Train Wheelset Research Articles

Defect Detection of Metro Wheel Set Tread Based on Image Recognition

In the operation of railway vehicles, the quality of bogies directly affects the operation quality and driving safety. Wheel set is one of the most important components in bogie, so the maintenance of wheel set is very important. For a long time, the detection of train wheel sets in China is still in the stage of manual measurement with backward technology and low efficiency. A new automatic detection method of wheel flange tread based on fuzzy neural network image processing technology is proposed in this paper. This method can accurately detect the defects of wheel flange tread. It collects the original image of the tested wheel set through the digital camera, inputs it into the computer, through certain calculation and processing, and compares it with the model established based on fuzzy neural network, so as to detect the defects of wheel flange and tread. First, the research status of wheel tread defect detection is summarized. Second, the basic principles of digital image technology are studied, the image processing models are confirmed, and the image processing method based on fuzzy neural network is established. Finally, eight wheel set treads are selected to carry out defect detection, and the analysis results show that the proposed method can obtain the better inspection precision.

A New Method of Wheelset Bearing Fault Diagnosis

During the movement of rail trains, trains are often subjected to harsh operating conditions such as variable speed and heavy loads. It is therefore vital to find a solution for the issue of rolling bearing malfunction diagnostics in such circumstances. This study proposes an adaptive technique for defect identification based on multipoint optimal minimum entropy deconvolution adjusted (MOMEDA) and Ramanujan subspace decomposition. MOMEDA optimally filters the signal and enhances the shock component corresponding to the defect, after which the signal is automatically decomposed into a sequence of signal components using Ramanujan subspace decomposition. The method’s benefit stems from the flawless integration of the two methods and the addition of the adaptable module. It addresses the issues that the conventional signal decomposition and subspace decomposition methods have with redundant parts and significant inaccuracies in fault feature extraction for the vibration signals under loud noise. Finally, it is evaluated through simulation and experimentation in comparison to the current widely used signal decomposition techniques. According to the findings of the envelope spectrum analysis, the novel technique can precisely extract the composite flaws that are present in the bearing, even when there is significant noise interference. Additionally, the signal-to-noise ratio (SNR) and fault defect index were introduced to quantitatively demonstrate the novel method’s denoising and potent fault extraction capabilities, respectively. The approach works well for identifying bearing faults in train wheelsets.

High-speed train wheel set bearing fault diagnosis and prognostics: evaluation of signal processing methods under multi-source interference

Despite the numerous studies on bearing fault diagnosis based on frequency domain or time-frequency domain analyses, there is a lack of a fair assessment on which method or methods are practically effective in identifying the fault frequencies of damaged bearings in noisy environments. Most methods were developed based on experiments with simple lab test rigs equipped with bearings having manufactured artificial defects, and the signal-to-noise ratio under lab conditions is too ideal to be useful for verifying the effectiveness of a signal processing method. The purpose of this study is to evaluate the effectiveness of advanced signal processing methods applied in a high-speed train operating environment with multi-source interference. In this work, the most advanced signal processing methods (including spectral kurtosis, deconvolution, and mode decomposition) are studied, and the shortcomings of each method are analyzed. Based on the characteristics of high-speed train wheel set bearings (HSTWSBs), the concept of fault characteristic signal-to-noise ratio (FCSNR) is put forward to quantitatively evaluate the fault periodicity intensity, and corresponding improved methods are proposed by combining the FCSNR with existing signal processing methods; all these methods consider the periodic characteristics and impact characteristics of the bearing fault. The simulation signal and actual signals of HSTWSB with natural defects help verify the effectiveness of the proposed methods. Finally, the advantages and disadvantages of the different signal processing methods are objectively evaluated, and the application scope of each method is analyzed and prospected. This study provides a reference and new ideas for the fault diagnosis of HSTWSB and other industrial bearings.

Drive-by deflection estimation method for simple support bridges based on track irregularities measured on a traveling train

Bridge deflection under train passage is a typical index of the bridge’s structural performance that can be measured from the ground with considerable effort and expense. This study develops a novel drive-by bridge deflection estimating method based on track irregularities measured on a moving train as a significantly efficient bridge deflection measuring method. First, theoretical analysis demonstrates that the difference in track irregularities observed by the first and last vehicles of a train is proportional to the bridge deflection that the train travels through when the dynamic response of the bridge is small and quasi-static response is predominant. Furthermore, it is stated that the proportionality coefficient is only determined by the train wheelset arrangement and the bridge span. Based on the different peaks of track irregularities observed at the first and last vehicle of a train, a unique method for estimating bridge deflection after canceling the track irregularity components other than bridge displacement has been devised. The accuracy verification by vehicle-bridge interaction simulation shows that the proposed method can be applied to track irregularities measured by a traveling train at a speed of <150 km/h with a standard deviation of <0.2 mm measurement error. Finally, the proposed method is applied to actual railway track irregularities as measured by the first and last vehicles of a train traveling at ∼ 100 km/h. The maximum deflections of five bridges estimated from the track irregularity are shown to be in good agreement with the on-site deflection measurement results, which were separately performed, with an error of <10%.

Influence of wheelset rotational motion on train collision response and wheelset lift mechanism

ABSTRACT When a train is in a nonbraking state during an actual collision, the wheelset rotational movement aggravates the wheel–rail impact and increases the wheel lift and risk of overriding. Based on the principle of equal initial collision kinetic energy, the idea of using wheelset translation instead of rotation is proposed. Moreover, the calculation formula for a mass equivalent coefficient is deduced. From the perspective of the dynamic transmission relationship between the car body, bogie, and wheelset, the mechanism of the wheelset lift are investigated. The research results show that without considering the rotational kinetic energy of wheelset, the difference of collision response can be effectively reduced by increasing the translational mass of wheelset by 10%. The vertical speed lift displacement and inclined lift displacement of the bogie have the largest contribution to the wheel lift. When the wheelset rotation is not considered, the wheel lift is significantly reduced.

High-speed train wheel set bearing fault diagnosis and prognostics: Research on acoustic emission detection mechanism

Acoustic emission (AE) technology is suitable for the condition monitoring and fault diagnosis of high-speed train wheel set bearings (HSTWSB) owing to its high frequency and high sensitivity. However, there has been no research on the mechanism of the relationship between HSTWSB state and AE signal. Therefore, this study explores the AE detection mechanism of HSTWSB state. First, based on the structure and load characteristics of HSTWSB, combined with the Hertz contact theory, a force analysis of the whole, local, and single roller is performed. Second, based on the assumption of random distribution of asperities, a contact model of the rough surface of the bearing is established, and the elastic deformation analysis of the asperities under load is performed based on the theory of elasticity. The corresponding relationship between the asperity deformation energy and root mean square (RMS) of the AE signal is obtained. Meanwhile, considering the influence of elastohydrodynamic lubrication (EHL) and signal attenuation on the deformation energy, an AE mathematical model of the HSTWSB is established. Accordingly, based on the characteristics of HSTWSB with different damage types, the influences of the load, deformation, speed, and defect morphology on the AE signals are thoroughly analyzed, and the improved models corresponding to different damage types are established, which extends the AE model to greater freedom. Finally, taking different damage types of the outer ring fault bearings as examples, the correctness of the proposed models is verified by comparing the theoretically calculated and experimental results. The proposed models can provide theoretical support for the application of AE to bearing condition monitoring.

Broad auto-encoder for machinery intelligent fault diagnosis with incremental fault samples and fault modes

Intelligent fault diagnosis (IFD) has been a widely concerned topic in the field of prognostics and health management. Existing machinery IFD approaches are generally developed based on the one-time learning manner. Therefore, they are powerless to deal with the data stream issue in which new fault samples and fault modes will be progressively collected for model training. To overcome this drawback, this paper proposes a broad auto-encoder (BAE) with incremental learning capabilities for on-line IFD of machinery. The BAE is constructed by stacking a series of auto-encoders in the width direction. Then, the output weight matrix of the BAE is calculated by the ridge regression algorithm. After that, the capabilities of sample-incremental learning and class-incremental learning are developed, so that the BAE can easily update itself to accommodate the new fault samples and fault modes without model retraining. With the two incremental learning capabilities, the BAE can be first trained using limited historical fault samples, and then incrementally learn new diagnosis knowledge from the newly coming fault samples and fault modes. In this way, the BAE will be more and more powerful over time. Finally, the proposed BAE is applied to diagnose faults for high-speed train wheelset bearings and disc components. The results show that the proposed BAE offers an efficient solution for machinery IFD to deal with the continuous data stream issue.

Running Safety of High-Speed Railway Train on Bridge During Earthquake Considering Uncertainty Parameters of Bridge

China’s railway network is wide, and some of them cross the seismic zone, and the ratio of high-speed railway (HSR) bridges is high. Therefore, the safety of trains on the bridge may be endangered in the event of an earthquake. Because the response of track–bridge system is sensitive to the randomness of bridge structural parameters during the earthquake, while the train wheelset is directly in contact with the track system, the running safety of train (RST) may be also sensitive to the randomness of structural parameters. In this paper, the model of train–bridge coupled system (TBCS) under earthquake was established, and the accuracy of the model was verified by test results. To efficiently calculate the safety performance of trains considering the randomness of structural parameters, the point estimation method (PEM) was used in this paper, and the applicability of PEM was proved by comparing with the calculation results of Monte Carlo simulation (MCS). Then, PEM was used to discuss the running safety performance of trains under different ground motion (GM) intensities, different train speeds, and different pier heights. Finally, based on the maximum probability, the GM intensity threshold of a bridge based on running safety is determined.

Yolov4 High-Speed Train Wheelset Tread Defect Detection System Based on Multiscale Feature Fusion

The Yolov4 detection algorithm does not sufficiently extract local semantic and location information. This study aims to solve this problem by proposing a Yolov4-based multiscale feature fusion detection system for high-speed train wheel tread defects. First, multiscale feature maps are obtained from a feature extraction backbone network. The proposed multiscale feature fusion network then fuses the underlying features of the original three scales. These fused features contain more defect semantic information and location details. Based on the fused features, a path aggregation network is used to fuse feature maps at different resolutions, with an improved loss function that speeds up the convergence of the network. Experimental results show that the proposed method is effective at detecting defects in the wheel treads of high-speed trains.

Multi-fault Classification of Train Wheelset System

Abstract The condition monitoring and faults diagnosis of wheelset are of great significance for the safe and stable operation of trains. Aiming at the axle crack, wheel flat, wheel out-of-roundness and their coupling faults in wheelset system, this paper proposes a method to diagnose these multiple faults simultaneously. Firstly, the dynamic model of a wheelset system with flat and out-of-roundness is established in SIMPACK software to analyze the fault characteristics of these two typical wheel damages. And the wheel flat and out-of-roundness are equivalent to displacement excitation and then loaded into the finite element (FE) model of the cracked wheelset. The vibration responses of the cracked wheelset with flat and out-of-roundness are solved in ABAQUS software. Then, the simulated multi-fault vibration data of wheelset are input into a Light-GBM model to realize the faults classification of axle crack, wheel flat, wheel out-of-roundness and their coupling faults. Finally, the wheelset vibration signals under the conditions of health, axle crack, wheel flat, and crack-flat coupling fault are collected from an experimental bench to verify the proposed method. This work can provide a basis for multi-fault classification and on-line health monitoring of trains.

High-speed train wheel set bearing fault diagnosis and prognostics: Fingerprint feature recognition method based on acoustic emission

Acoustic emission (AE) technology is suitable for the condition monitoring and fault diagnosis of high-speed train wheel set bearings owing to its high frequency and high sensitivity [1]. However, current AE diagnosis methods cannot consider both real-time characteristic and fault periodicity. To overcome these shortcomings, a fingerprint feature recognition method is proposed in this paper. First, the concept of dynamic threshold is proposed to ensure that the typical burst or hit-based AE signal can be accurately extracted under different speeds, loads, and damaged bearing states. Based on the dynamic threshold, a specific feature, namely the fingerprint feature, is defined to provide an instant visual pattern of the bearing fault. Second, a clustering significance index (CSI) is constructed, which can not only guide the intelligent selection of the dynamic threshold, but also help to realize the quantitative evaluation of the bearing damage state. Furthermore, this study combines hit statistics with the fault frequency to form a fault hit statistical spectrum. On this basis, a fault hit significance index (FHSI) is established for the quantitative judgment of the bearing damage state. Finally, the validity of the proposed methods was verified by testing under complex test conditions close to the actual line of a high-speed train, providing a valuable reference for the online monitoring of the bearing state under actual industrial conditions.

Offline and Online Measurement of the Geometries of Train Wheelsets: A Review

Train wheelset is one of the most important and costly-on-maintenance components in the rolling stock. Timely assessment of the wheelset wear can prevent the wheelset wear induced derailment, as well as enable the scheduling of optimal maintenance strategy, like when to perform reprofiling or when to replace, to maximize availability, minimize downtime, and minimize maintenance costs. The geometric parameters of wheelset that need to be constantly measured include wheel diameter and wheel profile. Various measurement methods, from the classical handheld gauges to machine vision-based sophisticated online noncontact vision systems, have emerged for the timely measurement of the geometrics of train wheelsets. Despite the significant development on train wheelset geometrics measurement methods in the past decades, a systematic review is not available yet. In this paper, we systematically review reported gauges and systems for measuring the geometrics of train wheelsets. Forty-six research articles and 13 granted patents published from 1990 to 2022 were reviewed. We provide a categorization of the measurement methods, and then detailly reported the gauges and systems based on offline and online categorization. At the end, the limitations of the state-of-the-art and the research prospects are given.

A Deformable Residual Attention Network for Defect Detection of Train Wheelset Tread

A Deformable Residual Attention Network for Defect Detection of Train Wheelset Tread

Self-supervised signal representation learning for machinery fault diagnosis under limited annotation data

Recently, convolutional neural networks (CNNs) have achieved remarkable success in machinery fault diagnosis. However, these methods usually require mass of manually labeled data, which is expensive and impractical. To this end, this paper explores the application of self-supervised learning (SS-Learning) paradigm in the field of machinery fault diagnosis, and proposes a new fault diagnosis framework based on self-supervised representation learning. This method can directly learn representative features that can be used for signal classification from unlabeled signals. In addition, it enables the network to have a deeper semantic understanding of vibration signals. In this way, the proposed method can significantly improve the performance of the diagnostic model in the case of limited labeled data. Furthermore, this paper deeply analyzes the mechanism behind the SS-Learning algorithm and the reasons for its excellent performance. The proposed SS-Learning algorithm is verified on three real fault diagnosis datasets high-speed train (HST) wheelset bearing dataset, CWRU dataset and motor bearing dataset). When there are only 50 labeled samples, the proposed SS-Learning algorithm achieves an accuracy of 85% on the motor dataset, which is 17.86% higher than the ordinary CNN. It is proven that the proposed method can provide a powerful supervision signal for feature learning of unlabeled samples and obtain quite competitive fault diagnosis performance with limited labeled samples.

Railway wheel tread damage and axle bending stress – Instrumented wheelset measurements and numerical simulations

ABSTRACT A combination of instrumented wheelset measurements and numerical simulations of axle bending stresses is used to investigate the consequences of evolving rolling contact fatigue (RCF) damage on a passenger train wheelset. In a field test campaign, stresses have been monitored using a wheelset with four strain gauges mounted on the axle, while the evolution of wheel tread damage (out-of-roundness) has been measured on regular occasions. The strain signals are post-processed in real time and stress variations are computed. Based on a convolution integral approach, the measured wheel out-of-roundness has been used as input to numerical simulations of vertical dynamic wheelset–track interaction and axle stresses. Simulated and measured axle stresses are compared for cases involving combinations of low or high levels of rail roughness and the measured levels of RCF damage. The study enhances the understanding of how wheel tread damage and track quality influence axle stress amplitudes.

Laser curve extraction of a train wheelset based on an encoder-decoder network.

An algorithm of laser curve segmentation for a train wheelset based on an encoder- decoder network is proposed. Aiming at the rich local features and simple semantic features of the train wheelset laser curve image, a neural network with shallow depth, high resolution, and good detail performance was designed. The proposed neural network makes full use of the dense connection mechanism and the upsampling module to enhance feature reuse and feature propagation. It can extract context semantic information at multiple scales with fewer parameters. Experimental results show that the encoder-decoder network has better performance than other neural networks in laser curve extraction of train wheelset. Based on the encoder-decoder neural network, mIOU, Recall, Accuracy, and F1_score of the laser curve dataset of the train wheelset, the score index reached 86.5%, 89.2%, 99.9%, and 85.0%, which can accurately extract the laser stripe of the train wheelset. Additionally, the encoder-decoder network can diminish the influence of noise on the extraction of laser fringes of a train wheelset to a certain extent. Therefore, it has good application in railway safety.

FEM based design of experiment for train wheelset diagnostics by laser ultrasonics

Laser-Ultrasonics Testing (LUT) is a Non-Destructive Technique (NDT) with good potential for application in the railway sector, nevertheless this technique is not yet used in practice because there are some practical difficulties to overcome. The possibility of measuring on a complete wheelset by bypassing the keying of the wheel will allow drastically reducing the inspection time but it has not yet been demonstrated. In fact, the attenuation of the signal in the path makes complex the interpretation of the generated waves. This paper aims at illustrating how the combination of simulated and experimental data allows to optimize the test setup for having output data of unambiguous interpretation. The main innovations presented in this paper are: (i) the possibility to work with low energy waves in the thermoelastic-ablative limit while maintaining satisfactory contrast levels for the purpose of defect detection and (ii) the implementation of a complete Finite Element Model (FEM) including the generation and propagation of waves in the solid domain and the propagation in air. This last step has not considered before in previous papers. The model allows to define the optimal experimental conditions to have a measured signal with an adequate signal-to-noise ratio (SNR > 6 dB) and to define an experimental procedure for defect detection reliable and comparable with current standards. This study lays the foundations for an innovative approach for train axle diagnostics which can be used during train extraordinary maintenance interventions. The laser ultrasonics system presented in this paper can be likely integrated in the pit lathe and exploited to monitor the railway wheels during their re-profiling phase, without having to remove them from the vehicle.

An investigation into the effects of modelling assumptions on sound power radiated froma high-speed train wheelset

In predicting rolling noise froma railway wheelset, some assumptionsare usually required to make the calculations less complicated and more efficient.In this study, the effects of some modelling assumptions on the prediction ofsound power radiated from a high-speed train wheelset are investigated by using the finite element and boundary element methods.Use is made of the axi-symmetry of the wheelset to allow the solution to be obtained in terms of a Fourier series in the circumferential direction. Compared with a moving-wheel formulation, the moving-roughness approach is shown to be sufficient provided that the wheelset receptance takes into account the effects of rotation and the rail receptance accounts for the effect of the moving load.Wheel/rail coupling should take account of lateral as well as vertical forces; neglect of the lateral interaction mayresult in significant overestimatesof the contribution of the axial modes of the wheelset, and thus the overall wheelset radiated noise.If the rotation of the wheelset is not considered, its radiated noise may be underestimated by up to 3 dBat frequencies above 2000 Hz.If the wheelset suspension is not included, the overall sound power would be overestimated at frequencies below 2000 Hz. Comparedwithsymmetric excitation of the wheelset, assuming incoherentexcitations for the left and right wheel/rail contacts will significantly affect the radiated noise in the frequencyrange below 2000Hz but has negligible effect on the important region above 2000 Hz. Finally, the contribution from the axle is shown to be significant below 1000 Hz.In terms of overall A-weighted level, the effect of the lateral wheel/rail interaction, the rotation of the wheelset and the inclusion of bearings and axleboxes should not be neglected.

Implementation of a novel algorithm of wheelset and axle box concurrent fault identification based on an efficient neural network with the attention mechanism

With the rapid development of urban rail transit in recent years, it becomes necessary to ensure the operation safety of train wheelset axle boxes. Aiming at the problems of large model size and long diagnosis time in traditional fault diagnosis methods, this paper proposed a novel model to identify concurrent faults in wheelset axle boxes based on an efficient neural network and the attention mechanism. The model was developed based on an improved Ghost bottleneck module sequence to achieve an efficient model. Furthermore, the use of the convolutional block attention module adaptively refined the feature map to improve the generalization of the model. Feature pyramid network was used to fuse the shallow and deep features in the network to improve the extraction ability of various size features. Experiments were carried out on wheelset axle box concurrent fault datasets. Model size, diagnosis speed and accuracy were used as evaluation indexes to compare the efficiencies of existing fault diagnosis methods and our proposed model. Experimental results revealed that the proposed algorithm effectively diagnosed wheelset axle box concurrent faults.

An approach for wheel flat detection of railway train wheels using envelope spectrum analysis

Due to the increasing demand for safer and faster rail transport, train wheelsets operating under high axle loads require more careful and reliable inspections and maintenance. During service, the train wheelsets are constantly operating under harsh conditions such as fatigue, thermal variation and impact. The wheel defects can induce damage to railway tracks or even derailments, increasing costs for both railway administrations and rolling stock operators. Therefore, early detection of wheel defects may prevent significant damages that could lead to service interruptions or derailments. The purpose of this research is to present an approach to detect the presence of wheel flats using an envelope spectrum analysis, as well as to test, discuss and analyse the sensitivity of the proposed approach to the unevenness of the track, the random position of the wheel flat impact occurrence and the severity of the flat. Subsequently, the sensitivity of the proposed method is tested to accurately detect a wheel flat when the signal is perturbed by different noise intensities. A wide range of 3D simulations based on a train–track interaction model has been performed for different train speeds and different types of flat geometries. From the obtained results, it is evident that envelope spectrum analysis is a capable tool and a cost-effective method that can be used to detect wheel flats along with the flat impact frequency for different train speeds in real-world conditions.