k-Wave

1. bwalker000
   

   Member
   Joined: Nov '15
   Posts: 4

   I am modeling the reflections from a piece of plastic in air. When I use the following material properties the simulation is unstable and diverges.

   % source / reflector (PLA)
   c_plastic = 2730;   % [m/s]         According to Olympus NDT
   rho_plastic = 1190; % [kg / m^3]    According to ONDA Tables
   
   c_air = 345;        % [m/s]         According to ONDA Tables
   rho_air = 1.293;    % [kg / m^3]    According to ONDA Tables

   However, if I increase the density of the air by a factor of ten, then the simulation is stable:

   % source / reflector (PLA)
   c_plastic = 2730;   % [m/s]         According to Olympus NDT
   rho_plastic = 1190; % [kg / m^3]    According to ONDA Tables
   
   c_air = 345;        % [m/s]         According to ONDA Tables
   rho_air = 10*1.293;    % [kg / m^3]    According to ONDA Tables

   This isn't a bad workaround since the speed of sound stays the same and the reflection coefficient largely stays the same (huge).

   Alternatively I can reduce the density of the plastic and get a stable solution:

   c_plastic = 2730;   % [m/s]         According to Olympus NDT
   %rho_plastic = 1190; % [kg / m^3]    According to ONDA Tables
   rho_plastic = 1/10*1190; % [kg / m^3]    According to ONDA Tables
   
   c_air = 345;        % [m/s]         According to ONDA Tables
   rho_air = 1.293;    % [kg / m^3]    According to ONDA Tables

   My suspicion is that the temporal sampling rate is somehow not properly determined for materials with these extreme mismatches. I would suspect that this can be overridden to resolve the issue. Can anyone comment on this hypothesis or tell me how to do an override of the temporal sampling rate?

   Thanks.

   Posted 2 years ago #

2. bencox
   

   Developer
   Joined: Mar '10
   Posts: 292

   Hi bwalker000,

   Your first plan of making the material properties of air larger, which as you say doesn't change much acoustically, is a trick we've used in the past. That's probably the best option, if you have to model such an interface. If the interface is planar it would be worth thinking about whether you can use an image source to model the reflection.

   If you want to you can set the time steps yourself using the setTime' method ofkWaveGrid', eg.

   kgrid = kWaveGrid(Nx, dx);'

   kgrid.setTime(Nt, dt)'

   Best wishes
   Ben

   Posted 2 years ago #

RSS feed for this topic

Reply

You must log in to post.