Application of back propagation neural network algorithms on modeling failure of B-737 bleed air system valves in desert conditions

Accurate life prediction of aircraft engine components is very critical because it has a direct impact on aircraft safety and on operators' profits. The engine bleed air system valves have considerably high failure rates when the engines are operated in desert conditions because of sand particles erosion and blockage. In this work, an Artificial Neural Network (ANN) model for the prediction of failure rate of the most important of these valves in Boeing 737 engines is developed and validated. A previously developed feed-forward back-propagation algorithm is implemented to train the ANN. The effects of changing the number of neurons in the input layer, the number of neurons in the hidden layer, the rate of learning, and the momentum constant are investigated. The model results are validated using comparisons with actual valves failure data from a local operator in Saudi Arabia, as well as comparisons with classical Weibull model results.