Defective Areas Identification in Aircraft Components by Bivariate EMD Analysis of Ultrasound Signals

Marco Leo, David Looney, Tiziana D’Orazio, Danilo P. MandicLecture Notes in Computer Science 2010 · DOI · OpenAlex

Abstract

This paper presents a novel signal processing framework for the non-destructive testing (NDT) of aircraft composite materials. By utilizing Bivariate Empirical Mode Decomposition (EMD) on complex-transformed ultrasound signals, the authors extract phase synchrony features that are significantly more sensitive to internal defects than traditional amplitude-based or wavelet-based methods. Experimental results on a honeycomb structure with artificial defects demonstrate that the proposed method identifies all defective regions, including deep-seated flaws that standard wavelet techniques frequently miss.

Introduction

Ensuring the structural integrity of aircraft is a critical safety challenge. Traditional NDT often relies on human operators to interpret ultrasound data, a process that is slow and prone to error, particularly when identifying internal cracks or delamination in complex composite materials. While Artificial Neural Networks (ANNs) are used for automated classification, their effectiveness depends heavily on the quality of the input features. Standard pre-processing methods like Wavelet Decomposition often struggle with the non-stationary and non-linear nature of ultrasound echoes, leading to false negatives in high-stakes environments.

Proposed Methodology

The researchers propose a multi-stage pipeline that shifts the analytical focus from signal amplitude to phase dynamics. The process involves converting real-valued ultrasound A-scans into a complex representation to capture localized phase information that is otherwise inaccessible.

Proposed Signal Processing Pipeline

As shown in the pipeline above, the method follows four key steps:

  1. Complex Transformation: Raw signals are converted via Discrete Fourier Transform (DFT).
  2. Bivariate EMD: The complex signal is adaptively decomposed into Intrinsic Mode Functions (IMFs). Unlike wavelets, which use a fixed basis, EMD is entirely data-driven, allowing it to better handle the "messy" signals found in composites.
  3. Phase Synchrony Analysis: The "temporal locking" between the real and imaginary components of the IMFs is measured. Internal defects disrupt these expected phase relationships, providing a highly sensitive indicator of material changes.
  4. Neural Classification: These synchrony features are fed into a three-layer Back Propagation neural network to classify each point as "Defective" or "Non-Defective."

Experimental Setup and Datasets

The study utilized a custom-built benchmark consisting of a honeycomb structure with a Nomex core and 48 plies of thickness, mimicking actual aircraft skin. To simulate real-world flaws, specific materials were inserted during manufacturing, including brass foil (to represent voids/delamination) and pressure-sensitive tape or dry peel ply (to represent inclusions).

Experimental Specimen Layout and Defect Types

The dataset comprised a spatial grid of 193x181 samples, with each sample containing 77 time-domain measurements, providing a comprehensive testbed for the proposed algorithm.

Key Results and Insights

The Bivariate EMD approach demonstrated significant superiority over traditional methods:

  • Detection Coverage: The EMD-based method successfully revealed all defective areas in the test sample.
  • Comparison with Wavelets: A standard wavelet-based approach failed to segment four major defective regions, missing approximately half of the defects located deeper within the material structure.
  • Sensitivity vs. Precision: While EMD was more sensitive at finding defects, the authors noted that wavelet-based approaches sometimes provided sharper geometric boundaries. However, in aviation maintenance, the priority is avoiding false negatives (missing a defect), making the EMD-based approach the preferred choice.

Conclusion

By treating ultrasound signals as complex-valued and applying adaptive Bivariate EMD, this research provides a robust method for identifying internal damage in aircraft components. The shift toward phase synchrony analysis allows for the detection of deep-seated flaws that traditional amplitude-based inspections might overlook, significantly enhancing the reliability of automated NDT systems.

Summary updated

BibTeX
@inproceedings{leo2010defective,
  author = {Marco Leo and David Looney and Tiziana D’Orazio and Danilo P. Mandic},
  title = {{Defective Areas Identification in Aircraft Components by Bivariate EMD Analysis of Ultrasound Signals}},
  booktitle = {Lecture Notes in Computer Science},
  year = {2010},
  doi = {10.1007/978-3-642-12159-3\_20},
}
Figure