Monte Carlo simulation for system damage prediction: an example from thermo-mechanical fatigue (TMF) damage for a turbine engine

For high performance/high cost systems like aircraft systems, there is a need to achieve real-time, and continual assessment of engine condition, and possibly, to extend the life of operation. A turbine engine component life is varied depending on its operating environments. It is almost impossible to predict a damage of component correctly since operating conditions might be different. The challenge of life management is to find a reasonable compromise between "safe life" and maximum usage of engine parts to reduce costs. This paper describes prediction problem of thermo-mechanical fatigue (TMF) damage of a cooled turbine stator. Prediction of TMF damage allows us to manage components of a turbine engine with satisfying "safe life" and costs reduction of management since we can predict the next TMF damages according to expected or randomly selected operating conditions. Neural networks with Monte Carlo simulation are used for predicting TMF damages in different ambient temperatures and airport elevation as our operating conditions for the study of damage prediction. The TMF damage model is provided by a NASA Glenn Research Center engine simulator. A neural network that represents an engine simulator is studied to the accuracy of the prediction. The neural network engine model is combined with Monte Carlo simulation for the study of the performance of damage prediction