Axle bearing faults are critical to the safety and reliability of high-speed trains (HSTs), making their early detection paramount. To address this, we have developed an enhanced fault detection approach that begins with rigorous testing on specialized benches. The methodology centers around a suite of machine learning classifiers designed to analyze key features extracted from vibration data collected on a test bench at the Complex Systems and Interactions Laboratory (CSI). The inputs to these classifiers include time, frequency, and time-frequency based statistical features. We then select the best classifier based on performance metrics such as accuracy. This approach enables precise identification and classification of bearing faults, providing a robust solution that can be tested and applied in high-speed railway environments with axle bearings.