Neural network approach for a combined performance and mechanical health monitoring of a gas turbine engine

Abstract Traditionally independent diagnostics methods were employed for health monitoring of system. These exhibited an overall satisfactory performance, but with a limited effectiveness range. A discipline that has emerged in recent years is that of an information (or data) fusion, which allows interweaving of different methods with different effectiveness ranges, to produce a wider and more reliable coverage of diagnosis. It is a multidisciplinary domain wherein, data from the various domain is blended together to arrive at a more reliable monitoring. The present paper brings out a Neural Network (NN) based approach for executing this task of combined health monitoring viz. mechanical and performance, with an example case study pertaining to a developmental power turbine. The various parameters used along with the trending methodologies both for steady state and transient operations are brought out. In addition, the influences of various parameters that can lead to deviations in the response are also discussed. The whole process of executing this task is put forward in a rather simple manner. The results accrued have been well corroborated with the findings on dismantling of the turbine.