An electrically driven pump with an actuator system without control valves is a candidate for eco-friendly electrohydraulic systems. The efficiency of a system combined with an electric servo motor and a variable displacement pump can be maximized if the optimal rotational speed and displacement are used. This study applied efficiency maps derived from regression and reinforcement learning to the control problem. It was confirmed that under certain conditions, the two proposed methods can achieve the same output power with higher efficiency than that of the conventional control method, which uses a fixed rotational speed or fixed displacement of the pump. The proposed control method using an efficiency map exhibited a stable performance although it required preliminary data collection. The other proposed method using reinforcement learning can be run without any knowledge about the environment, although it is sensitive to the noise in the measured value.
ABE et al. (Thu,) studied this question.