A Perturbation Approach to Modeling Eddy Current Signals of Compound Defects

Abstract

Multiple defects located close to each other can significantly complicate eddy current testing. This is particularly problematic in applications like railway track inspection, where detecting and evaluating head checks (surface cracks in close proximity) is crucial. This paper proposes a novel approach using perturbation theory to model the interaction between multiple defects. By simulating single defects and applying perturbation theory, we can generate synthetic data for compound defects, enabling the training of machine learning algorithms for automated defect detection. This approach offers several advantages, including reduced computational cost, improved accuracy, and enhanced flexibility. In the case of rail testing it makes the systematic generation of synthetic head check data feasible in the first place. We present the theoretical framework of the perturbation approach, and validate for surface cracks in the context of rail testing using laboratory experiments and numerical simulations.