Font Size: a A A

Subspace Identification Method Algorithms Software Development For 2-D Discrete Systems

Posted on:2017-11-10Degree:MasterType:Thesis
Country:ChinaCandidate:J X ChengFull Text:PDF
GTID:2348330491461010Subject:Control engineering
Abstract/Summary:PDF Full Text Request
With the development of two-dimensional discrete system theory and application, in the fields of image restoration, communication engineering, multidimensional filtering network, iterative learning controller design and optimization, performance evaluation, establishing accurate 2-D state-space model has the vital significance. Actually,2-D model is not the simple transplant of 1-D model, its evolution along with two freedom directions and mutual coupling between state vector, the new definition of eigenvalues, state response and system stability criterion have great differences between 1-D systems, thus the problem of 2-D system identification faces great challenges.Therefore, this study is devoted to subspace identification methods applied to the 2-D causal, recursive, separable-in-denominator model (abbreviated as "2-D CRSD model").2-D CRSD systems, as the subclass of 2-D systems, have good structural characteristics, which can present some possibilities of 2-D systems identification. Subspace identification methods have simple form and are especially suitable for MIMO systems, have the advantages of good robustness, without complex iterative optimization, simple calculation, which directly utilize measurement data to achieve consistent estimation of piant parameters.Aiming at this identification problem, the study, firstly, under the unifying formulation, three corresponding improved bias-eliminated subspace identification methods are proposed for 2-D CRSD systems in the combined deterministic-stochastic case. Based on principal component analysis (PCA) and instrumental variable, several novel 2-D CRSD closed-loop subspace identification methods are developed. Besides, instrumental variables are adopted to improve the identification accuracy. For online identification of 2-D CRSD systems, novel recursive subspace algorithms are proposed for both open and closed loop 2-D CRSD systems. The main goals of the proposed recursive subspace approaches are to lighten the burden of computation and storage, and improve the model accuracy, until achieve the requirements.Numerical examples and injection molding application have been used to verify the proposed identification approaches. The simulation examples demonstrate that when the random noise in an acceptable level, the identified model show good fitting with the real model, can predict the real output with high accuracy. In the presence of model and operation uncertainty or random disturbance in practical applications, the recursive method can continuously update the identified model, timely tracking the real parameters. Undoubtedly, the recursive identification methods are of great significance for the adaptive control, adaptive prediction, online tracking and adjusting control strategy or on-line fault diagnosis.
Keywords/Search Tags:2-D discrete system, subspace identification algorithm, closed-loop model, recursive identification methods, instrument variables, consistency and efficiency
PDF Full Text Request
Related items