Use this URL to cite or link to this record in EThOS:
Title: Robust change detection in automotive & aerospace systems
Author: Hermans, Filip J. J.
ISNI:       0000 0004 2746 9677
Awarding Body: University of Manchester
Current Institution: University of Manchester
Date of Award: 1997
Availability of Full Text:
Access from EThOS:
This thesis is a study of the use of fault detection in automotive and aerospace applications. It investigates three applications in particular. They are engine misfire detection, suspension monitoring and gas turbine sensor monitoring. The main aim is to produce robust monitoring algorithms for these systems and extract typical features which will facilitate the generation of monitoring solutions for similar but not directly related applications. To treat these problems, the thesis is divided into four main parts. They are as follows: Engine Misfire Detection: This part states the problem of monitoring the input of a continuous system, using measurable outputs, knowing that the input signal has a certain periodicity. The aim is to detect anomalies in the periodicity of the input. For the misfire application, the output is the engine crankshaft speed, the period the engine cycle and the anomaly in the periodic component the misfire. The anomalies are detected using signal processing methods based on recursive-vector least squares with discontinuous forgetting and generalised likelihood ratio. This detection algorithm augmented by a small look-up table is tested on real data representing a variety of driving conditions. Suspension Monitoring: The problem can be considered as that of monitoring the physical parameters of a continuous system. The monitoring is done based on a set of fixed models and some statistical evaluation. The models are formulated in discrete time using the h -operator. This leads to the generation of Kalman filter and Descriptor Kalman filter algorithms using the h-operator. For robustness and diagnostic reasons, a new criteria is proposed involving a sensitivity matrix. This allows simultaneous detection of several parameter changes independently. For the statistical evaluation, a new noise invariant cusum is introduced. The algorithm is finally tested on the tyre pressure and damper monitoring problem using simulations and real test rig and car data. Gas Turbine Sensor Monitoring: The aim is to monitor the sensors of non-linear time variant systems using a set of linear models i.e. robust monitoring. To incorporate robustness, the sliding mode observer is introduced and its fault detection capabilities investigated. This is done first using a simple but illustrative example and then using the real gas turbine data. Both illustrations show the advantages of the sliding mode observer. Sliding Mode Estimator: This part introduces the use of sliding mode principles for parameter estimation. This results in a sliding mode estimator. The tracking and decoupling capabilities of the sliding mode estimator are further compared with RLS. This is done using a simple example where the parameters are varying dependently, independently and/or abruptly. To substantiate the potential of the new estimator, the suspension monitoring problem is revisited with successful results.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available