Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.796435
Title: The application of state observer techniques to problems of system design and integrity in helicopter flight control
Author: Paterson, Charles Stewart
Awarding Body: University of Glasgow
Current Institution: University of Glasgow
Date of Award: 1990
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
Automatic flight control systems of modern aircraft, whether fixed wing or rotorcraft, have become increasingly complex and often involve the use of control activity which goes beyond the levels normally associated with human pilot operation. The sophisticated control laws employed frequently utilise the complete state vector, however, in practice not every state variable is available, either owing to the failure of its sensor or because it is impracticable to measure. The most feasible solution to the problem is therefore to use an estimate of the state vector produced from an observer. This thesis is concerned with the application of deterministic, continuous-time, linear, time-invariant system theory in the design of 'Luenberger' state observers for state estimation in the flight control systems of the single rotor helicopter. Observer design and system simulation were facilitated by using a complicated mathematical model of the helicopter. This model, which was provided by the Royal Aerospace Establishment, Bedford, is examined in detail and its limitations are discussed. Observer design methods are reviewed and two approaches, a method proposed by Gopinath and an observable canonical form method, are examined in detail. Due to numerical problems the Gopinath method is shown to be unsuitable, however it is demonstrated that the observable canonical form method is capable of producing accurate designs. Details of the software implementation of the canonical form technique are given and the results obtained and problems encountered, are analysed. Using this software, full and reduced order observers are designed for both eighth and fourteenth order system models. The performance of these observers are thoroughly assessed and it is shown that good estimates can be produced if the system states are 'clean', but that noise corrupted states result in poor estimates. To solve this problem a new form of observer --- the twin observer --- is introduced and it is demonstrated that with a precise model of the system, the twin observer can produce accurate, relatively noise free estimates of the system state. A review of instrument fault detection techniques is given and an observer based scheme, known as the Dedicated Observer Scheme, is selected for analysis with the twin observer and a fourth order, longitudinal system model. The advantages and disadvantages of this scheme are examined and possible solutions to some of the problems are proposed.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.796435  DOI: Not available
Share: