Hybrid control of hydraulic directional valves: Integrating physics-based and data-driven models for enhanced accuracy and efficiency
Tobias Glück, Amadeus Lobe, Adrian Trachte, Matthias Bitzer, Wolfgang Kemmetmüller
Abstract
In this paper, we tackle the challenge of accurately controlling the position of the valve spool in hydraulic 4/3 two-stage directional control valves utilized in mobile applications. The pilot valve's overlapping design often leads to a significant dead zone, negatively impacting positioning accuracy and necessitating a sophisticated controller design. To overcome these challenges, we introduce a control strategy founded on a control-oriented model. This model enables systematic compensation for the dead zone, pressure-induced flow fluctuations, and the solenoid's nonlinearities, optimizing the valve's operation for enhanced tracking performance, as verified by test bench measurements. Addressing the limitations inherent in traditional physics-based design methodologies, we suggest approximating the system's primary nonlinearities with a data-driven surrogate model. We propose a solution tailored for systems that rely on minimal sensor information. By merging the advantages of both physics-based and data-driven models, we formulate a hybrid control strategy. This comprehensive approach not only ensures high tracking performance but also has the potential to expedite the commissioning process for new valve variants.