| Title |
Deep Learning-Enhanced Wear Analysis of Fe-Cr Composite Coatings for Automotive Brake Applications
under Euro 7 Requirements |
| Authors |
박태준(Tae-Jun Park) ; 이계원(Gye-Won Lee) ; 김종일(Jong-Il Kim) ; 남산(Sahn Nahm) ; 오윤석(Yoon-Suk Oh) |
| DOI |
https://doi.org/10.3365/KJMM.2025.63.12.997 |
| ISSN |
1738-8228(ISSN), 2288-8241(eISSN) |
| Keywords |
Fe-Cr composite coating; HVOF; Wear mechanism; Deep learning; Tribology; Wear pattern classification; Euro 7 compliance |
| Abstract |
The increasing transition to electric vehicles and new Euro 7 standards has intensified the focus
on brake-derived particulate emissions. Current braking systems exhibit significant susceptibility to corrosion
with an accelerated wear rate. This susceptibility highlights the need for materials that can maintain stable
friction coefficients under various operating conditions. Therefore, this study investigated the application of
Fe-Cr-based composite coatings on gray cast iron substrates for brake-disc applications, while also developing
an automated wear pattern recognition methodology. High-velocity oxygen fuel-sprayed Fe-Cr-based
composite coatings were evaluated via ball-on-disk tests against Si3N4 and WC counterfaces under
standardized conditions (load: 20 N, sliding distance: 1000 m, speed: 100 RPM). Microstructural analysis
revealed that the composite coatings achieved significantly enhanced mechanical properties (hardness: 6.52?
7.51 GPa) compared to the gray cast iron substrate (1.98 GPa). Tribological testing demonstrated superior
wear resistance, with Composite B exhibiting a specific wear rate of 7.37 × 10?6 m3 N?1 m?1, representing a
three-fold improvement over that of conventional materials. Energy-dispersive X-ray spectroscopy analysis
identified two distinct tribochemical interactions: SiO2 and WO2 formation with the Si3N4 and WC
counterfaces, respectively. A deep ensemble convolutional neural network, trained on 500 scanning electron
microscopy images, achieved a superior classification performance (training: 0.947, validation: 1.000, test:
0.960 accuracy) in automated wear pattern recognition. This integrated approach, which combines materials
science with machine learning, provides an effective methodology for both material development and
automated wear analysis in tribological applications. |