Abstract | U ovome radu je razvijen eksperimentalni postav s umjetnim pneumatskim mekim mišićem i pripadnom didaktičkom opremom tvrtke Festo. Za takav sustav je izrađen simulacijski model na primjeru regulacije pomaka mišića kao dizalice tereta. Kako bi se usporedbom simulacije i eksperimenta dobila detaljnija analiza dinamike i perfomansi, koristile su se dvije različite regulacijske tehnike, PID i LQR.
Prije razvoja simulacijskog modela, dan je detaljan opis pneumatskog mišića, poput fizikalnih specifikacija, tehnologije izrade, ograničenja i područja rada. Nakon toga je dan pregled matematičkih modela sile, poput geometrijskog, energetskog, te empirijskog modela, koji su sastavni dio dinamičkih modela odnosno mehaničkih modela s obzirom na mehanizam u eksperimentalnom postavu i ostalu korištenu opremu. Također, razmotrene se razlike između različitih modela i aproksimacija različitih pristupa u modeliranju, gdje je konačno prihvaćen Sarosijev empirijski model sile, u svrhu izrade simulacijskog modela promatranog sustava.
Analizom postojećih regulacijskih tehnika na području pneumatskih mišića kao aktuatora, može se potvrditi složenost i vrijednost pojedinih regulacijskih tehnika. Najčešće korišteni su klizni regulator, te regulator s modelom predikcije, a koji mogu savladati nelinearnu promjenjivost dinamičkih koeficijenata uz utvrđenu robusnost na širok raspon pomaka, frekvencija i tereta.
Testiranjem naprednije regulacijske tehnike LQR, dobio se uvid u prednosti i nedostatke takvog regulatora, zbog čega je ova tehnika uspoređena s PID regulatorom, kao naprednija tehnika, jer se koristi zapis sustava u prostoru stanja. Osim toga, pokazalo se da LQR ima direktan utjecaj na pojedinu varijablu stanja, poput pomaka, brzine i tlaka, što je vidljivo na danim odzivima. To znači da indeks ponašanja koji se računa kroz minimizaciju težinske funkcije pripadnih matrica, diktira koje varijable je potrebno pojačati ili prigušiti ovisno o željenom ponašanju, tj. omjeru perfomansi i trošku takvog regulatora. |
Abstract (english) | In this study, an experimental setup with pneumatic artificial soft muscle is developed and with its related didactic equipment from the company Festo. For this system, a simulation model is designed based on an example of a displacement regulation as the load lifting device. A comparison between simulation and experiment can result with detailed dynamic and perfomance analysis, in which two different regulation techniques are used, PID and LQR.
Before the development of a simulation model, a detailed description of pneumatic muscle is provided, such as physical specifications, manufacturing technology, its limitations and operating work area. Behind which, a review of mathematical force models is given, such as geometrical, energetic, and empirical model, which are the integral part of the dynamic models regarding mechanical model with regard to the used experimental setup and equipment. Also, differences between different models and approximations in modelling approaches is considered, where as a final conclussion, Sarosi empirical force model is accepted, and with this force model, described system is developed.
By analysis of existing regulation techniques in the field of pneumatic muscles as actuators, complexity and value can be determined for each individual regulation technique. Often used, are the sliding regulator and model prediction regulator, in which they can adjust for the nonlinear change of the dynamic coefficients, by showing robustness for a wide range of displacement, frequency and loads. By testing the more advanced regulation technique LQR, insight in advantages and shortcomings is given, which is why it is compared with PID regulator, as more advanced technique, because it is using a state space representation of the system. Besides, it turned out that LQR has a direct effect on the individual state variable, such as displacement, speed and pressure, which is visible on the resulting responses. Which means that the behaviour index, which is calculated through the minimization of the weight function of associated matrices, dictates which variable need to be amplified or dampened depending on the desired behaviour, i.e. the ratio of performance and cost of such a regulator. |