Ternary diagnosis method for influence of basic deformation on ballastless track based on supervised learning

Through supervised learning technology and video data processing, a basic deformation diagnosis model for ballless tracks was established, which solved the problem of basic deformation monitoring under complex geological conditions, achieved efficient and accurate monitoring and early warning of ballless tracks, and improved railway operation safety.

CN119048959AActive Publication Date: 2024-11-29BEIJING JIAOTONG UNIV +3
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411159443.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-22
Publication Date
2024-11-29
Estimated Expiration
2044-08-22

AI Technical Summary

Technical Problem

Under complex geological conditions, traditional methods are difficult to effectively monitor and diagnose deformation of ballastless track foundations, affecting the safety and smoothness of the track structure.

Method used

Using a method based on supervised learning, a ternary diagnostic model is established by obtaining and processing ballastless orbital motion posture video data, combining motion posture video, vibration mode and vibration function to realize real-time monitoring and early warning of basic deformation, and continuously adjust it Training set to improve diagnostic accuracy.

Benefits of technology

It realizes non-contact, efficient and accurate basic deformation monitoring and early warning, reduces manual inspection costs, improves railway operation safety, and provides a scientific basis for engineering design and maintenance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048959A_ABST
    Figure CN119048959A_ABST
Patent Text Reader

Abstract

The invention provides a ternary diagnosis method for influence of basic deformation on a ballastless track based on supervised learning, and belongs to the technical field of ballastless track basic deformation monitoring and diagnosis based on machine learning, and the method comprises the steps: obtaining ballastless track motion posture video data to be recognized; and inputting to-be-identified ballastless track motion attitude video data into the trained ternary diagnosis model of the influence of the basic deformation on the ballastless track for processing, obtaining a result of identifying the influence of the basic deformation on the ballastless track, and performing early warning. According to the method, the motion attitude video of the track motion attitude under the driving condition is shot, and the key positions of the steel rail and the ballastless track plate are extracted from the motion attitude video to establish the post function and the oscill function, so that the service state and the vibration law of the track structure are obtained; and the service state of the track structure is diagnosed by comparing the track motion posture video and function with the track motion posture video and function of the ballastless track structure in a normal section under the driving condition, and part of classical diagnosis results are put into a training set for subsequent diagnosis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of ballastless track foundation deformation monitoring and diagnosis based on machine learning, and in particular to a ternary diagnosis method for the influence of foundation deformation on ballastless track based on supervised learning. Background Art

[0002] In sections with complex geological conditions, the ballastless track has a high stiffness, which has an extremely adverse effect on the track structure after being subjected to foundation deformation. Therefore, the prediction or diagnosis of foundation deformation is particularly important for indicators such as railway driving safety and smoothness. However, the foundation is buried under the track structure and is not exposed to the field of vision, so it is difficult to monitor it through traditional sensor deployment or non-contact measurement. As a type of artificial intelligence machine learning technology, supervised learning has gradually matured in the recognition technology and feature extraction technology of motion posture videos and functions. After training with a sufficiently large training set, the accuracy of the diagnosis can be guaranteed, and the accuracy will increase as the training set continues to expand. Summary of the invention

[0003] The object of the present invention is to provide a ternary diagnosis method and system for the influence of basic deformation on ballastless track based on supervised learning, so as to solve at least one technical problem existing in the above-mentioned background technology.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions:

[0005] In a first aspect, the present invention provides a ternary diagnosis method for the influence of basic deformation on ballastless track based on supervised learning, comprising:

[0006] Acquire video data of the motion posture of the ballastless track to be identified;

[0007] The motion posture video data of the ballastless track to be identified is input into the trained ternary diagnosis model of the influence of foundation deformation on the ballastless track for processing, and the results of identifying the influence of foundation deformation on the ballastless track are obtained and an early warning is issued; wherein, after the early warning information is obtained later, some samples with classic labels as early warnings are added to the training set to retrain the ternary diagnosis model of the influence of foundation deformation on the ballastless track, so as to subsequently refine the hierarchical identification label results; wherein, the motion posture video sub-model, the vibration posture function sub-model, and the vibration oscillate function sub-model of the ternary diagnosis model of the influence of foundation deformation on the ballastless track are trained by multiple motion posture video samples, function samples, and safety labels. When and only when the three sub-models are all judged to be safe, the ternary diagnosis model of the influence of foundation deformation on the ballastless track outputs the final "safety label".

[0008] Optionally, the posture function establishment process is as follows: the intersection points of the neutral axis and the symmetry axis of each section of the rail and the ballastless track plate are obtained from the motion posture video to form a curve, and a preliminary posture function is obtained by fitting in the established spatial Cartesian coordinate system, which is recorded as posture = f(x, y, z); after expanding the y direction, the final posture function is obtained, which is recorded as posture = p(x, e (y) ,z); the oscillate function is established by replacing the time parameter t in the motion posture video with posture = p(x,e (y) ,z) function, we get oscillate = g(t,e (y) ,z).

[0009] Optionally, multiple groups of motion posture videos of the ballastless track structure under normal geological conditions under driving conditions are shot and uploaded, and the motion posture videos, posture=p(x,e (y) ,z)、oscillate=g(t,e (y) ,z) is entered into the training set as a benchmark sample with the label of safe for learning the ternary diagnosis model of the influence of foundation deformation on ballastless track.

[0010] Optionally, after the diagnostic warning information is output, the foundation deformation state of the target section is manually investigated. If it is found that the foundation deformation reaches the operational safety threshold or is of great significance, it is regarded as a classic motion posture video sample, and the sample and the corresponding important significance label are placed in the training set to facilitate the subsequent clustering process, that is, the state of the foundation deformation of the target section can be refined and graded for identification and diagnosis; wherein the important significance includes but is not limited to: locomotive operation safety, comfort threshold, cracks in the ballastless track structure or damage and failure of the track structure, etc.

[0011] Optionally, the training method of the ternary diagnostic model of the influence of foundation deformation on ballastless track is supervised learning, including but not limited to neural network algorithm, random forest algorithm, support vector machine algorithm, etc.

[0012] In a second aspect, the present invention provides a ternary diagnosis system for the influence of basic deformation on ballastless track based on supervised learning, comprising:

[0013] An acquisition module, used for acquiring video data of the motion posture of the ballastless track to be identified;

[0014] The processing module is used for preprocessing such as enhancement and denoising of motion posture video samples, and the preprocessed motion posture video sample data is input into the trained ternary diagnosis model of the influence of foundation deformation on ballastless track for processing, so as to obtain the results of identifying the influence of foundation deformation on ballastless track and issue an early warning; wherein, after obtaining the early warning information in the subsequent period, classic samples are screened out after manual investigation, and classic labels are given, such as "affecting driving comfort but not driving safety" or "affecting driving safety" or "reaching the limit state of track structure", etc. The above classic motion posture video samples and corresponding labels are added to the training set to retrain the ternary diagnosis model of the influence of foundation deformation on ballastless track, so as to refine the hierarchical identification label results in the subsequent period; wherein, the motion posture video sub-model, the vibration posture function sub-model, and the vibration oscillate function sub-model of the ternary diagnosis model of the influence of foundation deformation on ballastless track are trained by multiple motion posture video samples, function samples, and safety labels. When and only when the three sub-models are all determined to be safe, the ternary diagnosis model of the influence of foundation deformation on ballastless track outputs the final "safety label".

[0015] Furthermore, the acquisition module includes: a set of non-parallel high-definition cameras and processing modules pre-installed near the measured section, which can obtain the motion posture video of the complete section. The high-definition camera is powered separately. When the power of the high-definition camera is less than 5%, the power alarm is triggered and a reminder to replace the power supply is issued. The processing module can realize the function of pre-processing and uploading the motion posture video after shooting. When the high-definition camera captures the vibration of the track, it triggers the start of filming. When the high-definition camera captures the stop of vibration of the track, it triggers the end of filming, thereby obtaining a single motion posture video sample.

[0016] In a third aspect, the present invention provides a non-transitory computer-readable storage medium, which is used to store computer instructions. When the computer instructions are executed by a processor, the ternary diagnosis method of the influence of basic deformation on ballastless track based on supervised learning as described in the first aspect is implemented.

[0017] In a fourth aspect, the present invention provides a computer device comprising a memory and a processor, wherein the processor and the memory communicate with each other, the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the ternary diagnosis method for the influence of basic deformation on ballastless track based on supervised learning as described in the first aspect.

[0018] In a fifth aspect, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes instructions for implementing the ternary diagnostic method for the influence of basic deformation on ballastless track based on supervised learning as described in the first aspect.

[0019] Beneficial effects of the present invention: by adopting motion posture video shooting technology and space curve establishment, function fitting and variable replacement, three-dimensional real-time monitoring and early warning of foundation deformation are realized. The method has the advantages of non-contact, high efficiency, high precision and wide applicability, which can greatly reduce the cost and labor intensity of manual inspection, improve the safe operation level of railways, and provide a scientific basis for guiding engineering design, maintenance and repair, safety assessment and other work.

[0020] Additional advantages of the present invention will be more clearly given in the following description or learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0022] Figure 1 The present invention is a flowchart of a ternary diagnosis method for the influence of basic deformation on ballastless track based on supervised learning according to an embodiment of the present invention.

[0023] Figure 2 It is a schematic diagram of the symmetry axis and neutral axis of the rail according to an embodiment of the present invention.

[0024] Figure 3 A schematic diagram of establishing a spatial Cartesian coordinate system according to an embodiment of the present invention.

[0025] Figure 4 This is a flow chart of the identification of a ternary diagnosis method for the influence of basic deformation on ballastless track based on supervised learning according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below by the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention.

[0027] It should be understood by those skilled in the art that unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art to which this invention belongs.

[0028] It should also be understood that terms, such as those defined in commonly used dictionaries, should be understood to have a meaning consistent with that in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless as defined herein.

[0029] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the 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 and / or groups thereof.

[0030] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. Different embodiments or examples described in this specification and features of different embodiments or examples may be combined and combined by those skilled in the art without contradiction.

[0031] To facilitate understanding of the present invention, the present invention is further explained below with reference to specific embodiments in conjunction with the accompanying drawings, and the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0032] Those skilled in the art should understand that the drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily necessary for implementing the present invention.

[0033] Example 1

[0034] In this embodiment 1, a system for diagnosing the influence of foundation deformation on ballastless track based on supervised learning is first provided, comprising: an acquisition module, used for acquiring video data of motion posture of ballastless track to be identified; a processing module, used for inputting the video data of motion posture of ballastless track to be identified into a trained ternary diagnosis model of the influence of foundation deformation on ballastless track for processing, obtaining the result of identifying the influence of foundation deformation on ballastless track and issuing an early warning; wherein, after the early warning information is obtained later, some samples with classic labels as early warning are added to the training set to retrain the ternary diagnosis model of the influence of foundation deformation on ballastless track, so as to refine the hierarchical identification label results later; wherein, the motion posture video sub-model, the vibration posture function sub-model and the vibration oscillate function sub-model of the ternary diagnosis model of the influence of foundation deformation on ballastless track are trained by multiple motion posture video samples, function samples and safety labels, and the ternary diagnosis model of the influence of foundation deformation on ballastless track outputs the final "safety label" only when and only when the three sub-models are all determined to be safe.

[0035] The acquisition module includes: a set of non-parallel high-definition cameras and processing modules pre-installed near the measured section, which can obtain the motion posture video of the complete section. The high-definition camera is powered separately. When the power of the high-definition camera is less than 5%, the power alarm is triggered and a reminder to replace the power supply is issued. The processing module can realize the function of pre-processing and uploading the motion posture video after shooting. When the high-definition camera captures the vibration of the track, it triggers the start of the video recording. When the high-definition camera captures the stop of the vibration of the track, it triggers the end of the video recording, thereby obtaining a single motion posture video sample.

[0036] In this embodiment, based on the above system, a method for diagnosing the influence of foundation deformation on ballastless track based on supervised learning is implemented, including: obtaining motion posture video data of ballastless track to be identified; inputting the motion posture video data of ballastless track to be identified into the trained ternary diagnosis model of the influence of foundation deformation on ballastless track for processing, obtaining the result of identifying the influence of foundation deformation on ballastless track and issuing an early warning; wherein, after obtaining the early warning information later, adding some samples with classic labels as early warnings to the training set to retrain the ternary diagnosis model of the influence of foundation deformation on ballastless track, so as to subsequently refine the hierarchical identification label results; wherein, the motion posture video sub-model, the vibration posture function sub-model, and the vibration oscillate function sub-model of the ternary diagnosis model of the influence of foundation deformation on ballastless track are trained by multiple motion posture video samples, function samples, and safety labels. When and only when the three sub-models are all determined to be safe, the ternary diagnosis model of the influence of foundation deformation on ballastless track outputs the final "safety label".

[0037] The process of establishing the posture function is as follows: the intersection points of the neutral axis and the symmetry axis of each section of the rail and ballastless track slab are obtained from the motion posture video to form a curve, and then a preliminary posture function is obtained by fitting in the established spatial Cartesian coordinate system, denoted as posture = f(x, y, z); after expanding the y direction, the final posture function is obtained, denoted as posture = p(x, e (y) ,z); the oscillate function is established by replacing the time parameter t in the motion posture video with posture = p(x,e (y) ,z) function, we get oscillate = g(t,e (y) ,z).

[0038] Shoot multiple sets of motion posture videos of ballastless track structures under normal geological conditions under driving conditions and upload them. (y) ,z)、oscillate=g(t,e (y) ,z) is entered into the training set as a benchmark sample with the label of safe for learning the ternary diagnosis model of the influence of foundation deformation on ballastless track.

[0039] After the diagnostic warning information is output, the foundation deformation status of the target section is manually investigated. If it is found that the foundation deformation has reached the operational safety threshold or is of great significance, it will be regarded as a classic sample and placed in the training set to facilitate the subsequent clustering process, that is, the status of the foundation deformation of the target section can be refined and graded for identification and diagnosis; wherein, the said important significance includes but is not limited to: locomotive operation safety, comfort threshold, cracks in the ballastless track structure or damage and failure of the track structure, etc.

[0040] The training method of the ternary diagnosis model of the influence of foundation deformation on ballastless track is supervised learning, including but not limited to neural network algorithm, random forest algorithm, support vector machine algorithm, etc.

[0041] Example 2

[0042] This embodiment 2 provides a ternary diagnosis method for the influence of basic deformation on ballastless track based on supervised learning, which is a technology that utilizes the motion posture of the track under driving conditions to diagnose the service status of the track. The core of this technology is to shoot a video of the track motion posture under driving conditions, and extract key positions of the rails and ballastless track plates from the motion posture video to establish posture functions and oscillate functions, thereby obtaining the service status and vibration laws of the track structure. By comparing the track motion posture video and function with the normal section ballastless track structure under driving conditions, the service status of the track structure is diagnosed, and some classic diagnosis results are placed in the training set for subsequent diagnosis.

[0043] The ternary diagnosis method of the influence of foundation deformation on ballastless track based on supervised learning includes the steps of obtaining the training set motion posture video samples, the ternary diagnosis model of the influence of foundation deformation on ballastless track, the motion posture video to be identified, the vibration posture function, and the vibration oscillate function. The ternary diagnosis model of the influence of foundation deformation on ballastless track is obtained by training multiple motion posture video samples, function samples, and safety labels, and includes three sub-models: motion posture video sub-model, vibration posture function sub-model, and vibration oscillate function sub-model. The motion posture video to be identified is input into the trained ternary diagnosis model of the influence of foundation deformation on ballastless track, the identification label is obtained, and the corresponding warning is issued. After the warning information is obtained later, some samples with the classic label of warning are added to the training set to retrain the model, so as to refine the hierarchical recognition label results later.

[0044] The posture function establishment process is as follows: the intersection points of the neutral axis and the symmetry axis of each section of the rail and ballastless track plate are obtained from the motion posture video to form a curve, and the initial posture function is obtained by fitting in the established spatial Cartesian coordinate system, which is recorded as posture = f(x, y, z). After expanding the y direction, the final posture function is obtained, which is recorded as posture = p(x, e (y) ,z).

[0045] The oscillate function is established as follows: replace the time parameter t in the motion posture video with posture = p(x, e (y) ,z) function, we get oscillate = g(t,e (y) ,z).

[0046] Shoot multiple sets of motion posture videos of ballastless track structures under normal geological conditions under driving conditions and upload them. (y) ,z)、oscillate=g(t,e (y) ,z) is entered into the training set as a benchmark sample with the label of safe for learning the ternary diagnosis model of the influence of foundation deformation on ballastless track.

[0047] The motion posture video samples in the training set and the motion posture videos to be identified are preprocessed after acquisition, namely, denoising and enhancement processing, so as to improve the quality of the motion posture videos and extract features.

[0048] The training method of the ternary diagnosis model of the influence of foundation deformation on ballastless track is supervised learning, including but not limited to neural network algorithm, random forest algorithm, support vector machine algorithm, etc.

[0049] The ternary diagnostic model of the impact of foundation deformation on ballastless track outputs the final “safety label” only when and only when all three sub-models are judged to be safe.

[0050] The method of obtaining the video samples of classic motion postures is as follows: after the diagnostic warning information is output, the foundation deformation state of the target section is manually investigated. If it is found that the foundation deformation reaches the operational safety threshold or has other important significance, it is regarded as a classic sample and placed in the training set together with the critical labels enriched after the manual investigation, which is convenient for the subsequent clustering process, that is, the state of the foundation deformation of the target section can be refined and graded for identification and diagnosis. Among them, the above-mentioned important significance includes but is not limited to: locomotive operation safety, comfort threshold, cracks in the ballastless track structure or damage and failure of the track structure, etc.

[0051] A set of non-parallel HD cameras and processing modules are pre-installed near the section to be tested, and the motion posture video of the entire section can be obtained. The HD camera is powered separately. When the power of the HD camera is less than 5%, the power alarm is triggered and a reminder to replace the power supply is issued. The processing module can realize the pre-processing and uploading functions of the motion posture video after shooting. When the HD camera captures the vibration of the track, it triggers the start of the video recording. When the HD camera captures the stop of the vibration of the track, it triggers the end of the video recording, thereby obtaining a single motion posture video sample.

[0052] Track motion posture refers to the dynamic deformation of the track vibration in the measured section from the time the high-definition camera captures the start of track vibration to the time the track stops vibrating.

[0053] This method can be applied to any ballastless track structure in areas prone to foundation deformation.

[0054] Example 3

[0055] Due to the characteristics of the ballastless track substructure hidden under the track structure, poor regularity of change, and difficulty in prediction, it is difficult to diagnose the foundation deformation using traditional monitoring methods. Figures 1 to 4 As shown, this embodiment provides a ternary diagnosis method for the influence of foundation deformation on ballastless track based on supervised learning, which reflects the service status by shooting the motion posture video of the track motion posture, and finally achieves the purpose of inverting the foundation deformation.

[0056] During the installation phase of each module: set up high-definition cameras and processing modules at the target locations to ensure that the high-definition cameras can obtain clear and complete track motion postures, and perform pre-processing operations such as denoising and enhancement in the processing modules. The more benchmark samples, the better, so that the ternary diagnostic model of the impact of foundation deformation on ballastless track can be fully learned and the diagnostic accuracy can be improved.

[0057] Upload the preprocessed motion posture video samples. In the early stage of ballastless track construction, it is assumed that there is no deformation of the lower foundation. Therefore, all the samples obtained at this stage are labeled as "safe".

[0058] Shoot a sufficient number of motion posture video samples, establish a spatial Cartesian coordinate system, and take the intersection of the neutral axis of the target section rail and ballastless track slab with the symmetry axis of the rail cross section as the origin, the train operation direction as the positive direction of the z-axis, and the vertical upward as the positive direction of the y-axis. Follow the "right-hand principle" to establish the x-axis. Obtain the neutral axis position of the target section rail and ballastless track slab in the motion posture video to generate their respective spatial position-mileage curves, and fit them into the vibration mode posture space function, denoted as posture = f(x, y, z). Among them, considering that the influence of foundation deformation on the ballastless track structure is mainly reflected in the influence in the y direction, in order to amplify this feature, and because the natural exponential function has the characteristics of fast growth rate and the derivatives of each order are consistent with the original function, the above function is corrected to: posture = p(x, e (y) ,z).

[0059] Among them, the reason for choosing the neutral axis as the main feature to represent the rail is that the shear stress at the neutral axis is 0, which can reduce the impact of the distortion of the rail and ballastless track slab on the model. The reason for choosing the symmetry axis to represent the characteristics of the rail and ballastless track slab is that the rail edge will suffer from wear, cracks, peeling, core damage, etc. during the operation of the track structure, and the ballastless track slab is prone to cracking and separation. Therefore, the symmetry axis position at the exact center is selected, the characteristics are obvious, and the difficulty of acquisition is low. The reason for choosing the natural exponential function for the magnification function in the y direction is that the growth rate of the natural exponential function increases with the increase of the independent variable, that is, it is a convex function, and the magnification effect on the y direction is significant. And the derivatives of each order are consistent with the original function, that is, the derivation operation will not change the characteristics. The original function and the derivatives of each order are important features of the posture function, which is extremely beneficial to the accuracy of the speedometer for subsequent feature extraction and recognition.

[0060] To take dynamic factors into account, the above posture=p(x,e (y) ,z) function is replaced by the time parameter t in the motion posture video, and the vibration oscillate = g(t,e (y) ,z) function, which reflects the change of rail vibration in the y direction with time. The reasons for selecting the independent variables of the vibration function are as follows: the foundation deformation mainly affects the vertical direction of the rail, so the y direction should be retained; although the change in the z direction is negligible relative to the length of the rail, it reflects the mileage of the rail and also needs to be retained; the x direction reflects the lateral direction of the rail, the foundation deformation has little effect on this direction, and the change over time is not large, so x is replaced by the time parameter t.

[0061] 80% of the motion posture video samples and safety labels are used as training sets to extract the characteristics of the track motion posture in the motion posture video, including: vibration frequency, vibration type. 80% of the above-mentioned posture function samples are used as training sets to extract dynamic features including: the expression of the function itself, the positive and negative of the first-order derivative, etc. 80% of the above-mentioned oscillate function samples are used as training sets to extract dynamic features including: the expression of the function itself, the positive and negative of the first-order derivative, etc. The supervised learning algorithm is used to train the ternary diagnostic model of the influence of foundation deformation on ballastless track, and the above-mentioned features and safety labels are used to train the feature vectors in the ternary diagnostic model of the influence of foundation deformation on ballastless track.

[0062] 20% of the motion posture video samples are used as a test set, and are input into the encoder of the preliminarily trained ternary diagnostic model of the influence of basic deformation on ballastless track to obtain a feature vector, which is then output through the decoder. Similarly, 20% of the posture function samples described in 3 are used as a test set, and are input into the preliminarily trained ternary diagnostic model of the influence of basic deformation on ballastless track to obtain a "safety label". 20% of the oscillate function samples described in 3 are used as a test set, and are input into the preliminarily trained ternary diagnostic model of the influence of basic deformation on ballastless track to obtain a "safety label". This model is sufficiently accurate if and only if all three are true at the same time.

[0063] After the basic deformation ternary diagnosis model is trained with the training set and verified with the test set, the corresponding "safety label" or "warning label" can be output after the motion posture video to be tested is input into the model. The motion posture video to be identified is preprocessed and the final posture=p(x,e (y) ,z) and oscillate=g(t,e (y) ,z), input the three simultaneously into the ternary diagnostic model of the influence of foundation deformation on ballastless track, and output "safety label" or "warning label". When the ternary diagnostic model of the influence of foundation deformation on ballastless track outputs "warning label", conduct field investigation to explore the state of foundation deformation at the target position. When the foundation deformation is the threshold of each stage or has special meaning, the corresponding motion posture video is given the corresponding foundation deformation label, and is regarded as a classic sample and placed in the training set, and used as a sample to retrain the ternary diagnostic model of the influence of foundation deformation on ballastless track. After a sufficient number of samples, the foundation deformation model is iterated from being able to output only "safety label" and "warning label" to being able to output "safety label" and foundation deformation reaching each threshold label. Finally, the function of inverting the deformation state of the lower foundation is realized through the track motion posture video and the posture function and oscillate function track service state.

[0064] Specifically, in this embodiment, after the ballastless track structure in the target section is constructed, it is assumed that there is no foundation deformation here, and a set of non-parallel high-definition cameras and processing modules are installed nearby. Under driving conditions, the installed high-definition cameras are used to obtain a sufficient number of preliminary track motion posture videos, which are uploaded after noise reduction and enhancement pre-processing to obtain safety labels and corresponding motion posture videos. Figure 3 As shown in the figure, a spatial Cartesian coordinate system is established. The intersection points of the neutral axis and the symmetry axis of each section of the rail and the ballastless track plate are extracted from the above motion posture video to form a curve, and the preliminary rail vibration shape posture space function posture=f(x,y,z) corresponding to the above motion posture video is obtained by fitting. After the increase and improvement in the y direction, the final posture=p(x,e (y) ,z). The established posture = p(x,e (y) ,z) and the time parameters in the motion posture video to establish oscillate = g(t,e (y) ,z) function. The training set is used to train the ternary diagnosis model of the influence of foundation deformation on ballastless track, and the neural network algorithm is selected to train the algorithm using the extracted motion posture video features. The diagnostic performance of the model is optimized by continuously adjusting the model parameters. After the model is verified and optimized, the preliminary training of the ternary diagnosis model of the influence of foundation deformation on ballastless track is completed, and it can be used to diagnose whether foundation deformation occurs, and output "safety label" or "warning label". The motion posture video to be identified is obtained and preprocessed by a high-definition camera, and is input into the ternary diagnosis model of the influence of foundation deformation on ballastless track that has been preliminarily trained to realize the diagnosis of foundation deformation, that is, output the label corresponding to the sample to be tested. After the model issues an early warning, the state of foundation deformation is manually investigated on site. If it is found that the foundation deformation reaches the various thresholds of safe operation or other important significance, the corresponding samples and corresponding labels after preprocessing are added to the training set as classic samples. At this time, the ternary diagnosis model of the influence of foundation deformation on ballastless track can identify the corresponding motion posture video samples to be identified under such foundation deformation conditions, give accurate early warnings, and output "early warning labels" with the state of foundation deformation.

[0065] In summary, the present embodiment provides a ternary diagnosis method for the influence of foundation deformation on ballastless track based on supervised learning, which realizes ternary real-time monitoring and early warning of foundation deformation by adopting motion posture video shooting technology and function fitting. This method has the advantages of non-contact, high efficiency, high precision, and wide applicability, which can greatly reduce the cost and labor intensity of manual inspection, improve the safe operation level of railways, and provide a scientific basis for guiding engineering design, maintenance and repair, safety assessment and other work.

[0066] Example 4

[0067] This embodiment 4 provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the above-mentioned supervised learning-based three-element diagnosis method of the influence of basic deformation on ballastless track is implemented. The method includes:

[0068] Acquire video data of the motion posture of the ballastless track to be identified;

[0069] The motion posture video data of the ballastless track to be identified is input into the trained ternary diagnosis model of the influence of foundation deformation on the ballastless track for processing, and the results of identifying the influence of foundation deformation on the ballastless track are obtained and an early warning is issued; wherein, after the early warning information is obtained later, some samples with classic labels as early warnings are added to the training set to retrain the ternary diagnosis model of the influence of foundation deformation on the ballastless track, so as to subsequently refine the hierarchical identification label results; wherein, the motion posture video sub-model, the vibration posture function sub-model, and the vibration oscillate function sub-model of the ternary diagnosis model of the influence of foundation deformation on the ballastless track are trained by multiple motion posture video samples, function samples, and safety labels. When and only when the three sub-models are all judged to be safe, the ternary diagnosis model of the influence of foundation deformation on the ballastless track outputs the final "safety label".

[0070] Example 5

[0071] This embodiment 5 provides a computer device, including a memory and a processor, the processor and the memory communicate with each other, the memory stores program instructions executable by the processor, the processor calls the program instructions to execute the above-mentioned supervised learning-based basic deformation on ballastless track ternary diagnosis method, the method comprising:

[0072] Acquire video data of the motion posture of the ballastless track to be identified;

[0073] The motion posture video data of the ballastless track to be identified is input into the trained ternary diagnosis model of the influence of foundation deformation on the ballastless track for processing, and the results of identifying the influence of foundation deformation on the ballastless track are obtained and an early warning is issued; wherein, after the early warning information is obtained later, some samples with classic labels as early warnings are added to the training set to retrain the ternary diagnosis model of the influence of foundation deformation on the ballastless track, so as to subsequently refine the hierarchical identification label results; wherein, the motion posture video sub-model, the vibration posture function sub-model, and the vibration oscillate function sub-model of the ternary diagnosis model of the influence of foundation deformation on the ballastless track are trained by multiple motion posture video samples, function samples, and safety labels. When and only when the three sub-models are all judged to be safe, the ternary diagnosis model of the influence of foundation deformation on the ballastless track outputs the final "safety label".

[0074] Example 6

[0075] This embodiment 6 provides an electronic device, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes instructions for implementing the above-mentioned three-element diagnosis method of the influence of basic deformation on ballastless track based on supervised learning, the method comprising:

[0076] Acquire video data of the motion posture of the ballastless track to be identified;

[0077] The motion posture video data of the ballastless track to be identified is input into the trained ternary diagnosis model of the influence of foundation deformation on the ballastless track for processing, and the results of identifying the influence of foundation deformation on the ballastless track are obtained and an early warning is issued; wherein, after the early warning information is obtained later, some samples with classic labels as early warnings are added to the training set to retrain the ternary diagnosis model of the influence of foundation deformation on the ballastless track, so as to subsequently refine the hierarchical identification label results; wherein, the motion posture video sub-model, the vibration posture function sub-model, and the vibration oscillate function sub-model of the ternary diagnosis model of the influence of foundation deformation on the ballastless track are trained by multiple motion posture video samples, function samples, and safety labels. When and only when the three sub-models are all judged to be safe, the ternary diagnosis model of the influence of foundation deformation on the ballastless track outputs the final "safety label".

[0078] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0079] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0080] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0081] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0082] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative work on the basis of the technical solution disclosed in the present invention should be included in the scope of protection of the present invention.

Claims

1. A ternary diagnosis method for the influence of foundation deformation on ballastless track based on supervised learning, characterized in that: include: Acquire video data of the motion posture of the ballastless track to be identified; The motion posture video data of the ballastless track to be identified is input into the trained ternary diagnosis model of the influence of foundation deformation on the ballastless track for processing, and the results of identifying the influence of foundation deformation on the ballastless track are obtained and an early warning is issued; wherein, after the early warning information is obtained later, some samples with classic labels as early warnings are added to the training set to retrain the ternary diagnosis model of the influence of foundation deformation on the ballastless track, so as to subsequently refine the hierarchical identification label results; wherein, the motion posture video sub-model, the vibration posture function sub-model, and the vibration oscillate function sub-model of the ternary diagnosis model of the influence of foundation deformation on the ballastless track are trained by multiple motion posture video samples, function samples, and safety labels. When and only when the three sub-models are all judged to be safe, the ternary diagnosis model of the influence of foundation deformation on the ballastless track outputs the final "safety label".

2. According to the supervised learning-based ternary diagnosis method for the influence of foundation deformation on ballastless track in claim 1, it is characterized in that: The process of establishing the posture function is as follows: the intersection points of the neutral axis and the symmetry axis of each section of the rail and ballastless track slab are obtained from the motion posture video to form a curve, and then a preliminary posture function is obtained by fitting in the established spatial Cartesian coordinate system, denoted as posture = f(x, y, z); after expanding the y direction, the final posture function is obtained, denoted as posture = p(x, e (y) ,z); the oscillate function is established by replacing the time parameter t in the motion posture video with posture = p(x,e (y) ,z) function, we get oscillate = g(t,e (y) ,z).

3. The ternary diagnosis method for the influence of foundation deformation on ballastless track based on supervised learning according to claim 2 is characterized in that: Shoot multiple sets of motion posture videos of ballastless track structures under normal geological conditions under driving conditions and upload them. (y) ,z)、oscillate=g(t,e (y) ,z) is entered into the training set as a benchmark sample with the label of safe for learning the ternary diagnosis model of the influence of foundation deformation on ballastless track.

4. The ternary diagnosis method for the influence of foundation deformation on ballastless track based on supervised learning according to claim 1 is characterized in that: After the diagnostic warning information is output, the foundation deformation status of the target section is manually investigated. If it is found that the foundation deformation has reached the operational safety threshold or is of great significance, it will be regarded as a classic sample and placed in the training set to facilitate the subsequent clustering process, that is, the status of the foundation deformation of the target section can be refined and graded for identification and diagnosis; wherein, the important significance includes but is not limited to: locomotive operation safety, comfort threshold, cracks in the ballastless track structure, or damage and failure of the track structure.

5. The ternary diagnosis method for the influence of foundation deformation on ballastless track based on supervised learning according to claim 1 is characterized in that: The training method of the ternary diagnosis model of the influence of foundation deformation on ballastless track is supervised learning, including but not limited to a neural network algorithm, a random forest algorithm or a support vector machine algorithm.

6. A ternary diagnosis system for the influence of foundation deformation on ballastless track based on supervised learning, characterized in that: include: An acquisition module, used for acquiring video data of the motion posture of the ballastless track to be identified; The processing module is used to input the motion posture video data of the ballastless track to be identified into the trained ternary diagnosis model of the influence of the foundation deformation on the ballastless track for processing, obtain the results of identifying the influence of the foundation deformation on the ballastless track and issue an early warning; wherein, after the early warning information is obtained later, some samples with classic labels as early warnings are added to the training set to retrain the ternary diagnosis model of the influence of the foundation deformation on the ballastless track, so as to subsequently refine the hierarchical identification label results; wherein, the motion posture video sub-model, the vibration posture function sub-model, and the vibration oscillate function sub-model of the ternary diagnosis model of the influence of the foundation deformation on the ballastless track are trained by multiple motion posture video samples, function samples, and safety labels. When and only when the three sub-models are all determined to be safe, the ternary diagnosis model of the influence of the foundation deformation on the ballastless track outputs the final "safety label".

7. The supervised learning-based ternary diagnosis system for the influence of foundation deformation on ballastless track according to claim 6 is characterized in that: The acquisition module includes: a group of non-parallel high-definition cameras and processing modules are pre-installed near the measured section, which can obtain the motion posture video of the complete section; the high-definition camera is powered separately, and when the power of the high-definition camera is lower than 5%, the power alarm is triggered and a reminder to replace the power supply is issued; the processing module can realize the pre-processing and uploading functions after the motion posture video is shot; when the high-definition camera captures the vibration of the track, it triggers the start of shooting, and when the high-definition camera captures the stop of the vibration of the track, it triggers the end of shooting, thereby obtaining a single motion posture video sample.

8. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by the processor, the ternary diagnosis method of the influence of basic deformation on ballastless track based on supervised learning is implemented as described in any one of claims 1 to 5.

9. A computer device, characterized in that: It includes a memory and a processor, the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the ternary diagnosis method for the influence of basic deformation on ballastless track based on supervised learning as described in any one of claims 1 to 5.

10. An electronic device, characterized in that: include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes instructions for implementing the ternary diagnostic method for the influence of basic deformation on ballastless track based on supervised learning as described in any one of claims 1 to 5.

Citation Information

Patent Citations

  • Sub-health diagnosis system and method for railway vehicle door system

    CN114510776A

  • High-speed railway track structure deformation damage monitoring system and method

    CN114537466A

  • Bearing fault intelligent diagnosis model, diagnosis method and system

    CN115420502A

  • Railway track detection method and system and railway track detection vehicle

    CN117789136A

  • Method of constructing three-dimensional model of terrain along railway track bed

    RU2726256C1