DISCUSSION ON THE APPLICATION OF WIRELESS ACTIVE SENSING UNIT

Transcription

1 4 th Worl Conference on Structural Control an Montorng 4WCSCM-26 DISCUSSION ON HE APPLICAION OF WIRELESS ACIVE SENSING UNI FOR SRUCURAL CONROL C.H. Loh an K.C. Lu Natonal awan Unversty, ape 06-7,awan Y. Wang, R.A. Swartz an J.P. Lynch Unversty of Mchgan, Ann Arbor, MI , USA Stanfor Unversty, Stanfor, CA , USA P.Y. Ln Natonal Center for Research on Earthquake Engneerng, awan Abstract Performance aspects of wreless sensors, nclung the relablty of the wreless communcaton channel for real-tme ata elvery, an ts applcaton to bulng structural control s explo n ths stuy. he applcaton of the MR-amper to actvely control a half-scale -story steel bulng excte by a shakng table s stue usng the wreless actve sensors. Wth one MR amper nstalle on each floor, structural responses urng sesmc exctaton are measu by wreless sensors an communcate to the MR amper s wreless actve sensng unt where an H 2 controller has been mplemente. he wreless actve sensor s responsble for the recepton of response ata, etermnaton of optmal control forces, an ssung of comman sgnals to the amper. Fnally, the control performance s scusse by usng both the w communcaton system an the wreless actve sensng unt. Introucton ratonal structural control technology has employe wre-base systems to o the communcaton for structural control an structural ata collecton. However, the nstallaton of these wre-base control systems can be expensve n labor, tme an cost. In orer to uce these monetary an tme expenses for the nstallaton of wre-base systems, new technologes n embee systems an wreless communcaton have been aopte n acaemc an nustry research for wreless sensng an montorng. he use of wreless communcaton for SHM ata acquston was llustrate by Strser an Kremjan [998]. More recently, Lynch et al. extene the work by embeng amage entfcaton algorthms nto wreless sensng unt [Lynch et al. 2004] an has alreay proven relable when use n leu of coaxal wrng n structural montorng systems [Wang et al. 2005; Lynch et al. 2006]. he avantages of usng wreless sensng unts for structural health montorng have been verfe whle n the structural control area, many challenges must stll be explo n greater etal. o captalze on low-cost sem-actve actuators nstalle n hgh ensty n a sngle structure, wreless communcaton s propose to mnmze the hgh-cost of coaxal wres. A prototype Wreless Structural Sensng an Control (WSSCon system has been propose [Lynch et al. 2006] for structure response mtgaton. he software wrtten to operate the wreless sensors uner the real-tme requrement of the control problem s presente n ths stuy. he promsng performance n applyng wreless communcaton an embee computng technology nto a real-tme feeback structural control system were presente. hs paper presents the expermental verfcaton of usng both fully centralze control an fully ecentralze control strateges wth the mplementaton of sem-actve control evces an aoptng a wreless control an sensng system. wo major research rectons are emphasze: ( evelop the theoretcal bass of fully Loh, Lu, Wang, Swartz, Lynch, an Ln

2 centralze an fully ecentralze control algorthms an mplement them on the WSSCon system for feeback control, an (2 expermental verfcaton of the wreless communcaton for structural control. Comparson on both the w control system an the wreless control system s mae. he Expermental Setup of WSSCon System WSSCon (Wreless Structural Sensng an Control System s a prototype system esgne for real-tme wreless structural sensng an feeback control [Wang et al an Lynch et al. 2006]. In the WSSCon system, wreless communcaton s use for the feeback of structural response ata to wreless sensors servng as the control kernel (.e. to calculate control solutons base on receve ata. It conssts of a wreless sensng unt wth the embee control software n the mcrocontroller an the acton boar. Fg. shows the basc archtecture of the wreless sensng an control unt. he wreless sensng unt s responsble for measurng the ynamc response of the structure an an actuaton boar s esgne to convert a gtal sgnal (6 bt to analog sgnal for control purpose. For the calculaton of control forces at each tme-step, the wreless sensng unt was also esgnate as the control kernel (terme the wreless control unt whch utlzes ts local embee computng resources to quckly process sensor ata, generate control sgnals, an apply control commans to structural actuators wthn the esgnate tmestep uraton. On each floor a WSSCon system s nstalle an connecte to the MR-amper va a VCCS (voltage to current converter. he operaton of the WSSCon system an the program flow can be explane from the server se an the embee coe se whch s summarze nto fve steps: Step : Boot up the system from both the PC server an embee system n the WSSCon system; Step 2: he server checks all the wreless sensng an control unts n the network through the wreless transcever (as shown n Fg. 2a for etal escrpton of step 2 an reset clock/counter the synchronzaton process s verfe; Step : From the embee coe, each wreless control unt broacasts a beacon to all other unts n the network sequentally announcng that a new tme step begns. he communcaton latency of each wreless transmsson nees to be carefully conse. For centralze control t s set to 20 mllsecons for each sensng unt to receve the sgnal from the other unt n ths stuy (as shown n Fg. 2b. For the case of centralze control, a total of 60 ms s neee for all the unts to receve all the sgnals from other unts. hen to calculate the control force whch 5 mllsecons may nee. So a total of 95 mllsecons s requ to conuct the centralze actve control. hs s the reason to set the samplng rate of 0Hz for centralze control; Step 4 an Step 5: Feeback the results to the PC server an ext the control program. he bulng control system nclues: Sensors (velocty meters an accelerometers, MR-ampers, VCCS (voltage converts to current system, an WSSCon system, as shown n Fg.. Because two fferent 0 ~ 5V Analog Sensors Sensng Interface Sensor Sgnal Dgtzaton 4-channel 6-bt Analog-to-Dgtal 4-channel 6-bt Analog-to-Dgtal Converter ADS84 Converter ADS84 SPI Port Control Interface Computatonal Core 28kB External SRAM CY6228B Parallel Port 8-bt Mcro-controller Amega28 SPI Port 6-bt Dgtal-to-Analog 6-bt Dgtal-to-Analog Converter AD5542 Converter AD5542 UAR Port Control Sgnal Generaton Wreless Communcaton 20kbps 24XStream Wreless Moule wth a 2.4GHz ranscever Structural Actuators that Accepts -5V ~ 5V Comman Sgnal Fgure : Har ware archtecture of wreless control unt (Lynch et al Loh, Lu, Wang, Swartz, Lynch, an Ln 2

3 (a (b Fgure 2: (a Synchronzaton of ata communcaton an check start from sever coe n Step 2, (b Data broacast an tme sharng among sensng unts for centralze control n Step. control strateges are use n ths stuy, the setup of the control system s also fferent for usng fferent control strateges. For e-centralze control, three nvual sub-systems are conse an each sub-system has ts own control evce. he velocty meters are nstalle at both ses of the amper to measure the relatve splacement between two ajacent floors (usng w system. hs sgnal wll fee nto the WSSCon system to calculate the control force. he floor acceleraton ata from each partcular floor (or sub-system s collecte wrelessly by the WSSCon system an a Kalman estmator wll be use to estmate the full state whch ncorporates the measu floor relatve velocty to calculate the control voltage. Base upon a measure of the shaft velocty of the amper an the estmate control force from the WSSCon system, a es comman voltage for control can be acheve. hrough VCCS (Voltage-Current Converter system the control current wll be estmate an fee nto the amper for control purpose. As for centralze control, each WSSCon system wll collect all of the floor acceleraton responses wrelessly nstea of just a sngle floor response. Both centralze an e-centralze control algorthms wll be ntrouce n the followng secton. Formulaton he Control Algorthm Sensng unt Accelerometer (-th floor Velocty meter Power Supply (24 Volt VCCS (convert voltage to current : 0~2amp -th floor Recor Relatve velocty (-th floor WSSCon Unt Calculate Control Force (MR-Moel Convert Force to Voltage Broacast Floor Acceleraton Acton Boar (6 bt D/A converter 0.0 V ~.2 V Fgure.: Control setup usng wreless sensng an control unt an ts connecton wth the MR-amper. Loh, Lu, Wang, Swartz, Lynch, an Ln

4 wo fferent control algorthms are use for the control experment: fully centralze an fully ecentralze control. In evelopng the fully centralze control algorthm, the entre structural system an all measurements are conse to generate the control gan together. Snce the full-state measurement s mpossble an can not be mplemente n the realstc structural control applcaton, a Kalman estmator, whch transforms measurement responses nto the full-state vector, wll be use to estmate the full-state responses of the structural system an to calculate the control gan. Conserng the lmte number of sensors n the structure, the Kalman estmator nees to be selecte to transform the lmte set of measurements nto the full-state measurement estmaton. he H 2 control algorthm wll be use. Frst the objectve functon, nclung the system matrx an the term of control evces, efne as: [ ] 6 6 [ ] [ ] [ ] ( k = 0 J = z k Q z k u k R u k Here, the weghtng matrces, Q an R, are relate to the full state an the control forces, respectvely. After transformng the functon by the proceures of varaton of methos, the new form of the functon s the Rcatt equaton: A P A A P B (2 R B P B B P A 2Q= P ( n whch the matrx P s the soluton of the Rcatt equaton, obtane by teratve calculaton. Conserng the lmte number of sensors n the structure, the Kalman estmator nees to be selecte to transform the lmte measurements nto the full-state measurement estmaton, as shown below: z [ k ] = A z [ B u[ L( y [ C z [ F u[ (, s Here, L s the soluton of the Rcatt equaton from the Kalman formulaton [Dyke et al. 996]. he control force can then be replace by the followng form u[ k ] = Gz = GAˆ cs [ k ] = G( A zˆ[ GL ms y[ B G LC c LF G z [ GLy where Âcs s the mofe system matrx n relatng to control. Fnally, Eq. (7 proves the estmate fullstate response ( z to estmate the control forces n the real response. Because the control force was generate usng the full-state response vector, t s calle the fully centralze control soluton. he fully ecentralze control emphaszes on the control of local system aroun the locaton of control evce by usng the response measurements from sensors aroun the local sub-system so as to generate the control force nstantly an effcently. From ths pont of vew, a complete structural system can be separate nto many sub-systems an each sub-system contans ts own sensor measurements an control evces. In ths stuy, conser three nepenent sub-systems n the structure, the state-space equaton an the measurement equatons become: [ (4 k ] = A y [ = ( C y [ = ( C 2 y [ = ( C, 6,2 6, 6 = ( B ( D ( D ( D u [ E && x,,22, u [ 2 g u [ u [ [ j= 2, j=, j=,2 ( D ( D, j ( D,2 j, j u [ ( F j u [ ( F j j u [ ( F, g && x [,2 g, g && x [ && x [ (5 Loh, Lu, Wang, Swartz, Lynch, an Ln 4

5 Here, R s not a matrx any more an may vary wth fferent control forces because every objectve functon focuses on one control force of ts own sub-system. Snce there are three control evces an three acceleraton measurements at each floor as feeback, therefore only three subsystems for ths ecentralze control metho are nvolve an the objectve functons are shown below: J = z [ Q u [ R u [ 6 6 k = 0 2 = k = 0 = 6 6 k = 0 J z [ Q u [ R u [ J z [ Q u [ R u [ Here, each objectve functon focuses on each control force so that the number of control gans s also three. he Rcatt equaton, whch s erve from each objectve functon, can be presente (, 6 (2 A (, 6 (, PA AP B R BP B B 6 PA 2Q = P (7 he control gan can be escrbe as: u[ = (2 R B, PB, B, PA = G 6 (8 hen the Kalman estmator nees to be mofe as: z [ k ] = A z [ ( B u [ L (( y [ ( C z [ ( F u [ (9 j, s j j j, It means that base on the measurement from nvual sub-system the estmaton of full-state s requ. where ncates the -th estmator whch s wth the same length of the full-state vector. An the control force can be obtane from u [ k ] = G ( z [ k ] (0 (6 Analyss of he Expermental Results From the shakng table test of the -story steel frame wth the mplementaton of one MR-amper on each floor, the control effectveness usng both w an wreless control systems are examne. he MRamper s nstalle n the mle of each floor system wth a V-type bracng system. Verfcaton on the communcaton of control comments s stue frst, an then the control effectveness usng fferent control strateges s scusse. In ths control experment three fferent communcaton systems are use: ( NCREE ata acquston system (Pacfc Seres 5500 Dgtal Contonng System wth samplng rate of 200Hzwth Smulnk for control, (2 WSSCon Centralze control system (wth samplng rate of 0Hz, ( WSSCon De-centralze control system (wth samplng rate of 50Hz. Detal escrpton of the three system s escrbe as follows: (a. NCREE Laboratory Control System: Combne the NCREE gtal contonng system, Smulnk an the DSpace I/O boar, the tratonal laboratory control test usng w system wll be use an the results wll be use as the benchmark moel for comparson. he samplng rate for ths control experment (ether centralze or ecentralze control s 200Hz an H 2 control algorthm wll be use. he result from ths control system wll be serve as a benchmark for comparson wth the other wreless control soluton. (b. Centralze control wth samplng rate of 0 Hz: In the wreless centralze control system the sensng unt on each floor wll collect all the acceleraton ata broacaste from all other sensng unts on fferent floors. If a hgh samplng rate s chosen, then ata wll be lost urng the communcaton, so a samplng rate of 0 Hz s use n ths case. (. De-centralze control wth samplng rate of 50 Hz: In the ecentralze control, each wreless sensng an control unt only receves sgnal from the sensor n each specfc subsystem. here s no nee to wat for the sgnals transmtte from other Loh, Lu, Wang, Swartz, Lynch, an Ln 5

6 sensng unts, so a hgher samplng rate of 50Hz can be use for ecentralze control. Data collecte usng WSSCon system an NCREE-w system wll be scusse so as to verfy the controllng capablty of the wreless control system. Valaton of the WSSCon system Comparson of the comman voltage nees to be verfe: one s the sgnal generate from the mcrocontroller (calculate from an embee program of the wreless sensng unt n gtal format an the other one s generate from the actuaton boar n analog form. Comparson on the recore acceleraton ata ether rectly from the sensor or broacaste from other sensng unts s also verfe. Comparson of the comman voltage collecte from the actuaton boar an from numercal smulaton s mae for both the centralze an ecentralze control. Comparson between the calculate amper force from the numercal theoretcal moel (sol lne an the estmate amper force from WSSCon wreless system (ash lne s also compa, as shown n Fgure 4. From all these valaton stues of the WSSCon system, t s prove that: the commen voltage s correct, there s no loss of ata urng the communcaton, an the actuaton boar can generate the control voltage correctly. (a (b (c Fgure 4: (a Example shows the measu relatve velocty ( st floor of the MR-amper, (b Comparson between the calculate amper force between the theoretcal moel (sol lne an the estmate amper force from wreless system (ash lne, (c Dece voltage to be fee nto the MR-amper. Valaton of wreless control effectveness Performance evaluaton of the MR-amper urng the control experment plays an mportant one on control effectveness. In centralze control the amper s hysteretc behavor s examne by usng both the wreless system an w system (NCREE system uner the same samplng rate of 0Hz s compa, as shown n Fgure 5a for centralze control. Fgure 5b shows the comparson for ecentralze control wth a samplng rate of 50 Hz. o verfy the control effectveness usng wreless sensng an control system, the control results are compa wth the w control system wth samplng rate of 200Hz. Comparson on the floor acceleraton an splacement usng w control system (200Hz an the wreless control system for case of centralze control (0Hz an ecentralze control (50Hz s also conucte. It s prove that the wreless sensng an control system can almost reach the same control effectveness as the w system. Comparson on the control effectveness among fferent control methos s also mae. Fgure 6a shows the comparson among uncontrol case, numercal smulaton of fully centralze control an wreless centralze control wth 0 Hz, an Fgure 6b s the comparson wth the ecentralze control usng wreless wth 50 Hz. he results Loh, Lu, Wang, Swartz, Lynch, an Ln 6

7 ncate that both centralze an ecentralze control usng wreless sensng system can reach almost the same control effectveness as the wre control system. (a (b Fgure 5: Comparson on force-stroke an force-velocty relatonshp of MR-amper usng both w an wreless control system; (a for centralze control wth samplng rate of 0Hz, (b for ecentralze control wth samplng rate of 50 Hz(usng EL Centro groun moton. Conclusons hs stuy examnes the potental of wreless communcaton an embee computng technologes for real-tme structural control applcaton. Base on the mplementaton of the prototype WSSCon system n the structure, both the centralze an ecentralze control methos are apple to control a -story steel structure wth an MR-amper n each floor subject to EL-Centro earthquake exctaton on shakng table. he followng conclusons are rawn:. he WSSCon concept has been emonstrate n ths stuy, the performance of ths system s shown to be superor to the case of sem-actve control system an nearly comparable to the w control system. he control effectveness usng ether centralze or ecentralze control shows relable performance of the system. 2. he latency lmtaton assocate wth wreless communcaton nees to be carefully examne. Dfferent control algorthm to be embee n the WSSCon system may have fferent latency whch has to be fgu out n orer to not loss the ata n communcaton.. In the ecentralze control esgn, only local sensor nformaton has been use to generate the control sgnal that s sent to the ampers of each control subsystem. hs e-centralze control algorthm can be carre out successfully for a large-scale structural system. he smulaton results emonstrate that ecentralze control proves almost the same control effectveness (nces as the centralze control.an avantage of the ecentralze control s the robustness n the face of falure of the control system urng a strong earthquake. Gven one subsystem fals, the other subsystems can be capable of compensatng accorngly an ensure sutable global performance of the system. 4. Fnally, the wreless sensor networks are a promsng technology capable of operaton for a real-tme communcaton an control system. he wreless feeback structural control wth embee computng can uce the cost by eracatng cables n the control system, usng a low-cost Loh, Lu, Wang, Swartz, Lynch, an Ln 7

8 mcrocontroller for the control kernel an a hghly flexble an aaptable system confguraton because of wreless communcaton. Fgure 6: Comparson on the acceleraton responses n each floor for un-control case, numercal smulaton an expermental results where expermental result s: (a for centralze control, (b for ecentralze control (EL Centro groun moton. Acknowlegements hs research s support by both Natonal Scence Councl uner grant No. NSC Z an Central Weather Bureau. he authors woul also lke to express ther grattue to NCREE techncans for ther assstance when conuctng the shakng table experments. References Dyke, S.J., Spencer, B.F., San, M.K. an Carlson, J. D., Moelng an control of magnetorheologcal ampers for sesmc response ucton, Smart Materals an Structures, 5, 996, Dyke, S.J., Spencer, B.F., Quast, P. San, M.K., Kaspaar, D.C. an Soong,.., Acceleraton feeback control of MDOF structures, ASCE, J. of Engneerng Mechancs, 22(9, 996, Ln, P.Y., P.N. Roschke, C. H. Loh an C. P. Cheng, Sem-Actve controlle base-solaton system wth MR ampers an penulum system,, WCEE, Vancouver, August, 2004, paper #69. Lynch, J. P., Sunararajan, A., Law, K. H., Kremjan, A. S., an Carryer, E., Embee amage etecton algorthms n a Wreless Sensng Unt for Attanment of Operatonal Power effcency, Smart Materals an Structures, IOP, Vol., No.4, 2004, Lynch, J. P., Wang, Y., Lu, K. C., Hou,. C., an Loh, C. H., Post-sesmc Damage Assessment of Steel Structures Instrumente wth Self-nterrogatng wreless Sensors, Proceengs of the 8 th Natonal Conference on Earthquake Engneerng, San Francsco, CA, USA, Aprl 8-22, Lynch, J. P., Wang, Y., Swartz, R. A., Lu, K. C. an Loh, C. H., Implementaton of a close-loop structural control system usng wreless sensor networks, submt to J. of Structural Control an Health Montorng (2006. Renz, B. an Serno, G, estng an moelng a sem-actvely controlle steel frame structure equppe wth MR ampers, Structural Control an Health Montorng, 2004 Spencer, B.F., Dyke, S.J., San, M.K. an Carlson, J. D., Phenomenologcal moel for magnetorheologcal ampers, J. Engneerng Mechancs, ASCE, 2, 997, Straser, E. G. an Kremjan, A. S., A moular, wreless amage montorng system for structures, Report No.29, John A. Blume Earthquake Engneerng Research Center, Department of Cvl & Envronmental Engneerng, Stanfor Unversty, CA 998. Wang, Y., Lynch, L. P., Law, K. H., Desgn of a low-power wreless structural montorng system for collaboratve computatonal algorthms, Proceengs of SPIE 0 th annual Int. Symposum on Nonestructve Evaluaton for Health Montorng an Dagnostcs, San Dego, CA, March 6-0, Loh, Lu, Wang, Swartz, Lynch, an Ln 8