Developing a novel Gaussian process model predictive controller to improve the energy efficiency and tracking accuracy of the pressure servo control system

Abstract

Pressure servo control plays a crucial role in the majority of industrial applications that use compressed air as their power source. How to reduce the energy consumption of the pneumatic system while ensuring the high accuracy of pressure control remains a key problem to be solved. This paper proposes a Gaussian Process (GP) 1 model predictive controller (MPC) applied to a pressure servo control system based on high-speed on-off valves. The smallest quantity of collected data is used to create the GP model. The entire model, which includes some predictive data, is produced once the generated model has been optimized. The GP model is combined with the MPC to perform comparative experiments using an industrial controller with a Field Programmable Gate Array (FPGA). The system responds quickly and with little tracking error in steady-state response studies as well as dynamic response experiments. The GP-MPC achieves the best outcomes in the dynamic comparison tests. The root mean square error (RMSE) is just 2.42 kPa, and the overshoot is less than 59.9% of the Proportional-Integral-Derivative (PID) controller and 68.5% of the sliding mode controller (SMC). Although the compressed air consumption of GP-MPC is basically the same as that of PID, it is significantly better than SMC, with a total saving of about 70.2%. All the experiments prove that the controller proposed in this paper can effectively improve the energy efficiency and tracking accuracy of the pressure servo control system.