Why do particles in the concrete cube compression test disperse directly?

Dear all,

In the Yade official examples, I found the concrete Brazilian splitting test [1], which is a simple process where two walls approach the specimen at a constant speed, resulting in tensile failure of the specimen. Below is the code:
-------------------------------------
# -*- encoding=utf-8 -*-
from __future__ import print_function
from yade import plot,pack
"""
A simple script of a Brazilian splitting test.
A cylinder with x axis form 0 to specimenLength.
Load in z direction by z-perpendicular walls.
Code strongly inspired by uniax.py
"""
# default parameters or from table
readParamsFromTable(noTableOk=True, # unknownOk=True,
 young = 24e9,
 poisson = .2,
 sigmaT = 3.5e6,
 frictionAngle = atan(0.8),
 epsCrackOnset = 1e-4,
 relDuctility = 30,
 intRadius = 1.5,
 dtSafety = .8,
 strainRate = 1,
 specimenLength = .15,
 specimenRadius = .05,
 sphereRadius = 3.5e-3,
)
from yade.params.table import *
# material
concreteId = O.materials.append(CpmMat(
 young = young,
 poisson = poisson,
 frictionAngle = frictionAngle,
 epsCrackOnset = epsCrackOnset,
 relDuctility = relDuctility,
 sigmaT = sigmaT,
))
# spheres
sp=pack.randomDensePack(
 pack.inCylinder((0,0,0),(specimenLength,0,0),specimenRadius),
 radius = sphereRadius,
 spheresInCell = 500,
 memoizeDb = '/tmp/packing-brazilian.db',
 returnSpherePack = True
)
sp.toSimulation()
# walls
zMin,zMax=[pt[2] for pt in aabbExtrema()]
wallIDs = O.bodies.append([wall((0,0,z),2) for z in (zMin,zMax)])
walls = wallMin,wallMax = [O.bodies[i] for i in wallIDs]
v = strainRate * 2*specimenRadius
wallMin.state.vel = (0,0,+v)
wallMax.state.vel = (0,0,-v)
# engines
O.engines=[
 ForceResetter(),
 InsertionSortCollider([Bo1_Sphere_Aabb(aabbEnlargeFactor=intRadius,label='is2aabb'),Bo1_Wall_Aabb()]),
 InteractionLoop(
  [Ig2_Sphere_Sphere_ScGeom(interactionDetectionFactor=intRadius,label='ss2sc'),Ig2_Wall_Sphere_ScGeom()],
  [Ip2_CpmMat_CpmMat_CpmPhys()],
  [Law2_ScGeom_CpmPhys_Cpm()],
 ),
 NewtonIntegrator(damping=.2),
 CpmStateUpdater(realPeriod=.5),
 PyRunner(iterPeriod=100,command='addPlotData()',initRun=True),
 PyRunner(iterPeriod=100,command='stopIfDamaged()'),
]
# stop condition
def stopIfDamaged():
 if O.iter < 1000: # do nothing at the beginning
  return
 fMax = max(plot.data["f"])
 f = plot.data["f"][-1]
 if f/fMax < .6:
  print("Damaged, stopping.")
  print("ft = ",max(plot.data["stress"]))
  O.pause()
# plot stuff
def addPlotData():
 # forces of walls. f1 is "down", f2 is "up" (f1 needs to be negated for evlauation)
 f1,f2 = [O.forces.f(i)[2] for i in wallIDs]
 f1 *= -1
 # average force
 f = .5*(f1+f2)
 # displacement (2 times each wall)
 wall = O.bodies[wallIDs[0]]
 dspl = 2*wall.state.displ()[2]
 # stress (according to standard brazilian test evaluation formula)
 stress = f/(pi*specimenRadius*specimenLength)
 # store values
 yade.plot.addData(
  t = O.time,
  i = O.iter,
  dspl = dspl,
  f1 = f1,
  f2 = f2,
  f = f,
  stress = stress,
 )
# plot dspl on x axis, stress on y1 axis and f,f1,f2 in y2 axis
plot.plots={'dspl':('stress',None,'f1','f2','f')}
O.dt = 0.
O.step(); # to create initial contacts
# now reset the interaction radius and go ahead
ss2sc.interactionDetectionFactor=1.
is2aabb.aabbEnlargeFactor=1.
# time step
O.dt = dtSafety*PWaveTimeStep()
# run simulation
plot.plot()
#O.run()
-------------------------------------------------------

Regarding this process, I have two questions, and I would appreciate any assistance or insights from the community.

1. According to the model's micro-scale parameters, this should simulate a concrete specimen, but the stress-displacement curve indicates that the displacement is too large, roughly an order of magnitude greater than real-world conditions. Why is there such a significant difference between micro-scale parameters and macroscopic behavior, and how should I adjust it? (I believe that as an example, the micro-scale parameters in this code are reasonably set, at least in the same order of magnitude.)
2. I would like to simulate a uniaxial compression test, so I changed the original specimen shape to a cube. However, the process becomes very strange, with a sudden increase in stress, and the specimen composed of particles eventually explodes. What is causing this? Did I miss setting some parameters?
  I have replaced the #spheres section of the original code with the following code：
# spheres
sp=pack.randomDensePack(
 pack.inAlignedBox((0,0,0),(specimenLength,specimenLength,specimenLength)),
 radius = sphereRadius,
 spheresInCell = 500,
 memoizeDb = '/tmp/packing-brazilian.db',
 returnSpherePack = True
)
sp.toSimulation()

Thanks
Antonia

[1] https://gitlab.com/yade-dev/trunk/-/blob/master/examples/concrete/brazilian.py?ref_type=heads