Target Detection and Characterization from Electromagnetic Induction Data

Habib Ammari, Junqing Chen, Chen, Zhiming, Josselin Garnier, Darko VolkovarXiv 2012 · DOI · arXiv

Abstract

The goal of this paper is to contribute to the field of nondestructive testing by eddy currents. We provide a mathematical analysis and a numerical framework for simulating the imaging of arbitrarily shaped small volume conductive inclusions from electromagnetic induction data. We derive, with proof, a small-volume expansion of the eddy current data measured away from the conductive inclusion. The formula involves two polarization tensors: one associated with the magnetic contrast and the second with the conductivity of the inclusion. Based on this new formula, we design a location search algorithm. We include in this paper a discussion on data sampling, noise reduction, and on probability of detection. We provide numerical examples that support our findings.

Introduction

Nondestructive testing (NDT) is a cornerstone of industrial safety and quality control, used to identify internal defects like cracks or impurities without compromising a component's integrity. Eddy current induction is a preferred method for metallic structures, yet the mathematical modeling of how small inclusions perturb electromagnetic fields remains a complex challenge. This is particularly true in the "eddy current regime," where the skin depth (the depth of current penetration) is comparable to the size of the object. This paper addresses this gap by providing a rigorous mathematical framework and a numerical approach for the detection and characterization of small conductive inclusions.

Mathematical Modeling: The Small-Volume Expansion

Central to this work is the derivation of a small-volume expansion formula. Unlike full Maxwell equations, which are computationally expensive and often unnecessary for small targets, this asymptotic formula describes the magnetic field perturbations measured away from a target.

Crucially, the authors introduce a novel conductivity polarization tensor alongside the standard magnetic polarization tensor. Together, these two tensors encapsulate how the target's geometry and electrical properties interact with the induced field. This dual-tensor approach provides a more complete mathematical "fingerprint" of the inclusion, showing that a target's shape and its conductivity are inextricably linked in the resulting data.

Localization and Signal Processing

To pinpoint the location of an inclusion, the authors employ the Multiple Signal Classification (MUSIC) algorithm, a technique borrowed from signal processing. By analyzing the data matrix generated from a transmitter-receiver array, the algorithm scans the search domain for the target's coordinates.

A key mathematical insight provided in the paper is that a single inclusion produces exactly three significant singular values in the data matrix. This provides a clear signature that the MUSIC algorithm can exploit for high-precision localization. Furthermore, the authors note a practical "imaginary advantage": because the background (unperturbed) field is real-valued while the perturbations are complex, focusing on the imaginary part of the data allows researchers to isolate the target more effectively from the background environment.

Noise Reduction and Detection Probability

Practical NDT is often hindered by environmental noise. To combat this, the paper proposes a Hadamard acquisition scheme. Instead of firing one source at a time, all sources fire simultaneously in specific positive/negative patterns. This increases the signal-to-noise ratio, a phenomenon known as the "Hadamard advantage."

To formalize the detection process, the authors apply Random Matrix Theory (specifically the Tracy-Widom distribution) to analyze the singular values of the data matrix. This allows for the mathematical prediction of the Probability of Detection (POD). Engineers can use this framework to set concrete detection thresholds based on a desired False Alarm Rate (FAR), knowing exactly how much noise a system can tolerate before a target of a certain size becomes undetectable.

Numerical Validation

The framework was validated using synthetic data from a simulated $16 \times 16$ array of coincident transmitters and receivers (256 total). The target—a small conductive sphere with a 1cm radius—was placed at the origin. Gaussian white noise was added to the synthetic signals to test the robustness of the localization and detection algorithms. The numerical results for the POD matched the theoretical predictions almost perfectly, confirming the reliability of the Random Matrix Theory approach.

Conclusion and Future Work

This research bridges the gap between theoretical electromagnetics and practical engineering. While the current mathematical proofs assume smooth boundaries for inclusions, the study provides a robust foundation for future work. Potential next steps include extending the model to handle multiple targets, non-smooth inclusions (like jagged cracks), and using multi-frequency tracking to better differentiate between a target's physical size and its conductivity. The authors suggest that these polarization tensors could eventually be used in "dictionary matching" for real-time target identification in industrial settings.

Summary updated

Abstract

The goal of this paper is to contribute to the field of nondestructive testing by eddy currents. We provide a mathematical analysis and a numerical framework for simulating the imaging of arbitrarily shaped small volume conductive inclusions from electromagnetic induction data. We derive, with proof, a small-volume expansion of the eddy current data measured away from the conductive inclusion. The formula involves two polarization tensors: one associated with the magnetic contrast and the second with the conductivity of the inclusion. Based on this new formula, we design a location search algorithm. We include in this paper a discussion on data sampling, noise reduction, and on probability of detection. We provide numerical examples that support our findings.
BibTeX
@misc{ammari2012target,
  author = {Habib Ammari and Junqing Chen and Chen, Zhiming and Josselin Garnier and Darko Volkov},
  title = {{Target Detection and Characterization from Electromagnetic Induction Data}},
  year = {2012},
  eprint = {1210.7915},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1210.7915},
}
Figure