A Gaussian process regression approach to model aircraft engine fuel flow rate

Abstract

The problem of building statistical models of cyber-physical systems using operational data is addressed in this paper, using the case study of aircraft engines. These models serve as a complement to physics-based models, which may not accurately reflect the operational performance of systems. The accurate modeling of fuel flow rate is an essential aspect of analyzing aircraft engine performance. In this paper, operational data from Flight Data Recorders are used to model the fuel flow rate. The independent variables are restricted to those which are obtainable from trajectory data. Treating the engine as a statistical system, an algorithm based on Gaussian Process Regression (GPR) is developed to estimate the fuel flow rate during the airborne phases of flight. The algorithm propagates the uncertainty in the estimates in order to determine prediction intervals. The proposed GPR models are evaluated for their predictive performance on an independent set of flights. The resulting estimates are also compared with those given by the Base of Aircraft Data (BADA) model, which is widely used in aircraft performance studies. The GPR models are shown to perform statistically significantly better than the BADA model. The GPR models also provide interval estimates for the fuel flow rate which reflect the variability seen in the data, presenting a promising approach for data-driven modeling of cyber-physical systems.