A improved Elman neural network with application to fault diagnosis of pneumatic valve actuator

Because of the inherent uncertainty, changeable, nonlinear characteristics, faults diagnosis is a very difficult task in the pneumatic actuator system. In addition, slow convergence speed exists in the original Elman network, so, a model, an improved Elman neural network, was presented to achieve fault diagnosis in this paper. The adjustable weights between the context units and output units are embedded in the improved model. So, they make the speed of convergences more quickly and make the detection of nonlinear dynamic system better. The severity of the common faults contains control valve faults, servomotor faults, etc. Through the simulation research based on MATLAB platform, the new method is efficient at diagnosing the pneumatic actuator system´s faults from the experimental findings.