Identification Of Non-Uniformly Sampled System Based On State-Space Model

The traditional approach is uniform sampling, which is the so-called single-rate sampling. However, with the increasing demand for network and communications in modern industry, more sensors or sub-control systems are introduced. Because of this demand, some difficulties appeared when the traditional single-rate system was used in the industrial system. With variety requirements of industrial control, multi-rate system has been widely applied in many areas. Non-uniform mul-tirate system is an extension of the general mul-tirate systems. It’s best to identify the non-uniform mul-tirate system on state space model directly, for it can not only solve MIMO systems, but also identify nonlinear time-varying systems.Based on the current research on non-uniform sapling systems, two identification algorithms were proposed for multivariable non-uniform sampling systems, namely, subspace identification method and hierarchical identification method. Subspace identification method is proposed in order to deal with the problem that model parameter are unknown, which offers numerically reliable state-space models for complex multivariable dynamical systems directly from sample data; The hierarchical identification method employs a Kalman filter to estimate the system parameters directly, and combining with the SVD decomposition. By changing the nature of the numerical calculation, this can overcome the defect of error accumulation and transport occur of recursive least squares method. In addition, the paper also demonstrates the convergence of singular value decomposition algorithm by the lemma and theorem.This paper describes non-uniform multirate systems in detail, and three non-uniform sampling system scenarios were elaborated. Besides, the establishment of state-space model of non-uniform sampling systems of various programs is also derived. On this basis, the two identification methods were proposed. By compared weather the identification system to track the original system or not, and the effect of estimation error convergence under different SNR values. The simulation results demonstrate the effectiveness of the proposed algorithm. Finally, we summarized the whole paper and proposed ideas for future work and research.