Hello,
I have a question and hope that you can help me.
We are trying to get into using automotive ultrasonic sensors for 3d environment mapping.
The idea is similar to medical 3d ultrasonic devices or how bats map the environment.
You send pulsed ultrasonic waves into different directions (the bat moves the head, we use
some sensors with well defined locations). Then sensors are taking turns, sending,
waiting for the echo, then the next one.
After a full iteration through all sensors it is possible to estimate a 3d map of the environment.
But how good this works is depending strongly on the location of the sensors, their orientation
and of course how many sensors are being used.
Thus we would like to get into some theory first. Since I used Elmer for other projects
I thought it might be possible to use it to simulate a use case (cars on a street, some houses along
the road, maybe some pedestrians).
Do you think it is possible? Pulsed waves? Which model could be used?
Kind regards,
Markus
Using Elmer for simulation of ultrasonic tomography system

 Posts: 40
 Joined: 07 Jun 2011, 11:24
 Antispam: Yes

 Posts: 253
 Joined: 25 Jan 2019, 01:28
 Antispam: Yes
Re: Using Elmer for simulation of ultrasonic tomography system
I would start with section III of the Elmer's model manual, Acoustics and Acoustic Waves. Basically the bat is sending out an acoustic wave and receiving it with an ear.

 Site Admin
 Posts: 3463
 Joined: 22 Aug 2009, 11:57
 Antispam: Yes
 Location: Espoo, Finland
 Contact:
Re: Using Elmer for simulation of ultrasonic tomography system
Hi Markus
It all boils down to wavenumber. Ultrasound is <2cm and hence if you want to study surroundings of cars you need at least 20 m which means at least 1000 waves. For each wave you need 10 elements. Make this 3D and you suddenly have 10^12 elements which cannot be done.
So I would look at ray tracing type of methods. Large wavenumbers and Helmoltz equation is a tough fit at least for standard FEM. There are methods that enrich the solution space with a set of plane waves (ultraweak formulation) and thereby can model propagating waves on a courser mesh. Elmer has been used more for problems with small wavenumbers such as modeling in losses in microscale.
Peter
It all boils down to wavenumber. Ultrasound is <2cm and hence if you want to study surroundings of cars you need at least 20 m which means at least 1000 waves. For each wave you need 10 elements. Make this 3D and you suddenly have 10^12 elements which cannot be done.
So I would look at ray tracing type of methods. Large wavenumbers and Helmoltz equation is a tough fit at least for standard FEM. There are methods that enrich the solution space with a set of plane waves (ultraweak formulation) and thereby can model propagating waves on a courser mesh. Elmer has been used more for problems with small wavenumbers such as modeling in losses in microscale.
Peter

 Posts: 40
 Joined: 07 Jun 2011, 11:24
 Antispam: Yes
Re: Using Elmer for simulation of ultrasonic tomography system
Hello Peter,
thanks a lot for your helpful reply.
Do you think it is also possible to use small wave numbers and just squeeze the solution along the time domain
to scale sonic to ultrasonic? Maybe I am missing something about reflectivity and spacial resolution though...
Basically I just need a certain number of "echos" to have a synthetic dataset where we can try to reconstruct the
environment.
thanks a lot for your helpful reply.
Do you think it is also possible to use small wave numbers and just squeeze the solution along the time domain
to scale sonic to ultrasonic? Maybe I am missing something about reflectivity and spacial resolution though...
Basically I just need a certain number of "echos" to have a synthetic dataset where we can try to reconstruct the
environment.

 Site Admin
 Posts: 3463
 Joined: 22 Aug 2009, 11:57
 Antispam: Yes
 Location: Espoo, Finland
 Contact:
Re: Using Elmer for simulation of ultrasonic tomography system
Hi
You could study the problem on low frequency and study what happens there. However, there is no way, as far as I know, to use the low frequency model to estimate the highfrequency outcome. Maybe there is.
The transient vs. harmonic choice should not have much impact on the required space resolution. Of course one transient timestep is a lot quicker to solve but you need so many steps that usually it pays off to solve the harmonic problem.
Peter
You could study the problem on low frequency and study what happens there. However, there is no way, as far as I know, to use the low frequency model to estimate the highfrequency outcome. Maybe there is.
The transient vs. harmonic choice should not have much impact on the required space resolution. Of course one transient timestep is a lot quicker to solve but you need so many steps that usually it pays off to solve the harmonic problem.
Peter