Continuous Systems Identification Method Research Based On Step Test

Identifying the transfer function model is very important for us to acquire the system characteristics and adjust the controller parameters, moreover, it is the basis of most controller designing approaches. In most real industrial process, the complex signal is not permitted. And the step signal is simple, easy to obtain and widely applied. Hence, continuous systems identification method research based on step test has important theoretical significance and application value. Delay time extensively exists in most chemical processes and has significant impact on the control performance. In order to estimate model parameters of continuous processes with delay time, this dissertation deep studies continuous system identification method from step test signal on the basis of studying abounding literatures. The main dissertation work is as following.1. Focusing on the particularity of the step signal and step response, some arguments or methods about identification are summarized by studying previous research achievements, such as identifiability, classical representative points method and the problem of how to transfer time domain data to frequency domain information.2. Guided by robust identification based on time integral approach, an improved identification method for integrating processes with delay time is proposed. Using inverse Laplace transformation, the transfer function of output is transformed to time-domain expression; then, on the basis of integral characteristics of exponential function, the linear regression equation is obtained via time integral and replacing the exponential part; and the least square method is used to estimate the model parameters. By using a large number of simulations, the effectiveness of these introduced methods are proved.3. Surrounding the parametrization problem of transfer function with delay time from frequency response, a novel frequency domain iterative approach is proposed. In this method, first order approximation of Taylor expansion is used to linearize the time delay part and the new linear regression equation is acquired by recombining the obtained equation. Then, the model parameters are acquired by the least square algorithm. Finally, the iterative algorithm is used to optimize the model parameters further. Moreover, the proposed mothed is extended to closed-loop identification and decentralized closed-loop parameter identification for multivariable processes by computing the frequency response value of the identified object. The simulation experiments show that the proposed method has a fast convergence rate and can greatly optimize the model parameters efficiently.Taken together, this paper tries to explore with some innovative method in foundation of previous research achievements. The research topic is discussed from the view of time domain and frequency domain and proposes novel identification method of model parameters. A large number of simulation studies reveal that the innovative approaches are feasible and the identification methods based on step test agree with the test requirement of real industrial process.