Particle Simulation in Magnetorheological Flows Norman M. Wereley

University of Maryland
GPU Summit
Particle Simulation in Magnetorheological Flows
Norman M. Wereley
Minta Martin Professor and Department Chair
[email protected]
+ contributions from graduate students, research staff, and collaborators
Smart Structures Laboratory
Dept. of Aerospace Engineering
University of Maryland
Adaptive Energy Absorption
Systems Using MR Fluids
Objective: To dissipate energy in vehicle
systems in order to protect occupants and
payloads from injurious vibration, repetitive
shock, crash and blast loads.
Sponsors:
General Motors, Boeing-Mesa
US Army, US Navy
Protective Seating: Impact, Crash, Blast
q
q
q
q
q
Events are rapid (< 50 ms)
Flight qualification of SH-60 seat with
MR vibration control and deploy
Develop lightweight compact MREAs for
adaptive crash safety
Verify MREA control strategies via test
Other vehicle applications
– Expeditionary Fighting Vehicle (EFV) semiactive seat technology
• Automatic adaptation b/w water-mode shock and
ground-mode vibrations
• Sea trials completed in 9/09
– Adaptive high speed watercraft seats
• Mark V SOC sea trials completed
– Adaptive Mine-blast attenuating seats
• Best “dynamic response index”
MR Fluids: Phase Transformation
Recipe
Magnetorheological Fluid

1 cup of oil (hydraulic)
1/2 cup of carbonyl iron powder
(heavy)
Mix well

Properties of MR Fluid:






High specific gravity
Yield stress: 60-100 kPa at full field
for high solids loading
Full Field: 1 Tesla, 1 Amp in coil at 23 Volts (under 3 Watts)
Temperature insensitive
Microstructure of MR fluids
MR Fluid: Bingham Plastic Behavior
Ferrous particle
Carrier fluid
N
S
N
Shear Stress,
Stress, ττ >> τ0y
Shear
v
S
NoField
FieldApplied
Condition
Optical Micrograph Image of Ferrous
Particle Chains in MR Fluid
Dimorphic MR Fluids
(Collaboration: R. Bell, & D. Zimmerman, PSU)
q
q
q
q
q
Morphology
Microwires with fixed diameter and
distribution of lengths (2-20
microns)
Spheres with narrow distribution of
diameters
Magnetic dipole reorients with
magnetic field
Key physics
– Yield stress
– Viscosity
– Sedimentation
• Smart Materials Structures (3/09)
– Elastic percolation
• Appl. Physics Letters (7/01/09)
Magnetorheology
Shear
Stress
kPa
60 wt% Iron with 15% Nanometer Scale Particles
MR Fluids - Too Simple a Model?
(Bingham Plastic Model)
q
Apparent
Viscosity
q
q
Bingham plastic MR fluid behavior
– Newtonian in absence of field
– Bingham plastic in presence of
field
Viscosity (µ) independent of field
Yield Stress (τy) dependent on field
Newtonian Fluid
Bingham Plastic
MR Dampers
q
q
q
MR fluids exhibit shear thinning at high shear rates
Yield stress changes as a function of magnetic field
MR fluid behaves approximately as a
– Bingham-plastic
– Field dependent yield force
– plus a viscous stress that is the product of viscosity and velocity
q
Use nondimensional analysis
– Ndim plug thickness as independent variable
– Ndim dynamic range as dependent variable
– Damper performance
MR Fluid Damper
Spring Configuration: Coil-over damper
Pneumatic Reservoir
Pneumatic Reservoir: N2 1.7 MPa
Floating Piston
Maximum Stroke: 9 cm
Spring Retainer
Overall Length: 8 cm
Bore Diameter: 4 cm
Piston Head
Piston Rod
Flux Return
Coils
Steel Core
Damper Performance is Controllable
q
q
q
Plug thickness varies
as a function of field
and yield stress
Change in plug
thickness is akin to
opening and closing a
mechanical valve
Leakage is added to
the flow path to
smoothen damper
response hence the
finite damping at plug
thicknesses
approaching the gap
in the valve
Protective Seating: Impact, Crash, Blast
q
q
q
q
Flight qualification of SH-60 seat with
MR vibration control and deploy
Develop lightweight compact MREAs for
adaptive crash safety
Verify MREA control strategies via test
Other vehicle applications
– Expeditionary Fighting Vehicle (EFV) semiactive seat technology
• Automatic adaptation b/w water-mode shock and
ground-mode vibrations
• Sea trials completed in 9/09
– Adaptive high speed watercraft seats
• Mark V SOC sea trials completed
– Adaptive Mine-blast attenuating seats
• Best “dynamic response index”
Control of Impact Loads
q
q
q
q
Want to use all available stroke for
every impact speed or payload mass
Payload mass varies from 105 lbs
(5th %tile female) to 225 lbs (95th
%tile male)
Impact speed varies from 0-15 mph
(airbags take over at higher speeds)
Must use an adaptive device like an
MREA
V0
payload mass
neutral line
δ
x
magnetorheological
energy absorbers
(MREAs)
reference line
fd
Impact
Plane
S
Magnetorheological Energy
Absorber (MREA) Can Adapt
q
q
q
Passive energy absorbers (EAs) cannot accommodate different occupant
weights (Left)
Use simple optimal control of a terminal trajectory to achieve soft landing for
every occupant weight (right)
Solution requires Lambert W functions
– Schottke diode equation
– MREA utilizes entire stroke
– Minimizes stroking load to the occupant
Extensive Sled Tests at GM
q
q
q
q
Optimal control technique uses all available
stroke (2 inches) regardless of impact speed
or occupant weight
Implemented in dSpace
MREA sized to accommodate stroking loads
needed for 5th female to 95th male
Speeds up to 15 mph
Flight Qualification
Ground Testing
q
Seat was tested to ensure that vibration
isolation system did not hinder seat integrity
during high onset rate crash events
– Seat maintained structural integrity during tests
q
q
Seat has been qualified for flight test in SH-60
Seahawk
Flight test completed…
Challenge Level of Current Design
Max Field-On Force vs. Max Velocity
Maximum Field-On Force (lb)
21000
Carlson et al, 2006
for Seismic Dampers
UMD
18000
Other Researchers
Aft
Damper
at Low
Sink Rate
15000
12000
Forward and Aft
Dampers at High
Sink Rate
Forward
Damper at
Low Sink
Rate
9000
6000
MD-500
Wereley et al, 2005 for
Impact Dampers
Ahmadian et al, 2004
for Impact Dampers
3000
Wereley et al, 2009
for Energy Absorbers
0
0
5
10
15
20
25
Maximum Piston Velocity (ft/s)
The dampers we are designing will require combinations of high force and high
velocity that have not been attempted in known past research.
MR Fluid Characterization:
High Shear Rates
q
q
q
q
q
Need data at high
shear rates up to
100K /s
Virtually all studies
< 1K /s
Lab-built Searle cell
magnetorheometer
can measure up to
25K / s
Apparent viscosity
vs. shear rate
Is the data good?
MR Fluid Characterization for Shear
Rates > 1000 / s
q
Exploit the Mason
number
– Ratio of viscous to
magnetic stress
– Klingenberg showed
curves collapse for low
shear rates
q
q
q
Do apparent viscosity vs.
Mason number curves
should collapse onto a
single curve?
Data is good
We are on the right track
to provide design data
for high shear rate
devices
MR Fluid Characterization
Yield stress persists at 25K /s
Microstructures at High Shear Rate
q
q
Bulk material perspective works for many applications
Need simulation capabilities at high shear
– Need CFD for pressure driven flow
– Need CFD for direct shear flow
q
Current state of the art simulates a few hundred to a few
thousand particles
CFD with MR Particle Interactions
q
q
q
Most simulations
limited to particle
counts in the low
thousands
To simulate practical
flow volumes, we need
millions of particles
Need a three order of
magnitude increase in
particle count!!!
CFD with MR Fluids
q
q
Dynamic simulation allows
for insight into chain
formation under shear
Goal: Simulate at
experimental volume scales
– Need N=1,000,000 particles
but state of the art is
N<10,000
q
Use Nvidia’s CUDA
environment to run code on
desktop GPU
– Delivers 100x increase in
computing performance
Chain Metrics
q
q
Visual observation of
chain formation is
difficult
Developed simulation
scale independent chain
metrics
– Chain length
– Connectivity
q
q
Demonstrated scale
independence
Shown shear response
of metrics is a function
of Mason number
Lamellar Sheet Formation
q
q
q
q
Static equilibrium
structure for MR fluids
are chains
Under shear, particles
form lamellar sheets
Experimentally
demonstrated in MR
fluids
Requires large volume
size to simulate
Lamellar Sheet Formation
of an ER fluid
Cao, J., Huang J., & Zhou, L.
(2006).
Simulation Output
Sherman, S. & Wereley,
N.M. (INTERMAG, 2012).
Million Particle Parallelized
Simulations Using CUDA
Future Challenges in MR – Still a
Rich Field of Research
q
20 years of MR fluid Research
– Still many problems to solve!
q
Demonstrate MR energy absorbers for crash protection systems
–
–
–
–
q
General Motors
U.S. Navy Air Warfare Center (H60 crew seat in 40 ft/s crash)
US Army Research Lab (HDL) Blast mitigation (against IEDs)
US ARMY AATD Active Crash Protection Systems in Rotorcraft
GPU simulations using CUDA enabled understanding of high shear rate
MR fluid behavior
–
–
–
–
Typically measured up to 1000 /s
Want up to 100,000 /s (we have measured to 25,000 /s)
Need GPU simulation because experiments are difficult at high shear rate
Need a full CFD capability in GPU
• Adaptive meshing for complex geometry
• Axisymmetric geometry
• Determine impacts of device on fluid flows
University of Maryland
GPU Summit
Particle Simulation in Magnetorheological Flows
Questions?