Identification of Hammerstein Model of Intelligence Sensor Based on Hybrid Neural Networks

An identification method based on hybrid neural networks for Hammerstein model is investigated in this paper to analyze the nonlinear dynamic system of intelligence sensor, and the corresponding algorithm is presented. In this model, the nonlinear dynamic characteristic of sensor is expressed by cascading a nonlinear static subunit (NLSS) with a linear dynamic subunit (LDS). According to the characteristic of the model, a PID nonlinear neural network (PID-NLNN) simulating the NLSS and a LDN linear neural network (LDN-LNN) simulating the LDS form a hybrid neural network (HNN), which is used to identify Hammerstein model. By means of the HNN approach, the parameter of the model can be identified and separated into two parts simultaneously, one part is the coefficient of the NLSS, the other is the coefficient of the LDS. The simulation has proved the efficiency of the proposed method.