Prediction of remaining useful life of an aircraft engine under unknown initial wear

Effectiveness of Condition Based Maintenance (CBM) strategy depends on accuracy in prediction of Remaining Useful Life (RUL). Data driven prognosis approaches are generally used to estimate the RUL of the system. Presence of noise in the system monitored data may affect the accuracy of prediction. One of the sources of data
noise is the presence of unknown initial wear in the samples. Present paper illustrates the effect of such initial wear on prediction accuracy and presents the guidelines to handle such initial wears. Two Artificial Neural Network (ANN) models are developed. First model is developed with the help of complete data; while the second model is developed after removing samples with abnormal initial wear and R control chart is used to screen the samples with abnormal initial wear. It is found that the presence of initial wear significantly affects the prediction accuracy. Also, it is found that RUL estimation for a unit with short history tends to produce great uncertainty. Hence, it is recommended that RUL prediction should be continuously updated with age of the unit to increase the effectiveness of CBM policy.