In the past 20 years, many studies have been performed on ballast layer inspection and condition evaluation with ground penetrating radar (GPR). GPR is a non-destructive means that can reflect the ballast layer condition (fouling, moisture) by analysing the received signal variation. Even though GPR detection/inspection for ballast layers has become mature, some challenges still need to be stressed and solved, e.g., GPR indicator (for reflecting fouling level) development, quantitative evaluation for ballast fouling levels under diverse field conditions, rapid GPR inspection, and combining analysis of GPR results with other data (e.g., track stiffness, rail acceleration, etc.). Therefore, this paper summarised earlier studies on GPR application for ballast layer condition evaluation. How the GPR was used in the earlier studies was classified and discussed. In addition, how to correlate GPR results with ballast fouling level was also examined. Based on the summary, future developments can be seen, which is helpful for supplementing standards of ballast layer evaluation and maintenance.

The rapid development of high-speed and heavy-haul railways caused rapid rail defects and sudden failure. This requires more advanced rail inspection, i.e., real-time accurate identification and evaluation for rail defects. However, existing applications cannot meet future demand. In this paper, different types of rail defects are introduced. Afterwards, methods that have the potential to achieve rapid accurate detection and evaluation of rail defects are summarized, including ultrasonic testing, electromagnetic testing, visual testing, and some integrated methods in the field. Finally, advice on rail inspection is given, such as synchronously utilizing the ultrasonic testing, magnetic flux leakage, and visual testing for multi-part detection. Specifically, synchronously using the magnetic flux leakage and visual testing technologies can detect and evaluate surface and subsurface defects, and UT is used to detect internal defects in the rail. This will obtain full rail information, to prevent sudden failure, then ensure train ride safety.

To simulate ballast performance accurately and efficiently, the input in discrete element models should be carefully selected, including the contact model and applied particle shape. To study the effects of the contact model and applied particle shape on the ballast performance (shear strength and deformation), the direct shear test (DST) model and the large-scale process simulation test (LPST) model were developed on the basis of two types of contact models, namely the rolling resistance linear (RRL) model and the linear contact (LC) model. Particle shapes are differentiated by clumps. A clump is a sphere assembly for one ballast particle. The results show that compared with the typical LC model, the RRL method is more efficient and realistic to predict shear strength results of ballast assemblies in DSTs. In addition, the RRL contact model can also provide accurate vertical and lateral ballast deformation under the cyclic loading in LPSTs.

Ground penetrating radar (GPR) has been applied for ballast layer inspection for two decades, mainly for the analysis of ballast layer fouling levels. However, some issues that affect the inspection quality remain unsolved, such as issues involving the GPR equipment quality (antenna) and the correlation between the GPR indicator and fouling index. With the aim of solving these two issues, in this paper, we investigated the difference between the results of two different antennas, the GPR data processing technique, indicators for the fouling level (by GPR signal processing) and the correlation between the indicators and fouling index (obtained by sieving). The results show that the antenna quality determines the inspection quality. The indicators can reflect the ballast layer fouling level, and they correlate the best with the fouling index (obtained by the percentage of particles passing through a 5 mm sieve size). This study is helpful for the future modification of railway ballast maintenance standards.

Vehicle-mounted ground-penetrating radar (GPR) has been used to non-destructively inspect and evaluate railway subgrade conditions. However, existing GPR data processing and interpretation methods mostly rely on time-consuming manual interpretation, and limited studies have applied machine learning methods. GPR data are complex, high-dimensional, and redundant, in particular with non-negligible noises, for which traditional machine learning methods are not effective when applied to GPR data processing and interpretation. To solve this problem, deep learning is more suitable to process large amounts of training data, as well as to perform better data interpretation. In this study, we proposed a novel deep learning method to process GPR data, the CRNN network, which combines convolutional neural networks (CNN) and recurrent neural networks (RNN). The CNN processes raw GPR waveform data from signal channels, and the RNN processes features from multiple channels. The results show that the CRNN network achieves a higher precision at 83.4%, with a recall of 77.3%. Compared to the traditional machine learning method, the CRNN is 5.2 times faster and has a smaller size of 2.6 MB (traditional machine learning method: 104.0 MB). Our research output has demonstrated that the developed deep learning method improves the efficiency and accuracy of railway subgrade condition evaluation.