Control Optimization and Dynamic Programming-Informed Sizing for Novel Energy-Recovering Hydraulic Actuation
This study explores optimal control and sizing of a recently introduced efficient architecture for hydraulic actuation. Previous work established a physical model of the architecture posed as a single-input single-output (SISO) system in which the ratio of two hydraulic pump/motor swash plate angles served as the control input for regulating actuation speed. The architecture was heuristically sized and controlled within the context of a hydraulic elevator. High-fidelity simulations of the system demonstrated an upwards of 75% decrease in energy consumption compared to a throttling architecture. Monte Carlo simulations are now used to achieve optimal sizing of the system. Several uniformly random points in the design space are chosen and evaluated using Dynamic Programming, which provides both a deterministic and optimal value for energy efficiency of the system. Aggregation of evaluated points reveals a region within the three-dimensional space wherein the architecture is optimally sized for efficiency. Dynamic Programming is then used to inform efficient rule-based control strategies. Control techniques learned from Dynamic Programming suggest efficient operation of the system results through the maximization of pump/motor 1 displacement and the use of the auxiliary electric motor during retraction of the hydraulic cylinder. Dynamic Programming informed system achieved a 61% level of optimality. Additionaly, it exhibited a 21% improvement over a heuristically sized and controlled version. It is anticipated that optimal control and sizing guidelines presented are applicable within the context of other hydraulic actuation technologies for which the studied architecture may be used.