Rail Surface Research Articles

Impact of vertical girder deformations on the high-speed train operation performance

The deformation of bridge structures will cause deterioration of geometry on the rail surface during the service period, which will threaten the normal operation of train. A train–track–bridge coupling dynamics model was established. In order to simulate real operational conditions, initial geometric irregularity was taken into account, which was combined with rail irregularity caused by different vertical deformation of girder and was superimposed as external excitation. The probabilistic and statistical characteristics of operational performance indexes were analysed. The sensitivity of operational performance indexes to different deformation modes was studied. The vertical deformation thresholds of girder based on probability guarantee rate were obtained. The results show that unloading rate of wheel and wheel/track vertical force are more sensitive to single vertical rotation deformation mode. Vertical acceleration of the train and sperling index are more sensitive to bilateral vertical rotation deformation mode. Unloading rate of wheel is more sensitive to single vertical translation deformation mode. In the process of enhancing daily operation, single vertical rotation and multiple vertical translation should receive significant attention.

A frequency-domain simulation method for the evolution of rail corrugation in curves

Rail corrugation remains a major problem in many railways and metros, leading to high noise and vibration levels as well as potential damage to the track and vehicle. A frequency domain simulation method is proposed for the evolution of the rail surface roughness, which can simulate the formation and development of rail corrugation in curves. A coupled vehicle-track dynamic model is established in the frequency domain to calculate the dynamic vertical and lateral wheel-rail contact forces and the associated vibration. The wheel-rail contact state is determined by using a multi-body dynamics model. The frictional-work wear model is used to calculate the roughness change caused by wheel-rail dynamic interaction. The method is verified by comparing the results with those of on-site testing on a 300 m radius curve of a metro line. It is shown that the main causes of rail corrugation are associated with the peaks in the vertical wheel-rail forces in the frequency domain, generated by the interaction between the wheel and rail. At short wavelengths the corrugation is associated with the wave reflections in the rail between adjacent axles. The phase relationship between the vertical wheel-rail force and the initial roughness has a strong influence on the roughness growth trend, leading to preferential growth in certain wavelength bands. The wheel-rail contact state and friction parameters affect the rate of development of the corrugation. Finally, a corrugation index is proposed to allow quick evaluation of the wavelength and growth rate of corrugation.

3D rolling contact finite element analysis of high-speed railway turnout considering ratchetting effect

The present work introduces a three-dimensional finite element model for the wheel-turnout-sleeper system, incorporating a cyclic constitutive model that accurately captures the nonlinear hardening behavior of rail steel under cyclic loadings. At the same time, a mixed implicit-explicit finite element method is employed to investigate the transient rolling contact behavior of wheel-turnout. Through dynamic analysis, the transient rolling contact status of the wheel-turnout, stress and strain responses of the turnout, and wheel-rail forces are investigated under varying crossing speeds and number of cycles. It is found that during the wheelset crossing, the contact stress remains relatively stable on the stock rail, and multi-point contact occurs at the wing rail and long point rail, and the contact stress on the surface of the long point rail is more significant. With an increase in crossing speed in the same cycle, there is a gradual rise in contact stress on the long point rail face. At the same crossing speed, with the increase in the number of cycles, the contact position between the wheelset and the long point rail gradually moves back, and the top width of the contact position gradually expands. Meanwhile, the fluctuation range of vertical displacement of the wheel center diminishes, and the wheel flange moves away from the long point rail. Additionally, a progressive accumulation of ratcheting strain is observed on the rail, and the ratcheting strain on the long point rail significantly exceeds that on the stock rail. The research results will help to provide a reference for the optimization and design of the critical cross-section profile of the long point rail and will have guiding significance for the maintenance and repair of turnouts.

Deep Learning (Fast R-CNN)-Based Evaluation of Rail Surface Defects

In current railway rails, trains are propelled by the rolling contact between iron wheels and iron rails, and the high frequency of train repetition on rails results in a significant load exertion on a very small area where the wheel and rail come into contact. Furthermore, a contact stress beyond the allowable stress of the rail may lead to cracks due to plastic deformation. The railway rail, which is the primary contact surface between the wheel and the rail, is prone to rolling contact fatigue cracks. Therefore, a thorough inspection and diagnosis of the condition of the cracks is necessary to prevent fracture. The Detailed Guideline on the Performance Evaluation of Track Facilities in South Korea specifies the detailed requirements for the methods and procedures for conducting track performance evaluations. However, diagnosing rail surface damage and determining the severity solely rely on visual inspection, which depends on the qualitative evaluation and subjective judgment of the inspector. Against this backdrop, rail surface defect detection was investigated using Fast R-CNN in this study. To test the feasibility of the model, we constructed a dataset of rail surface defect images. Through field investigation, 1300 images of rail surface defects were obtained. Aged rails collected from the field were processed, and 1300 images of internal defects were generated through SEM testing; therefore, a total of 1300 pieces of learning data were constructed. The detection results indicated that the mean average precision was 94.9%. The Fast R-CNN exhibited high efficiency in detecting rail surface defects, and it demonstrated a superior recognition performance compared with other algorithms.

ISRM: introspective self-supervised reconstruction model for rail surface defect detection and segmentation

The problems of intrinsic imbalance of the sample and interference from complex backgrounds limit the performance of existing deep learning methods when applied to the detection and segmentation of rail surface defects. To address these issues, an introspective self-supervised reconstruction model (ISRM) is proposed, which only requires normal samples in the training phase and incorporates the concept of self-supervised learning into an introspective autoencoder. The training framework of ISRM first extracts general features using a pretrained Feature Extractor. Subsequently, a Feature Transformer transfers the features to the target domain. Next, a synthetic defect embedder embeds Bessel-Gaussian random defects into the feature space. Finally, the asymmetric autoencoder reconstructs the rail surface features back into image space. The transformation of pretrained features into target-oriented features helps mitigate domain bias. Since defects exhibit higher commonality in the feature space relative to the image space, embedding synthetic defects into the feature space effectively improves training efficiency. Moreover, the adversarial training architecture enhances the clarity of reconstructed images. The impact of core parameters on the model performance is analyzed through ablation experiments. The results from comparative experiments demonstrate that ISRM achieves 98.5% and 97.2% accuracy on defect detection and segmentation tasks, respectively, reducing the error rate by 11.8% and 3.4% compared to the current state-of-the-art model.

Evaluations and enhancements of fatigue crack initiation criteria for steels subjected to large shear deformations

While large accumulated plastic deformations occur in the rail surface layer where rolling contact fatigue cracks initiate, many available Low Cycle Fatigue (LCF) crack initiation criteria focus on small plastic strains. Accordingly, this paper evaluates available fatigue crack initiation criteria for highly shear-deformed R260 steels, reflecting the conditions in the surface layer of rails. Furthermore, modified crack initiation criteria are suggested. The evaluation is based on three different experiments: Large shear strain increments under varying axial loading (predeformation), strain-controlled LCF tests after some predeformation, and axial High Cycle Fatigue (HCF) experiments. For the predeformation, Finite Element (FE) simulations, with a large-strain plasticity model for cyclic and distortional hardening, provide predictions of the local stress and strain histories. A cross-validation procedure is used to assess the accuracy and reliability of both established and modified fatigue crack initiation criteria. The proposed modifications to one of the criteria show an improved fit to the experimental data. However, there is a tendency to overfitting, which can be improved by including more experimental data.

Rail surface defect detection using a transformer-based network

The detection of Rail Surface Defects (RSDs) plays a critical role in railway track maintenance. Traditional image processing methods exhibit limitations due to their intricate design and insufficient robustness, thereby restricting their broader applications. Recently, deep learning-based RSD detection methods have drawn great attention. However, these methods predominantly rely on Convolutional Neural Networks (CNN), neglecting the hierarchical linkages amongst disparate features, which impedes the refined portrayal of RSDs. To address these issues, we propose RailFormer, a novel system leveraging the capabilities of Transformer-based networks for the precise and efficient detection of RSDs. The encoder in RailFormer incorporates overlapped patch merging, efficient self-attention, and a Mix-feed Forward Network (FFN), all meticulously designed to bolster feature fusion from both global and local perspectives. Additionally, we have implemented a Criss-Cross attention module within the decoder to facilitate RSD detection and manage computational complexity. In this study, the proposed RailFormer and four other models including SegFormer, Swin Transformer, ViT, and UNet are trained and compared. We employ the widely used public RSD datasets RSDD, encompassing both Type-I and Type-II RSDD images and a customized RSD dataset, as a basis for performance comparison. The training outcomes and visualization results show that RailFormer achieves the highest mean Intersection over Union (mIoU) and superior visualization performance on the RSDD and the customized RSD datasets. These results demonstrate the superiority of RailFormer and underline its potential for future deployment in railway track inspection applications.

IoT Based Railway Track Faults Detection and Localization using Acoustic Analysis and Signalling

Ensuring the safety and efficiency of rail transportation is paramount, necessitating timely inspection of railway tracks to prevent potential hazards. This research introduces an innovative approach by implementing an Internet of Things (IoT)-based autonomous railway track fault detection system. Various track components, such as rail surface defects, broken sleepers, missing fasteners, and irregular ballast levels, can lead to hazardous situations if not properly monitored. This master's thesis proposes an efficient method for classifying track components using image processing techniques and deep learning algorithms. Specifically, it focuses on detecting missing fasteners by analyzing images captured by a line camera. Experimental results demonstrate the method's effectiveness and robustness in complex environments. Future work aims to develop a comprehensive model capable of classifying all track components simultaneously. Image processing is utilized to extract individual components such as fasteners, rail, ballast, and sleepers, with the model trained to classify fastener states. This research contributes to advancing railway inspection methods, mitigating operational costs, preventing accidents, and enhancing passenger experience

Investigating the stick-slip vibration behavior of a locomotive with adhesion control in a curve

In order to explain the wheel-rail stick-slip vibration phenomenon of alocomotive in a curve, a co-simulation dynamic model taking into account adhesion control was established to reproduce the locomotive wheel-rail curve stick-slip vibration behavior, and the effect of parameters such as creep threshold, descent slope, rail surface condition and track curve radius on the stick-slip vibration behavior was measured by the traction force and the overall dispersion of adhesion coefficient. The results illustrate that the wheel-rail curve stick-slip vibration is caused by the dynamic traction force fluctuation under the adhesion control, and the increase of creep threshold will lead to the decrease of tractive force fluctuation amplitude and the decrease of stick-slip vibration intensity, which will increase the adhesion utilization rate, However, the increase of descent slope, the decrease of track curve radius and wheel-rail friction coefficient have the opposite effect on stick-slip vibration behavior. This phenomenon can be eliminated by improving the rail surface condition, expanding the track curve radius and lowering the locomotive traction force.

Rail surface spot irregularity effects in vehicle-track interaction simulations of train-track-bridge interaction

The presence of rail surface damage leads to increased vibrations on railway bridges as wheels cross the localized irregularities. These short wavelength irregularities appear frequently on the top of rail in railway tracks and, along with the increased axle loading, fatigue failure accelerates, and ultimately, serviceability decreases. Compromised bridge safety and reliability is countered by an increase in inspections as bridge structures continue to age and deteriorate. In recent years, the effects of rail surface spot irregularities on the condition of railway bridges have attracted the attention of the industry and the research community. This study has two objectives. Firstly, it introduces the methodology, algorithms, and models for Vehicle Track Interaction (VTI) simulations developed to study the effects of Rail Surface Spot Irregularities (RSSI) on system response. The proposed VTI simulator solves the wheel-rail interaction problem in a Staggered Time Marching (STM) form. Long- and short-wave surface irregularities are accounted for. A multi-point wheel-rail contact algorithm is developed and implemented for properly capturing the wheel contact points in view of the relative size of the wheel and the rail surface irregularities. The proposed method is verified analytically. Subsequently, this paper investigates and quantifies the effects of the shape and location of RSSI on the vibrations of the bridge structure and the train for different train speeds. It is shown that RSSI affects wheel-rail interactions forces significantly and the induced vibrations substantially stimulate bridge velocity and acceleration. The effects are more pronounced for specific RSSI locations and train speeds.

Possibility of Monitoring the Dynamic Impact of the Rolling Stock on the Track Superstructure Using Strain Gauges and Vibrometry

One of the priority tasks in development of railway transport is to raise its efficiency by increasing the time between repairs for both the track superstructure and the rolling stock. Solving this problem is impossible without timely, complete, and reliable information about the dynamics of interaction of the rolling stock with the infrastructure. To organise monitoring of dynamic processes, completely different types of primary transducers are used: resistance strain gauges, fibre-optic sensors, dynamometers (force sensor pads), accelerometers, acoustic emission sensors. This implies relevance of the scientific and technical problem of comparative testing of sensors of different types to assess the information content of their signals and justify the criteria for choosing primary transducers when solving specific monitoring problems.The objective of the study is to comparatively test removable resistive strain sensors, optical polarisation strain sensors and accelerometers under a passing train and to evaluate their information content to control rail depression and detect defects on the running surface of wagon wheels.The study using finite element modelling and a physical model of a rail as of a beam on an elastic foundation, substantiates the relationship between longitudinal deformations and vertical accelerations of the rail foot. Comparative tests of removable strain-resistive and optical polarisation strain sensors and accelerometers were held on an experimental section of track under a passing train. A signal processing algorithm has been developed and the equivalence of strain gauges and vibrometers for determining the depression of a rail under a passing train has been proven. Acomparison has been made of the pulse components of the signals of deformation and vibration acceleration during movement of a wheel with a defect on the running surface, and the requirements for the frequency characteristics of the sensors and their mounting on the rail surface have been substantiated.

Optimising digital signal processor‐based defect detection in smart manufacturing with lightweight convolutional neural networks

AbstractIndustrial defect detection is an important part of intelligent manufacturing, and Internet of things (IoT)‐based defect detection is receiving more and more attention. Although deep learning (DL) can help defect detection reduce the cost and improve the accuracy of traditional manual quality inspection, DL requires huge computational resources and is difficult to be simply deployed on IoT devices with limited computational power and memory resources. Digital signal processor (DSP) is an important IoT device with small size, high performance and low energy consumption, which has been widely used in intelligent manufacturing. In order to perform accurate defect detection on DSP, the authors proposed various optimisation strategies and then used a parallel scheme to scale the model to execute on multiple cores. The authors’ method evaluated it on Northeastern University Surface Defect Dataset, Magnetic Tile Defect Dataset, Rail Surface Defect Dataset and Silk Cylinder Defect Dataset, and the experimental results showed that the authors’ method obtains faster speeds without accuracy loss compared to running the same Convolutional Neural Networks model on a mainstream desktop CPU. This means that the authors’ method can realise efficient and accurate defect detection on IoT devices with limited computational power and memory resources, which opens up new possibilities for future development in the field of smart manufacturing.

Rough surface rolling contact fatigue crack stress intensity factor calculation for modern rail steels

For premium grades (and rail surface treatments such as ‘laser cladding’) there is little plastic damage expected during typical operating conditions, and smooth surface contact models predict a long life. However, damage still develops, including rolling contact fatigue, affecting premium steels from multiple suppliers. A model is described predicting stress intensity factors and indicative crack growth rates for cracks driven by surface roughness stresses. Accompanying field data quantify the asperity separation and tip radii for rails from metro, mixed traffic and a freight line in ground or unground conditions. The results predict that in rough surface contact small surface cracks (50–100 μm, exceeding microstructural dimensions) can grow faster than crack removal by wear, even when smooth surface models predict no growth. The model has application in identifying how rail grinding marks or wheel roughness may be an important factor in determining rail life through raised stresses at the rail surface (leading to wear), and increased stress intensities and growth rates for incipient cracks.

Analysis of rolling contact fatigue and wear characteristics of aging-treated laminar plasma strip quenching rails

The application of laminar plasma discrete quenching (LPQ) technology to enhance the wear resistance of railway tracks has been a notable advancement. Nonetheless, it has been observed that LPQ-treated rails experience a reduction in rolling contact fatigue (RCF) resistance. Our previous study found that aging treatment can improve the toughness of laminar plasma spot quenched rails. However, the microstructure as well as the mechanical properties of strip quenching are significantly different from those of spot quenching. To tackle this concern, the study delves into the wear and RCF damage mechanisms of aging-treated laminar plasma strip quenching (ALPQ) rails. The findings reveal that the aging process induces the precipitation of supersaturated solid solutions within the hardened layer, consequently elevating the RCF resistance of ALPQ rails. Following appropriate aging treatment, nearly complete elimination of fatigue cracks on the rail surface is achieved.

Study on the Formation Mechanism for High Rail Corrugation

This research employed a tracking investigation and numerical study to examine the initiation of high rail corrugation. Beginning with the trial operation of Metro Line 6 in Chengdu, China, we conducted a follow-up investigation for 2½ years that recorded the rail surface wear morphology of a 600-m radius curved track. The field observation results show that high rail corrugation occurred 12–17 months after the subway was officially put into operation, whereas the low rail of the same track did not show any corrugation. The distribution of high rail corrugation was discontinuous. Due to wear, the profiles of the rails and wheels changed. Two points of contact may occur between the leading outer wheel and high rail, with one contact point at the gauge corner and the other in the vicinity of the railhead (hereafter referred to as two-point contact). We developed a finite-element model that used the measured worn profiles to predict the self-excited frictional vibrations of a leading wheelset–rails system under the condition of two-point contact. We predicted the occurrence trend of self-excited frictional vibrations via the complex eigenvalue analysis. The simulation results show that in the condition of two-point contact, the self-excited frictional oscillation produced at an outer wheel–high rail interface can induce high rail corrugation. The use of straight or double S-shaped web plate wheels and an increase in railpad damping can help to reduce high rail corrugation.

Rail Surface Integrity Analysis in Laboratory

This study aims to assess the effect of grinding on the surface integrity of rail materials. Two grinding processes (preventive and corrective) were employed along with the corresponding usage scenarios throughout a series of tests. For the purposes of testing, samples in the form of discs were utilised to assess the surface integrity of rails. Testing initially involved introducing some wear on the discs followed by the appropriate grinding process and then further testing to monitor the disc’s performance. More specifically the preventive grinding was utilised in a scenario where rail discs endured normal usage prior to grinding and corrective grinding process was employed in rails discs that had been run for extensive cycles (2.5 times longer compared to normal usage) to generate significant wear before grinding. After the maintenance process all the specimens were further examined to assess their post-grinding performance. The experiments revealed that the aggressive single grinding pass methodology followed in a corrective maintenance approach could lead to undesirable results as the increased heat input, high surface roughness as well as the initial cracks from the prolonged pre-grinding running period could have an adverse effect on the RCF life. This was demonstrated by an almost triple wear rate compared to the “normal” usage scenario during the post-grinding testing.

Asymmetrical Contrastive Learning Network via Knowledge Distillation for No-Service Rail Surface Defect Detection.

Owing to extensive research on deep learning, significant progress has recently been made in trackless surface defect detection (SDD). Nevertheless, existing algorithms face two main challenges. First, while depth features contain rich spatial structure features, most models only accept red-green-blue (RGB) features as input, which severely constrains performance. Thus, this study proposes a dual-stream teacher model termed the asymmetrical contrastive learning network (ACLNet-T), which extracts both RGB and depth features to achieve high performance. Second, the introduction of the dual-stream model facilitates an exponential increase in the number of parameters. As a solution, we designed a single-stream student model (ACLNet-S) that extracted RGB features. We leveraged a contrastive distillation loss via knowledge distillation (KD) techniques to transfer rich multimodal features from the ACLNet-T to the ACLNet-S pixel by pixel and channel by channel. Furthermore, to compensate for the lack of contrastive distillation loss that focuses exclusively on local features, we employed multiscale graph mapping to establish long-range dependencies and transfer global features to the ACLNet-S through multiscale graph mapping distillation loss. Finally, an attentional distillation loss based on the adaptive attention decoder (AAD) was designed to further improve the performance of the ACLNet-S. Consequently, we obtained the ACLNet-S ∗ , which achieved performance similar to that of ACLNet-T, despite having a nearly eightfold parameter count gap. Through comprehensive experimentation using the industrial RGB-D dataset NEU RSDDS-AUG, the ACLNet-S ∗ (ACLNet-S with KD) was confirmed to outperform 16 state-of-the-art methods. Moreover, to showcase the generalization capacity of ACLNet-S ∗ , the proposed network was evaluated on three additional public datasets, and ACLNet-S ∗ achieved comparable results. The code is available at https://github.com/Yuride0404127/ACLNet-KD.

An Improved YOLOv8 Algorithm for Rail Surface Defect Detection

An Improved YOLOv8 Algorithm for Rail Surface Defect Detection

Multi-Scale Rail Surface Anomaly Detection Based on Weighted Multivariate Gaussian Distribution

Multi-Scale Rail Surface Anomaly Detection Based on Weighted Multivariate Gaussian Distribution

Research on Image Preprocessing Algorithm for Rail Surface Recognition

Research on Image Preprocessing Algorithm for Rail Surface Recognition