Fully Coupled FSI with ADINA DMP
With the availability of multi-core machines, clusters, and cloud computing facilities, parallel processing is widely used to reduce solution times. This enables analysts to solve very large computational models to obtain, for example, deeper insight into the performance of their designs.
Parallel processing has been available in ADINA for many years. A new distributed memory parallel solver is now available in ADINA for CFD and fully coupled FSI problems. ADINA DMP can be used on a single multi-core machine or on a cluster across multiple blades. As both the assemblage and solution of the equations is performed in parallel, the distributed solution capabilities in ADINA significantly reduce the overall solution time.
The objective of this Brief is to demonstrate this distributed solution capability for a fully coupled FSI problem. The problem is solved using 16 cores on a single machine.
We consider the problem shown in Figure 1 below, where turbulent flow is analyzed over a shell structure. The flow is three-dimensional as the flow goes both around and over the shell. Since a recirculation region forms downstream of the shell, it is important that the fluid domain is sufficiently large so that the recirculation region does not exit the domain. The K-ε model is used for the turbulent flow simulation.
In the first analysis, the shell is considered to be rigid; 13 million FCBI-C elements are used in the fluid domain. In the second analysis, the shell is considered to be highly flexible such that the fluid flow significantly deforms the shell. The movie above shows the large deformations for this case. A fully coupled FSI analysis is carried out in this simulation.
The steady-state pressure and velocity plots are shown, respectively, in Figures 2 and 3 below.
As expected the results show that a stronger wake is formed when the shell is rigid.
The distributed solution capability of ADINA is clearly a powerful tool for the solution of very large CFD and FSI problems. As ADINA DMP performs both the assemblage and solution of the equations in parallel, excellent scalability is observed. The performance of ADINA DMP as a function of processors used will be the subject of a future Brief.