A Method for Obtaining High-Frequency Fatigue Load Spectrum of Ballastless Track Slabs Based on Deep Clustering

Abstract

This invention provides a method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering, comprising: firstly, conducting load force system analysis of the ballastless track slab; based on different working conditions and a vehicle-track coupled dynamics model, analyzing the spatiotemporal distribution characteristics of track slab stress; and based on deep clustering, setting strain and its gradient control conditions to extract the most unfavorable location in the spatial distribution of track slab response. Based on the track slab strain and support reaction force dataset, using an LSTM or GRU network architecture to predict the fastener support reaction force, thereby achieving track slab load prediction research based on a calibration-free method. Secondly, arranging strain measurement points on the track slab surface in the field to provide field load data for track slab load prediction. Thirdly, statistical analysis of the load distribution based on the rainflow counting method, selecting multiple sine functions combined with inverse Fourier transform and Monte Carlo method to achieve curve fitting and random sampling of load curve parameters.

Description

Technical Field

[0001] This invention relates to the field of load spectrum technology, and in particular to a method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering. Background Technology

[0002] With the increasing operating speed of high-speed railways, ballastless track structures exhibit several typical defects under long-term cyclical train and environmental loads, such as rail surface corrugation, fastener breakage, interlayer separation, and surface concrete cracks. These defects significantly impact driving comfort and safety. Furthermore, these defects lead to reduced track structure lifespan and increased operating costs. Therefore, lifespan assessments of existing ballastless track structural components are necessary. However, fatigue life analysis of ballastless track structures faces two prominent challenges: calculating long-term damage based on vehicle-track coupled dynamics simulations may present significant computational costs; and current track structure strength and fatigue designs do not consider the impact of different speed levels and load conditions on load coefficients, potentially hindering simulation of actual operating loads. Therefore, it is crucial to further clarify the load characteristics of ballastless tracks and propose a more reasonable and efficient dynamic load application method suitable for track slab fatigue analysis.

[0003] Load spectra are fundamental for fatigue testing of structural specimens, full-scale structural fatigue testing, and structural fatigue life simulation. They can reflect the stress conditions of a track structure during its operational service, simulating the load-time history of the track structure throughout its entire life cycle. They effectively reflect the changes in the service condition of a track structure over time. Currently, load spectra are mainly used in my country for transportation equipment, such as aircraft and rolling stock.

[0004] Currently, a clear and effective fatigue load spectrum has not been established for ballastless track structures. Existing fatigue loading methods are derived solely from code values ​​and the field experience of technicians, considering only a single factor and relatively simple load application methods. Regarding the selection of load frequencies, for example, the loading frequency of support reactions is calculated using simple formulas, considering few factors and failing to accurately reflect the on-site structural load and stress distribution. Consequently, it cannot provide an effective basis for assessing load and structural stress state. Summary of the Invention

[0005] The embodiments of the present invention provide a method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering, which is used to solve the technical problems existing in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution.

[0007] A method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering includes: S1. Based on the service status and environment of the ballastless track under actual operating conditions, conduct load force analysis of the ballastless track slab; S2. By selecting different operating conditions and vehicle-track coupled dynamics models, an analytical model is established that considers the influence of different service states of track structure, vehicle parameters and different driving speeds, and the stress time-domain curve under vehicle operating conditions is extracted. S3. Set strain and its gradient control conditions, and use deep clustering methods based on K-means and DBSCAN, combined with principal component analysis dimensionality reduction method, to extract the most unfavorable location of the track slab response spatial distribution. At the extracted most unfavorable location of the track slab response spatial distribution, identify stress concentration areas and fatigue sensitive points. S4. Based on deep learning methods, a multi-layer time-series prediction model is constructed using LSTM or GRU recurrent neural network architecture to predict the reaction force of fastener support points. By introducing attention mechanisms and residual connections, the learning ability of the multi-layer time-series prediction model to long-term dependencies is improved, enabling the multi-layer time-series prediction model to perform track slab load prediction based on the non-site calibration method. S5. Arrange strain measurement points on the surface of the track slab on site, deploy measurement equipment, optimize sensor configuration based on stress concentration areas and fatigue sensitive points identified in step S3, carry out track slab load testing and load data acquisition through the multi-layer time series prediction model constructed in step S4, and establish a complete on-site detection database. S6. Statistical analysis of fatigue load characteristics based on rainflow counting method, and identification of the amplitude, mean and frequency distribution characteristics of stress cycles of ballastless track slabs in a complete field test database; S7. Based on the amplitude, mean and frequency distribution characteristics of stress cycles of the ballastless track slab identified in step S6, the load curve parameters of the track slab are fitted by combining multiple sine functions with inverse Fourier transform and different service conditions and influencing factors. S8. Using the Monte Carlo method, random sampling is performed on the load curve parameters of the track slab. Considering the probability distribution characteristics of parameters such as load amplitude, frequency, and phase, a high-frequency fatigue load spectrum of ballastless track covering different operating conditions is established through large-sample statistical analysis. This high-frequency fatigue load spectrum for ballastless track, covering different operating conditions, is used for maintenance work on ballastless track structures.

[0008] Preferably, step S1 includes: extracting the main load type and directional characteristics of the ballastless track based on its service status and environment under actual operating conditions.

[0009] Preferably, step S2 includes: selecting different operating conditions and vehicle-track coupled dynamic models, considering the multi-factor influence of train speed, axle load variation, and track irregularity excitation, establishing a multi-degree-of-freedom vehicle-track system dynamic simulation model, and obtaining the spatial and temporal distribution characteristics of track slab stress and the stress time-domain curve under spatial location.

[0010] Preferably, step S3 includes: Based on the deep clustering method, cluster analysis is performed on the stress time-domain curves at different locations of the track slab, and the unfavorable locations of the spatial distribution of the track slab response are initially divided according to the stress at different spatial locations of the track slab. By combining strain threshold and strain gradient control conditions, the differences in stress response of track slab are further clarified, and the test point at the most unfavorable stress position on the track slab surface with a large stress value and a relatively small strain gradient is selected. Based on the test point located at the most unfavorable stress position on the ballastless track slab surface with a large stress value and a relatively small strain gradient, and combined with the spatial distribution characteristics of the track slab strain, the track slab strain test was conducted on the in-service line for verification.

[0011] Preferably, step S4 further includes: The fastener support reaction force is set as the direct load of the ballastless track slab. Based on the test verification results, the data set of track slab strain and corresponding fastener support reaction force is obtained. Based on the multi-layer time series prediction model, the fastener support reaction force is predicted by the measured track slab strain data. Improve model accuracy by optimizing the neural network architecture; The ballastless track slab body is used as a force sensor to measure the reaction force of the fastener support point, which is used to accurately measure the reaction force of the support point. Based on the obtained track slab strain test data, the fastener support reaction force is predicted using a multi-layer time-series prediction model.

[0012] Preferably, step S6 includes: Based on the predicted data of fastener support reaction force, the statistical characteristics of fatigue load are analyzed using the rainflow counting method. The load mean, amplitude and statistical quantity distribution are extracted, and the distribution of track slab load is obtained through analysis.

[0013] Preferably, step S7 includes: Based on the fastener support reaction force dataset affected by different service conditions, train speeds and vehicle parameters, multiple sine functions combined with inverse Fourier transform are selected as the fitting curves of the support reaction force to conduct fitting analysis of the support reaction force under different influencing factors. A parameter fitting method is selected to fit the parameter distribution state of the load curve parameters under different influencing factors, and to realize the mathematical expression of the support reaction load parameters from the perspective of probability distribution. Based on the completed load parameter fitting results, combined with the Monte Carlo parameter random sampling method, and considering load distribution factors, the load on the track slab is realistically reproduced.

[0014] Preferably, the sub-step of realistically reproducing the track slab load based on the completed load parameter fitting results, combined with the Monte Carlo parameter random sampling method, and considering load distribution factors, further includes: Through fitting formula

[0015] Fit the load curve; where, For dynamic loads, F Based on the load, For load amplitude coefficient, For the load frequency coefficient, This is the load phase offset coefficient.

[0016] As can be seen from the technical solutions provided by the embodiments of the present invention above, the present invention is a method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering. The present invention is mainly applicable to the formation and compilation of fatigue load spectra of ballastless track slabs, thereby forming a load application method suitable for evaluating the cyclic stress performance of ballastless tracks. The process for forming this high-frequency fatigue load spectrum of ballastless tracks is as follows: First, a load force system analysis of the ballastless track slab is conducted. Based on different working conditions and the vehicle-track coupled dynamics model, the spatiotemporal distribution characteristics of track slab stress are analyzed. Based on the deep clustering method, strain and its gradient control conditions are set, and the most unfavorable position of the track slab response spatial distribution is extracted. Based on the track slab strain and support reaction force dataset, an LSTM or GRU network architecture is used to predict the fastener support reaction force, thereby realizing track slab load prediction research based on a calibration-free method. Finally, on-site strain measurement points are arranged on the track slab surface to provide on-site load data for track slab load prediction. Statistical analysis of load distribution was conducted using the rainflow counting method. Multiple sine functions were selected, combined with inverse Fourier transform and Monte Carlo methods to achieve curve fitting and random sampling of load curve parameters. The goal of this high-frequency fatigue load spectrum for ballastless track is to effectively simulate cyclic loads under complex operating environments, including different irregularities, periodic irregularities caused by temperature gradients, and wheel out-of-roundness. Finally, basic load spectrum parameters such as load loading curves, load frequencies, and load coefficient distributions were proposed as load input conditions to guide fatigue experiments and simulations of ballastless track structural components, thereby enabling fatigue damage analysis and lifespan studies of the main structural components of ballastless track.

[0017] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and will become apparent from the description or may be learned by practice of the invention. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a technical roadmap for the method of obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering, provided by the present invention.

[0020] Figure 2 The figure shows the results of batch calculation of track slab stress for the method of obtaining high-frequency fatigue load spectrum of ballastless track slab based on deep clustering provided by the present invention.

[0021] Figure 3 The figure shows the deep clustering analysis results of the track slab stress in the method for obtaining the high-frequency fatigue load spectrum of ballastless track slab based on deep clustering provided by the present invention.

[0022] Figure 4 The signal spectrum diagram of the measured load data of the method for obtaining the high-frequency fatigue load spectrum of ballastless track slab based on deep clustering provided by the present invention.

[0023] Figure 5 The image shows a comparison of the spectrum information before and after signal filtering in the method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering provided by the present invention.

[0024] Figure 6 The image shows a comparison of the time-domain signals before and after filtering in the method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering provided by this invention.

[0025] Figure 7 This is a schematic diagram of the final correction signal after data processing in the method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering provided by the present invention.

[0026] Figure 8 A schematic diagram of a track slab load prediction model based on bidirectional LSTM, which is provided by the method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering in this invention.

[0027] Figure 9 This is a schematic diagram illustrating the influence of the load frequency coefficient on the method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering provided by the present invention.

[0028] Figure 10 This is a schematic diagram of partial coefficient fitting for the method of obtaining the high-frequency fatigue load spectrum of ballastless track slab based on deep clustering provided by the present invention.

[0029] Figure 11 This diagram illustrates the influence of different conditions on the load partial factor of the method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering provided by the present invention.

[0030] Figure 12 This is a schematic diagram of the load curve reproduction result of the method for obtaining the high-frequency fatigue load spectrum of ballastless track slab based on deep clustering provided by the present invention.

[0031] Figure 13 This is a schematic diagram of the load distribution under complete train operation for the method of obtaining high-frequency fatigue load spectrum of ballastless track slab based on deep clustering provided by the present invention.

[0032] Figure 14 This is a schematic diagram of the simulation results of track slab stress in the method for obtaining the high-frequency fatigue load spectrum of ballastless track slab based on deep clustering provided by the present invention. Detailed Implementation

[0033] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0034] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or couplings. The term “and / or” as used herein includes any and all combinations of one or more of the associated listed items.

[0035] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as herein.

[0036] To facilitate understanding of the embodiments of the present invention, the following will provide further explanation and description with reference to the accompanying drawings and several specific embodiments. These embodiments do not constitute a limitation on the embodiments of the present invention.

[0037] See Figure 1 This invention provides a method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering, comprising the following steps: S1. Based on the service status and environment of the ballastless track under actual operating conditions, conduct load force analysis of the ballastless track slab; S2. By selecting different operating conditions and vehicle-track coupled dynamics models, an analytical model is established that considers the influence of different service states of track structure, vehicle parameters and different driving speeds, and the stress time-domain curve under vehicle operating conditions is extracted. S3. Set strain and its gradient control conditions, and use deep clustering methods based on K-means and DBSCAN, combined with principal component analysis dimensionality reduction method, to extract the most unfavorable location of the track slab response spatial distribution. At the extracted most unfavorable location of the track slab response spatial distribution, identify stress concentration areas and fatigue sensitive points. S4. Based on deep learning methods, a multi-layer time-series prediction model is constructed using LSTM or GRU recurrent neural network architecture to predict the reaction force of fastener support points. By introducing attention mechanisms and residual connections, the learning ability of the multi-layer time-series prediction model to long-term dependencies is improved, enabling the multi-layer time-series prediction model to perform track slab load prediction based on the non-site calibration method. S5. Arrange strain measurement points on the surface of the track slab on site, deploy measurement equipment, optimize sensor configuration based on stress concentration areas and fatigue sensitive points identified in step S3, carry out track slab load testing and load data acquisition through the multi-layer time series prediction model constructed in step S4, and establish a complete on-site detection database. S6. Statistical analysis of fatigue load characteristics based on rainflow counting method, and identification of the amplitude, mean and frequency distribution characteristics of stress cycles of ballastless track slabs in a complete field test database; S7. Based on the amplitude, mean and frequency distribution characteristics of stress cycles of the ballastless track slab identified in step S6, the load curve parameters of the track slab are fitted by combining multiple sine functions with inverse Fourier transform and different service conditions and influencing factors. S8. Using the Monte Carlo method, random sampling is performed on the load curve parameters of the track slab. Considering the probability distribution characteristics of parameters such as load amplitude, frequency, and phase, a high-frequency fatigue load spectrum of ballastless track covering different operating conditions is established through large-sample statistical analysis.

[0038] This high-frequency fatigue load spectrum for ballastless track, covering different operating conditions, is used to predict and evaluate the fatigue life of ballastless track structures. It can guide the maintenance work of ballastless track and enable the timely replacement of ballastless track structural components.

[0039] In a preferred embodiment of the present invention, the process of performing load analysis on the ballastless track system in step S1 may specifically include: extracting the main load types and directions, etc., as the basis for the next load analysis. Regarding the track slab, as a typical reinforced concrete structure, the track slab mainly bears the train load transmitted from the upper fasteners to the rail support platform during structural use; therefore, the vertical force at the rail support platform location should be considered as the main load type.

[0040] In step S2, actual load conditions need to be selected based on the actual operating line conditions. To cover complex operating environment conditions, typical cross-sections of ballastless track under different service conditions are selected for measuring point arrangement and load testing. This requires considering different track structure service years and the condition of the ballastless track itself. As service time increases, track structure stiffness and track smoothness will change, potentially leading to increased track structure stiffness and deterioration over time. Other influencing factors are also considered, such as rail cross-section type, longitudinal relative position of fasteners and track slabs, relative position of track slabs and lower base plates, and different geographical locations, to analyze their impact on load values ​​and statistical distribution characteristics. This increases the reliability and applicability of the load spectrum establishment results. Different operating conditions and vehicle-track coupled dynamic models are selected to analyze the spatiotemporal distribution characteristics of track slab stress and extract stress time-domain curves at different spatial locations. To cover complex on-site conditions, an analytical model was established considering the influence of different service states of the track structure, vehicle parameters, and different operating speeds. Stress time-domain curves were extracted under vehicle operating conditions to serve as the data foundation for further analysis of the most unfavorable stress locations on the track slab. Batch extraction of track slab stress was then performed. Figure 2 As shown.

[0041] In some embodiments, step S3 includes: setting strain and its gradient control conditions, and extracting the most unfavorable location in the spatial distribution of the track slab response based on a deep clustering method. First, cluster analysis is performed on the stress-time domain curves at different locations on the track slab, based on the established deep clustering method, to initially divide the track slab according to the stress at different spatial locations. Then, the differences in the stress response of the track slab are further clarified by combining strain thresholds and strain gradient control conditions, etc. Finally, the test point at the most unfavorable stress location on the track slab surface with a large stress value and a relatively small strain gradient is selected. The deep clustering results are as follows: Figure 3 As shown.

[0042] Then, strain measurement points were arranged on the surface of the track slab to conduct track slab load tests and acquire load data. Based on the most unfavorable stress location of the track slab obtained by the aforementioned deep clustering method, and combined with the spatial distribution characteristics of the track slab strain, track slab strain tests were conducted on the in-service line.

[0043] Next, typical load section environmental data collection and processing are performed. After acquiring the field-measured data, spectral analysis is required for each data point. By analyzing the main frequency components of the collected data, noise is filtered and processed using different methods. Considering the influence of temperature and humidity conditions and electromagnetic interference in the field operating environment, zero-point offset processing and baseline fluctuation correction are necessary for the load data. Furthermore, the acquisition time is usually longer than the actual effective data time, necessitating the truncation of effective data segments. Simultaneously, different sets of data collected at the same location and at the same speed level are stitched together to increase the amount of data for analyzing individual data points, creating data that facilitates the next step of load spectrum compilation. Signal spectrum analysis is as follows... Figure 4 As shown, the spectral information of the signal before and after filtering is compared as follows: Figure 5 As shown, the time-domain signal comparison before and after signal filtering is as follows: Figure 6 As shown, the final corrected model number is as follows: Figure 7 As shown.

[0044] In step S4, based on deep learning methods, LSTM or GRU network architectures are used to predict the fastener support reaction force. The fastener support reaction force is the direct load on the track slab. Based on data obtained from field measurements of track slab strain and corresponding fastener support reaction forces, LSTM (Long Short-Term Memory Neural Network) and GRU (Gated Recurrent Neural Network) are used to predict the fastener support reaction force from the measured track slab strain data. Model accuracy is improved through model optimization of the neural network architecture. This overcomes the problem that traditional support reaction force testing methods rely on iron pads and rail strain measurements, both of which require on-site calibration. By using the track slab itself as a force sensor for measuring the fastener support reaction force, accurate measurement of the support reaction force is achieved. The established model architecture is as follows: Figure 8 As shown.

[0045] The datasets for track slab strain and corresponding fastener support reactions are constructed based on data obtained from on-site tests of track slab strain and corresponding fastener support reactions conducted on service lines. This dataset is primarily designed to provide relatively reasonable data support for modeling, thus requiring accurate model data and necessitating model validation using on-site measured datasets. This dataset will subsequently be used for simulation model validation. When employing a multi-layer time-series prediction model, the dataset will mainly originate from batch calculations of the simulation model to meet the basic data volume requirements.

[0046] Optimizing neural network architecture models can be achieved using existing techniques, such as neural network architecture model optimization.

[0047] Statistical analysis of fatigue load characteristics was conducted using the rainflow counting method. After acquiring a certain amount of support reaction force data, the statistical characteristics of the fatigue load were analyzed using the rainflow counting method. The load mean, amplitude, and their statistical distribution were identified, thus providing a basis for analyzing the load distribution on the track slab as the computational foundation for the next fitting analysis. For the time-domain data that had undergone load data type conversion, load cyclic counting analysis was performed using the commonly used rainflow counting method to preliminarily statistically analyze the one-dimensional and two-dimensional load distributions. This transformed the time-domain load data into load statistical results containing load amplitude and mean, facilitating the next step of analysis. Simultaneously, the distribution characteristics of the load amplitude were analyzed, using either a normal distribution or a Weibull distribution to describe the load time-domain characteristic distribution. The influence of the load frequency coefficient and the fitting results of some parameters are shown below. Figures 9-11 As shown.

[0048] Track slab load curve fitting is achieved using inverse Fourier transform and sine curve superposition to fit the track slab load curve parameters. Based on a dataset of fastener support reaction forces under different service conditions, train speeds, and vehicle parameters, multi-level sine curves are selected as the fitting curves for the support reaction forces to study the fitting of support reaction forces under different influencing factors. This dataset is mainly obtained through batch simulation calculations based on simulation models. This method can provide sufficient load data support, thus meeting the basic data volume requirements. The simulation dataset enables the prediction of the effect of track slab strain on fastener support reaction forces.

[0049] A parameter fitting method was selected to fit the parameter distribution of the load curve parameters under different influencing factors, thereby realizing the mathematical expression of the support reaction load parameters from the perspective of probability distribution. Based on the completed load parameter fitting results, combined with the Monte Carlo parameter random sampling method, and considering load distribution factors, the load on the track slab was realistically reproduced. The load reproduction results are as follows: Figures 12-11 As shown.

[0050] When fitting the load curve, factors such as operating speed and other different working conditions can influence the load curve based on the baseline curve. Therefore, different coefficient values ​​are used to reflect the influence of different conditions on the load curve. Thus, the load coefficients are determined and refined. Simultaneously, to effectively characterize the influence of the load coefficient and frequency, the following formula is used for preliminary simulation of the load curve, characterizing the effects of the load coefficient and frequency. After refining the formula for different load characteristics, the effect of the fitting formula is further improved. As shown in the following formula:

[0051] In the formula, For dynamic loads, Based on the load, For load amplitude coefficient, For the load frequency coefficient, This is the load phase offset coefficient. , and All parameters are determined after parameter fitting. The above formulas aim to fit and refine the actual field load curves using sine curves of different frequency bands. This achieves full coverage of high-frequency and low-frequency loads, thereby enabling comprehensive simulation of high-frequency excitation and reproducing the high-frequency, velocity, and strain responses under field service conditions.

[0052] Finally, based on the established load spectrum and load reproduction effect, and using the finite element simulation model, the load spectrum is used as the load application condition to simulate and obtain results such as track slab stress. Figure 14 As shown, the feasibility of the load spectrum can be verified.

[0053] In summary, this invention is a method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering. This invention is mainly applicable to the formation and compilation of fatigue load spectra of ballastless track slabs, thereby forming a load application method suitable for evaluating the cyclic stress performance of ballastless tracks. The process for forming this high-frequency fatigue load spectrum of ballastless tracks is as follows: First, the load force system of the ballastless track slab is analyzed. Based on different working conditions and the vehicle-track coupled dynamics model, the spatiotemporal distribution characteristics of track slab stress are analyzed. Based on deep clustering, strain and its gradient control conditions are set, and the most unfavorable position of the track slab response spatial distribution is extracted. According to the track slab strain and support reaction force dataset, the fastener support reaction force is predicted using an LSTM or GRU network architecture, thereby realizing track slab load prediction research based on a calibration-free method. On-site track slab surface strain measurement points are arranged to provide on-site load data for track slab load prediction. Statistical analysis of load distribution is performed based on the rainflow counting method. Multiple sine functions are selected, combined with inverse Fourier transform and Monte Carlo method to achieve curve fitting and random sampling of load curve parameters. The primary goal of this high-frequency fatigue load spectrum for ballastless track is to effectively simulate cyclic loads under complex operating environments, including varying degrees of irregularity, periodic irregularities caused by temperature gradients, and wheel out-of-roundness. Ultimately, fundamental load spectrum parameters such as load loading curves, load frequencies, and load coefficient distributions are proposed as load input conditions to guide fatigue experiments and simulations of ballastless track structural components. This allows for fatigue damage analysis and lifespan studies of key structural components of ballastless track, and provides guidance for the optimized maintenance of ballastless track structures.

[0054] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of one embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing the present invention.

[0055] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present invention.

[0056] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for apparatus or system embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The apparatus and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0057] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for obtaining the high-frequency fatigue load spectrum of ballastless track slabs based on deep clustering, characterized in that, include: S1. Based on the service status and environment of the ballastless track under actual operating conditions, conduct load force analysis of the ballastless track slab; S2. By selecting different operating conditions and vehicle-track coupled dynamics models, an analytical model is established that considers the influence of different service states of track structure, vehicle parameters and different driving speeds, and the stress time-domain curve under vehicle operating conditions is extracted. S3. Set strain and its gradient control conditions, and use deep clustering methods based on K-means and DBSCAN, combined with principal component analysis dimensionality reduction method, to extract the most unfavorable location of the track slab response spatial distribution. At the extracted most unfavorable location of the track slab response spatial distribution, identify stress concentration areas and fatigue sensitive points. S4. Based on deep learning methods, using LSTM or GRU recurrent neural network architecture, a multi-layer time-series prediction model is constructed to predict the reaction force of fastener support points. By introducing attention mechanisms and residual connections, the learning ability of multi-layer time series prediction models to long-term dependencies is improved, enabling multi-layer time series prediction models to perform track slab load prediction based on non-site calibration methods. S5. Arrange strain measurement points on the surface of the track slab on site, deploy measurement equipment, optimize sensor configuration based on stress concentration areas and fatigue sensitive points identified in step S3, carry out track slab load testing and load data acquisition through the multi-layer time series prediction model constructed in step S4, and establish a complete on-site detection database. S6. Statistical analysis of fatigue load characteristics based on rainflow counting method, and identification of the amplitude, mean and frequency distribution characteristics of stress cycles of ballastless track slabs in the complete field detection database; S7. Based on the amplitude, mean and frequency distribution characteristics of stress cycles of the ballastless track slab identified in step S6, the load curve parameters of the track slab are fitted by combining multiple sine functions with inverse Fourier transform and different service conditions and influencing factors. S8. Using the Monte Carlo method, random sampling is performed on the load curve parameters of the track slab. Considering the probability distribution characteristics of parameters such as load amplitude, frequency, and phase, a high-frequency fatigue load spectrum of ballastless track covering different operating conditions is established through large-sample statistical analysis. This high-frequency fatigue load spectrum for ballastless track, covering different operating conditions, is used for maintenance work on ballastless track structures.

2. The method according to claim 1, characterized in that, Step S1 includes: extracting the main load types and directional characteristics of the ballastless track based on its service status and environment under actual operating conditions.

3. The method according to claim 2, characterized in that, Step S2 includes: selecting different operating conditions and vehicle-track coupled dynamic models, considering the multi-factor influence of train speed, axle load variation, and track irregularity excitation, establishing a multi-degree-of-freedom vehicle-track system dynamic simulation model, and obtaining the spatial and temporal distribution characteristics of track slab stress and the stress time-domain curve under spatial location.

4. The method according to claim 3, characterized in that, Step S3 includes: Based on the deep clustering method, cluster analysis is performed on the stress time-domain curves at different locations of the track slab, and the unfavorable locations of the spatial distribution of the track slab response are initially divided according to the stress at different spatial locations of the track slab. By combining strain threshold and strain gradient control conditions, the differences in stress response of track slab are further clarified, and the test point at the most unfavorable stress position on the track slab surface with a large stress value and a relatively small strain gradient is selected. Based on the test point located at the most unfavorable stress position on the ballastless track slab surface with a large stress value and a relatively small strain gradient, and combined with the spatial distribution characteristics of the track slab strain, the track slab strain test was conducted on the in-service line for verification.

5. The method according to claim 4, characterized in that, Step S4 also includes: The fastener support reaction force is set as the direct load of the ballastless track slab. Based on the test and verification results, the data set of track slab strain and corresponding fastener support reaction force is obtained. Based on the multi-layer time series prediction model, the fastener support reaction force is predicted by the measured track slab strain data. Improve model accuracy by optimizing the neural network architecture; The ballastless track slab body is used as a force sensor to measure the reaction force of the fastener support point, which is used to accurately measure the reaction force of the support point. Based on the obtained track slab strain test data, the fastener support reaction force is predicted using the multi-layer time-series prediction model.

6. The method according to claim 5, characterized in that, Step S6 includes: Based on the predicted data of fastener support reaction force, the statistical characteristics of fatigue load are analyzed using the rainflow counting method. The load mean, amplitude and statistical quantity distribution are extracted, and the distribution of track slab load is obtained through analysis.

7. The method according to claim 6, characterized in that, Step S7 includes: Based on the fastener support reaction force dataset affected by different service conditions, train speeds and vehicle parameters, multiple sine functions combined with inverse Fourier transform are selected as the fitting curves of the support reaction force to conduct fitting analysis of the support reaction force under different influencing factors. A parameter fitting method is selected to fit the parameter distribution state of the load curve parameters under different influencing factors, and to realize the mathematical expression of the support reaction load parameters from the perspective of probability distribution. Based on the completed load parameter fitting results, combined with the Monte Carlo parameter random sampling method, and considering load distribution factors, the load on the track slab is realistically reproduced.

8. The method according to claim 7, characterized in that, The sub-step of realistically reproducing the track slab load based on the completed load parameter fitting results, combined with the Monte Carlo parameter random sampling method, and considering load distribution factors, further includes: Through fitting formula Fit the load curve; where, For dynamic loads, F Based on the load, For load amplitude coefficient, For the load frequency coefficient, This is the load phase offset coefficient.

Images