Probabilistic Matching of Turbofan Engine Performance Models to Test Data

Abstract

This paper describes the development of an improved method for reliable, repeatable, and accurate matching of engine performance models to test data. The centerpiece of this approach is a minimum variance estimator algorithm with a priori estimates which addresses both deterministic and probabilistic aspects of the problem. Specific probabilistic aspects include uncertainty in the measurements, prior expectations on model matching parameters, and noise in the power setting parameters. The algorithm is able to produce optimal results using any number of measurements and model matching parameters and can therefore take advantage of all measured data to produce the best possible match. This improves on current matching algorithms which require that the number of measured parameters be equal to the number of model matching parameters. This algorithm has been implemented in the Numerical Propulsion System Simulation (NPSS) and tested on a generic high-bypass turbofan model typical of those used in commercial service. The baseline engine model and simulated test data are described in detail. Several exercises are discussed to illustrate results available from this algorithm including the matching of a typical power calibration data set and matching of a typical production engine data set.