#### Location

La Jolla, CA

#### Start Date

1-1-1987 12:00 AM

#### Description

Conventional eddy current NDT methods rely for their operation on the interaction of quasi-static electromagnetic fields with flaws in the specimen under test. The physics of such interactions are described completely by a parabolic diffusion equation 1 ∇X(1μ∇XA)=−σ∂A∂t+JS as holography and tomography in the analysis of data from eddy current probes [1]. Ideally, one would desire an analytical closed formjsolution of equation (1) in terms of the material parameters μ(r) ) and σ(r), so that one has a direct method for solving the inverse problem or imaging problem. The nature of the defect characterization problem in eddy current NDT and the difficulties involved in the analytical modeling of realistic test geometries are described at length in [2,3]. Simulation of nonlinear, practical problems with arbitrary defect shapes are generally done using numerical techniques such as the finite difference and finite element methods. However, the major drawback of numerical models is lack of a closed form solution, which renders them inadequate for directly solving the inverse problem. This paper presents a new approach to the general inverse problem in NDT of defect imaging, that uses the finite element model iteratively for estimating the test specimen parameters {o} as shown in Figure 1.

#### Book Title

Review of Progress in Quantitative Nondestructive Evaluation

#### Volume

6A

#### Chapter

Chapter 4: Image Analysis, Signal Processing and AI

#### Section

Artificial Intelligence

#### Pages

899-906

#### DOI

10.1007/978-1-4613-1893-4_103

#### Copyright Owner

Springer-Verlag US

#### Copyright Date

January 1987

#### Language

en

#### File Format

application/pdf

An AI Approach to the Eddy Current Defect Characterization Problem

La Jolla, CA

Conventional eddy current NDT methods rely for their operation on the interaction of quasi-static electromagnetic fields with flaws in the specimen under test. The physics of such interactions are described completely by a parabolic diffusion equation 1 ∇X(1μ∇XA)=−σ∂A∂t+JS as holography and tomography in the analysis of data from eddy current probes [1]. Ideally, one would desire an analytical closed formjsolution of equation (1) in terms of the material parameters μ(r) ) and σ(r), so that one has a direct method for solving the inverse problem or imaging problem. The nature of the defect characterization problem in eddy current NDT and the difficulties involved in the analytical modeling of realistic test geometries are described at length in [2,3]. Simulation of nonlinear, practical problems with arbitrary defect shapes are generally done using numerical techniques such as the finite difference and finite element methods. However, the major drawback of numerical models is lack of a closed form solution, which renders them inadequate for directly solving the inverse problem. This paper presents a new approach to the general inverse problem in NDT of defect imaging, that uses the finite element model iteratively for estimating the test specimen parameters {o} as shown in Figure 1.