A methodology is proposed and developed for the simulation of post-surge condition in a multi-stage compressor that is part of a gas generator system that also includes the combustor and turbine and ducts. Given the essentially one-dimensional nature of surge, the approach basically consists of coupling single blade passage multi-stage RANS CFD simulations of the compressor for with 1D equations modelling the behaviour of the other components applied as dynamic boundary conditions. This method allows for the simulation of the flow behaviour inside a multi-stage compressor during surge and, by extension, for the prediction at the design phase of the time variation of aerodynamic forces on the blades and of the pressure and temperature at bleed locations inside the compressor used for turbine cooling. The main advantages of this method over existing methods are its relatively modest computational time and resource requirements and the fact that it does not require any empirical data input beyond what is used in standard CFD simulations.
The method is implemented in a commercial CFD code (ANSYS CFX) and applied to three compressor geometries with distinct features. Simulations on a low-speed (incompressible) three stage axial compressor allows for a validation with experimental data, which shows that the proposed methodology captures the surge behaviour of the system very well both qualitatively and quantitatively. This comparison also highlights the strong dependence of the surge cycle frequency on the volume of the downstream plenum (combustion chamber). Subsequently, the addition of a low-speed centrifugal compressor to the previous compressor is used to demonstrate the adaptability of the approach to a multi-stage axial-centrifugal configuration, yielding qualitatively realistic surge results. Finally, application of the method to an industrial transonic compressor geometry demonstrates the tool on a mixed flow-centrifugal compressor configuration operating in a highly compressible flow regime. A comparison of predicted versus measured shaft loading amplitude during surge is highly promising.