Gas-path Diagnostics and Prognostics for Aero-engines Using Fuzzy Logic and Time Series Analysis

Reducing the direct operating-costs is now crucial in order to ensure competitive advantages for airlines and manufacturers, and so effective advanced engine-condition monitoring methodologies are necessary. Hence gas-path diagnostics and prognostics methods are reviewed and the specifications for such effective tools deduced, together with their pertinent future prospects. First, the considerable value that a preliminary observability study adds to the diagnostics process was recognised. A secure procedure has been devised: it is capable of (i) the identification of the severity of correlations between any two of the available measurements, as well as the correlations between any two of the component changes, (ii) the identification of more complex correlations that involve more than two changes in performance parameters, and (iii) the quantification of the quality of the system observability through a pertinent parameter. This enables comparisons among a significant number of measurement set selections. The core of the research is a novel gas-path diagnostics (GPD) method that uses fuzzy logic in order to provide secure quantification of the gas-path component faults. A fuzzy diagnostics system was set up for the Rolls-Royce Trent 800 engine that relies on an extensive statement of fuzzy rules generated using an engine model to achieve a quantitative solution through a non-linear approach, which is competent to achieve (i) SFI (single fault isolation) in the presence of noisy data, (ii) tuning over a known global deterioration level for all the performance parameters (baseline) computed for the previous flight, (iii) partial MFI (multiple fault isolation) with up to 2 degraded components (i.e. 4 performance parameters) considerably faulty at the same time, (iv) SFI while isolating systematic errors in the measurements (biasses). A bias-tolerant system was devised by means of the NOT logical operator and a new formulation of the fuzzy rules that includes the location of the bias. An innovative prognostics framework was devised, which uses ARIMA models and regression models respectively for short and long term investigations, to compute forecasts and the associated prediction intervals, which are aimed at assisting the prognostics decision-making process. This is strictly related to the diverse business intentions: in this study safety and economic related applications are investigated. For example, the optimisation of the TBO (time between overhauls) considering maintenance cost and additional fuel cost due to the deterioration is studied and the potential cost savings for the operators highlighted. HMP 1.1 for performance analysis was developed: it is a health-monitoring andprognostics framework consisting of three modules that perform respectively observability study, gas-path diagnostics and prognostics. The substantial benefits that can be achieved with such a tool, relative to the enhanced maintenance planning and improved mission scheduling, are discussed in the thesis via applications to the Trent 800 engine.