Research On Identification Methods Of Hammerstein Systems Under Non-ideal Data Condition

Most systems in industrial production exhibit some degree of nonlinear characteristics,and modeling the production process using the system’s mechanism costs a lot of manpower.In recent years,the block-structured Hammerstein model has been widely used to describe practical nonlinear systems.However,In process of the actual acquisition of sampling data,due to the changes in the environment,artificial errors,and the existence of sensor hardware restrictions,the obtained data sets often have outliers,time-varying dynamic disturbances,dual-rate sampling,and other non-ideal conditions.If the influence of the above factors is ignored and the Hammerstein system is directly identified,the recognition results will have a large estimation error,or even fail to meet the requirements.Therefore,the identification algorithm of the Hammerstein system under the conditions of studying non-ideal data has important theoretical value.The contents of this thesis are as follows:(1)For the Hammerstein system affected by outliers,two robust recursive identification algorithms based on auxiliary models are proposed.In order to overcome the sensitivity of the traditional two-norm criterion function to outliers,by introducing the logarithmic mechanism into the criterion function and combining the idea of the auxiliary model,the logarithmic stochastic gradient algorithm and the logarithmic recursive least squares algorithm based on the auxiliary model are respectively proposed.The main idea of the proposed method is to suppress the effect of outliers by constructing weight factors.Compared with the auxiliary model recursive least squares algorithm,the proposed algorithm can obtain stable identification results and higher parameter estimation accuracy,and has stronger robustness.(2)For the Hammerstein system affected by dynamic disturbance,a robust hierarchical stochastic gradient identification algorithm with forget factor and a weighted robust hierarchical least squares identification algorithm are proposed.In order to realize the real-time tracking of the dynamic disturbance,it is regarded as an independently changing time-varying parameter,and the extended identification model is further obtained.Then,with the help of the principle of hierarchical identification,the extended model with dynamic perturbation is decomposed into two sub-models,and a robust hierarchical stochastic gradient algorithm with forget factor is derived,and the identification performance of the algorithm is improved by using the multi-innovation method.Considering the slow convergence speed of the stochastic gradient algorithm,the robust hierarchical least squares algorithm is further used for identification,and the weighted idea is used to modify the algorithm,and a weighted robust hierarchical least squares identification algorithm is obtained.Finally,the simulation experiments show that the proposed algorithm can not only achieve high precision parameter identification results,but also track dynamic disturbances in real-time(3)For the dual-rate Hammerstein systems with interferences of non-zero mean Gaussian noises.Firstly,by using the polynomial transformation technique,the dual-rate Hammerstein system is converted into a model that can be directly identified based on the dual-rate sampled data,and the recursive least squares algorithm is utilized for identification.Considering that the identification result obtained by the recursive least squares algorithm is biased.In order to obtain unbiased parameter estimates,based on the principle of bias compensation,parameters in the bias term are solved by introducing a non-singular matrix and an extended information vector,thus a bias compensation based recursive least squares algorithm is derived.The numerical simulation results show that the bias compensation recursive least squares algorithm can effectively compensate for the identification results of the recursive least squares algorithm and obtain unbiased estimation of parameters.