Problem
Research questions and friction points this paper is trying to address.
acoustic inverse problem
non-destructive imaging
simulation-based learning
boundary measurements
inclusion reconstruction
Simulation-Based Imaging: Learning Acoustic Inverse Problems from Simulated Data
🤖 AI Summary
This work addresses the inverse problem in acoustic non-destructive imaging by proposing an end-to-end deep learning approach driven solely by simulated data to reconstruct the three-dimensional location and geometry of internal inclusions from sparse boundary pressure measurements. High-fidelity nodal discontinuous Galerkin methods are employed to generate training data, which are then used in conjunction with a 2D convolutional neural network to achieve efficient 3D voxelized reconstruction. The proposed method establishes, for the first time, a practical acoustic imaging system entirely reliant on synthetic data, shifting complexity from hardware to software and substantially reducing deployment costs. Remarkably, it achieves 96% imaging fidelity using only 24 sensors—17% of a full array—and exhibits robustness to noise, with reconstruction error increasing by merely 13% under 5% additive noise while accurately recovering target structures.
📝 Abstract
We introduce Simulation-Based Imaging (SBI), a framework for non-destructive acoustic imaging in which machine learning models trained entirely on simulated data serve as real-time solvers for the acoustic inverse problem. A high-fidelity nodal Discontinuous Galerkin forward solver generates large training datasets by randomizing inclusion geometry within a unit-cube domain; a 2D convolutional neural network then learns a direct mapping from boundary pressure measurements to a 32 by 32 by 32 voxel reconstruction of the interior. The trained model reliably recovers inclusion position and size from 144 boundary sensors with no prior knowledge of inclusion count or geometry. Reconstruction error degrades by only 13% under 5% additive measurement noise, and just 17% of the sensor array (24 of 144 sensors) suffices for quality within 4% of full coverage. These results establish SBI as a viable proof-of-concept imaging device whose complexity resides in software rather than hardware, opening a path toward cheap, portable, deployable imaging systems.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Simulation-Based Imaging
acoustic inverse problem
Discontinuous Galerkin
data-driven reconstruction
non-destructive imaging
🔎 Similar Papers
No similar papers found.
