Research On Idenfication Of Hammerstein Model

The identification of Hammerstein model belongs to a branch of nonlinear system identification. Hammerstein model consists of a non-linear blocks and a linear blocks, which connect in series to describe the non-linear characteristics. In many non-linear experimental models, compared with other non-linear model structure. Hammerstein model is simpler, easier to identify, less amount of calculation and better reaction of process characteristic, so it is used widely.This paper introduces the Hammerstein model, and then the main research focuses on a class of Hammerstein model which dynamic linear blocks is CARMA or called ARMAX model, such models are a class of colored noise in stochastic systems. This paper is in two different ways to research:according to identification algorithm: 1. Iterative algorithm.2. Recursive algorithm, according to non-linear part of the method:1. Non-linear part is the power function 2. Non-linear part is as a function of dead-zone. This paper puts forward a new idea of identification-a minimal probability approach. In order to obtain effectiveness of the different methods, as well as the results of comparison of different methods for each question, we have carried out a simulation study to obtain meaningful conclusions. The main contents of this paper are as follows:1. Introduce stochastic systems structure, the conception of noise-signal ratio and the definition of the error criteria.2. Describe the structure and mathematical description of Hammerstein model and its extended model.3. According to classification of non-linear blocks, introduce Hammerstein model identification methods specifically. According to the simulation results, the pros and cons of different methods are analyzed. 4. Introduce a minimum probability approach, then use simulation to study their effectiveness, and the future of this approach is put forward at last.