POSITION CONTROL OF A PNEUMATIC SYSTEM USING ADAPTIVE INTELLIGENT METHODS
MetadataShow full item record
Behrad Dehghan: Position Control of a Pneumatic System using Adaptive Intelligent Methods. M.A.Sc. Thesis, Queen’s University, June, 2012. A large body of research is devoted to the development of advanced control techniques to improve the positioning performance of pneumatic systems, which are known to be highly nonlinear systems. Although model based controllers show good results, the requirement for a system model makes these methods difficult to implement. So-called intelligent algorithms, such as neural networks and fuzzy rule based controllers, are attractive since they do not require a model. The performance of these controllers can be enhanced by adding an adaptive mechanism to adjust controller parameters in a continuous on-line fashion. The objective of this thesis was to explore different adaptive intelligent controllers for position control of a pneumatic system. The application was the x-axis and z-axis of a pneumatic gantry robot. They were tested independently for their ability to track step and sine wave trajectories. The rodded x-axis cylinder was an example of a short stroke low friction application. The rodless z-axis cylinder was an example of a long stroke high friction application. Five different controllers were tested: 1) PID, 2) Fuzzy, 3) PID+Adaptive Neural Network Compensator (ANNC), 4) ANNonly and 5) Fuzzy Adaptive PID (FAPID). Results with FAPID and PID+ANNC showed improvement in tracking performance over PID by 60% for the rodded and 35% for the rodless cylinder. This level of improvement was expected given the adaptive nature of the controller. Unfortunately, both required significant effort to setup and tune. In order to reduce the tuning effort, a second adaptive mechanism was added to FAPID, to adjust output weights. Results with adaptive PID and modified FAPID (MFAPID) showed further improvement performance over PID by 87% for the rodded and 70% for the rodless cylinder (in addition to being easier to tune). To provide a measure of robustness, experiments were conducted at two supply pressures and three tracking frequencies. The fact that MFAPID was able to improve performance for both cylinders, is considered further evidence of its robustness. MFAPID is considered novel for two reasons: 1) fuzzy rule set is reduced in size relative previous work and 2) addition of an adaptive mechanism for output weights is new.