Vibration analysis of gas turbines by intelligent knowledge base systems

Market pressure within the aero industry, and the changing requirement of airline operators, motivates considerable effort to continuously improve the design of the modem day gas turbine. Fundamental to the development programme of a new gas turbine is the monitoring of vibrations of the engine during its development tests. This entails manual inspection and interpretation of thousands of diagrams depicting the vibration history observed by several hundred strain gauge transducers, located on rotating and stationary compressor blades. This thesis describes an Expert System, developed during a three year research contract (co-funded by Rolls-Royce and SERC), that has been devised to assist in the analysis task. Features, depicting various forms of vibration phenomena, are located on each diagram by image processing techniques and interpreted by stored engineering knowledge representing the experience of vibration engineers. Thus, the system is capable of scrutinising vibration data, obtained from engine tests, and classifying that data as acceptable or in need of further study. The thesis provides an overall description of the problem of vibration in gas turbines and describes the image processing techniques and methods of reasoning employed to interpret the data.