Lighthill Tensor Loading

Example files on this page

  • acoustic.e — Salinas_rtest/training/exampleproblem/transient/load/lighthill/fuego/hemisphere.exo

  • fluid.i — fuego_rtest/fuego/mesh_deformation_file/fluid.i

  • Fuego mesh — fuego_rtest/fuego/mesh_deformation_file/fluid_mesh.g

  • hemisphere.jou — Salinas_rtest/training/exampleproblem/transient/load/lighthill/fuego/hemisphere.jou

  • input — Salinas_rtest/training/exampleproblem/transient/load/lighthill/fuego/acoustic_nodeset.inp

17. Lighthill Tensor Loading#

In this section we provide the steps for applying the Lighthill tensor as a load in a Sierra/SD acoustics simulation. The Lighthill tensor captures the noise generated by unsteady convection in fluid flow simulation. In this work, we use the Sierra/TF incompressible thermal fluids code Fuego to simulate a small chamber, shown in Figure 17.1, that undergoes a sinusoidal pumping motion in the x-direction. The air moving in and out of the chamber produces turbulence that is captured by the Lighthill tensor computed during the Fuego simulation. The divergence of the Lighthill tensor is handed off to Sierra/SD and is used as an acoustic source term for far-field acoustic noise modeling in the larger semi-circular domain shown in Figure 17.2.

Figure 17.1 a) Fuego mesh of fluids domain where sideset 2 (green) is absorbing, sideset 4 (blue) undergoes the pumping motion, and all other sides shown in red are fixed. Sideset 2 shown in green will be tied to the larger Sierra/SD domain shown in Figure . b) Fuego mesh shown on z-plane. c) Fuego interpolation mesh for output of the divergence of the Lighthill Tensor. Domain dimensions are also shown in c).#

These simulations are part of the Sierra test suite and provide regression testing for both the Sierra/SD and Fuego parts of Lighthill noise modeling. Lighthill loading has also been verified in Sierra/SD for a 1-D waveguide with documentation provided in the Sierra/SD verification manual. The input for this example is attached.

Producing the Lighthill load and applying it in Sierra/SD is a \(5\) step process. The initial steps produce the divergence of the Lighthill tensor from a Fuego CFD simulation. Questions about these initial steps should be directed to the Sierra Thermal Fluids team.

The final steps involve preparing the Fuego output for use in Sierra/SD and then running the Sierra/SD simulation and are found in the input deck. Questions about the final steps should be directed to the Sierra/SD team.

Figure 17.2 a) Sierra/SD domain for acoustic noise propagation. The yellow block is the Fuego output domain containing divT and the red block is the additional domain acoustic.e for the Sierra/SD simulation. The pink sideset will interface with infinite elements. b) Sierra/SD tetrahedron mesh coarsened out from the Fuego mesh.#

17.1. Mesh Deformation For Fuego#

This section describes the process of producing a deformation field used to drive the Fuego simulation. Questions about Aria should be directed to the Sierra/TF team.

In this example, Aria is used to produce a sinusoidal displacement in the x-direction on sideset 4, shown in blue in Figure 17.1a.

Begin String Function displacement
  Value is "0.02*sin(1000*3.14159*t)"
End

Additionally, the displacement in the y- and z-direction is fixed, and the simulation is terminated at \(t\)=6e-3s. The output of the Aria simulation, displacements.e, is then used as input for Fuego.

17.2. Fuego Simulation#

This section describes the process of running Fuego to produce the divergence of the Lighthill Tensor. Questions about Fuego should be directed to the Sierra/TF team.

The Fuego input file fluid.i is executed with the following command:

mpirun -np 8 fuego -i fluid.i

The Fuego simulation is terminated at \(t\)=3e-3s. The Fuego simulation is discretized by the tetrahedron mesh shown in Figure 17.1b. The Fuego simulations writes the divergence of the Lighthill tensor out to the coarser hexahedron mesh shown in Figure 17.1c as nodal data. This data is written to acoustic.e.8.[0-7] and provides the loading for the Sierra/SD simulation.

17.3. Processing Fuego output for Sierra/SD#

This section describes the steps required to run a Sierra/SD simulation using the Fuego output. Questions about this section should be directed to the Sierra/SD team.

The first step is to join the partitioned Fuego files back together using the epu Seacas tool:

epu -auto acoustic.e.8.0

The above Fuego simulation writes the divergence of the Lighthill tensor out as nodal data with the variable names: divT_x, divT_y, divT_z. The Fuego domain is much smaller than the Sierra/SD domain. If these two domains were joined together into a single Exodus file, nodal data of divT=0 would be created on the larger Sierra/SD domain. To circumvent this unnecessary storage of divT data on the Sierra/SD mesh, we convert the Fuego divT data to nodeset data using the ejoin Seacas tool:

ejoin -output acoustic_nodeset.exo \
      -convert_nodal_to_nodesets all \
      acoustic.e

which produces the output file acoustic_nodeset.exo.

17.4. Mesh for Sierra/SD#

The Sierra/SD simulation will use the Fuego divT data as a source term to model noise propagation in a larger domain. For this example we join the smaller Fuego mesh containing the interpolated divT data to a larger semi-circular domain, see Figure 17.2a. A cubit journal file for creating the semi-circular mesh is contained in hemisphere.jou. This mesh must contain sidesets (sideset 5 in the cubit journal file) that will be tied to sideset 2 in the Fuego output mesh, shown in green in Figure 17.1a. This mesh also contains sideset 6 on the exterior of the semicircular domain which will be used for applying absorbing boundary conditions via infinite elements. The two separate meshes are joined together with the ejoin Seacas tool:

ejoin -output acoustic_nodeset_distribution_factors.exo \
      hemisphere.exo acoustic_nodeset.exo

This produces the full meshed domain shown in Figure 17.2b for the Sierra/SD simulation.

17.5. Sierra/SD simulation#

This Sierra/SD simulation will be described in this section. Lighthill loading causes Sierra/SD to use the acceleration potential form of the acoustic equation. The Sierra/SD simulation is terminated after t=0.06s, which is twice as long as the Fuego simulation. For the final 0.03s of the simulation there will not be any available Fuego produced divT data to be read in for Lighthill Loading. For this case, the final divT data read in at t=0.03s will be applied for the remainder of the simulation, which produces a warning to this effect.

Some Lighthill specific portions of the attached Sierra/SD input file are:

  1. The Lighthill loading is applied as a function load the LOADS section with the Function described in FUNCTION 1. Lighthill loading is described in the User’s and verification manuals.

  2. Tied data ties together the Fuego and Sierra-SD domains. Sidesets must be defined on these surfaces when they are created in Cubit. It is difficult to add a sideset to a mesh after it contains nodal data, i.e. The sidesets needed to tie the meshes together must be defined on the mesh used for Fuego output before the Fuego simulation is run.

  3. Infinite elements are used on sideset 6 to absorb the pressure waves.