How to achieve a twisting/bending-caused bond break in Yade?

Thanks Bruno Chareyre, that solved my question.

Hi Jasper,

I think you can achieve what you want with the JCFPM. In fact, you can reproduced the effect of bending/twisting resistance by increasing the interaction radius when using the JCFPM. The effect is "similar" in terms of material behavior/strength. Also, as you know, in such context (breakage of samples due to impact), fractures/fragments will develop as a result of the presence of flaws in the material (pre-existing fractures and/or weakness planes), and the JCFPM was made for that purpose and I think it might be relevant for your study.

Also, be aware that the way the numerical sample breaks is directly related to its mechanical properties (and the loading conditions as well). Maybe you could not achieve a "representative" simulation because you chose a non adequate set of parameters?

ANW, please find below a "cartoon" simulation that might be useful for you.

Luc

###

from yade import pack, plot

################# SIMULATIONS DEFINED HERE

#### Material microproperties
intR=1.2 # allows near neighbour interaction (can be adjusted for every packing)

#### material definition
def sphereMat(): return JCFpmMat(type=1,density=3000,young=5e9,poisson=0.2,frictionAngle=radians(15),tensileStrength=5e6,cohesion=5e7)
def wallMat(): return JCFpmMat(type=0,density=3000,young=1e9,poisson=0.2,frictionAngle=radians(15),tensileStrength=0,cohesion=0)

#### create the specimen
pred=pack.inCylinder((0,-0.5,0.5),(0.,0.5,0.5),0.25)
#O.bodies.append(pack.regularHexa(pred,radius=0.025,gap=0.,material=sphereMat))
O.bodies.append(pack.randomDensePack(pred,radius=0.025,rRelFuzz=0.2,useOBB=True,spheresInCell=3000,memoizeDb='/tmp/gts-triax-packings.sqlite',returnSpherePack=False,color=(0.9,0.8,0.6),material=sphereMat))

R=0
Rmax=0
Rmin=1e6
nbSpheres=0.
for o in O.bodies:
 if isinstance(o.shape,Sphere):
   nbSpheres+=1
   R+=o.shape.radius
   if o.shape.radius>Rmax:
     Rmax=o.shape.radius
   if o.shape.radius<Rmin:
     Rmin=o.shape.radius
Rmean=R/nbSpheres

print('nbSpheres=',nbSpheres,' | Rmean=',Rmean, ' | Rmax/Rmin=', Rmax/Rmin)

O.bodies.append(geom.facetBox((0,0,0),(2,2,1),Quaternion((1,1,0),pi/20.),wallMask=16,color=(1,1,1),wire=False,fixed=1,material=wallMat))

################# ENGINES DEFINED HERE

O.engines=[
    ForceResetter(),
    InsertionSortCollider([Bo1_Sphere_Aabb(aabbEnlargeFactor=intR,label='Saabb'),Bo1_Facet_Aabb()]),
    InteractionLoop(
        [Ig2_Sphere_Sphere_ScGeom(interactionDetectionFactor=intR,label='SSgeom'),Ig2_Facet_Sphere_ScGeom()],
        [Ip2_JCFpmMat_JCFpmMat_JCFpmPhys(cohesiveTresholdIteration=1,label='interactionPhys')],
        [Law2_ScGeom_JCFpmPhys_JointedCohesiveFrictionalPM(recordCracks=True,Key='test',label='interactionLaw')]
    ),
    GlobalStiffnessTimeStepper(active=1,timeStepUpdateInterval=10,timestepSafetyCoefficient=0.9,defaultDt=utils.PWaveTimeStep()),
    NewtonIntegrator(damping=0.4,gravity=(0,0,-100)),
    VTKRecorder(iterPeriod=2000,initRun=True,fileName='fallingBlock-',recorders=['facets','spheres','jcfpm','cracks'],Key='test',label='vtk')
]

################# PREPROCESSING

#### manage interaction detection factor during the first timestep and then set default interaction range
O.step();
### initializes the interaction detection factor
SSgeom.interactionDetectionFactor=-1.
Saabb.aabbEnlargeFactor=-1.

#### coordination number verification
numSSlinks=0
numCohesivelinks=0
for i in O.interactions:
    if isinstance(O.bodies[i.id1].shape,Sphere) and isinstance(O.bodies[i.id2].shape,Sphere):
      numSSlinks+=1
    if i.phys.isCohesive :
      numCohesivelinks+=1

print ("nblinks=", numSSlinks, " | nbCohesivelinks=", numCohesivelinks, "|| Kcohesive=", 2.0*numCohesivelinks/nbSpheres)

#### delete floating particles
Kmin=3
erased=0
for o in O.bodies:
    if isinstance(o.shape,Sphere):
        nbCont=0
        for i in O.interactions.withBody(o.id) :
            if not i.isReal : continue
            if i.phys.isCohesive:
                nbCont+=1
        if nbCont<Kmin:
            erased+=1
            O.bodies.erase(o.id)
print 'nb of floating particles (erased) = ', erased

################# SIMULATION REALLY STARTS HERE
O.run(14000)

@Luc
Do you mean that JCFpm has upper bounds for torques at interactions?
I thought rolling friction in the CohesiveFrictional law was the only concrete example of such thing atm.
Bruno

@Bruno

Yes, you are right, JCFPM does not consider rolling/twisting resitance in its formulation. But, increasing the number of bonds between particles (by allowing near neighbor bonding) increases their degree of interlocking. My interpretation is that the macroscopic consequence of this enhanced interlocking is"similar" to what you would obtain with a rolling resistance since the movements (i.e. translation and rotation) of the particles are limited. This, of course, only applies when the material is mostly cohesive (when particles are bonded). But, to be honest, I don't have any quantitative proof of this supposed macroscopic "similarity".

Luc

I once added rolling/twisting resistance to JCFpm. It was not too difficult: just add all the necessary members/if statements as shown in cohesivefrictional. But in the end, it didn't make a difference while studying wing cracks. I get the feeling rolling and twisting might play negligible roles in cohesive materials for DEM (or maybe the interaction range dominated as luc is pointing out?). Meanwhile, rolling and twisting seems to be indispensable in non-cohesive DEM models for obtaining shear bands.

Hi Jasperb,

I see your problem by chance when I am reading topics related to JCFPM.

I used to make some dynamic simulation for SHPB test by a commercial DEM software and now I am learning Yade for its extensibility.

I agree with Robert, the effects of rolling/twisting are negligible for your study. And in my opinion, the distinct characteristics of JCFPM is also not relevant to your study.

For your problem, I suggest you to use clumps in Yade or define clusters of particles by yourself to establish your sample.

hope this will help you.

Cheers
Yang