Abstract: Joint friction has a significant impact on the dynamical behavior of the mechanism. In scenarios where heavy loads or mechanisms have ball hinges, components are commonly connected by sliding joints, whose joint friction are related to the reaction force, and therefore the reaction force cannot be eliminated in the dynamical equations in generalized coordinate form. This leads to a forward dynamics described by differential-algebraic equations with too many unknown variables, and the real-time solution of the dynamical equations cannot be easily achieved. To address this problem, a surrogate model based fast solution method for dynamics of mechanisms with friction is proposed. Firstly, the smooth dynamical equations with continuous Stribeck friction model related to the reaction force are formulated. Secondly, the surrogate model is constructed by using a single hidden layer neural network to simplify the forward dynamics equations into ordinary differential equations, which greatly reduces the unknown variables that need to be solved. Finally, the effectiveness and superiority of the proposed method are verified by the hydraulic lifting system of the segment assembly robot. The numerical solutions prove that the proposed method significantly improves the calculation efficiency with guaranteed accuracy, with a maximum error of 0.023 4 mm and an average saving of 99.89% of the solution time.

Key words: mechanism dynamics, friction model, surrogate model, fast solution