On-line risk assessment model for aero engine using LS-SVM

The Least Square Support Vector Machine (LS-SVM) is applied in aero engine risk assessment with gas path fault diagnosis. Firstly, the deviations of engine cruise data from flight data recorder are analyzed, and sample data for modeling is obtained. The architecture of fault diagnosis model is established. Secondly, model selection is achieved using Pattern Search method; a real time risk assessment model based on LS-SVM algorithm is composed. Finally, by decoding ACARS report, real time cruise data set is acquired, and the diagnosis model is adopted in processing real time data set. Assessing results of engine gas path are shown. Moreover, the accuracy comparison with RBF ANN shows this method is suitable for risk assessment of gas turbine engine.