Indirect Dynamic Recurrent Fuzzy Neural Network and Its Application in Identification and Control of Electro-Hydraulic Servo System

As the affine nonlinear system is characterized by differential relations between the states, an adaptive dynamic recurrent fuzzy neural network (ADRFNN) taking only the measurable states as its inputs and describing the system's inner dynamic relation by its feedback matrix is proposed to evaluate the unknown dynamic nonlinear functions including nonlinearity, parameter uncertainty and load disturbance. The adaptive laws of the adjustable parameters and the evaluation errors' bounds of ADRFNN are formulated based on lyapunov stability theory. Also the stable indirect ADRFNN controller (ADRFNNC) with gain adaptive VSC (GAVSC) for the estimation errors by ADRFNN and the load disturbance are synthesized. It can overcome the shortcomings of the structural expansion caused by larger number of inputs in traditional adaptive fuzzy neural networks (TAFNN) taking all states as its inputs. The application in electro-hydraulic position tracking system (EHPTS) shows that it has an advantage over the TAFNN controller (TAFNNC) in steady characteristics of system. Furthermore, the chattering of the system's control effort is weakened and so the system possesses greater robustness.