Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.373072
Title: Short term spectral estimation with applications
Author: Balmer, Leslie
ISNI:       0000 0001 3441 8649
Awarding Body: University of Warwick
Current Institution: University of Warwick
Date of Award: 1986
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
This thesis investigates the effect of finite record length on spectral estimation with particular reference to the estimation of frequency response functions and to the application of frequency response methods to the problem of adaptive control. Conventional analysis of the statistics of frequency response estimation has concentrated on the statistics of the auto and cross spectral estimates involved. The approach adopted here is to attribute the errors in the estimates to transient terms caused by the finite record lengths of input and output data. It is shown that these transient terms can be separated into a term correlated with the input and an uncorrelated term. The first term causes bias and the second can be regarded as an uncorrelated external disturbance. Expressions for the bias of the estimate are obtained both in open-loop and closed-loop configurations. An expression for the variance of the estimator is obtained for the open-loop case only, but confidence intervals on the estimates are derived by both configurations. A new parametric method of identification is investigated, the novelty being in the introduction of additional parameters to account for the transient terms. It is shown that the bias due to the finite record length can be removed completely in the noise free case. Methods similar to those used for parametric identification in the time domain are used to reduce the bias when noise is present. Results obtained for non-parametric and parametric identification are applied to the area of adaptive control. Methods of applying frequency domain techniques to the problem of adaptive control are investigated. It is shown that in certain areas such an approach can have advantages over the corresponding time domain methods. The results obtained, both in the area of frequency response estimation and in the field of adaptive control, are verified by simulation studies.
Supervisor: Not available Sponsor: Science and Engineering Research Council
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.373072  DOI: Not available
Keywords: QA Mathematics ; QC Physics ; TA Engineering (General). Civil engineering (General) ; TJ Mechanical engineering and machinery
Share: