RADIOSS FSI at NASA Langley: Water Impact of 20 inch Sphere - Nasa langley

Innovation Intelligence®

NASA Langley Radioss Benchmark

Water Impact of 20-inch Sphere

Antoine Segnegon - Altair

John Brink - Altair

Greg Vassilakos – NASA Langley

May 16, 2012

Copyright © 2012 Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.

Americas Solver Task Force

Formed in 2011 with the purpose of helping existing and potential

customers adopt and migrate to Altair Solver Technology

Team Members:

John Brink – Director – Radioss Bulk

Antoine Segnegon – Radioss Block Explicit

Andrew Dyer – MotionSolve

Jaideep Bangal – AcuSolve

Abigail Arrington – AcuSolve


Introduction

• In 2011 NASA Langley completed construction of a Hydro Impact Basin

• The Hydro Impact Basin is next to an existing gantry


Introduction

• Allows water impact testing for both horizontal and vertical velocities

• Simulation is used in conjunction with the basin to evaluate water

landings for space vehicles


Objective of Study

• NASA Langley engineers wanted to determine the suitability of Radioss

for water impact simulation

• Blind Prediction—test results not available until after simulation

completed

• The problem given was a 20-inch hemisphere dropped from 5 and 10 ft


Responses of Interest

• Deceleration of the sphere

• Pressure on the surface of the sphere

Test Set-Up

Test Article Configuration

Pressure Transducer Locations


Options to Consider for Radioss Set-up

For Fluid Structure Interaction (FSI) simulation in Radioss, the following

options and parameters need to be considered:

• ALE or SPH

• Mesh Density and Spatial Discretization

• Material Modeling

• Initial Conditions

• Boundaries

• Gap of Interface Between Structure and Fluid

• Stiffness of Interface Between Structure and Fluid


Initial Set-up for Blind Prediction

• ALE (Arbitrary Lagrangian Eulerian) formulation used

• Expected to be less CPU costly based on experience

• Mesh Density and Spatial Discretization

• Followed guidelines from experience for mesh density and relative mesh sizing

• Sphere is modeled 20 mm x 20 mm

• Fluid mesh is modeled to have surface half the size of the sphere and half again

through the thickness

Lagrangian

ALE 5



• Material Modeling

• Choose Law 51 (Multi-Material Solid, Liquid, Gas) for fluid which exhibits less

diffusivity than older multi-material models

• Use Law 1 (Linear Elastic) for sphere

• Initial Conditions

• Added gravity to the problem with /GRAV

• Did not include initial hydrostatic pressure

• Boundaries

• Used quarter model with symmetries

• Used infinite boundary on top of air surface Quarter Model

~900k elements

Air Infinite

Boundary

Rigid

Sphere

Water Elements

Air Elements



• Gap of Interface Between Structure and Fluid

• Use Type 18 interface (Penalty-stiffness based, coupled Eulerian-Lagrangian) with

recommended Gap = 1.5 Lc, where Lc is the characteristic length of the fluid

(Lc=Volume/(Largest surface Area)=10x10x5/10x10)

• Stiffness of Interface Between Structure and Fluid

• The interface stiffness, Stfac, changes with mesh size and velocity and is

recommended to be:

Stfac = (ρ * V2 * Sel)/GAP

ρ - the highest fluid density in the model

V – impact velocity of the Lagrangian mesh

Sel - Surface area of Lagrangian impact element


Animation of Results – 5 Foot Drop

Animation of Results

Observation Regarding Fluid Mesh for ALE

The finite element model representing the entire cylindrical tank showed

unacceptable behavior at the contact area (top row of pictures) with the first

mesh attempted. The mesh of the fluid was redone in the impact area to avoid

“butterfly” mesh. The second row of images show the re-mesh fixed the issue.

Leakage at

Interface “Butterfly”

mesh

Orthogonal

mesh

Cleaner

Response


Comparison to Test Results – 5 Foot Drop

• Accelerations match the test results very closely

• The pressure trace does not capture the test magnitude

• The simulation ran for just over 2 hours on a 16-CPU machine

Note: Test acceleration results shown have been filtered with a 180 Hz Butterfly Filter; there is no filtering of simulation results

The pressure gage data from the tests is unfiltered except for the high frequency (4300 Hz) analog anti-aliasing filter that exists in the Data Acquisition System


Comparison to Test Results – 10 Foot Drop

• Again, accelerations match the test results very closely

• The pressure trace does not capture the test magnitude

• The reduced model ran in 45 minutes on a 24-CPU machine (4x faster)




Investigation of Pressure Profile

• J.P.D. Wilkinson’s paper “Study of Apollo Water Impact, Final Report,

Volume 4, Comparison with Experiments”, May 1967

• Pressure profile for water impact shows very steep pressure gradient

and very localized effect


Refine Mesh of 5 Foot Drop

• The mesh is refined so that the sphere has ~10mm x 10mm elements

• The fluid mesh is also refined to maintain proper mesh size ratios to

5mm x 5mm and 2.5mm in the impact direction


Mesh Refinement Results

• Accelerations peaks match well

• Pressures improve but at larger CPU cost (X 5)




Conclusions

• Doing a blind prediction resulted in very good correlation using

standard established modeling practices for Radioss FSI

• Pressure traces, however show a lower peak magnitude in simulations;

refining the mesh appears to help, but further refinement is necessary to

see if the pressures converge to test results

• It is unclear whether it is necessary to capture the exact pressure trace

to model the structural response of the test article

• Including the effect of the tank in the simulation was unnecessary; using

infinite boundaries gave very similar results at much less cost

• It is recommended to use an orthogonal mesh in the fluid in the area of

impact to avoid fluid to structure contact issues

• Radioss can be used confidently for water impact problems