Development and Validation of a Civil Aircraft Engine Simulation Model for Advanced Controller Design

Abstract

Computer simulation is a powerful tool for the mechanical and control system design of gas turbines. A high fidelity computer simulation can be used as a substitute for a real engine in many applications. For example, it is possible to simulate critical transients that must be avoided on the actual plant due to the risk of damage. Turbine engines can be modeled at various levels of detail, from full 3D descriptions of the gas path (e.g., NASA’s numerical propulsion system simulation (NPSS) (1)) that can require distributed computers or supercomputers to simplified algebraic equations (2) and even simple overall transfer functions. It is generally accepted that a 1D simulation is sufficient for accurate dynamic performance modeling and, therefore, controller design. The simulation method used in this study is known as an aerothermal transient performance model (3). This method avoids iterative calculations by arranging the component equations to follow the direction of the gas path and by introducing storage volumes between components to account for the unsteady balance of mass at the compressor discharge, at the combustion chamber, and between the turbines (4). Relative to an iterative model, there is some loss of accuracy, but this is negligible from the point of view of engine control system development and is offset by a superior execution time, particularly if a small time step is used. If a constant time step is chosen for numerical integration, then this method can provide a model with a predictable run time. If the simulation is designed to run in real time, then it can also be used with real hardware, although, clearly, the computer program outputs have to be generated at least as fast as the predicted physical phenomena for the model to run side by side with an engine (5). In this paper, we present a full aerothermodynamic model of a two-spool, high-bypass turbofan engine with an unmixed exhaust together with a switched, gain-scheduled aeroengine controller with bumpless transfer and antiwindup. The engine simulation in conjunction with this controller achieves dynamic performance representative of that of a real aeroengine. Model implementation is in the MATLAB-SIMULINK® environment. Full flight-envelope validation of both the model and the controller has been performed with the assistance of Alstom Aerospace, with the exception of engine startup as this is not within the scope of the model.