In computational fluid dynamics, besides physical modelling and numerical methods, the appropriate specification of boundary conditions is crucial for solution quality. For turbomachinery flows in contrast to exterior aerodynamics, the specification of boundary conditions becomes even more important because in- and outflow boundaries are typically located close the flow regime of interest due to geometric limitations. The trend to more compact engine designs with smaller axial gaps and higher aerodynamic blade loadings additionally increases the relevance of boundary condition specification.
Figure 1: Open boundary surfaces for the simulation of flows in turbo machinery components
In flow simulations inflow and outflow surfaces pose as artificial boundaries since they truncate a domain which is infinite in reality. Thus, perturbations which originate inside the computational domain, such as wakes or potential field perturbations of a blade, should propagate through the boundary surface unaffectedly. Boundary condition formulations, as they are widely spread in exterior aerodynamics or many commercial CFD tools, often produce spurious reflections at these open boundaries. This can result in poor prediction of blade pressure distributions, shock losses and operating points. Moreover, in aeroelasticity and aeroacoustics, artificial reflections of pressure perturbations can be massively detrimental to the prediction of aerodynamic damping or sound fields.
Therefore, non-reflecting boundary condition methods are available in TRACE. These are based upon a modal treatment and characteristic analysis of the linearized Euler equations, exploiting the typical periodicity of turbomachinery flows. Thus the wavelike behavior of the conservation equations is taken into account and it becomes possible to prescribe flow states on boundary surfaces such that no spurious perturbations are generated and transported into the computational domain.
These non-reflecting boundary conditions are further developed and improved in the course of the sponsorship. Beside the parallelisation of these methods in order to further accelerate TRACE when using many CPUs, activities towards better convergence and stability properties are undertaken within the sponsorship. Furthermore, the existing methods are unified in order to provide consistent boundary condition formulations to the user independently of the solution method employed (non-linear time domain, time linearized frequency domain or harmonic balance).