1. A Model for Rockfall Test
●Artificial rock model for rockfall test
Five iron balls (diameter 11mm, weight 5.48g) are connected by adhesive. x y 2 1 3 4 z 5
Plane Side Opposite side
The model is easily simulated by DEM

2. Experiment Condition
Slope (Wood)
Magnetic switch
rock
Height 30cm ↑ 4 5 1 2
100cm
← front
50cm
30 degree
＜side view＞
Initial Direction
Set the rock model to same direction, then drop it by magnet switch in order to drop the rock model in same condition
From slope foot, Y
From center, X
drop
＜plane view＞

3. Distribution of distance
We repeat the experiment 300 times
(drop of the rock model)
X-direction (from center)
　Mean μx =0.4cm
　St.Dev. =19.2cm
Y-direction (from foot)
　Mean μY ＝68.0cm
　St.Dev.σY＝24.2cm X
Center line Y
slope
foot
From slope foot, Y (cm)
slope
drop
slope
top
＜plane view＞
From center, X (cm)

4. Numerical Simulation by DEM
●Input parameter
●DEM model
↑up
・Spring constant (normal, shear)
　　2.40×105 N/m
・Damping constant (normal)
　　4.54×101 N・s/m
・Damping constant (shear)
　　5.80×101 N・s/m
・Friction angle between slope and particles
　　24.7 degree
They are determined
from simple test (rebound height),
and material property.
←front
All conditions, such as
initial position, height of drop,
etc., are same as the experiment

5. Perturbation of initial position in DEM
Rotate the model around axis X, Y, Z
with uniform random number
Case 1 : -1 to 1, Case 2 : -3 to 3, Case 3 : -5 to 5 (degree)
An example of DEM result
Trajectory of the rock model
Magnetic switch ↑ 4 5 1 2
X in
plane view
From center, X (cm)
Side view
Height (cm)
← front x z y
Horizontal location (cm)

6. Comparison of Experiment and DEM
Rotate the model around axis X, Y, Z, with uniform random number
Case 1 : -1 to 1 degree
Experiment
DEM
From slope foot, Y (cm)
From center, X (cm)

7. Comparison of Experiment and DEM
Rotate the model around axis with uniform random number
Case 1 : -1 to 1, Case 2 : -3 to 3, Case 3 : -5 to 5 (degree)
Cumulative curves are almost same irrespective of case 1-3 (uncertainty of position)
Output uncertainty does not depends on input uncertainty
From center, X
From slope foot, Y
Cumulative Distribution
DEM, Case 1
DEM, Case 2
DEM, Case 3
Experiment
　From center, X (cm) or From slope foot, Y (cm)

8. DEM result by Compiler A and B
Highly nonlinear simulation is sensitive to small change of any condition. What happens when the compiler is changed?
All conditions are same except the COMPILER!
Case 1 : -1 to 1 (degree)
Compiler A
Compiler B
Plane View
Side view
From center, X (cm)
Height (cm)
From slope foot, Y (cm)
Compiler A
Compiler B
Horizontal location (cm)
From center, X (cm)
Treatment of last digit of significant figures depends on compiler, option.

9. DEM result by Compiler A and B
Rotate the model around axis X, Y, Z, with uniform random number
Case 1 : -1 to 1 (degree)
From center, X
From slope foot, Y
Cumulative Distribution
Compiler A
Compiler B
Experiment
From center, X (cm) or From slope foot, Y (cm)

10. Concluding Remarks Rockfall test
The places where the rock model stop have large uncertainty
Simulations of the rockfall test by DEM
Good agreement in terms of distributions of the place where the model stop
Uncertainty level of initial condition does not affect the distribution
A compiler of DEM program affect each simulation, but does not affect the distribution 10 10