ICFD
ICFD:Summary of Recent and Future Developments
Since its release in R7 the Incompressible CFD solver (ICFD) has been rapidly improving and increasing its functionality. In this paper a summary of the latest and current developments will be presented. The focus will be on four topics. First the steady state solver and its coupling capabilities for fluid-structure interaction (FSI) or conjugate heat transfer (CHT) will be presented. In second place the recent modifications to the boundary layer mesh generation will be introduced where some default parameters have changed. The possible implications of these changes in the solution will be mentioned. Third a short introduction to coupling ICFD with LS-OPT® for shape optimization will be presented. The idea is to use ANSA to morph the surface mesh driven by LS-OPT to provide an optimal solution. Finally some of the current developments will be enumerated like immersed interfaces, periodic boundary conditions, porous media through shell elements for parachute simulation, etc. These developments will be part of future LS-DYNA® releases.
The demand for coupled multiphysics analysis has been steadily increasing in the past few years. In the wake of this many commercial solvers have rushed to provide engineers with solutions which many times carry the load of a stiff legacy implementation. Thus codes that were originally designed to deal with purely CFD analysis face the challenge of connecting their results to other solvers sometimes for other vendors to produce multiphysics results. This ends up in a cumbersome process of co-simulation which leaves much of the coupling burden to the engineer.
In LS-DYNA the ICFD module has been designed from the beginning to be a multiphysics solver that provides accurate and scalable CFD solutions and at the same time integrates easily to the other physics modules in LS-DYNA. Continuous work is devoted to integrate to newer modules, improve the coupling algorithm, increase efficiency and simplify the coupling process without assuming that the user is a CFD or Structural analysis engineer.
Two examples will be presented to illustrate these ideas in the field of FSI and CHT.
Fluid structure interaction on a ground vehicle
The first example addresses the FSI analysis of a ground vehicle where the roof panel deformation will be studied. The sketch of the model is shown in figure 1.
This model has been solved using three different approaches. The first approach is a non-linear transient coupling where the structural and fluid solver are strongly coupled and the solution is performed implicitly. This should results in the most accurate type of coupling and thus it should be considered as the reference solution. The second approach uses a steady state potential flow solver with a subsequent non-linear structural solution at the end using the pressure field from the potential flow solution. Using the potential flow solver assumes that the flow in the region of the domain under consideration is attached to the walls and laminar. In the real case the flow is of course not laminar but it is attached and so the pressure values are expected to be within a reasonable range to the Navier-Stokes solution to be used in the prediction of the structural displacement. In the third approach the force computed in the non-linear coupling is written to an LS-DYNA input deck for each load segment in the structural model. So just by including this file in the structural input deck a user can run s structural only analysis using the force computed in the expensive non-linear analysis. This results in a very fast run that allows engineers to easily change and test their designs until a model is ready for a new non-linear run. Figure 2 shows a comparison of the flow velocity for the Potential Flow vs Navier- Stokes solver.
The area of interest is the roof of the vehicle. In Figure 3 the result of the three different approaches is depicted. It is observed that the three solutions predict very similar displacements. It is also interesting noting the running time for each approach which is the reason that balances the potential lack of accuracy of the simpler models.
Cooling problem
This second example involves a Conjugate Heat Transfer problem where fluid is used to cool down a tool in a stamping simulation. The setup of the model is sketched in Figure 4.
In cooling problems fluid flows in a pipe inside a die at a lower temperature than the die is. As the fluid runs along the pipe it will slowly heat up as the die cools down. In the absence of a heat source the fluid and the die will reach the same temperature as the fluid inflow temperature. The velocity of the flow plays an important role in maintaining the appropriate temperature profile inside the pipes. In Figure 5 the velocity profile for the Navier-Stokes and the Potential flow solver are compared. The largest differences are at corners and areas of detachment/recirculation.
In Figure 6 the temperature profiles are shown for both Navier-Stokes and Potential flow solutions. Both profiles are in very good agreement.
Figure 6. Temperature profile at steady state for Navier-Stokes and Potential flow solver.
Shape Optimization using LS-OPT
One of the latest achievements related to ICFD has been using LS-OPT and the meshing package ANSA (by BETA-CAE) for shape optimization. The idea is to modify an initial geometry to optimize a functional. For instance in the case of ground vehicle aerodynamics the vehicle geometry is changed to maximize the down- force to drag ratio. In Figure 7 an initial geometry of a generic vehicle is presented. The optimization process will focus on the tail of the vehicle.
The idea is to use ANSA morphing capabilities to change the geometry using some parameters obtained from LS-OPT. Figure 8 shows the morphing boxes and the parameters to modify together with the LS-OPT optimization loop.
One of the benefits of using LS-DYNA for the optimization process is that the volume mesh is built at run time. So the morphing needs to take place only on the surface of the vehicle greatly simplifying the process.Morphing a volume mesh is expensive and less reliable especially in the presence of boundary layer mesh. The final results or shown in Figure 9 which show the predicted and most optimal configurations.
New Developments
In this section some of the current latest developments will be briefly presented. These developments are currently not part of the release version of LS-DYNA although a beta version is expected at some point this year.
Periodic Boundary Conditions
Periodic boundary conditions are often used in numerical methods to represent a large domain by modeling just a fraction of it. These boundary conditions are heavily used in rotating systems like in turbomachinery. The implementation involves the use of linear constraints that guarantee the continuity and conservation of flow across the periodic boundaries. Figure 10 is a setup example of a problem with periodic boundaries.
Sliding mesh is a technique that allows the simulation of transient rotating mechanisms without re-meshing. When sliding meshes are used typically the domain is split into at least two volume meshes. One mesh will have the rotating components and the other the rest of the domain. The interface between the two volume domains is the sliding mesh. The solutions on both domains are solved simultaneously by using linear constraints. Figure 11 shows an application example.
Immersed interface
Immersed interfaces simplifies the pre-processing of complex geometries. The goal is to have a hybrid immersed interface / body fitted mesh depending on the types of flow in different parts of the model.This new approach is based on discontinues finite element approximations which provide sharp interfaces and allow structural contact.A typical example is shown in Figure 12 which involves a lobe pump where the contact between the lobes makes it challenging for the more classical re-meshing approach.
About the authors
Dr. Facundo Del Pin received his PhD. from University of California Berkeley. After joined LSTC in 2008, he was involved in research and development of CFD.
Rodrigo R Paz received PhD. computational fluid mechanics and parallel computation from Université du Littora, Argentinal. After joined LSTC in 2013 he was involved in development of ICFD and FSI functions.
Chienjung Huang received PhD. in computational fluid mechanics and FSI from University of California, Los Angeles. After joined LSTC in 2017 he was involved in application & verification and technical support of LS-DYNA.
- Previous:Lsdyna_multi-physics2_
- Next:ALE