Out-of-Roundness Wheel Damage Identification in Railway Vehicles Using AutoEncoder Models

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(5), P. 2662 - 2662

Published: March 1, 2025

This study presents a comparative analysis of three AutoEncoder (AE) models—Variational (VAE), Sparse (SAE), and Convolutional (CAE)—to detect quantify structural anomalies in railway vehicle wheels, such as polygonization. Vertical acceleration data from virtual wayside monitoring system serve input for training the AE models, which are coupled with Hotelling’s T2 Control Charts to differentiate normal abnormal component behaviors. The results indicate that SAE-T2 model outperforms its counterparts, achieving 16.67% higher accuracy than CAE-T2 identifying distinct conditions, although 35.78% computational cost. Conversely, VAE-T2 is outperformed 100% analyzed scenarios when compared conditions while also exhibiting 21.97% average Across all scenarios, methodology consistently provided better classifications wheel damage, showing capability extract relevant features dynamic signals Structural Health Monitoring (SHM) applications. These findings highlight SAE’s potential an interesting tool predictive maintenance, offering improved efficiency safety operations.

Language: Английский

Fabrication of high-aspect-ratio 5 μm ultra-small pitch indium bump arrays by vacuum thermal deposition under variable rate

Infrared Physics & Technology, Journal Year: 2025, Volume and Issue: unknown, P. 105869 - 105869

Published: April 1, 2025

Language: Английский