Positioning control of hydraulic cylinder with unknown friction using on/off directional control valve

Abstract

In this study, a prediction-based positioning control scheme is proposed for the hydraulic cylinder controlled by a solenoid operated on/off directional control valve. The discrete-valued input, low switching frequency and significant delay of directional control valve make the control problem very complex. Only a discrete-valued control input can be used here; meanwhile, the input has switching frequency constraint and time-delay. Existing methods such as pulse-width modulation control and sliding-mode control are not suitable for this problem, because chattering may arise due to the control input restrictions. The newly proposed prediction-based positioning control scheme consists of two parts: a switching law based on predictions of future states and learning algorithms which learn required predictive models online. According to accurate predictions, the switching law can control the hydraulic cylinder to target position in an optimal way, and chattering is avoided. Meanwhile, the required predictive models are identified by a generalized growing and pruning for radial basis function network and a recursive least square estimation algorithm in real time. Essentially, the problems caused by input restrictions, time-delay and model uncertainty are solved by the accurate identifications to some well-designed models. The control scheme is verified by physical experiments. Fast and accurate positioning control can be achieved for the hydraulic cylinder with unknown nonlinear friction.