TWO-TANK LIQUID LEVEL CONTROL USING A BASIC STAMP MICROCONTROLLER AND A MATLAB-BASED DATA ACQUISITION AND CONTROL TOOLBOX

Transcription

1 TWO-TANK IQUID EVE CONTRO USING A BASIC STAMP MICROCONTROER AND A MATAB-BASED DATA ACQUISITION AND CONTRO TOOBOX Anshuman Panda, Hong Wong, Vkram Kapla, and Sang-Hoon ee Department of Electrcal and Computer Engneerng Department of Mechancal, Aerospace, and Manufacturng Engneerng Polytechnc Unversty, Brooklyn, NY Introducton A varety of PC-based data acquston and control (DAC) boards are currently avalable n the market. These DAC boards can be broadly classfed n two categores: ) hgh-end DAC boards, whch provde a wde range of advanced hardware capabltes along wth a sophstcated software envronment, and ) low-end DAC boards, whch are prmarly used for data acquston of a few selectve sgnals whle usng propretary software. Wth the emergence of Matlab [] as a wdely used scentfc computng tool n ndustry and academa, many users seek a DAC platform that can nterface wth and explot advanced computng capabltes of Matlab to perform hardware n the loop experments. Moreover, wthn the last decade, new developments n automated code generaton programs have allowed users to utlze nteractve con-based control system smulaton tools such as Smulnk [] for real-tme control. In partcular, usng the Smulnk block lbrary and Real-Tme- Workshop (RTW) along wth Smulnk block lbrares for vendor-specfc DAC boards, one can generate C code from Smulnk-based feedback control dagrams for real-tme controller mplementaton on PC-based DAC boards. The ablty to rapdly and effcently desgn and mplement complex control algorthms usng an con-based programmng envronment enables control desgners to enhance productvty. Unfortunately, the use of Matlab as a software envronment for DAC tasks frequently requres the use of expensve DAC boards (over $,) whch nclude advanced hardware features (e.g., hgh samplng rates and hgh resoluton analog to dgtal converters) that a typcal user may not utlze to the fullest potental. Thus, the use of exstng PC-based DAC hardware and software solutons may be uneconomcal for certan users (especally educators) nterested n desgnng and buldng experments that requre a sophstcated software nterface wth mnmal hardware costs. In contrast to the PC-based DAC boards, mcrocontrollers are nexpensve devces (costng only a few tens of dollars) whch are wdely used for embedded computng n hobby, academc, and ndustral projects. However, a majorty of mcrocontrollers requre programmng usng one or other embedded programmng varants of hgh-level programmng languages (e.g., Basc, C, Java, etc.). In addton, many low-cost mcrocontrollers do not allow floatng-pont numercal computatons that may be needed to mplement advanced feedback control algorthms. In recent research [3], we have developed a low-cost DAC platform whch allows mcrocontrollers to be programmed by Matlab and Smulnk thus provdng an nexpensve tool for data acquston and control tasks. Ths platform s well suted for tasks that requre graphcal user nterface and/or advanced computatonal capabltes, but do not requre strngent hardware performance. It uses the advanced computng power of Matlab, the graphcal user nterface of Smulnk, and

2 Parallax Inc. s Basc Stamp (BS) mcrocontroller [4] to provde an envronment whch allows users to mplement sophstcated control desgns usng nexpensve hardware. Furthermore, t provdes a lbrary of BS functons for Smulnk, allowng for drect nteracton between user-defned Smulnk block dagrams and sensors/actuators connected to the BS. The framework of reference [3] s sgnfcant for several reasons. Frst, t overcomes the two lmtatons common to most mcrocontrollers, vz., the lack of advanced con-based software nterface for effcent development of control algorthms and the lack of a graphcal user nterface (GUI) to allow ntutve nteracton. In partcular, our DAC toolbox overcomes these two lmtatons by extendng the capabltes of Smulnk to low-cost mcrocontrollers. Second, our DAC platform s very economcal snce t requres the use of only an off-the-shelf BS mcrocontroller (under $5) and obvates the need for overhead software such as RTW requred for most PC-based DAC systems, thus further lowerng the cost of acqurng such a system. Ths affords an opportunty to students to conduct ndustry-style rapd control prototypng and hardware n the loop experments. Thrd, the ablty to nterface and program mcrocontrollers usng the ntutve graphcal programmng envronment of Smulnk provdes the flexblty and versatlty of equppng a wde array of undergraduatelevel laboratores (physcs, measurement systems, feedback control, and mechatroncs) at an economcal cost usng our DAC platform. Fourth, our mcrocontroller-based DAC system s nherently portable due to ts small sze and low power requrement, extendng ts benefts to students who can acqure ther personal DAC system for capstone desgn projects and for expermental research. In ths paper, we llustrate several mportant features of the aforementoned DAC platform [3] by performng lqud level control of a coupled, two-tank system. qud level control s ubqutous n ndustral applcatons, e.g., food processng, effluent treatment, power generaton plants, pharmaceutcal ndustres, water purfcaton systems, and ndustral chemcal processng. Thus, the study of lqud level control encompasses many engneerng dscplnes, e.g., electrcal, chemcal, cvl, and mechancal. In ths paper, we employ the classcal proportonal-plus-ntegral (PI) controller for lqud level control of a coupled, two tank system [5]. To effectvely utlze ths control methodology, the system parameters, e.g., tank dmensons and pump characterstcs, must be known exactly. However, ths may not be feasble snce system parameters may change due to the effects of scalng n tanks and orfces, agng of pump, and wear. Thus, to determne the system parameters, we conduct an expermental system dentfcaton study whch uses the nherent nonlnear dynamcs of the two tank system. Ths paper s organzed as follows. Frst, we outlne the basc features of the Matlab data acquston and control platform [3]. Next, we descrbe the coupled, two-tank system model and formulate system dentfcaton and control desgn objectves. Then we outlne a method to perform system dentfcaton. In the next secton we desgn a PI controller. Then we perform expermental valdaton of the results of earler sectons. Fnally, we draw some concludng remarks. Matlab Data Acquston and Control Platform In ths secton, we brefly outlne the salent features of a software platform [3] that enables the use of an nexpensve mcrocontroller such as BS for data acquston and control tasks. Ths framework allows Matlab and Smulnk to seamlessly nteract wth the BS mcrocontroller;.e., usng the features of Smulnk, ths software platform allows for: ) automatc generaton of PBasc code (the natve programmng language for the BS mcrocontroller) for a varety of sensors and

3 actuators that can be connected to the BS mcrocontroller; ) programmng of the BS mcrocontroller; and ) data communcaton between the BS mcrocontroller and Matlab. In partcular, ths framework provdes a lbrary of Smulnk functons, whch allows for drect nteracton between user-defned Smulnk block dagrams and sensors/actuators connected to the BS mcrocontroller. See reference [3] for a detaled overvew of ths software platform. Component Overvew The PC-based data acquston and control platform of reference [3] s composed of two man components, hardware and software. The hardware requred for ths platform s the BS mcrocontroller and user selectable sensors and actuators. The software requred for ths platform s Matlab wth seral communcaton capablty and Smulnk. We refer to the software mplementaton of ths PC-based data acquston and control platform as the BS data acquston and control toolbox. The BS mcrocontroller s used as a hardware nterface that relays sensor data to and actuator data from a PC. Furthermore, the BS mcrocontroller communcates wth the PC usng a seral communcaton port. For further detals on the BS mcrocontroller, see the Parallax webste [4]. The BS data acquston and control toolbox supports several dfferent types of nput and output sgnals (see reference [3] for a comprehensve lst). However, n ths paper, only analog nput and output sgnals between ±5 Volts are consdered. To convert an analog nput sgnal to a dgtal format for use on a PC, we utlze the TC9 analog to dgtal converter (AD) IC, whch s natvely supported n the BS data acquston and control toolbox. The TC9 AD IC s a -bt AD and s manufactured by near Technology Inc. It can support up to 8 sngle nput channels, whch can be used as 4 dfferental nputs. For addtonal detals on the TC9 AD IC, see webste of near Technology []. Analogously, to convert an actuaton sgnal, whose representaton s n a dgtal format, to an analog output sgnal, we utlze the MAX537 dgtal to analog converter (DA) IC, whch s natvely supported n the BS data acquston and control toolbox. The MAX537 DA IC s a -bt DA and s manufactured by Dallas Semconductor Inc. It can support up to 4 sngle output channels, whch can be used as dfferental outputs. For addtonal detals on the MAX537 DA IC, see webste of Dallas Semconductor [7]. Matlab s the prmary software package used n the BS data acquston and control toolbox. In addton to provdng functonalty for the Smulnk toolbox, the Matlab scrptng language allows for hgh level development of functons that can nterface wth specfc hardware, n partcular the seral communcaton port. Reference [3] explots ths capablty by usng bult-n seral communcaton functons provded by Matlab (supported by versons. and hgher) to communcate data wth the BS mcrocontroller. Furthermore, the BS data acquston and control toolbox utlzes the graphcal user nterface capabltes of Smulnk to allow users to construct block dagrams nterfacng wth sensors and actuators. Smulnk Block Dagram Desgn usng BS Toolbox Overvew The BS data acquston and control toolbox allows users to mplement control desgns and to develop data acquston and control panels wthn Smulnk block dagrams. Ths toolbox provdes a Smulnk custom lbrary, whch contans blocks that nterface wth sensors and actuators connected to the BS. See Fgure for a complete lstng of sensor and actuator blocks currently avalable n the BS lbrary. To use sensors and actuators avalable from the BS lbrary, users must desgn a control panel or control algorthm usng a predefned, blank, Smulnk template model fle. The key property of ths template model fle s the ncluson of functons used to program the BS mcrocontroller just before the Smulnk block dagram s executed. When a user saves ths

4 model fle and runs the Smulnk block dagram, a set of predefned operatons are ntated before the Smulnk block dagram s executed. These operatons are as follows: ) defne global varables to ntate seral communcaton; ) generate PBasc code based on the sensor and actuator blocks used n the Smulnk block dagram to be excuted on the BS mcrocontroller; ) program the BS mcrocontroller va the seral communcaton port; and v) start the Smulnk block dagram f the BS mcrocontroller was successfully programmed, otherwse stop the block dagram and dsplay an error message. IOBlock: The man functon of ths block s to: ) ntate seral communcaton between the BS mcrocontroller and Matlab; ) send and receve data between BS and Matlab; and ) termnate ths seral communcaton lnk. Asde from seral communcaton, another mportant task of IOBlock s to provde the samplng perod of a gven Smulnk block dagram. Ths feature allows blocks that requre the samplng perod to be used properly, e.g., an ntegrator block or a tmer block (clock). The samplng perod s denoted as the amount of tme requred for one cycle of the Smulnk block dagram to execute. The samplng perod provded by the IOBlock s an averaged samplng perod that s calculated by averagng the tme taken to run a specfed number of cycles of the Smulnk block dagram. The number of cycles used for averagng s user defnable, such that the user can adjust the resoluton of the samplng perod,.e., a larger number of cycles wll provde a fner resoluton of the samplng perod compared wth a smaller number of cycles. However, ths procedure of obtanng a samplng perod does not provde the exact samplng perod for each Smulnk block cycle; thus, ths BS data acquston and control toolbox does not enforce any real-tme requrements,.e., enforcng a specfc samplng perod for the Smulnk block dagram s not permssble. AtoD_TC9 block: Ths block receves voltage data from an TC9 AD IC. The TC9 AD IC s a -bt AD (-bt plus an addtonal sgn bt) that has 8 sngle nput channels, whch can be used as 4 dfferental nputs. Furthermore, ths IC s controlled by the BS va the seral perpheral nterface (SPI). Fgure : BS Sensor and Actuator Block brary In ths paper, we use only the blocks labeled IOBlock, AtoD_TC9, and DtoA_MAX537 from Fgure. The followng are bref descrptons of these blocks: DtoA_MAX537 block: Ths block sends voltage data suppled by the Smulnk block dagram to the MAX537 DA IC. The MAX537 DA IC s a -bt DA (-bt plus an addtonal sgn bt) that has 4 sngle output channels, whch can be used as dfferental outputs. Furthermore, ths IC s controlled by the BS va the seral perpheral nterface (SPI).

5 Coupled Two-Tank System Model Fgure shows a two degree-of-freedom, state-coupled, two-tank system. Ths system conssts of two tanks wth orfces and lqud level sensors at the bottom of each tank, a pump, and a lqud basn. The two tanks have the same dameters and can be ftted wth dfferng dameter outflow orfces. In ths expermental setup, the pump provdes the nfeed to Tank and the outflow of Tank becomes the nfeed to Tank. The outflow of Tank s empted nto the lqud basn. The followng condtons, wth regard to the system dynamc model, are used to descrbe the levels of lqud n Tanks and : ) the lqud level n Tanks and, denoted as ( t ) and ( t), respectvely, are measured usng two pressure sensors; ) ( t ) s always greater than cm and less than 3 cm; ) ( t ) s always greater than cm; v) the desred lqud level n Tank s greater than cm and less than cm for t > ; and v) the voltage appled at the nput termnals of the pump, denoted by V p( t ), s between and Volts. The nonlnear mathematcal model descrbng the lqud levels of Tanks and s gven by [5] t) A ( t) + BV ( ), () ( p t t) C ( t) D ( ), () ( t a K p where A g, B a, C g, and A A A a D g, a s the cross-sectonal area of A the outflow orfce at the bottom of the Tank, A s the cross-sectonal area of the Tank, K p s the pump constant, g s the gravtatonal acceleraton, and,. Fgure : Two Degree-of-Freedom, State- Coupled, Two-Tank System System Identfcaton Objectves In ths paper, we assume that the system parameters ABC,,, and D or equvalently a, A, K p, and g, for,, are not known exactly. As prevously dscussed, the uncertanty about system parameters may arse due to corrosve buldup n the lqud level system or the deteroraton of the pump motor. Thus, a control desgn based on the dealzed values of the system parameters may lead to performance deteroraton for the closed-loop system. In a subsequent secton, we outlne a seres of experments, whch explot the nonlnear dynamcs of () and (), to numercally estmate the system parameters A, BC,, and D.

6 Control Objectves Gven a desred, constant, lqud level for Tank, denoted as d, desgn a control nput V p( t) such that lm ( t. Ths objectve s to be t ) d acheved wth the assumpton that ( t ) and ( t ) can be measured. Furthermore, the control desgn s to utlze estmates of the system parameters. To quantfy ths control objectve, we defne the lqud level trackng error for Tanks and,.e., ( ) and ( ), as follows: t t () t () t d () t () t d, (3), (4) where d s some desred, constant lqud level for Tank, whch usng () s characterzed as a d d. a nearzed Error System Model To facltate the desgn of a controller to acheve the aforementoned control objectves, we begn by descrbng the lnearzed error system model (see reference [5] for detals) ( t) α ( t) + βu( t), (5) () t α () t + β (), t () where u() t Vp () t Vpd and V pd s the desred pump voltage, whch usng () s characterzed gd as V pd a. Furthermore, the parameters Kp α, β, α,and β n (5) () are defned as, respectvely, A α, β B, 4d (7) D C α, and β. 4 4 d d Next, we develop the transfer functon models for the lnearzed error system of (5) () by takng the aplace transform of (5) and () and rearrangng terms to yeld ( s) β, u ( s ) s α ( s) β, ( s ) s α where ( s ) [ ( t )], ( s ) [ ( t )] u ( s ) [ u ( t )] (8) (9), and wth denotng the aplace operator [8]. System Identfcaton In ths secton, we outlne an expermental method to determne the parameters of the coupled two-tank system gven n () and (). Specfcally, motvated by the nonlnear dynamcs of () and (), we outlne a set of experments that yeld the system parameters A, BC,, and D. In experment, we frst obtan numercal values for the ratos A and C. In B D experment, we obtan the numercal value for the parameter A, whch usng the numercal value of A obtaned n experment yelds B B. In experment 3, we obtan the numercal value for the parameter D usng the same procedure as n experment, whch usng the numercal value of C obtaned n experment D yelds C. Experment In ths experment, we obtan expressons that relate the ratos A and C to steady state B D condtons of the pump and the lqud level of Tanks and. Specfcally, we begn by notng that f a fxed voltage s appled to the pump, then the lqud level of Tanks and wll reach steady state after a suffcent amount of tme. Ths characterstc can be proven by drectly solvng the dfferental equatons () and ()

7 gven that V p( t ) s constant. Next, the soluton of () and () under steady state condtons,.e., wth ( t) and ( t), yelds A Vpss, () B ss C ss, () D respectvely, where ss, ss, V pss are the steady state lqud level of Tanks and, and the steady state pump voltage, respectvely. From () and (), by applyng a fxed voltage V pss to the pump and drectly measurng the lqud level of Tanks and after a suffcent amount of tme,.e., numercal values of Experment ss ss and ss A and B, we can determne C. D In ths experment, we fll Tank wth a fxed amount of lqud and record the tme hstory of the lqud level n Tank as t s dscharged. Usng ths tme hstory, we can determne a relatonshp between the system parameter A and the lqud level n Tank at a gven tme nstant. To do so, we solve the dfferental equaton () wth V ( t) and obtan p t) ( t) A( t ), () ( t where t s the ntal tme and ( t ) s the ntal lqud level n Tank. Note that equaton () s only vald when the rght hand sde of () s nonnegatve,.e., pror to full dscharge of Tank snce once Tank s fully dscharged the rght hand sde of () becomes negatve. To determne when the lqud n Tank s fully dscharged, we set ( t) n () and solve for the correspondng tme, resultng n * ( t) t + t, (3) A where t * s the tme of full dscharge of Tank. Usng (3), the analytc expresson for the lqud level n Tank s gven by ( t) ( t) A( t t) t < + t A ( t) t + t () t. A (4) Thus, f we choose a tme nstant t wth ( t) t < + t, or equvalently when the A lqud level n Tank s stll dschargng, then () can be rewrtten to yeld the parameter A as follows ( t) ( t ) A. (5) ( t t ) Next, substtutng (5) nto (), we can determne the parameter B as follows ( t) ( t ) ss B. ( t t) () Vpss Note that the determnaton of the numercal value for B n () requres the use of the expermental data obtaned from experments and,.e., the steady state values ss, Vpss from experment and t, t, ( t ), ( t) from experment. Experment 3 In ths experment, we fll Tank wth a fxed amount of lqud and then obtan the tme hstory of the lqud level n Tank as t s dschargng. Usng ths tme hstory, we can obtan a relatonshp between the system parameter D and the lqud level n Tank at a gven tme nstant. Smlar to experment, usng the soluton of (), we obtan a

8 relatonshp between the system parameter D and the lqud level n Tank as follows ( t) (ˆ) t D, (7) (ˆ t t ) where tˆ s a tme nstant before Tank s fully dscharged. Now, substtutng (7) nto (), we can determne the parameter C as follows ( t) (ˆ) t ss C. (8) (ˆ t t) ss Note that the determnaton of the numercal value for C n (8) requres the use of the expermental data obtaned from experments and 3,.e., the steady state values ss, ss from experment and t ˆ, t, (ˆ), t ( t) from experment 3. Proportonal-Plus-Integral (PI) qud evel Control In ths secton, we outlne a PI lqud level control desgn that s based on the lnearzed model of (8) (9). As descrbed prevously, the lqud level control problem conssts of controllng the lqud level of Tank to a desred, constant lqud level d. From (9), to control the lqud level n Tank, we must use the lqud level of Tank,.e., (), t as a pseudo control nput. Ths necesstates control of lqud level of Tank to a desred lqud level d, whch serves as a command sgnal for the Tank controller u (t), or equvalently V ( t). Referrng to Fgures 3 and 4, we desgn PI controllers for Tank,,, as follows KI C ( s) KP +, (9) s where K P s the proportonal gan and K I s the ntegral gan. To ad n determnng the p controller gans K P and K I, the desred closedloop responses for Tank are specfed n terms of a settlng tme T s and percent overshoot o performance crteron. It then follows from reference [8] that the settlng tme and percent overshoot crteron can be related wth the closed loop dampng rato ξ and natural frequency ω for Tank, as follows n ξ b, () b + π 4 ωn, () T ξ o where b log. Thus, usng () (), the controller gans K P and K I that acheve the desred closed-loop responses for Tank are obtaned as follows [5] K s ( ξ ω + α ) P n, () β ω n K I, (3) β where α and β gven n (7) depend on the parameters ABC,,, and D, whch are obtaned va system dentfcaton. Expermental Valdaton Ths example explores the expermental system dentfcaton and lqud level control of a two tank, lqud level system usng the BS mcrocontroller and the BS data acquston and control toolbox of reference [3]. The twotank, lqud level system s commercal laboratory equpment developed and marketed by Quanser Consultng Inc. [9].

9 Fgure 3: Smulnk Block Dagram of the PI Controller for Tanks and

10 Fgure 4: PI Controller for Tanks and Fgure 5: Calbraton and System Identfcaton Smulnk Block Dagram

11 As descrbed earler, the lqud levels of Tanks and are measured usng pressure sensors, where the output sgnal of the sensor s a voltage n the range of to 5 Volts correspondng to the lqud level n Tanks and. A calbraton gan for the pressure sensors s then determned,.e., the gan that converts the sensor output voltage to centmeters of lqud s cm computed as. The pressure sensor 4. Volt analog sgnals are suppled to the TC9 AD and are utlzed n the Smulnk block dagram, as shown n Fgures 3 and 5. Usng the MAX537 DA, the BS supples a controlled voltage sgnal to the pump, whch controls the lqud level of Tanks and. System Identfcaton In performng experment, as outlned earler, we appled a constant 7 Volt sgnal to the pump V pss, whch produced a steady state lqud level n Tank of ss.75 cm and a steady state lqud level n Tank of ss.4 cm. Next, we performed experment outlned earler by fllng Tank wth.4 cm of water,.e., ).4 cm, ( t and obtaned the lqud level of Tank as t was dscharged. Fgure (a) shows the lqud level tme hstory of Tank. From Fgure (a), the ntal tme of dscharge t s.5 sec and the tme of full dscharge t * s sec. To compute the system parameter A, we chose t sec, whch corresponds to a lqud level n Tank of ( t ) cm. Usng (5) and (), A and B are computed numercally to be.3978 and.83, respectvely. Next, we performed experment 3 outlned earler, by fllng Tank wth 5.8 cm of water,.e., ) 5.8 cm, and obtaned the ( t lqud level of Tank as t was dscharged. Fgure (b) shows the lqud level tme hstory of Tank. From Fgure (b), the ntal tme of dscharge t s 59.5 sec and the tme of full dscharge * t s 7 sec. To compute the system parameter D, we chose t ˆ sec, whch corresponds to a lqud level n Tank of ( t ) 7.8 cm. Usng (7) and (8), C and D are computed numercally to be.45 and.3938, respectvely. Next, we used the lqud level tme hstores gven n Fgure to obtan numerous tme nstances t and tˆ, correspondng to many estmates of the system parameters ABC,,, and D. We then averaged, ndvdually, the estmates of ABC,,, and D and obtaned the followng: A. 438, B.5, C. 448, D For comparson, manufacturer specfcatons of the Tank and cross-sectonal areas and orfces areas produce the nomnal system parameters A.58, B.99, C.58, and D.58. PI qud evel Control Next, we mplement the PI control archtecture dscussed earler to control the lqud level of Tank. Referrng to Fgure 3, two AD blocks labeled Tank qud evel and Tank qud evel, are used to mport the sgnals ( t ) and ( t ). These sensor sgnals are then used for the PI controller of Tanks and, wheren the control sgnal V p( t) s sent to a DA block labeled Pump Voltage for output to the pump. In ths experment, we specfy the desred lqud level of Tank to be d cm. Next, we desgn the control gans KP, KI,, usng () (3) gven a set of settlng tme and overshoot performance crteron. The resultng control gans are shown n Table. Fgures 7 9 show the tme hstory of ( t ) and ( t ) usng the control gans gven n Cases 3 gven n Table. Note that the controller desgn for the system s obtaned by lnearzng the nonlnear model near the predetermned steady state level of Tank. Therefore, the response s not guaranteed to satsfy the mposed constrants.

12 evel for Tank (cm) 8 evel for Tank (cm) Tme (seconds) Tme (seconds) (a) (b) Fgure : (a) Experment ( ) Response and (b) Experment 3 ( ) Response t t Table : Settlng Tme and Percent Overshoot Requrements and PI Controller Gans T s o T s o K P K I K P K I Case sec.5% 3 sec.5% Case 35 sec.5% 5 sec.5% Case 3 sec.5% 7 sec 3.5% Actual evel for Tank Desred evel for Tank 8 5 evel for Tank (cm) 9 evel for Tank (cm) Tme (seconds) (a) Tme (seconds) (b) Fgure 7: Response of (a) ( ) and (b) ( ) usng the PI Controller Gans of Case t t

13 4 4 Actual evel for Tank Desred evel for Tank 8 5 evel for Tank (cm) 9 evel for Tank (cm) Tme (seconds) (a) Tme (seconds) (b) Fgure 8: Response of (a) ( ) and (b) ( ) usng the PI Controller Gans of Case t t 4 4 Actual evel for Tank Desred evel for Tank 8 5 evel for Tank (cm) 9 evel for Tank (cm) Tme (seconds) (a) Tme (seconds) (b) Fgure 9: Response of (a) ( ) and (b) ( ) usng the PI Controller Gans of Case 3 t t

14 Concluson In ths paper, we utlzed a Matlab and Smulnk based data acquston and control platform [3] to control a two degree-of-freedom, state-coupled, two-tank system. We provded a self contaned bref overvew of the DAC platform, outlned a comprehensve method to perform system dentfcaton, and provded a desgn of a set-pont trackng controller to control the lqud level of the two-tank system. In addton, we expermentally valdated the system dentfcaton method, whch produced estmates of the two-tank system parameters. Fnally, we expermentally obtaned the response of the set-pont trackng controller, whch showed good trackng performance usng several sets of performance crteron. The paper llustrates that the ntegraton of a low-cost BS mcrocontroller and the Matlab and Smulnk based DAC toolbox can be used for data acquston, expermental system dentfcaton, and advanced feedback control. Acknowledgements Ths work s supported n part by the Natonal Scence Foundaton under an RET Ste grant 7479 and a GK Fellows grant 3378 and the NASA/NY Space Grant Consortum under grant References. Onlne: s/matlab/, webste of MathWorks Inc., developer and dstrbutor of Matlab.. Onlne: s/smulnk/, webste of MathWorks, Inc., developer and dstrbutor of Smulnk. 3. A. Panda, H. Wong, V. Kapla, and S. ee, Matlab Data Acquston and Control Toolbox for Basc Stamp Mcrocontrollers, Proceedngs of the IEEE Conference on Decson and Control, pp , San Dego, CA,. 4. Onlne: product_dbs-ic, webste of Parallax Inc., developer and dstrbutor of the Basc Stamp (BS-IC) mcrocontroller (access lnk for BS-IC product nformaton). 5. H. Pan, H. Wong, V. Kapla, and M.S. dequeroz, Expermental Valdaton of a Nonlnear Backsteppng qud evel Controller for a State Coupled Two Tank System, Journal of Control Engneerng Practce, Vol. 3, pp. 7 4, 5.. Onlne: etal.jsp?navidh,c,c55,c,c58,p48, webste of near Technology Inc., developer and dstrbutor of the TC9 AD (access lnk for product nformaton). 7. Onlne: vew.cfm?qv_pk5, webste of Dallas Semconductor Inc., developer and dstrbutor of the MAX 537 DA (access lnk for product nformaton). 8. R.N. Bateson, Introducton to Control System Technology, Prentce-Hall, Upper Saddle Rver, NJ, Onlne: ml/products/fs_product_challenge.asp?lang_ codeenglsh&pcat_codeexprot&prod_co der-posserv, webste of Quanser Consultng Inc., developer and dstrbutor of the DC motor test-bed (access lnk for product nformaton). Bographcal Informaton Anshuman Panda was born n New Delh, Inda. He s currently pursung a dual B.S/M.S. degree n Electrcal Engneerng and expects to graduate n June 7. He s a member of Tau Beta P. He has worked as a teachng and research assstant wth responsbltes n the area of mechatroncs.

15 Hong Wong was born n Hong Kong, Chna. In June of and, he receved the B.S. and M.S. degrees, respectvely, n Mechancal Engneerng from Polytechnc Unversty, Brooklyn, NY. He s a member of P Tau Sgma and Tau Beta P. He worked for the Ar Force Research aboratores n Dayton, OH, durng the summers of and. He s currently a doctoral student at Polytechnc Unversty and expects to graduate n June 7. Hs research nterests nclude control of mechancal and aerospace systems. Engneerng from Sung Kyun Kwan Unversty, Seoul, Korea, n 99 and the M.S. degree n Mechancal Engneerng from Polytechnc Unversty, Brooklyn, NY, n. From 99 to 997, he worked for Samsung Engneerng Co., td. n Korea. He s currently contnung research at Polytechnc Unversty as a doctoral student. Hs research nterests nclude lnear/nonlnear control and mechatroncs. Vkram Kapla s an Assocate Professor of Mechancal Engneerng at Polytechnc Unversty, Brooklyn, NY, where he drects an NSF funded Web-Enabled Mechatroncs and Process Control Remote aboratory, an NSF funded Research Experence for Teachers Ste n Mechatroncs that has been featured on WABC-TV and NY News, and an NSF funded GK- Fellows project. He has held vstng postons wth the Ar Force Research aboratores n Dayton, OH. Hs research nterests are n cooperatve control; dstrbuted spacecraft formaton control; lnear/nonlnear control wth applcatons to robust control, saturaton control, and tme-delay systems; closed-loop nput shapng; spacecraft atttude control; mechatroncs; and DSP/PC/ mcrocontroller-based real-tme control. He receved Polytechnc s Jacob s Excellence n Educaton Award and 3 Dstngushed Teacher Award. In 4, he was selected for a three-year term as a Senor Faculty Fellow of Polytechnc Unversty s Othmer Insttute for Interdscplnary Studes. He has edted one book and publshed four chapters n edted books, book revew, 39 journal artcles, and 8 conference papers. He has mentored 7 hgh school students, 38 hgh school teachers, undergraduate summer nterns, and seven undergraduate capstone-desgn teams. In addton, he has supervsed three M.S. projects, two M.S. thess, and three Ph.D. dssertatons. Sang-Hoon ee was born n Seoul, Korea. He receved the B.S. degree n Mechancal