Aircraft engine condition monitoring: stochastic identification and neural networks

The performance of complex systems such as the aircraft gas turbine engine deteriorates in time due to the degradation or failure of its components. Condition monitoring systems have been developed to provide advanced warning of impending failure of components. By correctly predicting that a component is failing it can be replaced at an appropriate time thereby saving time and money for the operator of the system. These condition monitoring systems use various approaches and techniques to evaluate system parameters and make judgements on the condition of various components. This paper focuses on the two general approaches being investigated for condition monitoring systems: static pattern analysis approach and the dynamical systems approach. Both techniques are applied to real engine data and their performance results given