Introduction

ADflow is a multi-block and overset structured flow solver initially developed at Stanford University under the sponsorship of the Department of Energy Advanced Strategic Computing (ASC) Initiative. It solves the compressible Euler, laminar Navier-Stokes and Reynolds-Averaged Navier-Stokes (RANS) equations using a second-order finite volume discretization. It currently features three solvers: multigrid, approximate Newton-Krylov, and Newton-Krylov, more details of which are under solvers. Users control the CFD solver via a solver option dictionary. More details are in options.

Although its primary objective in this program was to compute the flows in the rotating components of jet engines, ADflow has been developed as a completely general solver, and it is therefore applicable to a variety of other types of problems, including external aerodynamic flows. ADflow is a parallel code, suited for running on massively parallel platforms. The parallelization is hidden as much as possible from the end user, i.e. there is only one grid file, one volume solution file, and one surface solution file. The only thing the end user needs to do is to specify the number of processors they want to use via the mpirun (or equivalent) command.

A summary of the various features that can be found in ADflow is given below:

  • Compressible, URANS flow solver with various turbulence modeling options (Spalart-Allmaras, k-\(\omega\), SST, v2-f)

  • Multi-block structured approach with arbitrary connectivity. One-to-one mesh point matching with subfacing (C-0 multiblock) and point mismatched abutting meshes at block interfaces (C-1 multiblock) are allowed. CGNS I/O (mesh and solution) as well as native, MPI-IO parallel I/O option with back and forth conversion utilities.

  • Massively parallel (both CPU and memory scalable) implementation using MPI.

  • ALE Deforming grid implementation using IDWarp

  • Interface to conservative and consistent load and displacement transfer for aeroelastic computations.

  • Multigrid, Runge-Kutta solver for the mean flow and DD-ADI solution methodology for the turbulence equations.

  • Central difference discretization (second order in space) with various options for artificial dissipation.

  • Adaptive wall functions for poor quality meshes.

  • Unsteady time integration using second- or third-order (in time) backwards difference formulae (BDF) or a time-spectral approach for time-periodic flows.

  • Fully parallel, scalable pre-processor responsible load balancing.