Data-Driven Modeling of Critical Engine Parameters and Intelligent Monitoring Applications Using Flight-Test Data Mining

This study proposes a data-driven framework for intelligent engine monitoring through the integration of Least Squares Support Vector Regression (LS-SVR) and variable-bandwidth kernel density estimation. Flight test data mining is employed to establish a high accuracy surrogate model for engine Exhaust Gas Temperature(EGT). A probabilistic characterization of the deviation between LS-SVR predicted EGT and flight test measurements is achieved via variable bandwidth kernel density estimation, deriving the probability density function of prediction errors. Based on this distribution, a 95% probability interval is constructed around the model predictions. Experimental results demonstrate that 95% of observed EGT values fall within this interval under normal operating condition, while excursion beyond this range provide statistically significant indicator of engine anomaly. The proposed methodology also offers extensibility to other critical engine parameters—such as fan speed (N1) and core speed (N2)—thereby establishing a comprehensive intelligent monitoring system for flight test engine.