Direct Identification Approach Of Continuous Hammerstein Model

Industrial application background of subject was complex mechanical system dynamics modeling and design. Funding by the National Science Fundamental Project which named"Direct identification of nolinear continuous-time dynamic parametric models via wavelet modulating functions"(Project No:50875028), we developed research topic that was direct identification of continuous Hammerstein with measurement noise. Focusing on the following aspects:1. Direct identification algorithm of modulation least square of continuous Hammerstein was introduced. Introducing separation methods of nonlinear parameters, we use it to decompose product terms of reforming model parameters, then estimation parameters of continuous Hammerstein model were gained.2. Direct identification algorithm of modulation least square of continuous Hammerstein with measurement noise was introduced. Least-square characteristic of noise modulation was studied, so knowing its parameter estimation was biased parameter estimation. Modulation noise was processed by the generalized noise model, then we propose modulation generalized least squares algorithm to get the unbiased estimation of system.3. In high noise circumstance, modulation generalized least squares parameter estimation appear distortion, so algorithm of time-window wavelet de-noise was proposed to reduce noise of the high data on line, then estimating parameters of model by de-noise data. Studying the application of time-window wavelet de-noising in linear continuous dynamics model and the continuous Hammerstein model, we found that the de-noise data also has low noise, so proposed time-window wavelet de-noise modulation auxiliary variable method and time-window wavelet de-noise modulation generalized least square algorithm to obtain the high accuracy unbiased estimation of corresponding model parameters.4. Linear continuous dynamic parameter model and continuous Hammerstein model were establish for HAGC pressure closed-loop system of cold rolled flat machine. Using the proposed algorithm to identify the corresponding model parameters, we can get better identification result of parameters. Application example shows that the proposed method is effective and practicalResearch shows that: under the interference of measurement noise, this paper studies identification algorithm estimated continuous Hammerstein model parameters that have good anti-disturbance capacity. This algorithm can adapt to large industrial system model identification...