CONTROL OF A TIME DELAY SYSTEM USING NEURAL PREDITION ANDGAIN SCHEDULING

摘要

Consideration is given to positioning of a pneumatic servo-actuator using a hybrid control scheme withmodel prediction and gain scheduling. Time delay and steady-state error exist in the system due to the inherent staticfriction and air compressibility, which considerably limit the system positioning performance. The non-linear dynamicsof the system is modelled using a neural network of the group method and data handling (GMDH), a self-structuredand feedforward type neural network. A gain-scheduling algorithm is proposed and a hybrid PI control scheme ofneural prediction and gain scheduling considered improving the control behaviour. The effect of time delay is mostsignificant at small stroke actuation of fast response demand, and the pneumatic servo-actuator dynamics is lessdamped around the mid-stroke than any other position. Experiments were carried out and results show that the timedelay effect to the actuator positioning is eliminated and the steady-state error is significantly lowered by using thehybrid control scheme.