Table 2 Comparison of the selected rail assets status evaluation studies

Li and He [60]

Prediction of the remaining useful life of train wheels and trucks

Data from three types of detectors (wheel impact load detector, machine vision systems, and optical geometry detectors)

Random forest

Niu et al. [61]

Anomaly detection of rail surface

Rail surface images

Adaptive pyramid graph and variation residuals

Shim et al. [62]

Anomaly detection of wheel flats

Wheel flat signals

Signal processing and deep learning

Li et al. [63]

Early fault detection of trucks and train wheels

Data from multiple types of detectors

Support vector machine

Sun et al. [64]

Fault diagnosis in railway track circuits

Short-circuit current signals

Multi-class support vector machine

Wan et al. [66]

Anomaly detection of train wheels

Vibration signal collected with a pair of fiber Bragg grating sensors

Unsupervised learning algorithms

This paper

Asset status evaluation

Automatic train supervision (ATS) logs, assets parameters, and maintenance data

One-class support vector machine