only be changed using menu driven display withkeypad. Therefore, in order to implement the Advanced BWFControl strategy, the programming feature of Distributed Controlsystem (DCS) system is employed. DCS is an integrated digital con-trol system designed for process control of chemical plants whichconsisting of data acquisition system and microprocessor basedcontroller at the back end and a PC based Operator Workstation(OWS) at the front end. All the plant online field instrument signalsare hooked to data acquisition cards of DCS where they are ampli-fied and digitized for further processing by DCS controller. The DCScontroller then converts the raw digital signals form of field instru-ment signals into an instrument data by applying necessary filters,scaling and linearization logic. This instrument data is then sentfrom DCS controller to OWS in real time for operator visualizationand control. Manipulation of raw instrument data and generation ofcontrolled output by the DCS Controller is achieved using ControllerProgramming. The programming is done using standard mathe-matical, logical and control blocks available in DCS software andinter-wiring these blocks as is done in MATLAB Simulink software.For shifting the control from dedicated SCHENCK controller toDCS controller following modifications were carried out.1. The INTECONT PLUS controller was configured to transmit thebelt load signal instead of feed rate signal by modifying itsparameter E01and its PI control function was disabled 20. Thisbelt load signal was connected to one channel of Analogue Input(AI) card of DCS as shown in Fig. 11. Also, a DCS instrument tag,WIT-101A, was created to indicate belt load in kg/m on DCS.2. A separate speed transmitter which accepts the proximity sensorpulse output was employed and its 420 mA DC signal was againconnected to a separate DCS AI card channel. Belt speed indicatorBS-101A tag was also created to display belt speed in m/s on DCS.3. The speed and weight signals were connected to multiply blockalong with a constant PARAMETER (having value of 3.6) to obtainthe BWF feed rate indication in MT/hr.4. The feed rate signal was connected to a DCS standard PID con-troller block and belt weigh feeder controller WIC-101A wascreated. The derivative action of PID block was disabled toimplement a conventional PI controller. The output of the PIDcontroller is connected to Analogue output channel which pro-vides the output signal to motor SCR drive as shown in Fig. 11.The system was then started and operated in manual mode withstep input. Based on the step response, the DCS PI controller wastuned manually to have minimum overshoot. (Kp= 0.75, Ti= 15 s).Fig. 11. Modified hardware set-up implementation of PI control in DCS for control of plant BWF.N.P. Mahajan et al. / Journal of Process Control 62 (2018) 5565 635.2. Implementation of PI fuzzy logic with cascaded adaptiveBWF controller in DCSThere is no programmable Fuzzy logic block available in stan-dard DCS software which allows users to create their own fuzzylogic membership functions, rule base and de-fuzzification logic.However, calculation blocks such as CALC/ACTION are there whereuser defined programs can be written using “Expression Editor”.Math functions, logical operators, constants, parameter referenceand I/O reference values can be used in the expressions which arestructured texts similar to ANSI C statements 19.The Advanced belt weigh feeder controller was implementedusing the ACTION block. In the program ERROR value of PI con-troller was referenced. The fuzzy logic program was implementedusing primarily IFTHEN ELSEEND IF expressions to developthe membership functions for error and rate of change of error andfuzzy rule base. A part of fuzzy rules implemented in DCS calcu-lation block as per the DCS Expression Editor syntax is depictedbelow.The values of input membership functions PL, PS, NL, NS, andso on were selected based on past operational data of the systemas in case of model simulation in MATLAB Simulink. The values ofoutput membership functions PVL, PML, and so on (PVL1, PML1for Integral time) were then selected based on operator experienceand simulation results. The GAIN and RESET time (Kpand Ti) ofDCS PI controller were tuned online as per the fuzzy rules shown inTable 3. Membership functions of Kpand Ti were built in the range0.23 and 525 s respectively. The variables wt and wt1 are theweights calculated based on degree of membership and adaptivegain correction factor.After implementing of the Advanced BWF controller in DCS, theloop was put on AUTO with an initial set-point of 5 T/h and setpoint was subsequently changed step wise to 10 MT/h, 20 MT/hand 12.5 MT/hr after every 2.5 min. The loop performance w.r.t. set-point changes was captured using DCS trend window data (with1 s update) which was exported to Excel and plotted as depictedin Fig. 12(b). The belt load was initially adjusted to 55 kg/m andthereafter the system experienced the normal load disturbancesencountered during regular plant operation although it was triedto maintain it within same band by adjusting the hopper gate. After10 min the loop was taken in manual and again the controller waschanged to conventional PI control by disabling calculation blockcontaining the fuzzy program. The loop was put on AUTO with orig-inal GAIN and RESET values and performance was again monitoredfor the same step changes in set-point for the next 10 min as shownin Fig. 12(a). This was done so that the bulk material characteris-tics, plant load and other disturbance factors remain similar to thatTable 5Improvements achieved with Advanced BWF controller implemented in DCS.Sr. No. ControlParametersPI Controller Advanced BWFController1 Maximum Deviation w.r.t. set point 2.8 MT 1.4 MT2. Period of Deviation w.r.t set point 156 s 96 s3 IAE 47.4 31.2in the first case and performance of both controllers are comparedunder similar environments.6. Results and discussionDuring the period of testing, the belt load was varying between55 and 78 kg/m and this was causing intermittent drift in actualfeed rate from set feed rate. Comparing the DCS trends capturedin Fig. 12(a) and (b), it is observed that Advanced BWF controllerin DCS provides a tighter control and doesnt allow the actual feedrate to deviate much from the operator set-point both in terms ofquantum and period. The improvements achieved with AdvancedBWF controller are summarized in Table 5 below. The IAE figurefor well-tuned DCS PI controller is 47.4 while that for AdvancedBWF controller in DCS is 31.2 resulting in improvement of 34.17%in control accuracy of belt weigh feeder machine. This is superiorto 1530% improvement achieved in previous research 3.Most of the previous researches on belt weigh feeders 14have not performed detailed modelling of belt weigh feeder sys-tem and the focus was only on controller design. In the presentresearch, mathematical modelling has been attempted as it pro-vides good insights into the variables affecting the dynamics of thesystem although complete non-linearity cant be exactly modelled.Even then, the simulation of the model with different PI controllerparameters at different operating points provides the universe ofdiscourse and facilitates design of Fuzzy Logic Controller. Also theopen loop response at different operating points provided goodinsight into the changing gains of the system at different operat-ing points. This has facilitated design of adaptive control action toachieve good stability. Such kind of rigorous testing is not possiblein a running plant.Most of the DCS system allows users to create their own pro-gram in the calculation blocks. However, the same feature has beeninnovatively utilized here to create a nonlinear PI Fuzzy Logic withcascade adaptive controller. The DCS controller executes the calcu-lation block program and updates the outputs in each scan period.However, there is no specific function in DCS to evaluate the changeof error signal which is required to developing fuzzy rules. Thisproblem was overcome by using ERROR/scan time signal instead of?e signal. Since DCS execution is very fast, this did not have anyeffect in performance and in fact the DCS based Advanced controllerresponse was better than the one developed in MATLAB Simulink.7. ConclusionBelt Weigh Feeder (BWF) systems exhibit non-linear behaviourdue to frequent changes in bulk material property, system inertiaand motor friction because of which its open loop gain keeps ofchanging at different operating points. Considering the changinggains of the weigh belt feeder system, the performance of con-ventional PI controller is not optimal for achieving accurate speedcontrol. The limitations of PI controller can be overcome by modi-fying the conventional PI controller such that its gains namely theProportional gain Kpand Integral gain Kigets continuously tunedbased on PI Fuzzy Logic with cascade adaptive (Advanced BWF)controller. This type of controller can be implemented in plant DCSsystem with an innovative use of calculation block available in the64 N.P. Mahajan et al. / Journal of Process Control 62 (2018) 5565Fig. 12. (a) Response of BWF with conventional DCS PI controller (b) Response of BWF with Advanced BWF Controller.DCS software. The results obtained after implementing AdvancedBWF controller for an actual plant BWF indicate improved responseof machine on all control parameters namely rise time, overshoot,set point tracking, belt load disturbance rejection and the controlaccuracy of the system improves by around 34% based on calcula-tion of Integral of Absolute Error (IAE).Appendix A. Supplementary dataSupplementary data associated with this article can be found,in the online version, at /10.1016/j.jprocont.2017.12.010.References1 T.Y. Sun, M.C. Yang, S.J. Tsai, J.S. He, An improved flush material belt weighfeeder system via fuzzy logic controller and adaptive neural networks,FUZZ-IEEE, in: IEEE International Conference, Korea, August 2024, 2009, pp.12591263.2 T.Y. Sun, M.C. Yang, S.J. Tsai, J.S. He, Adaptive neural network controller for theflush material belt weigh feeder, in: Proceedings of International JointConference on Neural Networks, Atlanta, Georgia, USA, June 1419, 2009, pp.30183023.3 Yanan Zhao, Emmanuel G. Collins Jr, Fuzzy PI control design for an industrialweigh belt feeder, IEEE Trans. Fuzzy Syst. 11 (3) (2003) 311319.4 E.G. Collins Jr., Y. Zhao, R. Millett, A genetic search approach to unfalsified PIcontrol design for a weigh belt feeder, Int. J. Adapt. Control Signal Process. 15(5) (2001) 519534.5 Virgala Ivan, Peter Frankovsk y, Mria Kenderov, Friction effect analysis of aDC motor, Am. J. Mech. Eng. 1 (1) (2013).6 N. Muruganantham, S. Palani, Hybrid fuzzy-PI controller based speed controlof PMBLDC motor using soft switching inverter, Eur. J. Sci. Res. 78 (2) (2012)259272.7 N. Muruganantham, S. Palani, Novel soft switching inverter for brushless DCmotor using fuzzy logic, Int. J. Electr. Eng. 4 (5) (2011) 601616.8 E.G. Collins Jr., C. Fan, Automated PI tuning for a weigh belt feeder viaun-falsified control, in: Proc. the 38th Conference on Decision & Control,Phoenix, Arizona USA, 1999.9 T. Sato, S. Kitano, T. Yamamoto, N. Araki, Y. Konishi, Implementation ofperformance-adaptive PI control on a weigh feeder, Int. J. Adv. Mechatron.Syst. 3 (3) (2011) 181187.10 N.P. Mahajan, S.B. Deshpande, S.G. Kadwane, Optimization of weigh feederperformance using modelling and simulation, in: Icwet 2016, Mumbai, Feb2829, IET Digital Lib, 2018, /10.1049/cp.2016.1115 (ISBN:978-1-78561-307-4).11 Takao Sato, Yasuo Konishi, Toru Yamamoto, Yasuo Konishi, A des
- 温馨提示:
1: 本站所有资源如无特殊说明，都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
2: 本站的文档不包含任何第三方提供的附件图纸等，如果需要附件，请联系上传者。文件的所有权益归上传用户所有。
3.本站RAR压缩包中若带图纸，网页内容里面会有图纸预览，若没有图纸预览就没有图纸。
4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
5. 人人文库网仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对用户上传分享的文档内容本身不做任何修改或编辑，并不能对任何下载内容负责。
6. 下载文件中如有侵权或不适当内容，请与我们联系，我们立即纠正。
7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
人人文库网所有资源均是用户自行上传分享，仅供网友学习交流，未经上传用户书面授权，请勿作他用。
川公网安备: 51019002004831号