[CIG-SEISMO] How to choose time step and other parameters to avoid floating-point error

[Chunyang Xiao]

Hi there,


I am an undergraduate student majoring in mechanical engineering. Recently when I changed the speed of P-wave (Vp from the section 'velocity and density models' in Par_file) in SPECFEM2D, an error occurred.


Program received signal SIGFPE: Floating-point exception - erroneous arithmetic operation.

Backtrace for this error:

#0  0x7F209C4FDE08

#1  0x7F209C4FCF90

#2  0x7F209C14D4AF

#3  0x4192BA in read_regions_ at read_regions.f90:99

#4  0x418CEE in read_parameter_file_ at read_parameter_file.F90:88

#5  0x42521C in MAIN__ at meshfem2D.F90:403

Floating point exception (core dumped)

This happened a few times also when changed other parameters, such as time step.  A similar error happened to Mr. Moritz Wehler before, and Dimitri's answer on Wed May 27 07:40:46 PDT 2015 was


It seems you simply forgot to also divide the time step by two to remain

stable when you increased mesh resolution by two:

  *** Max stability for P wave velocity (must be below about 0.50 or so)

= 0.972936358376379

I was curious where we could see the 'Max stability for P wave velocity' for a given case. As I turned to the SPECFEM2D user manual (version 7.0) for help, I saw chapter 4.6 (How to choose the time step) say:


You can directly compare these values with the value given in sentence ‘Max stability for P wave velocity’ in file output_solver.txt to see whether you set the correct ∆t in Par_file or not. For elastic simulation, the CFL value given in output_solver.txt does not consider the V p /V s ratio......

That makes me more confused because I can find neither the sentence ‘Max stability for P wave velocity’ nor the file output_solver.txt. The file named output_solver.txt only exists in the folders of a few examples I did not use or run.

Also, in Ms. Parvathy Sindhu Nair's question on Mon Feb 15 14:25:50 PST 2016, she and Dimitri mentioned CFL, which seems to matter but does not appear in my Par_file.

this is either the time step that is a bit too high (unlikely, since the CFL you used (0.44) seems fine), or if the amplitude of the source you used is very high......

Would anyone kindly let me know how I can choose appropriate parameters (time step, speed, etc.) to avoid such floating-point errors and where I can see my CFL, output_solver.txt, and max stability for P wave?

Sincerely,
Charles Xiao
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/cig-seismo/attachments/20180501/43777f42/attachment.html>