Tribology dataset: Fe-Cr Composite Coatings Wear Analysis SEM Images and Friction Data
Description
This dataset comprises 500 high-resolution SEM micrographs of Fe-Cr composite coatings after ball-on-disk wear tests, along with matched friction-coefficient time-series CSV files for each class (0–4). It supports automated wear-pattern classification using deep learning and quantifies tribological behavior under controlled conditions (20 N load, 100 RPM, 1000 m sliding distance) .
Files
Steps to reproduce
1. Tribological testing: Perform ball-on-disk wear tests per ISO 20808 (20 N, 100 RPM, 1000 m) in a 33 ± 2 °C, 50 ± 10 % RH environment. 2. Surface analysis: Capture SEM images (JEOL JSM-6390, secondary electron mode, 300× magnification) and record friction coefficients at 1 kHz sampling. 3. Data preprocessing: - Split images into train/validation/test (64 %/16 %/20 %), stratified by class. - Convert to 100×100 px PyTorch tensors, normalize (ImageNet stats), grayscale, apply random sharpness (×2) and autocontrast. 4. Model pipeline: Use the provided PyTorch code to extract friction-time features, train an ensemble of five VGG-style CNNs (seed 2020), and evaluate via majority-voting ensemble prediction.