Implicit Time-Stepping for High-Fidelity Simulation of Moving Bodies with Spectral/hp Element Method
Please login to view abstract download link
The high-order spectral/hp element method discretisation of the incompressible Navier-Stokes (NS) equations within the Nektar++ framework is tailored for Direct Numerical Simulation (DNS) and under-resolved DNS of turbulent flows. It robustly handles moving geometries through a moving reference frame (MRF) formulation with absolute flow variables for further stability. As such, the incompressible NS MRF algorithm alleviates dynamic remeshing and its associated computational costs and complications. Using absolute velocities expressed in the moving frame coordinates further strengthens solver stability and overcomes issues arising from relative velocity formulations. However, the current implementation is based on explicit treatment of nonlinear convective terms which pose serious restrictions for turbulent flow simulations, particularly over a long period of time. Large timescales are ubiquitous in many industrial applications such as the wind energy sector, which typically involve high Reynolds numbers, therefore also demanding dense meshes for scale-resolved simulations. To address this, the present work extends the MRF to support an implicit time-stepping that has previously been implemented and demonstrated in Nektar++ but is limited to stationary geometries. Implementing an implicit treatment of advection terms relaxes the CFL constraint, preventing the time-stepping from compounding the already substantial costs associated with the mesh, enabling the MRF solver for highfidelity, industrially relevant simulations. To this end, the implicit Velocity Correction Scheme (VCS) is first derived within the moving reference frame, leading to a pressure Poisson equation and an advectiondiffusion-reaction equation for each velocity component. As the derivation is within the MRF, additional terms arising from Coriolis and centrifugal effects are present. Potential stability implications arising from these terms are discussed. Subsequently, numerical tests are presented using both implicit and explicit timestepping schemes. These tests demonstrate the accuracy of the new scheme and detail the cost reduction in the MRF.
