Suitability of performance adaptation methods for updating the thermodynamic cycle model of a turboprop engine

Abstract

A model of the thermodynamic cycle can be used in various fields, such as engine operation and control optimization, diagnosis, and maintenance. It is therefore important to develop a thermodynamic cycle model with high accuracy, and for this purpose, studies have been conducted to increase the accuracy of the model using measurement data. This study was conducted to identify the most effective method of updating the thermodynamic cycle model based on measurement data for the turboprop engine. A performance adaptation approach was applied to update the thermodynamic cycle model using measured data, and adaptation factors were used to adjust the component performance and sensor values. We considered three methods of calculating the adaptation factors and generating functions to apply to an existing engine model, which was a non-adapted model. The first was based on the Newton-Raphson (NR) method, in which the factors were generated as a function through regression curve fitting (RCF). In the second method, adaptation factors were calculated using a genetic algorithm (GA)-based optimization technique, and were generated as a function using RCF. The third method involved optimizing the function of the adaptation factor through GA. The accuracy and calculation cost of each method were compared using steady-state data measured using an in-flight monitoring sensor for the engine condition. All three methods yielded better prediction accuracy than the existing engine model; particularly, the first approach, based on the NR and RCF, showed the highest improvement in accuracy and the lowest number of required calculations. The maximum absolute values of the relative errors for the fuel mass flow rate, low-pressure shaft power, compressor exit pressure, turbine exit temperature, and exhaust gas temperature for the adapted thermodynamic cycle engine model with the NR and RCF were 4.29%, 2.83%, 1.36%, 1.80%, and 3.49%, respectively. As a result, we believe that the NR and RCF can be effectively used to calculate new adaptation factors for added measurement data points and to create a new function including the added factors.