Yade

CHOLMOD error, Vh==NULL!!

Asked by azim

Dear all,
I have a question as the same as [1] and [2]. I face this problem many times. I need a practical solution for that. because I have a variety of simulations and some of them are so different from the others, and may one single solution not be the overall solution.

After filling a cube and imposing hydraulic gradient, I let my simulation run 10000 iterations, but it breaks with segmentation fault. I'm not sure if it is for particles overlapping (like in [2]) or inconsistent definition of boundaries (like in [1]) or something new. The problems in [1],[2] don't have a solution in them and are solved privately.

I can show a script if it is needed.

[1] https://answers.launchpad.net/yade/+question/241071
[2] https://answers.launchpad.net/yade/+question/634175

Azim

Question information

Language:
English Edit question
Status:
Solved
For:
Yade Edit question
Assignee:
No assignee Edit question
Solved by:
Robert Caulk
Solved:
Last query:
Last reply:
Revision history for this message
Robert Caulk (rcaulk) said : 		#1

Hello,

Please post an MWE and the full error. Also, consider describing the methods you have already used to try and solve this problem. Are there any scenarios that do not yield the error? If so what part of the script are you changing? Have you checked if you have overlapping spheres? Have you checked your boundary conditions?

Cheers,

Robert

Revision history for this message
Meng Qing Xiang (642229461-k) said : 		#2

CHOLMOD is used to solve the function K*u=F in finite element method. I think you may need to install suitesparse properly.

Revision history for this message
azim (mirzavand) said : 		#3

Hi Meng Qing Xiang,
I have yadedaily installed from packages. my ubuntu is 16.02 LTS.
I tried "sudo apt-get install libsuitesparse-metis-dev" as instructed in installation procedure. it says that ' libsuitesparse-metis-dev' has no installation candidate.

Revision history for this message
azim (mirzavand) said : 		#4

Hi Robert,
I am providing a MWE but before that I need to know answers of following questions:

1) about boundary conditions:
my boundaries are build as the same as oedometer.py[1].
first of all, walls are added to the simulation with ids= [0,1,2,3,4,5]:
walls=aabbWalls([mn,mx],thickness=0,oversizeFactor=3,material='walls')

then, particles are added to the simulation with ids [6,7,8,...,10005].
** The errors (Vh=Null) are in range of particle ids and not of walls.

2) about overlappings:
I don't know how to check them!!
[1] https://github.com/MirzaAzim/trunk/blob/master/examples/FluidCouplingPFV/oedometer.py

Thanks
Azim

Revision history for this message
Robert Caulk (rcaulk) said : 		#5

How is your script different from oedometer.py? Does oedometer.py run on your machine without error?

Revision history for this message
azim (mirzavand) said : 		#6

Hi Robert,
the original oedometer.py works fine in my machine. But when I make a new PSD and change elastic parameters, the error comes up.
Azim

Revision history for this message
Best Robert Caulk (rcaulk) said : 		#7

Decrease the time step by an order of magnitude:

O.dynDt=False
O.dt=something/10

Revision history for this message
azim (mirzavand) said : 		#8

Thanks Robert Caulk, that solved my question.