Identification Of Transfer Function Model For Industrial Processes With Time Delay Under Load Disturbance

System identification has been widely used in dynamic response modeling for industrial processes,advanced control system design,and online self-tuning optimization.In industrial control systems,the adopted internal model control methods,model predictive control,and Smith predictor etc,are generally based on the process dynamic response transfer function model to design the controller and tune the system performance parameters.For many industrial process operations,there are the input delay and output response delays.Concerning the increasingly used computer-aided sampling control systems,it is a basic premise to precisely estimate the integer type delay parameter for effectively identifying the model parameters and designing the control systems.Besides,load disturbances with unknown dynamics are often encountered in engineering applications and identification tests in practical applications.Using the existing identification methods developed by assuming no disturbance or white noise to identify such a system will provoke non-negligible or even large estimation error.Hence,it has theoretical significance and engineering application value to study the identification of industrial processes with time delay subject to load disturbance.The main content and contributions of the thesis include:(1)Identification of the output error model(OEM)for sampling systems with time delay under colored noise.An interpolation method in combination with a one-dimensional searching algorithm is proposed in the thesis to estimate the delay parameter,such that the mixed integer programming problem with non-convex feature of simultaneously identifying the integer type delay parameter and the real type linear model parameters could be solved.Meanwhile,the instrumental variable(IV)method and auxiliary model(AM)method are used to eliminate the influence from colored noise.Therefore,consistent estimation of the OEM parameters could be realized.Moreover,based on the stochastic process theory,consistent convergence analysis is given for the proposed identification method with a proof.(2)Identification of the OEM parameters for sampling systems with time delay subject to unknown time-varying load disturbance.The output response arising from such disturbance is viewed as a time-varying parameter to estimate together with the system model parameters.By introducing an adaptive forgetting factor,a matrix consisting of adaptive forgetting factors is constructed to update the covariance matrix of parameter estimates,so as to improve the tracking estimation performace on time-varying load disturbance response and the consistent estimation convergence on time-invariant system model parameters.A one-dimensional searching algorithm is used to determine the integer type delay parameter.At the meantime,consistent convergence analysis and proof are given for the proposed identification algorithm.(3)Identification of the OEM parameters for dual-rate sampling systems with time delay subject to unknown time-varying load disturbance.By constructing an AM to predict the missed data caused by the dual-rate sampling,the dual-rate identification problem is converted to a esearch problem in a framework of unified sampling.Besides,the output response arising from disturbance is viewed as a time-varying parameter for estimation.Based on the separation strategy,two recursive least-squares(RLS)identification algorithms are developed to alternatively estimate the system model parameters and the load disturbance output response.According to different characteristics of these two types of parameters,two adaptive forgetting factors are introducedto improve the consistent estimation convergence on system model parameters and the tracking estimation performace on time-varying disturbance response.Moreover,sufficicent conditions are given for unbiased estimation against a constant type disturbance.(4)Identification of the Hammerstein nonlinear systems with time delay subject to unknown time-varying load disturbance.Based on the overparametrization method and the separation strategy,two RLS identification algorithms are established to iteratively estimate the model parameters and time-varying disturbance response.The multi-innovation strategy is adopted to augment the dimension of the innovation matrix,so as to improve the estimation accuracy on the time-invariant system model parameters.The time-varying disturbance response is estimated in a quick tracking manner by using the single innovation.Meanwhile,consistent convergence analysis and proof are given for the proposed identification algorithm.