Simulation-Driven Universal Surrogates of Coupled Mechanical Systems: Real-Time Simulation of a Forestry Crane

Abstract

Preparing simulation-driven surrogates for a coupled mechanical system can be challenging because the associated mechanical and actuator dynamics demand high-fidelity numerical solutions. Introduced here is a surrogate-assisted universal actuator approach that can more easily make better predictions outside the training data for a coupled mechanical system. A UHS is proposed as an alternative to the standard method, which uses lumped fluid theory to define hydraulically actuated and coupled mechanical systems. The UHS has been developed by using a 1 D cylinder model with an approximated force model at the position and velocity levels. It was tuned to approximate predictions made for the same mechanism based on lumped fluid theory. CMA-ES was used to compare results and further tune the UHS during the training process to minimize differences between the standard and UHS approaches. By taking a surrogate-assisted monolithic approach, which reduces the number of differential equations, the UHS promises better computational performance leading to real-time simulation solutions. Using it to predict the behaviors of the simple four-bar mechanism and the forestry crane validated its robustness. The focus was on numerical accuracy and computational efficiency. The maximum PN-RMSE between the states of the approximated force model and lumped fluid theory were approximately 2.04 % and 6.95 %, respectively. Reaching a numerical solution using the new method was approximately 52 times faster than when using the standard lumped fluid theory method. The UHS can be applied in simulation, optimization, control, state and parameter estimation.