A variable structure control based on neuron networks compensation has been developed. The control results applied to a hydraulic servo system showed that the proposed control can improve its adaptation and robustness. Good dynamical and static state performances can be obtained as well. It has valuable reference in practical engineering applications.