Auto-tuning numerical method for acoustic wave simulation using analytical solution

Area of Science:

• Computational physics
• Medical imaging
• Acoustic wave propagation

Background:

• Numerical wavefield simulation is crucial for designing Ultrasound Computed Tomography (USCT) systems.
• Existing simulators may introduce errors due to unoptimized numerical models.
• Optimal design requires precise calibration of physical parameters.

Purpose of the Study:

• To propose an auto-tuning numerical method (ATNM) for optimizing USCT simulation parameters.
• To ensure computed wavefields converge to analytical solutions derived from acoustic scattering theory.
• To minimize errors in numerical modeling for accurate USCT system design.

Main Methods:

• Developed an auto-tuning numerical method (ATNM) to optimize parameters like grid size and Courant-Friedrichs-Lewy (CFL) number.
• Employed a genetic algorithm (GA) for automatic calibration of numerical wavefield simulations.
• Investigated the optimal design of the perfectly matched layer (PML) absorption coefficient for USCT.

Main Results:

• The ATNM successfully calibrated numerical wavefield simulations.
• Preliminary tests demonstrated minimization of mean relative error (MRE) between k-Wave simulation and analytical models.
• The optimized PML absorption coefficient design was identified for USCT.

Conclusions:

• The auto-tuning numerical method (ATNM) effectively calibrates wavefield simulations for USCT.
• The calibrated simulator shows generalizability across different physical domains.
• This method enhances the accuracy and reliability of USCT system design through optimized numerical modeling.