Efficient Model Assisted Probability of Detection Estimations in Eddy Current NDT with ACA-SVD Based Forward Solver

Area of Science:

• Nondestructive Testing (NDT)
• Computational Electromagnetics
• Signal Processing

Background:

• Model-assisted probability of detection (MAPoD) is vital for NDT system capability assessment.
• Traditional simulation models for MAPoD can be computationally intensive.
• Accurate crack sizing and localization are key challenges in NDT.

Purpose of the Study:

• To develop an efficient 3D eddy current nondestructive evaluation (ECNDE) forward solver for improved MAPoD estimations.
• To enhance the performance of adaptive cross approximation (ACA) algorithms using singular value decomposition (SVD) in ECNDE.
• To validate the proposed solver's robustness and efficiency against established benchmarks.

Main Methods:

• Implementation of a 3D ECNDE forward solver based on the Boundary Element Method (BEM).
• Integration of Singular Value Decomposition (SVD) as a recompression technique within the ACA algorithm.
• Utilizing the solver to calculate probability of detection (PoD) curves for surface flaws, considering liftoff variations.

Main Results:

• The proposed ACA-SVD based BEM solver demonstrates significant efficiency improvements for ECNDE forward problems.
• Robustness and accuracy were confirmed by comparing simulated coil impedance variations with analytical, semi-analytical, and experimental data.
• Successful calculation of PoD curves for rectangular surface flaws on a finite thickness plate.

Conclusions:

• The developed ACA-SVD based BEM fast ECNDE forward solver is a highly effective tool for MAPoD estimations.
• This method offers a computationally efficient alternative for quantifying NDT system performance.
• The approach provides a reliable simulation model for crack detection capability analysis in NDT.