Variations on the Kalman filter for enhanced performance monitoring of gas turbine engines

Since their advent in the 1940's, gas turbines have been used in a wide range of land, sea and air applications due to their high power density and reliability. In today's competitive market, gas turbine operators need to optimise the dispatch availability (it i.e., minimise operational issues such as aborted take-offs or in-flight shutdowns) as well as the direct operating costs of their assets. Besides improvements in the design and manufacture processes, proactive maintenance practices, based on the actual condition of the turbine, enable the achievement of these objectives. Generating dependable information about the health condition of the gas turbine is a requisite for a successful implementation of condition-based maintenance. In this thesis, we focus on the assessment of the performance of the thermodynamic cycle, also known as Module Performance Analysis. The purpose of module performance analysis is to detect, isolate and quantify changes in engine module performance, described by so-called health parameters, on the basis of measurements collected along the gas-path of the engine. Generally, the health parameters are correcting factors on the efficiency and the flow capacity of the modules while the measurements are inter-component temperatures, pressures, shaft speeds and fuel flow. Module performance analysis can be cast as an estimation problem that is characterised by a number of difficulties such as non-linearity of the system and noise and bias in the measurements. Moreover the number of health parameters usually exceeds the number of gas-path measurements, making the estimation problem underdetermined. This thesis starts with a survey of the state-of-the-art in module performance analysis. We then propose enhancements to a monitoring tool for steady-state data developed by Dr. P. Dewallef during his thesis at the Turbomachinery Group. Specifically, the improvements concern the fault detection and isolation tasks, respectively handled by a hypothesis testing and a sparse estimator. As a complement, we define metrics for the selection and analysis of sensor--health parameter suites based on the Information Theory. In a second step, we investigate the feasibility and the benefit that could be expected from the processing of data collected during transient operation of a gas turbine. We also discuss the impact of modelling errors on the estimation procedure and propose a solution that makes the health assessment robust with respect to modelling errors. The theoretical developments are evaluated on the basis of simulated test-cases through a series of metrics that gauge the estimation accuracy and the performance of the fault detection and isolation modules