Train Noise Research Articles

Improving knowledge of the acoustic factors involved in railway noise annoyance: first results of a pilot field survey of sixty-two local residents

Railway transportation contributes to the objectives of decarbonization but also generates negative externalities, including noise. Energy noise indicators used to characterize population exposure do not adequately reflect the repetitive nature of railway noise peaks. The GENIFER pilot study aims to test a protocol designed to characterize railway noise events according to the instantaneous perceived annoyance when the train is passing, in order to improve understanding of the influence of acoustic factors on annoyance. The first phase of the survey was carried out in 2023 among 62 residents of a pilot site. An electronic device was used to collect around 5,000 ratings, ranging from 1 to 10, assessing the instantaneous annoyance induced by train noise at passing trains. The site instrumentation included sixteen sound level meters and two video recording systems, enabling annoyance ratings to be associated with the acoustic characteristics of railway noise events. A questionnaire aimed at identifying co-determinants of long-term annoyance was also administered to participants. Feedback on the field implementation of this survey and initial results concerning acoustic measurements, instantaneous annoyance ratings and questionnaire responses will be presented.

A framework in three different project stages to predict ground-borne noise of trains in railway tunnels

Trains passing through tunnels cause noise and vibrations that negatively affect nearby residents. Developing a model to effectively predict ground-borne noise in nearby buildings faces challenges in the early project phases due to limited available data. This study suggests a framework for three stages based on precision and available information: location stage, planning stage, and construction stage. The first two stages correspond to determining the location and designing the railway track. The third stage involves the construction of the railway tunnel where more detailed information may be acquired by measurements on site. The prediction model presented here is formulated as a source term and correction terms considering train type, track type, track treatment, train speed, distance attenuation, foundation coupling, and floor-to-floor attenuation. Moreover, instead of using safety factors, which may cause unnecessary increase in project costs, the concept of combined uncertainty is applied here, using the estimated standard deviation of each term and the root sum of squares. Consequently, a comprehensive ground-borne noise prediction model adapted to various stages and handling uncertainties is proposed.

Aircraft automatic detection on Station7hz interface using signal processing and artificial intelligence - ACOUDET - First method

The global project is to design a automatic noise event detection and classification system in real time using signal processing and artificial intelligence (AI). The first phase focuses on transport noise (aircrafts, trains, road vehicles, etc.). Secondly, the plan is to extend detection to construction site noise and any natural or life sources. This paper presents the first method developed to detect audible aircrafts pass-by. It focused on the development of a prototype detection algorithm, its results in situ conditions, and insights on future developments to create a robust aircraft and train noise detection system. The data collection method, aircraft audio signal particularities and a first detection method are described. The robustness of the detections is evaluated and limitation induced by real time detection are discussed. Thanks to the proprietary monitoring system Station-7hz. it has been possible to detect aircrafts in different types of sound environments in a straightforward manner, showing better results at night due to a quieter environment. Future developments based on a hybrid algorithm include AI models for audibility assessment and unwanted event filtering during aircraft pass-by. Aircraft's presence is detected via radar data and the detection of maximum acoustic levels correlated with aircraft parameters.

Analytical modelling techniques for efficient broadband noise prediction of high-speed trains

Analytical modelling techniques are presented for broadband noise prediction of high-speed trains. For interior noise problems, train body structures are modelled exactly by the dynamic stiffness method (DSM); whereas the interior acoustic cavities of the train body are modelled by the SDSM where the boundary conditions are described by modified Fourier series with a rapid convergence rate. Finally, the vibro-acoustic coupling model of the train body structures is constructed. For exterior noise problems, exterior train noise models are first formed by experimental data considering the Doppler effect; then both the DSM and theoretical attenuation formulas are used to build noise radiation models for different external environment conditions. The performance of the method is evaluated by predicting the interior sound field distribution of typical cross sections of the train body and the external radiated noise in a tunnel, or in an environment where the acoustic barrier or buildings along railways exist. Numerical results are in good agreement with the those obtained by the finite element method. Thus, the proposed methods can achieve broadband noise prediction highly accurately and efficiently, which can also serve as the benchmark for predicting the radiated noise generated by high‐speed trains.

Enhanced railway sound event detection using YAMNet and classification criteria

The segmentation-by-classification method has become a popular way to detect sound events. It uses neural networks, trained to detect sound sources on short signals (one tenth of a second to a second), or long-term signals divided into a succession of short segments. In this paper, we focus on detecting train pass-bys from long term signals, where it is assumed that railway noise is higher than ambient noise. Using the neural network YAMNet, we show that a processing stage is necessary to improve the segmentation of such events and reduce the high false positive rate. The applied criteria are related to the nature of a train pass-by, lasting several seconds and with broad frequency band. Around 90% of events are detected with proper boundaries. However, we observe that YAMNet is not designed to distinguish railway vehicles from other vehicles. False positive rate remains high whenever other vehicles travel near the measurement location. Additional AI models are proposed to perform this specific distinction. Their efficiency depends on train type, as recent passenger train noise resembles roadway noise. Otherwise, false positive rate decreases below 10%, and the railway noise contribution estimate aligns with the reference level, within a 0.5 dB(A) margin on average.

Robustness enhancement in neural networks with alpha-stable training noise

With the increasing use of deep learning on data collected by non-perfect sensors and in non-perfect environments, the robustness of deep learning systems has become an important issue. A common approach for obtaining robustness to noise has been to train deep learning systems with data augmented with Gaussian noise. In this work, the common choice of Gaussian noise is challenged and the possibility of stronger robustness for non-Gaussian impulsive noise is explored, specifically alpha-stable noise. Justified by the Generalized Central Limit Theorem and evidenced by observations in various application areas, alpha-stable noise is widely present in nature. By comparing the testing accuracy of models trained with Gaussian noise and alpha-stable noise on data corrupted by different noise, it is found that training with alpha-stable noise is more effective than Gaussian noise, especially when the dataset is corrupted by impulsive noise, thus improving the robustness of the model. Moreover, in the testing on the common corruption benchmark dataset, training with alpha-stable noise also achieves promising results, improving the robustness of the model to other corruption types and demonstrating comparable performance with other state-of-the-art data augmentation methods. Consequently, a novel data augmentation method is proposed that replaces Gaussian noise, which is typically added to the training data, with alpha-stable noise.

Numerical and experimental study on the aerodynamic noise characteristics of 600-km/h high-speed maglev trains

Abstract High-speed maglev trains represent a new mode of transportation; however, their 600-km/h traveling speed causes serious aerodynamic noise problems, with very little research on the aerodynamic noise of such trains. To clarify the aerodynamic noise source and radiation characteristics of high-speed maglev trains, both large eddy simulations and wind tunnel experiments with a 1:8 scale three-train model were conducted to obtain the aerodynamic noise source spectrum and spatial distribution characteristics of the maglev train at different speeds; the contribution of the dipole noise source was also analyzed. The results revealed that the train shape has a significant impact on the frequency distribution of the far-field aerodynamic noise below 1 kHz, with the highest sound power concentrated on the train head, especially in the areas around the track joint and streamlined sections. The surface dipole sound source energy increases linearly with speed and is mainly distributed at the bottom of the train head and tail train body. A short-streamlined train exhibited the highest ratio, especially in the trailing edge area, up to 1.42×1011 dB. The far-field sound pressure level of the short-streamlined maglev train is significantly higher, with a maximum difference of about 10 dB(A). The distribution pattern of far-field sound pressure level is influenced by the train type, reflected in the tail train and wake area. Different frequency intervals have different relationships between the acoustic pressure level and frequency; long-streamlined models may enhance broadband features in the sound field distribution of high-speed maglev trains. These findings could be used to guide the aero-acoustic design of maglev trains with a designed speed of 600 km/h.

Euclidean-Distance-Preserved Feature Reduction for efficient person re-identification

Person Re-identification (Re-ID) aims to match person images across non-overlapping cameras. The existing approaches formulate this task as fine-grained representation learning with deep neural networks, which involves extracting image features using a deep convolutional network, followed by mapping the features into a discriminative space through another smaller network, in order to make full use of all possible cues. However, recent Re-ID methods that strive to capture every cue and make the space more discriminative have resulted in longer features, ranging from 1024 to 14336, leading to higher time (distance computation) and space (feature storage) complexities. There are two potential solutions: reduction-after-training methods (such as Principal Component Analysis and Linear Discriminant Analysis) and reduction-during-training methods (such as 1 × 1 Convolution). The former utilizes a statistical approach aiming for a global optimum but lacking end-to-end optimization of large data and deep neural networks. The latter lacks theoretical guarantees and may be vulnerable to training noise such as dataset noise or initialization seed.To address these limitations, we propose a method called Euclidean-Distance-Preserving Feature Reduction (EDPFR) that combines the strengths of both reduction-after-training and reduction-during-training methods. EDPFR first formulates the feature reduction process as a matrix decomposition and derives a condition to preserve the Euclidean distance between features, thus ensuring accuracy in theory. Furthermore, the method integrates the matrix decomposition process into a deep neural network to enable end-to-end optimization and batch training, while maintaining the theoretical guarantee. The result of the EDPFR is a reduction of the feature dimensions from fa and fb to fa′ and fb′, while preserving their Euclidean distance, i.e.L2(fa,fb)=L2(fa′,fb′). In addition to its Euclidean-Distance-Preserving capability, EDPFR also features a novel feature-level distillation loss. One of the main challenges in knowledge distillation is dimension mismatch. While previous distillation losses, usually project the mismatched features to matched class-level, spatial-level, or similarity-level spaces, this can result in a loss of information and decrease the flexibility and efficiency of distillation. Our proposed feature-level distillation leverages the benefits of the Euclidean-Distance-Preserving property and performs distillation directly in the feature space, resulting in a more flexible and efficient approach. Extensive on three Re-ID datasets, Market-1501, DukeMTMC-reID and MSMT demonstrate the effectiveness of our proposed Euclidean-Distance-Preserving Feature Reduction.

Symmetry Breaking in Geometric Quantum Machine Learning in the Presence of Noise

Geometric quantum machine learning based on equivariant quantum neural networks (EQNNs) recently appeared as a promising direction in quantum machine learning. Despite encouraging progress, studies are still limited to theory, and the role of hardware noise in EQNN training has never been explored. This work studies the behavior of EQNN models in the presence of noise. We show that certain EQNN models can preserve equivariance under Pauli channels, while this is not possible under the amplitude damping channel. We claim that the symmetry breaks linearly in the number of layers and noise strength. We support our claims with numerical data from simulations as well as hardware up to 64 qubits. Furthermore, we provide strategies to enhance the symmetry protection of EQNN models in the presence of noise. Published by the American Physical Society 2024

Comparison of train noise and vibration measurements with U.S. Department of Transportation model predictions

Wayside noise and ground-borne vibration measurements of freight and light rail train operations were carried out in Denver, CO in support of the Burnham Yard Transportation Planning Study being pursued by the Colorado Department of Transportation (CDOT). This study was intended to evaluate various track layout alternatives for the Consolidated Main Line (CML) freight railroad alignment and enhancement of Regional Transportation District (RTD) light rail operations while maintaining options for Front Range Passenger Rail (FRPR) within Burnham Yard and the surrounding areas. The objective of the measurements was to compare measured train noise and vibration levels with predictions using models sanctioned by the U.S. Department of Transportation (USDOT) in order to validate the use of these models to assess the noise and vibration impacts of the project alternatives. The models included the Federal Railroad Administration (FRA) Chicago Rail Efficiency And Transportation Efficiency (CREATE) freight noise model, the FRA Horn Noise Model, and the models included in the Federal Transit Administration (FTA) Transit Noise and Vibration Impact Assessment Manual. Based on a comparison of measured and modeled results, it was concluded that the USDOT models could be used to assess noise and vibration from train operations in the Burnham Yard Study Area.

A bi-fidelity DeepONet approach for modeling hysteretic systems under uncertainty

This study proposes a multi-fidelity paradigm for developing surrogates of degrading hysteretic systems under uncertainty through the use of deep operator networks (DeepONets). Instead of attempting to directly train a DeepONet on the original response, this study adopts a residual modeling approach wherein the DeepONet is trained on the discrepancy between the original (high-fidelity) data source and a relatively simpler (low-fidelity) representation of the system. Within these examples, a conventional Bouc-Wen model is treated as a “low-fidelity” representation given that it is free of any further assumptions about the nonlinear behavior, while the “high-fidelity” data is generated from different structures with various forms of complex hysteretic behavior. The results of this study show that the proposed multi-fidelity approach consistently outperforms standard surrogates trained on only the original datasets considering a variety of systems with unknown parameters. The results also show that the difference in performance grows as training data becomes more scarce, a critical consideration for many real-world engineering systems, and that the proposed multi-fidelity approach maintains its performance edge even when controlling for training time and noise in the training data.

Interior noise prediction of inter-coach space of high-speed maglev trains based on wavenumber decomposition on aerodynamic excitation

Current research on noise and vibration control of high-speed maglev trains pays more attention to far-field noise, while the level of interior noise has a direct impact on the ride comfort and should be placed equal weight on. In this paper, inter-coach space, one of the main pressure fluctuation sources of a specific type of high-speed maglev train with a design speed of 600 km·h−1, is taken as the research object. The turbulent and acoustic components of wall pressure fluctuations (WPF) are separated based on a wavenumber-frequency analysis approach, and then each component is applied as different forms of source input to the vibroacoustic model, namely, finite element method-boundary element method (FEM-BEM) and statistical energy analysis (SEA) for low- and high-frequency ranges respectively, to investigate the contribution of both components to interior acoustic cavity in all frequency range. It can be seen quantitatively from the results that the amplitude of turbulent component is generally much higher than that of the acoustic one, but it can be vice versa when it comes to the interior response. The conclusion drawn in this paper are able to provide guidance for future researches on more targeted interior noise control of high-speed maglev trains.

Prediction of Noise Level in Ota Metropolis Using Artificial Neural Network

Capturing noise data is laborious, time-consuming, expensive and dangerous due to the exposure of the investigator to the menace. Also, appropriate software, computational skill and computational time are also required before the captured data could be of any use. In this work, an artificial neural network (ANN) was deployed to learn and train noise data in Ota Metropolis. Data were captured from forty-one (41) locations for the morning, afternoon and evening in Ota Metropolis. ANN with Levenberg Marquardt algorithm and architectural configuration of 2-21-9 (input-hidden neuron-output) was used to predict noise descriptors for Ota Metropolis with 73% accuracy. The two input variables were the latitude and longitude of the location measured in degrees while the nine output variables are the noise descriptors such traffic noise index (TNI), noise pollution level (LNP), and average equivalent noise levels (LAeq) computed for each selected location for morning, afternoon and evening periods. The results could be used in mobile applications, Google Earth and other platforms to guide residence dwellers, travellers, industrialists and technocrats in selecting travelling routes, choice of apartment location and use of appropriate personal protective equipment (PPE) in unavoidable noisy locations.

Reconstruction of elasticity modulus distribution base on semi-supervised neural network

Accurate reconstruction of tissue elasticity modulus distribution has always been an important challenge in ultrasound elastography. Considering that existing deep learning-based supervised reconstruction methods only use simulated displacement data with random noise in training, which cannot fully provide the complexity and diversity brought by in-vivo ultrasound data, this study introduces the use of displacement data obtained by tracking in-vivo ultrasound radio frequency signals (i.e., real displacement data) during training, employing a semi-supervised approach to enhance the prediction accuracy of the model. Experimental results indicate that in phantom experiments, the semi-supervised model augmented with real displacement data provides more accurate predictions, with mean absolute errors and mean relative errors both around 3%, while the corresponding data for the fully supervised model are around 5%. When processing real displacement data, the area of prediction error of semi-supervised model was less than that of fully supervised model. The findings of this study confirm the effectiveness and practicality of the proposed approach, providing new insights for the application of deep learning methods in the reconstruction of elastic distribution from in-vivo ultrasound data.

Effect of Ward Noise Management Combined with Meditation Training on Stroke Rehabilitation Patients.

To study the value of ward noise management combined with meditation training in stroke rehabilitation patients. According to the retrospective analysis method, 150 stroke patients hospitalized in the rehabilitation center of a Tangshan Workers' Hospital from July 2020 to December 2023 were selected as study objects. They were divided into three groups, namely the control group (routine rehabilitation care, n = 50), observation group A (meditation training, n = 50), and observation group B (meditation training and ward noise management, n = 50) according to whether they received ward noise management and meditation training. The general demographic data, Fatigue Severity Scale (FSS), Pittsburgh Sleep Quality Index (PSQI), and the Short Form 36 (SF-36) were collected. Chi-square test and analysis of variance were used to analyse the data. The baseline data of the patients in each group were not statistically significant (P > 0.05). Before treatment, no difference in the FSS, PSQI, SF-36 scores and environmental noise level between the groups (P > 0.05) was observed. After management, the scores of SF-36 in observation group B were higher than those in the control group and observation group A (P < 0.05) except for somatic pain. Other indicators in observation group B were lower than those in the control and observation group A (P < 0.001). Ward noise management and meditation training can effectively reduce patients' fatigue, significantly reducing ambient noise levels, promoting the improvement of life quality, and improving sleep quality.

Frictional vibration analysis of train braking system considering wheel-rail attachment and multi-body friction

Train vibration and noise have always been problems that affect the service life of relevant components and passenger comfort. This paper focuses on the vibration of the train braking system and establishes a torsional vibration model of the train braking system considering wheel-rail attachment and multi-body friction, which includes three rotational degrees of freedom of the brake disc, brake block, and wheelset. The periodic vibration and chaotic bifurcation of the braking system under low-speed working conditions are studied by numerical calculation. An optimization strategy is proposed for the chaotic bifurcation behavior, which eliminates the chaotic interval at low speeds through the reasonable matching of particle parameters, so as to ameliorate the brake groan problem of the train.

Mitigating the Impact of False Negative in Dense Retrieval with Contrastive Confidence Regularization

In open-domain Question Answering (QA), dense text retrieval is crucial for finding relevant passages to generate answers. Typically, contrastive learning is used to train a retrieval model, which maps passages and queries to the same semantic space, making similar ones closer and dissimilar ones further apart. However, training such a system is challenging due to the false negative problem, where relevant passages may be missed during data annotation. Hard negative sampling, commonly used to improve contrastive learning, can introduce more noise in training. This is because hard negatives are those close to a given query, and thus more likely to be false negatives. To address this, we propose a novel contrastive confidence regularizer for Noise Contrastive Estimation (NCE) loss, a commonly used contrastive loss. Our analysis shows that the regularizer helps make the dense retrieval model more robust against false negatives with a theoretical guarantee. Additionally, we propose a model-agnostic method to filter out noisy negative passages in the dataset, improving any downstream dense retrieval models. Through experiments on three datasets, we demonstrate that our method achieves better retrieval performance in comparison to existing state-of-the-art dense retrieval systems.

FFLDGA-Net: Image retrieval method based on Feature Fusion Learnable Descriptor Graph Attention Network

Image retrieval aims to retrieve and return the image in the database that is most similar to the query image. However, the performance of image retrieval models is often hindered by the limited dimensionality of images, which lacks depth information about objects. To address this issue, we propose a novel image retrieval model called FFLDGA-Net (Feature Fusion-based Learnable Descriptor Graph Attention Network). This model aims to overcome the absence of depth information in images by fusing feature information from both image data and point cloud data. First, we introduce the LDGA-Net, which effectively improved the model’s ability to mine hard samples and negative samples. Then, we combine the multi-scale route convolution module with a one-dimensional path aggregation network to fuse point clouds and image features at multiple scales, and establish a high-dimensional relationship between features and low-dimensional features. To mitigate training noise, we incorporate a soft label strategy tailored to the dataset’s characteristics. Our experimental results on two benchmark datasets demonstrate the significant improvements achieved by FFLDGA-Net in image retrieval performance.

Low-frequency cabin noise of rapid transit trains

Abstract Rapid transit or mass rapid transit (MRT) is a high-capacity public transport designed to carry a large number of passengers, especially during the peak hours. They are becoming very popular in major cities and some deem the presence of the rapid transit system in a city as a symbol of modern development and essential feature of urban life. As the rapid transit system expands, the traveling time on a rapid transit train may increase due to longer journey and cabin noise has become an environmental concern for the passengers. In the present study, we would attempt to do a more detailed study of the effect of viaduct height, in particular viaducts of different heights on the cabin noise of various rapid transit systems. The present study examined and benchmarked the cabin noise in terms of both dB(A) and dB(C) for four different rapid transit systems, namely part of the East-West line including the Tuas-West extension on elevated tracks with very high viaduct of the Singapore MRT System; part of Paris Line 2 from Anvers to Belleville station including a stretch of elevated track on viaduct; part of the Piccadilly line of London from Heathrow Airport to Green Park station with a stretch on surface ground; and finally part of Chongqing Line 3 from Gongmao to Lianglukou station across the Yangtze river. It was found that the cabin noise would be dominated by low-frequency content and would be better reflected if the measurements were presented in dB(C), especially for trains running on elevated tracks of greater height.

Fault Diagnosis of Fuel Cells by a Hybrid Deep Learning Network Fusing Characteristic Impedance

Establishing an accurate fault diagnosis system for the proton exchange membrane (PEM) fuel cell is essential for ensuring stable performance and delaying degradation. In this paper, a novel fault diagnosis method fusing characteristic impedance for the PEM fuel cell based on a hybrid deep learning network by combing the residual network (ResNet) and long short-term memory (LSTM) is proposed. Specifically, the characteristic impedance that can reflect internal dynamics loss is fused as the feature input alongside other commonly vehicular measurement signals and decoded to form a feature matrix. The feature matrix is then transferred to realize 25 categories of fault detection, including different degrees of membrane drying, flooding, and air starvation. The results showed the accuracy of ResNet-LSTM can reach 99.632% with a good balance of computational burden and is higher than that of a single LSTM, ResNet, and convolutional neural network, as well as traditional machine learning methods because such the hybrid structure can make full use of feature learning capability of ResNet and the time series analysis capability of LSTM. Moreover, the proposed hybrid framework is further validated and compared under different proportions of training samples, noise levels, and input signals.