yade-users team mailing list archive

[Léo Sénique <question679748@xxxxxxxxxxxxxxxxxxxxx>]

New question #679748 on Yade:
https://answers.launchpad.net/yade/+question/679748

Hello,

I'm trying to simulate a viscous flow using a lubrication law, but I'm running into an issue with two of the available resolution methods. I was initially using a modified version of the latest yadedaily release (2018.02b-3905f5add2~bionic), but I have found that this problem still occurs with the original version of this release.
I've simplified the script to try to pinpoint the exact issue:

from yade import pack,plot
import math
import random as rand
import numpy as np
from scipy.stats import gaussian_kde

O.dt=1e-6
slope = math.radians(31)
length = 9
width = 2
height = 2
gravityVector = Vector3(9.81*sin(slope),0.0,-9.81*cos(slope))
O.materials.append(FrictMat(density=2500,frictionAngle=math.radians(35.),poisson=0.2,young=1e7,label='Frict'))
O.materials.append(FrictMat(density=2500,frictionAngle=math.radians(0.),poisson=0.2,young=1e7,label='FrictWall'))

DepBox = O.bodies.append(aabbWalls([(0.,0.,-0.1),(length,width,2*height)],thickness=0.0,material='FrictWall',oversizeFactor=1.0,color=[104/255., 111/255., 140/255.]))
partCloud = pack.SpherePack()
partCloud.makeCloud(minCorner=(length/4.*3.,0.,0.),maxCorner=(length,width,0.8*height), rMean=0.05,num=1000)
spIds = partCloud.toSimulation(material='Frict',color=[205./255,175./255,149./255])

theta_method = 0.55
resolution = 2

O.engines=[
	ForceResetter(),
	InsertionSortCollider(
	[
		Bo1_Sphere_Aabb(aabbEnlargeFactor=1.5),
		Bo1_Box_Aabb(),
		Bo1_Wall_Aabb()
	],
	verletDist=0.1,
	allowBiggerThanPeriod=False),

	InteractionLoop(
	[
		Ig2_Sphere_Sphere_ScGeom6D(interactionDetectionFactor=1.5),
		Ig2_Box_Sphere_ScGeom6D()
	],
	[
		Ip2_FrictMat_FrictMat_LubricationPhys(eta=2,eps=0.001),
	],
	[
		Law2_ScGeom_ImplicitLubricationPhys(
						activateNormalLubrication=True,
						activateTangencialLubrication=True,
						activateTwistLubrication=True,
						activateRollLubrication=True,
						theta=theta_method,
						resolution=resolution,
						warnedOnce=True,
						maxSubSteps=20,
						SolutionTol=1e-6,
						debug=True)
	]),

	DragEngine(Rho=1200,Cd=0.47, label='Drag', dead=False),
	NewtonIntegrator(gravity=gravityVector, damping = 0.4, label='newton'),
]
Drag.ids=spIds
Drag.Rho=1200.
Drag.Cd=0.47

This runs fine when using resolution=0  (Iterative exact resolution with substepping) in the lubrication law options. 

However, when resolution=1 is used (Newton-Rafson dimentionless resolution), the following error fills the console entirely several times over:
WARN  /data/trunk/pkg/dem/Lubrication.cpp:132 NRAdimExp_integrate_u: Max Substepping reach: results may be inconsistant F=nan

A different error appears with resolution=2 (Dichotomy dimentionless resolution), filling the console as well:
WARN  /data/trunk/pkg/dem/Lubrication.cpp:199 DichoAdimExp_integrate_u: Max iteration reach: d_left=nan F_left=nan d_right=nan F_right=nan
WARN  /data/trunk/pkg/dem/Lubrication.cpp:163 DichoAdimExp_integrate_u: Wrong direction
ERROR /data/trunk/pkg/dem/Lubrication.cpp:175 DichoAdimExp_integrate_u: Initial point problem!! d_left=nan F_left=nan d_right=nan F_right=nan

Is something wrong with my script?
I'm also not entirely sure I understand the difference between those methods and if I should just use the one that works. Is there a difference in precision?

Thanks,

Léo

-- 
You received this question notification because your team yade-users is
an answer contact for Yade.