Continuous-time Model Identification From Discrete-time Sampled Data

System models play an irreplaceable role in modern control theory, and system identification is a key method to obtain the system models. Due to the popularity of digital computers, one paid much attention on the identification of discrete time mod-els. Continuous-time models have many advantages versus discrete time models, such as its parameters own physical meanings and it can deal with non-uniformly sampled date directly. Recently, the direct identification of continuous-time models is becoming a new research hotspot thanks to the developing of computers.This dissertation studies the identification of continuous-time models in time do-main and frequency domain respectively. Simulations and experiments are designed for the validation of the algorithms proposed.This dissertation utilizes two methods to deal with the differential operator, name-ly modulating function method and filter method. Firstly, these two methods are derivate theoretically and the procedures of them are presented. Secondly, we apply iteration method to optimize the bandwidth and searching method to identify the time delay of the system. Lastly, we carry on simulations which prove the effectiveness of those methods.This dissertation also deals with system identification in frequency domain when the amplitude response is lost. A new system is built so that the phase response data can be transferred into a series of complete frequency response data. When the new input and output equations are built, the subspace identification method is utilized. The zeros and poles can be calculated from the characters of the state matrix precisely. Numerical simulation and experiment show that this method can be very precise when dealing with identification problems from phase response data.