Abstract
There exists several problems such as low detection efficiency and strong subjectivity in evaluation for damage of aircraft brake moving disc based on personal experiences. This paper proposes a new evaluation method based on the image segmentation. This study reviews the backgrounds and literature of related researches and analyzes the various categories and features of brake moving disc. The image data set of damage area has been semantically segmented base on U-Net model and the model for quantitative analysis of the rust area and the calculation model for maximum radial width in shedding area have also been constructed. Finally, the image data of brake moving disc on Cessna 525 has been used to validate these models. The results show that: (1) the image segmentation of different categories of brake moving discs based on U-Net model works well and the average accuracy, average recall rate, average pixel accuracy and average ratio of Intersection over Union are 91%, 91%, 90% and 87% respectively. (2) the evaluation result for damage of brake moving disc is consistent with the empirical judgement form mechanics, thus verifying that the proposed algorithm is reasonable and feasible. The research results provide a reference basis for mechanical engineers to scientifically and quantitatively determine the degree of damage to aircraft braking moving disc and replacement needs.
Introduction: The Challenge of Manual Inspection
In aviation maintenance, the "moving disc" of a brake assembly is a critical component responsible for converting kinetic energy into heat. These discs, often composed of carbon/carbon composites, are subject to extreme thermal and mechanical stress, leading to surface rust and material "shedding" (flaking).
Currently, determining the airworthiness of these discs relies heavily on the visual inspection and empirical judgment of maintenance mechanics. This manual process is inherently subjective, time-consuming, and prone to inconsistency. Missing a critical defect poses a significant safety risk, while overly conservative replacements lead to unnecessary maintenance costs. The industry requires a shift from qualitative visual checks to objective, quantitative measurements.
Methodology: Deep Learning and Geometric Modeling
To address these issues, researchers developed a two-stage automated evaluation pipeline. The first stage utilizes a U-Net semantic segmentation model to classify image pixels into four distinct categories: rust, shedding, the normal disc surface, and steel sheet components. The model was trained on a dataset of 330 images from Cessna 525 brake discs, utilizing Mosaic data augmentation to ensure robustness against varying lighting and orientations.
The second stage involves quantitative geometric modeling to translate segmented pixels into physical dimensions. The system calculates the total area ratio of damage and, more critically, the maximum radial width of shedding areas. Because aviation standards often set strict limits on material loss (e.g., 0.25 inches or 6.4mm), the authors employed Bezier curves and least squares fitting to model the disc as concentric ellipses. This allows the algorithm to precisely measure the depth of material loss relative to the disc's geometry.
Results and Quantitative Analysis
The U-Net model demonstrated high performance, achieving a Mean Intersection over Union (mIoU) of 87% and an average accuracy of 91%. By bridging the gap between pixel-level classification and physical measurement, the system provides a direct comparison against safety standards.
In practical validation, the system successfully identified discs exceeding safety thresholds. For example, one sample was flagged with a 28.05% shedding area and a 10.60mm radial width—well beyond the 6.4mm limit. Conversely, the system identified samples that were within tolerances but might have been prematurely replaced under subjective manual inspection, highlighting the potential for significant cost savings.
Conclusion and Future Directions
This research provides a scientific basis for quantitatively determining the damage degree of aircraft brake discs. While the current methodology is a successful proof-of-concept, future work will focus on developing an integrated intelligent prototype for hangar-floor use. Expanding the dataset to include more aircraft types and automating the image capture process through robotic or borescope integration will further enhance the system's utility in real-world aviation maintenance environments.
Summary updated