Font Size:
Inverse scattering technique using deep learning for 3-D scalar wave propagation
Last modified: 2023-06-24
Abstract
The use of ultrasonic non-destructive testing is widespread in providing evidence of safety for structural materials.Accurately identifying the position, size, and shape of defects in materials is important for structural monitoring and health diagnostics.The methods for defect shape reconstruction in materials have been developed by several researchers for several years.The inverse scattering technique is known as an effective method for reconstructing a defect and has been applied to many engineering problems.However, in general, the inverse scattering formulation is mathematically very complicated andreconstructing a defect using it is relatively time-consuming.The inverse scattering technique is not as commonly used in practical engineering applications compared to the synthetic aperture focusing technique (SAFT), which is the most widely used method for defect detection and imaging.
In recent years, the deep learning has attracted attention in many engineering fields.Deep learning is a type of machine learning that involves training artificial neural networks to learn from large amounts of image data. It has proven to be particularly effective for solving image classification problems.The use of deep learning for ultrasonic non-destructive testing has been considered as an effective way to reduce the labor required by inspectors.Meng. at el. applied the deep learning to develop an defect imaging system for composite materials.Saitoh et al. solved an inverse scattering problem for a scatterer in an infinite region using deep learning.This method uses deep learning of B-scope images at receiver points to predict the size and position of a defect.However, it has a limitation as it is only applicable for 2-D antiplane wave propagation.
In this research, we extend the inverse scattering technique using deep learning from 2-D to 3-D case.Image data for deep learning are created by using the convolution quadrature time-domain boundary element method.The proposed inverse scattering method is verified through several numerical examples.
In recent years, the deep learning has attracted attention in many engineering fields.Deep learning is a type of machine learning that involves training artificial neural networks to learn from large amounts of image data. It has proven to be particularly effective for solving image classification problems.The use of deep learning for ultrasonic non-destructive testing has been considered as an effective way to reduce the labor required by inspectors.Meng. at el. applied the deep learning to develop an defect imaging system for composite materials.Saitoh et al. solved an inverse scattering problem for a scatterer in an infinite region using deep learning.This method uses deep learning of B-scope images at receiver points to predict the size and position of a defect.However, it has a limitation as it is only applicable for 2-D antiplane wave propagation.
In this research, we extend the inverse scattering technique using deep learning from 2-D to 3-D case.Image data for deep learning are created by using the convolution quadrature time-domain boundary element method.The proposed inverse scattering method is verified through several numerical examples.
An account with this site is required in order to view papers. Click here to create an account.