Hello!
I'm interested in sound propagation from liquid to solid and vice versa.
I understand the new functions pstdElastic are meant for solids while kSpaceFirstOrder is meant for liquids. Is there a way built in to have those simulations interact in a way that allows for pressure or stress waves to travel from one to the other?
Thank you very much for this powerful toolbox!
Marvin Willam
Hi Brad!
Thank you for your reply. I used a test simulation in 2D to have a look at the behaviour of pstdElastic for liquids before applying it to more complex volumes and I encountered a problem:
I put a stress source (disc) into a grid of water. Then I created a burst signal and applied it to the source like this:
source.sxx = input_signal;
source.syy = input_signal;
It looked like it would work fine - I was trying to achieve homogenous radiation - but as the pressure waves reached the corners of my grid, everything inside the medium but the PML seemed to explode, reaching high pressure values. Is there a way to mess up your PML so that this happens? I tried turning it off and it was a completely different result.
Thanks for your help!
Marvin
Hi Marvin,
That certainly sounds suspicious. If you post your code, I can take a look.
Also, have you updated to k-Wave version 1.1.1? This corrected a bug in the pstdElasticcodes that might affect stability.
Brad.
Hi Brad,
the new version does not seem to solve the issue. I would be glad if you took a look at my code :)
Best regards,
Marvin
clear all;
DATA_CAST = 'double';
MASK_PLANE = 'xy';
USE_STATISTICS = true;
PML_X_SIZE = 10;
PML_Y_SIZE = 10;
Nx = 100;
Ny = 100;
c_water = 1500; % [m/s]
c_plastic = 2230; % [m/s]
c_poly = 1560; % [m/s]
c_air = 343; % [m/s] @ 20°C
rho_water = 998; % [kg/m^3]
rho_plastic = 1330; % [kg/m^3]
rho_poly = 1100; % [kg/m^3]
rho_air = 1.22; % [kg/m^3]
freq = 15e3; % [Hz]
lambda = c_water/freq;
x = 1;
dx = x/Nx; % [m]
dy = dx; % [m]
kgrid = makeGrid(Nx, dx, Ny, dy);
tr_pos = [round(0.5*Nx), round(0.5*Ny)];
tr_rad = 10.8e-2/(2);
tr_pos_abs = [tr_pos(1)*dx, tr_pos(2)*dy]
ITC = makeDisc(Nx, Ny, tr_pos(1), tr_pos(2), round(tr_rad/dx));
medium.sound_speed_compression = c_water*ones(Nx, Ny);
medium.sound_speed_shear = zeros(Nx, Ny);
medium.density = rho_water*ones(Nx, Ny);
t_end = 10e-4; % [s]
kgrid.t_array = makeTime(kgrid, c_water*1.5);
delta_t = kgrid.t_array(2) - kgrid.t_array(1);
ncycles = 2;
input_signal = toneBurst(1/delta_t, freq, ncycles);
source.sxx = input_signal;
source.syy = input_signal;
source.s_mask = ITC;
sensor.mask = zeros(Nx, Ny);
sensor.mask(50, 100) = 1;
input_args = {'PlotScale', 'auto','PMLSize', [PML_X_SIZE, PML_Y_SIZE],...
'DataCast', DATA_CAST, 'DataRecast', true,'PlotPML', true};
sensor_data = pstdElastic2D(kgrid, medium, source, sensor, input_args{:});
Hi Marvin,
It seems to be a plot scale issue. With auto plot scaling, the limits are constantly adjusted, so it looks like there are waves remnant in your grid after the pulse reaches the PML, even though these are very small.
If you change your code to 'PlotScale, [-1, 1] it should work as expected. Also, I would advise using a slightly fatter PML (perhaps 20 grid points).
Hope that helps,
Brad.
Hi Brad,
thank you for your help :)
Marvin
You must log in to post.