Enhanced model-based iterative learning control and experiment research for multi-axis hydraulic road simulator

Abstract

The multi-axis hydraulic road simulator is an indispensable test rig in vehicle laboratory durability testing. Offline iterative learning control (ILC) based on frequency domain model is widely used to replicate the load conditions experienced by vehicles in the proving ground. Due to the strong nonlinearity of the hydraulic system and test specimens, classical offline ILC requires numerous iterations to generate the required driving signals. This process is not only time-consuming but also inevitably causes pre-damage to the specimens. To expedite the generation of driving signals, an enhanced model-based iterative learning control (EMILC) is proposed that combines the Quasi-Newton optimization algorithm with model-estimated optimal gains. The control strategy employs the Broyden-class optimization method to update the impedance matrix, reducing modeling errors and enhancing the robustness of iterations. Simultaneously, optimal learning gains are obtained during the iteration process based on the estimated model, thereby reducing damage to the specimens. Experimental results, replicating real spindle forces collected from Wheel Force Transducers (WFT) in the proving ground, demonstrate that EMILC improves convergence speed and tracking accuracy compared to existing offline ILC methods. This novel approach is not limited to durability testing and can be extended to other control systems requiring repetitive tracking tasks.