Neural adaptive dynamic surface control of an electro-hydraulic loading system for rail grinders

Electro-hydraulic loading system (EHLS) has been widely utilized in rail grinding industry for rail maintenance. Accurate tracking of the desired grinding force is critical to keep the rail surface at expected level. However, uncertain nonlinear friction, unmodeled nonlinearities of hydraulic systems, and high-frequency motion disturbances caused by random rail corrugations usually deteriorate the force tracking performance. To resolve this problem, a neural adaptive dynamic surface control strategy is designed. To efficiently address the uncertainties and disturbances of the single-rod EHLS, a radial basis function neural network (RBFNN) is developed and trained by adequate measurement data to approximate the nonlinear friction, then two nonlinear disturbance observers that integrate the RBFNN are constructed to estimate the matched and unmatched disturbances. Furthermore, the neural network weights are updated adaptively via the tracking error, which enhance the dynamic learning ability of the neural network. And, dynamic surface control (DSC) instead of traditional backstepping method is applied to construct the composite controller to avoid the “explosion of complexity”. Simulation and experimental results of force tracking prove the effectiveness and potential application value of the designed control strategy.