ADVANCED CONTROL WITH NI CRIO PLATFORM

cRIO je netipičen industrijski krmilnik zaradi svoje rekonfigurabilne vgrajene strukture. Namen magistrske naloge je seznaniti se s cRIO arhitekturo in videti, kje so meje cRIO krmilnika. cRIO krmilnik smo testirali na treh aplikacijah, in sicer magnetna levitacija, regulacija ravnovesja valja na letvi in na sistemu vezanih tlačnih posod. Magnetna levitacija je odptozančno netabilni proces, ki ima relativno hitro dinamiko. Zaradi tega smo regulacijski algoritem implementirali na FPGA, ki deluje stabilno okrog ene delovne točke. Sistem valja na letvi nima zahtevne dinamike tako kot magnetna levitacija in je zaradi tega kaskadna regulacijska struktura implementirana v Scan načinu delovanja. Pri kaskadni strukturi je zunanja zanka po položaju valja, medtem ko je notranja zanka po hitrosti valja. Na sistemu vezanih tlačnih posod smo uporabili adaptivno vodenje, in sicer metodo adaptacije povratnozančnega ojačanja. Algoritem adaptacije povratnozančnega ojačanja zaobjema identifikacijski algoritem v realnem času, in sicer metodo najmanjših kvadratov.cRIO is atypical indutrial controller because of his reconfigurable embedded structure. The goal of master thesis is to get to known the cRIO architecture and to discover its limitations. cRIO controller has been tested on three aplications, magnetic levitation, ball and beam system and pressure vessel system. Magnetic levitation is open loop unstable process, which has relatively fast dynamics. Because of that control algoritam was implemented on FPGA and it has been worked stable around one working point. Dynamics of ball and beam system is not so demanding as magnetic levitation system and because of that cacscade control was implemented in Scan mode. Cascade control consists of two control loops, outer ball position loop and inner velocity ball loop. On pressure vessel system adaptive control was implemented. Type od adptive system, which was implemented is adaptation of closed loop gain. For adaptive control, real time parametar identification is essential, in our case recursice least square method has been used