⚙️🧩 How does ADINA REALLY solve FSI?

 Monolithic or partitioned? 🤔

Direct solver or iterative solver?
And if iterative — what kind of preconditioner?

I dug into the ADINA literature to separate what is actually documented from what is often assumed.

🔥 The answer may surprise CFD/FEA engineers.

#ADINA #FSI #CFD #FEA #ANSYS #Simulation #Engineering #Multiphysics

 I checked the older ADINA papers and the currently available Bentley/ADINA information. The answer is actually quite clear on monolithic vs. partitioned, but the exact preconditioner for the fully coupled monolithic FSI matrix is much less clearly documented publicly.

What the literature establishes

1. ADINA supports both approaches.
   ADINA has a simultaneous/direct (monolithic) FSI formulation and an iterative/partitioned FSI formulation. The ADINA literature explicitly describes both. (ScienceDirect)

2. In the monolithic approach, fluid and structural equations are assembled into one coupled nonlinear system and solved together, using Newton–Raphson iterations. This is particularly attractive for strongly coupled FSI, such as large structural deformation or very stiff structures coupled to highly compressible fluids. (Massachusetts Institute of Technology)

3. In the partitioned approach, the fluid and structural equations are solved separately and iterated at the interface. Interestingly, Degroote, Bathe and Vierendeels specifically used ADINA to compare partitioned and monolithic FSI algorithms and state that ADINA can perform both. (ScienceDirect)

4. ADINA has historically supported both direct and iterative linear solvers. An ADINA paper from 2003 explicitly states that both iterative and direct solution procedures can be used for the fully coupled system. (ScienceDirect)

5. The older ADINA solver literature documents CG + incomplete Cholesky for symmetric positive-definite structural systems and BiCG/GMRES + incomplete LU for nonsymmetric/indefinite fluid systems. (ScienceDirect)

But there is an important caveat

I would not state that "ADINA's monolithic FSI solver uses ILU" based on the publicly available literature.

The papers establish that ADINA has preconditioned iterative solvers, including ILU for fluid equations, but that does not necessarily mean that the complete monolithic FSI Jacobian

$$\begin{bmatrix} K_f & K_{fs}\\ K_{sf} & K_s \end{bmatrix}$$

is solved using an ILU preconditioner.

In fact, older published ADINA FSI work frequently reports the use of a direct sparse solver for the monolithic/full Newton solution. For example, an ADINA 8.2 FSI study explicitly says that its FSI simulations could only be solved using a direct sparse solver. (PubMed Central (PMC))

So the safest technical answer to the original question is:

ADINA supports both monolithic/simultaneous and partitioned/iterative FSI formulations. The monolithic formulation uses Newton–Raphson iterations for the nonlinear coupled equations. ADINA has both direct sparse and iterative linear-solution capabilities, but publicly available ADINA documentation does not clearly establish that a specific ILU/IC/block preconditioner is used for the complete monolithic FSI Jacobian. The published ADINA literature more clearly documents direct sparse solution for monolithic FSI and ILU/IC-preconditioned iterative methods for the individual fluid/structural systems.

That distinction is important.

The current Bentley product information also explicitly says ADINA FSI supports direct FSI coupling or iterative FSI coupling, including displacement compatibility and traction equilibrium. (Bentley)

If you want to investigate this at the implementation level, the most useful references I found are:

• Zhang & Bathe — “Direct and iterative computing of fluid flows fully coupled with structures” — probably the most directly relevant paper for your question. (Massachusetts Institute of Technology)

• Zhang et al. — “Recent development of fluid–structure interaction capabilities in the ADINA system” — detailed description of ADINA's FSI architecture and solution procedures. (ScienceDirect)

• Tan & Bathe — “Studies of finite element procedures—the conjugate gradient and GMRES methods in ADINA and ADINA-F” — useful specifically for understanding ADINA's iterative solvers and preconditioning. (ScienceDirect)

• Degroote, Bathe & Vierendeels — “Performance of a new partitioned procedure versus a monolithic procedure in fluid–structure interaction” — particularly valuable because ADINA was used for the direct comparison. (ScienceDirect)

If you're asking this because you're considering ADINA versus ANSYS Fluent + Mechanical for a strongly coupled FSI problem, the solver architecture becomes especially interesting: ADINA's native monolithic FSI is fundamentally different from the usual Fluent–Mechanical partitioned coupling, and I can break down the two global matrices and convergence mechanisms side-by-side.