Neural-based adaptive grinding force tracking control for pneumatic end-actuator with uncertain dynamic model constraints

Abstract

Precision component manufacturing requires precise grinding force control to achieve high surface quality and maintain process stability. However, controlling pneumatic grinding end-effector poses significant challenges due to strong nonlinearity, time-varying parameters, friction, and model incompleteness. To address these issues, this study proposes an adaptive force tracking controller based on neural networks. Within an adaptive backstepping framework, it integrates a Luenberger state observer with an online radial basis function neural network to compensate for unknown and varying dynamic characteristics in real time. Based on a Lyapunov-based strict stability proof, the designed controller ensures finite-time boundedness of all signals in the closed-loop system and convergence of the contact force tracking error to zero within a small neighborhood. Grinding experiments demonstrate that the proposed control method exhibits excellent tracking accuracy and fast response speed under both static and dynamic grinding conditions. This contributes to maintaining surface finish and enhances the precision and operational reliability of pneumatic grinding systems in complex machining scenarios.