Defect Detection Algorithm for Wing Skin with Stiffener Based on Phased-Array Ultrasonic Imaging

Chuangui Wu, GuiLi Xu, Yimeng Shan, Xin Fan, Xiaohui Zhang, Yaxing Liu2023 · DOI

TL;DR

This paper introduces a novel phased-array ultrasonic imaging algorithm designed for the rapid and accurate detection of structural defects in aircraft wing skins, particularly in the stress-concentrated "R region" where stiffeners are present. By optimizing existing full-matrix imaging techniques for speed and noise reduction, the algorithm achieves a detection sensitivity of 1 mm and a resolution of 0.5 mm within a 0.5-second imaging frame, making it suitable for real-time aerospace inspection.

What Problem Does This Paper Address?

Aircraft wings are critical components, and their structural integrity is paramount for flight safety. The "R region" of wing skins, where they meet stiffeners, is a high-stress area prone to fatigue damage, leading to internal defects like cracks and voids. Traditional non-destructive testing (NDT) methods have limitations in terms of speed, resolution, or ability to characterize complex structures. While phased-array ultrasonic imaging offers advantages, existing algorithms often suffer from slow imaging speeds and susceptibility to noise artifacts, hindering their application in real-time engineering scenarios. This paper aims to bridge this gap by developing an improved phased-array ultrasonic imaging algorithm that is both fast and robust for detecting defects in these challenging wing structures.

Methods

The core of this work lies in enhancing existing phased-array ultrasonic imaging techniques. The authors start with the full-matrix-full-focusing (FMC-TFM) algorithm, known for its high imaging quality, but acknowledge its computational burden. To address this, they propose several key improvements:

  • Sparsity-Based Full-Focusing: A sparsity-based approach is introduced to reduce data redundancy and improve real-time performance. This involves optimizing data acquisition and processing by leveraging the reciprocity of sound fields and sparse sampling theory.
  • Adaptive Beamforming: To combat noise artifacts, an adaptive beamforming method is employed. This technique dynamically adjusts aperture weights based on echo signal characteristics, effectively suppressing clutter and off-axis signals while enhancing desired signals. Specifically, they investigate Minimum Variance Distortionless Response (MVDR) and Coherent-Factor (CF) beamforming.
  • Equivalent-Path-Length Pseudo-Echoes Dynamic Elimination: A novel dynamic pseudo-image removal scheme is developed to eliminate erroneous local images caused by non-unique time delays, particularly in near-surface regions. This method uses a bilateral root-mean-square calculation (B-RMSM) to distinguish true defects from artifacts.
  • GPU Parallel Acceleration: The entire processing pipeline is accelerated using a GPU parallel computing framework, significantly reducing imaging time.

Datasets

The paper primarily relies on simulated data for algorithm development and validation. The authors created a simulated environment using Matlab FieldII to model a linear phased-array probe and various defect scenarios. They simulated point defects, horizontal line defects, and inclined line defects with different depth distributions. For experimental validation, they used customized R-area aluminum alloy specimens designed to mimic real-world wing structures with simulated defects.

Key Insights and Ideas

  • The Trade-off Between Speed and Quality: The paper highlights the inherent challenge in ultrasonic imaging: achieving high resolution and low noise often comes at the cost of imaging speed. The proposed algorithm directly tackles this by optimizing for both.
  • Sparsity as a Key to Efficiency: By exploiting the inherent redundancies in phased-array data and employing sparse sampling strategies, significant computational gains can be achieved without sacrificing critical imaging information.
  • Adaptive Noise Suppression is Crucial: Real-world inspection environments are noisy. The adaptive beamforming techniques proved essential for isolating defect signals from background clutter and interference.
  • Pseudo-Echoes are a Significant Challenge: The development of a specific method to eliminate pseudo-echoes underscores the complexity of imaging near-surface structures with phased arrays, where signal reflections can create misleading images.
  • GPU Computing is Essential for Real-Time: The reliance on GPU acceleration demonstrates its indispensable role in making advanced imaging algorithms practical for real-time applications.

Key Results

The developed algorithm achieved the following key performance metrics:

  • Detection Sensitivity: 1 mm
  • Resolution: 0.5 mm
  • Imaging Time: Approximately 0.5 seconds per frame (within a 30 mm × 30 mm detection area).
  • Noise Artifact Suppression: Significantly reduced noise and artifacts, leading to clearer imaging of defects.
  • Defect Characterization: Accurately reconstructed and displayed defect situations in the R-region structure.
  • API Index: The FMC-TFM algorithm (as a baseline) had an API of 4.4898, while the optimized algorithm's performance, though not explicitly quantified with a single API value for the final combined method, demonstrated superior imaging quality and artifact reduction in simulations and experiments.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that while the algorithm significantly improves real-time performance, the complexity of defect detection in the R-area stringer wall panel and the need for rapid scanning still present challenges in obtaining characteristic information quickly and stably.
  • Ensuring accuracy and robustness against missed and false detections remains an ongoing technical problem.
  • The experimental validation was performed on customized specimens, and further testing on actual aircraft components would be beneficial.
  • Future Work:
    • The authors suggest that further improvements are needed to address the challenges of rapid and stable defect information acquisition, avoiding missed and false detections, and ensuring the overall accuracy and robustness of the detection results.
  • While the paper focuses on the R-region, extending the algorithm's applicability to other complex wing structures could be a future direction.

Summary updated

BibTeX
@misc{wu2023defect,
  author = {Chuangui Wu and GuiLi Xu and Yimeng Shan and Xin Fan and Xiaohui Zhang and Yaxing Liu},
  title = {{Defect Detection Algorithm for Wing Skin with Stiffener Based on Phased-Array Ultrasonic Imaging}},
  year = {2023},
  doi = {10.3390/s23135788},
}
Figure