Performance analysis of a new energy-efficient variable supply pressure electro-hydraulic motion control method

Electro-hydraulic actuation is used in many motion control applications due to its high power density, excellent dynamic response and good durability. However fluid power actuation has been shown to be very energy inefficient, with an average efficiency for fluid power systems across all industries of 22% in the USA. This is a very significant problem, given that 3% of the energy used by mankind is transmitted in this way. The key challenge for researchers is to reduce energy losses in hydraulic actuation systems without increasing weight, size, and noise, and without reducing speed of response. Conventional high performance electro-hydraulic motion control systems use a fixed supply pressure with valve-controlled actuators (FPVC). This is inherently inefficient due to the need to use a valve to throttle the flow required by each actuator in the system down to match its load pressure. In this paper, a new load-prediction based method is proposed, in which the supply pressure is varied to track the pressure required by any actuator branch. By implementing this model based approach using a high response servomotor-driven pump, it is shown that the dynamic response remains excellent. The load model not only allows feedforward control for servomotor speed based on the motion demand, but also feedforward for the control valves to supplement conventional proportional-integral feedback control. The new variable supply pressure valve-controlled (VPVC) method is investigated in simulation and experimentally using a two-axis hydraulic robot arm supplied by an axial piston pump. The performance has been rigorously compared with the same robot arm using,a fixed supply pressure and proportional-integral joint position control. Experimental results showed that up to 70% hydraulic power saving was achieved, and that the dynamic tracking errors for VPVC were about half that for FPVC as a result of using feedforward control.