I use your published program fastFWI to run some simulations. And compare received data with 2D analytical results, modeling frequency is 90 kHz. The first model I used is a 1 meter by 1 meter homogeneous model, with transducers deployed along a circle and each transducers acted as both transmitters and receivers
The real part of received data at transducer locations obtained by finite difference modeling (FDM) and 2D analytical green functions are shown as follows
It can be seen that FDM data and analytical results matched very well.
However, when I increase the size of model to 2 meter (x direction) by 3 meter (y direction). And use parallel array setup, as shown in the following figure (grid spacing 2mm)
The real part of FDM wavefield and analytical wavefield (obtained by green functions) are shown as follows
It shows that with the increase of model size, the analytical results (homogeneous model) and FDM results begin to differ significantly. Such discrepancy influence the imaging results, leading to the algorithm failure at some circumstances. But I cannot figure out the possible reasons.
What is the possible reason that causes discrepancy between the FDM modeling wavefield and analytical results? Is such discrepancy caused by modeling grid size or some other reasons?
Dear Dr Vladimir Kazei
I use your published program fastFWI to run some simulations. And compare received data with 2D analytical results, modeling frequency is 90 kHz. The first model I used is a 1 meter by 1 meter homogeneous model, with transducers deployed along a circle and each transducers acted as both transmitters and receivers
The real part of received data at transducer locations obtained by finite difference modeling (FDM) and 2D analytical green functions are shown as follows
It can be seen that FDM data and analytical results matched very well.
However, when I increase the size of model to 2 meter (x direction) by 3 meter (y direction). And use parallel array setup, as shown in the following figure (grid spacing 2mm)![1-3](https://github.com/vkazei/fastFWI/assets/54341310/769d3c66-ba18-4c10-8389-117883bf1a71)
The real part of FDM wavefield and analytical wavefield (obtained by green functions) are shown as follows![1-4](https://github.com/vkazei/fastFWI/assets/54341310/e2146881-5e67-417e-8fa1-e40119fe1824)
It shows that with the increase of model size, the analytical results (homogeneous model) and FDM results begin to differ significantly. Such discrepancy influence the imaging results, leading to the algorithm failure at some circumstances. But I cannot figure out the possible reasons.
What is the possible reason that causes discrepancy between the FDM modeling wavefield and analytical results? Is such discrepancy caused by modeling grid size or some other reasons?
Hope to hear from your answers!