Font Size: a A A

Recursive Least Squares Type Identification Methods For A Class Of Non-uniformly Sampled-data Systems

Posted on:2010-07-12Degree:MasterType:Thesis
Country:ChinaCandidate:Y J LiuFull Text:PDF
GTID:2178360278975396Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Non-uniformly sampled-data systems exist widely in modern industry.According to the sampling theory,periodically non-uniform sampling is more flexible than uniform sampling. In chemical process control,sampling data for some variables may be non-uniform because the limits of hardwares such as sensors etc.,which results in the non-uniformly sampled data (multirate) systems.For example,some variables which cannot be measured online are often obtained by laboratory analysis,which is usually non-uniform in time.Therefore,it is significant to study the identification problems of non-uniformly sampled systems both in theory and applications.In this thesis,the recursive least squares type identification methods for a class of non-uniformly sampled-data systems are derived based on the least squares principle.The main work of this thesis contain follows:1.For the periodically non-uniform updating and uniform sampling scheme and for different noise disturbances,the identification models for a class of non-uniformly sampled data systems are derived by using the model transformation methods,and are divided into two categories:equation error type non-uniformly sampled systems and output error type non-uniformly sampled systems.2.Referring to the derivation of the recursive least squares algorithm,defining different information vector and parameter vector,and using the idea of replacing the unmeasurable noise terms with their estimates,the recursive least squares algorithm for non-uniformly sampled ARX systems,the recursive extended least squares algorithm for non-uniformly sampled ARMAX systems and the recursive generalized extended least squares algorithm for non-uniformly sampled ARARMAX systems are derived.Simulation results indicate that the algorithms are effective.Convergence properties of the former two algorithms are studied.3.For non-uniformly sampled output error systems,two identification methods are presented. one is a recursive least squares identification algorithm based on the bias compensation technique for the unbiased estimates and the other is an auxiliary model-based recursive least squares algorithm using the auxiliary model idea.Convergence rate and estimation accuracy of the proposed algorithms are verified by simulation examples and the convergence properties of the auxiliary model based recursive least squares algorithm is studied by using the martingale convergence theorem.4.For non-uniformly sampled output error systems with colored noises,By combining the auxiliary model idea,extended least squares and generalized least squares methods,an auxiliary model based recursive extended least squares algorithm is presented for the non-uniformly sampled output error moving average systems and an auxiliary model based recursive generalized extended least squares algorithm is proposed for the non-uniformly sampled output error auto-regressive moving average systems.Simulation results show the merits of the algorithms.Finally,the difficulties encountered during the study,e.g.,the performance analysis of some proposed algorithms,are not be solved.The future topic of the identification for non-uniformly sampled systems is simply outlined.
Keywords/Search Tags:system identification, least squares, multirate systems, parameter estimation, non-uniformly sample, auxiliary models
PDF Full Text Request
Related items