FSI Benchmark: Techniques and Validation

Simulation by Andreas Hessenthaler in CHeart 
Motivation

Increasingly, fluid mechanics, solid mechanics and Fluid-Solid Interaction (FSI) are being proposed as tools that can be used to assist clinical medicine to address biomedical engineering problems.  In the current landscape, there have been a myriad of potential methodologies proposed which have, in some cases, been compared to 2D numerical benchmarks.  However, to date no validation test has been performed to effectively assess how well these techniques can predict the interaction of 3D deformable nonlinear materials with flow at regimes similar to those encountered in bioengineering.

In the fields of computational fluid dynamics and solid mechanics, there exist a number of benchmark problems that have been investigated and extensively used to validate simulation codes. In the case of fluid-structure interaction, researchers have published experimental data and computational benchmark results to verify fluid-structure interaction algorithms and implementations (e.g. Experimental and numerical study on a laminar fluid-structure interaction reference test case by Gomes et. al., and proposal for numerical benchmarking of fluid-structure interaction between an elastic object and laminar incompressible flow by Hron and Turek).

Building from these works, we have developed an experiment to provide (1) a platform for comparing FSI algorithms applied to a real world problem, (2) validation of different numerical approaches.  Design of our experiment focused on providing a test problem for numerical convergence analysis, avoiding geometric singularities (where possible) and providing easy modulation through a range of Reynolds numbers.  This proposed benchmark experiment will provide testbed for the comparison and validation of state-of-the-art FSI techniques.

The Experiment

The FSI phantom was designed to have two inlets which merge smoothly into a single outlet. Within this regular geometry, an elastic silicone filament is attached in the wake caused by the two inlets.  A glycol solution is pumped through the phantom (with a higher viscosity than water) at different rates through either inlet allowing for a variety of flow conditions.

Measurements of filament motion and flow are acquired using magnetic resonance imaging (anatomical scan, accelerated CINE and phase contrast MRI), providing detailed quantitative data through the phantom.  Pressure measurements are also taken within the flow field providing quantitative data on pressure waveforms.  This data provides:

(1) Flow velocity vectors in multiple slices throughout the model (PC MRI).

(2) Detailed motion of the solid body at multiples slices through the solid (accelerated CINE).

(3) Pressure at specific locations within the model (pressure catheter / anatomical scan).

(4) Geometric information and spatial orientation (anatomical scan)

The Benchmark Challenge

The experimental setup detailed above will serve as a platform for an FSI benchmark challenge, examining the performance (accuracy – efficiency) of different numerical approaches.  Challenge participants will be provided with the relevant information required to simulate the problem.  This includes:

- CAD geometric model of both the flow domain and solid model

- Inflow profile information for both inlets (3D vectors)

- Fluid density and viscosity measurements

- Solid density and stress / strain measurements (neo-Hookean solid)

- Pressure measurements at specific downstream locations

- Description of the experimental setup

This data will be provided in two phases corresponding to steady (Phase I) and transient inflow conditions (Phase II).  Participants will be asked to provide quantitative predictions of the solid deformation and flow behavior based on their model results for both Phase I and Phase II.  Participants will also be asked to submit details on the numerical setup (mesh size information, system rank, solve time, number of cores used, etc).  These initial results will be anonymized and distributed along with the experimental results to all participants.  Participants will then be asked to submit a final set of results which will be used to compare the performance of different numerical approaches.

COMING SOON: PHASE II PROBLEM DEFINITION

We are currently compiling the last portions of data for the Phase II experiment.  An updated document will be emailed to participants and posted here.

Take part in the FSI Benchmark challenge

All researchers engaged in FSI are encouraged to participate in the challenge.  Only three steps are required to participate.

(1) Contact us here to inform us of your interest.  We will then add you to the Challenge mailing list, provide links to the data, etc.

(2) Submit your preliminary results (June 1, 2015).  These results will be anonymized and shared to all participants accompanied by the experimental results.

(3) Submit your final results (Sept 1, 2015).  These results will be used to examine the accuracy and performance of different FSI methodologies in a final publication (authored jointly by all participants).

In addition to these mandatory steps, we also encourage participation in:

(1) The mini-symposia FSI Benchmark: Techniques and Validation (MS10) at CMBE in Paris 2015.  Results of the initial stage of the challenge will be unveiled at the conference prior to email distribution to participants and will be a good opportunity to discuss results with other participants.

(2) The special issue FSI Benchmark: Techniques and Validation within the International Journal For Numerical Methods in Biomedical Engineering.  Participants will be invited to contribute a manuscript detailing their methodology and analysis approach for the benchmark.  This issue will also be the source for a paper providing details of the experimental setup and results as well as a jointly published paper comparing results of the challenge.

 

TIMETABLE (bold items are mandatory for participants)

Jan 20, 2015: Respond with Expression of Interest

Jan 21, 2015: Distribution of FSI Challenge Data and Instructions

Phase I: Steady Inlet Boundary Conditions

 

Feb 15, 2015:  CMBE Conference abstract deadline (Extended to March 7)*

*abstracts should focus on the methods applied (results not required)

March 1, 2015: Distribution of FSI Challenge Data and Instructions

Phase II: Unsteady Inlet Boundary Conditions

June 1, 2015:  Submission of Initial FSI Challenge Results

June 29-July 1, 2015:  CMBE Conference, Presentations and Results

Sept 1, 2015:   Final Submission of FSI Challenge Results

Sept 31, 2015:  Paper deadline for FSI Challenge Special Issue

 

Experimental Team

Experimental design and measurements were conducted by a team of researchers.

- Nicolas Gaddum, Kings College London

- Ondrej Holub, Kings College London

- Andreas Hessenthaler, University of Stuttgart

- Ralph Sinkus, Kings College London

Organizers

- David Nordsletten, Kings College London

- Oliver Roehrle, University of Stuttgart

- Johan Hoffman, KTH Royal Institute of Technology

- Thomas J. R. Hughes, University of Texas, Austin

Please contact us here if you are interested in participating in our 3D FSI forward-prediction challenge and/or have any questions regarding the geometry, setup, boundary conditions, etc.