Parameter identification for nonlinear pneumatic cylinder actuators

Pneumatic actuators exhibit highly nonlinear characteristics due to air compressibility, significant friction presence and the nonlinearities of control valves. The unknown nonlinear parameters can not be directly measured once the actuators have been manufactured and assembled, which causes a great difficulty in pneumatic system modeling and control. A learning algorithm has been developed in this paper to identify the unknown pneumatic system parameters. The algorithm is initially developed and tested using the data generated by simulations. Then the algorithm has been extended onto the parameter identification using the data obtained from the real system measurement. The results revealed the characteristics of uneven distribution of friction parameters which are position, velocity, moving direction dependent. The results obtained in the paper can provide the manufacturers with the observation to the characteristics inside pneumatic cylinders.