Rail damage early warning method and device, computer equipment and storage medium

Abstract

The invention relates to a rail damage early warning method and device, computer equipment and a storage medium, and the method comprises the steps: carrying out the preprocessing under the condition that an initial ultrasonic echo signal of a to-be-detected rail is collected, and obtaining a to-be-detected ultrasonic echo signal; inputting the ultrasonic echo signal to be detected into a pre-trained track injury identification model to obtain an injury identification result output by the model; the injury identification result comprises injury positions, injury types of the injury positions and injury probabilities of the injury types; determining damage early warning information of each damage position based on the damage identification result, sending the damage early warning information to track damage early warning equipment, and performing sound and picture early warning on the damage condition of each damage position based on the damage early warning equipment; the sound and picture early warning comprises the step of displaying a target injury identifier of each injury position in a simulation picture, a target injury type of each injury position, an injury probability of each target injury type and a played target injury early warning audio in the simulation picture of the track to be detected; the early warning efficiency and effect can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a track damage early warning method and device, computer equipment and storage medium. BACKGROUND

[0002] In the field of railway transportation, the health status of the steel rail is crucial to ensure the safety of railway operation. In the related art, the common rail flaw detection method mainly uses a small instrument with a sliding shoe to detect flaws. The flaw detection worker pushes the small instrument on the rail and relies on manual identification of the B-scan ultrasonography graph to find the internal damage of the rail.

[0003] However, this manual damage identification method has many drawbacks. Since the flaw detection result is greatly affected by the flaw detection level of the flaw detection personnel, different experience and skill level personnel may miss or misjudge the damage, resulting in that some internal diseases of the rail cannot be found in time, which brings potential risks to the safety of railway transportation.

[0004] Therefore, there is an urgent need for a method that can overcome the shortcomings of the prior art, analyze rail damage diseases while detecting, and timely and effectively warn related damage diseases. SUMMARY

[0005] Therefore, it is necessary to provide a track damage early warning method and device that can improve the analysis and early warning accuracy of rail damage diseases, computer equipment, computer readable storage medium and computer program product.

[0006] In a first aspect, the present application provides a track damage early warning method, which comprises:

[0007] In the case of collecting the initial ultrasonic echo signal of the to-be-tested track, the initial ultrasonic echo signal is preprocessed to obtain a to-be-tested ultrasonic echo signal;

[0008] The to-be-tested ultrasonic echo signal is input into a pre-trained track damage identification model to obtain a damage identification result output by the track damage identification model; the damage identification result includes damage positions in the to-be-tested track, at least one damage type corresponding to each damage position, and a damage probability corresponding to each damage type;

[0009] Based on the damage identification result, damage early warning information of each damage position is determined, and the damage early warning information is sent to a track damage early warning device to perform audio and video early warning on the damage of each damage position in the to-be-tested track based on the damage early warning device;

[0010] The audio-visual warning includes the following displayed in the track damage warning device, which displays a simulated image of the track to be tested: the target damage identifier corresponding to each damage location in the simulated image, the target damage type corresponding to each damage location, the damage probability corresponding to each target damage type, and the target damage warning audio played by the track damage warning device.

[0011] In one embodiment, after obtaining the damage identification result output by the track damage identification model, the method further includes:

[0012] For each of the damage locations, the damage probability corresponding to each damage type at that damage location is compared with a preset damage probability threshold for each damage type;

[0013] If the comparison result indicates that the probability of damage corresponding to the damage type is equal to or greater than the corresponding damage probability threshold, the damage type is determined as the target damage type for the relevant damage location.

[0014] In one embodiment, the method further includes:

[0015] For each target damage type at each damage location, the damage probability of each target damage type is compared with the damage probability of each preset warning level. Based on the comparison results, the warning intensity of each target damage type is determined, and the warning intensity with the highest intensity is determined as the first warning intensity of the damage location.

[0016] For each of the aforementioned damage locations, the number of the target damage types is compared with the preset range of each warning level, and a second warning intensity for the damage location is determined based on the comparison result;

[0017] The target warning intensity for the damaged location is determined based on the highest of the first and second warning intensities at the damaged location.

[0018] In one embodiment, the method further includes:

[0019] Based on the target warning intensity of each damage location, a target damage warning audio is determined for each damage location; wherein, for different target warning intensities, different target damage warning audios are pre-set to match the intensity ranges; different target damage warning audios can be set to have different audio types and / or different audio volumes.

[0020] In one embodiment, the method further includes:

[0021] Based on the target warning intensity at each of the damage locations, a target damage identifier suitable for each of the damage locations is determined; wherein, for different target warning intensities, different target damage identifiers are pre-set to match different intensity ranges; different target damage identifiers can be set to have different identifier types and / or different identifier colors.

[0022] In one embodiment, the preprocessing of the initial ultrasonic echo signal to obtain the ultrasonic echo signal to be measured includes:

[0023] The initial ultrasonic echo signal is linearly normalized to obtain the ultrasonic echo signal to be measured.

[0024] The initial ultrasonic echo signal is obtained by analog-to-digital conversion of the acquired analog ultrasonic echo signal of the track to be tested based on a preset quantization accuracy. The initial ultrasonic echo signal is a digital ultrasonic echo signal.

[0025] In one embodiment, the method further includes:

[0026] The initial ultrasonic echo signals of multiple samples in the training sample feature set of the track are used to train the initial track damage identification model, thus obtaining a pre-trained track damage identification model.

[0027] Secondly, this application also provides a track damage early warning device, the device comprising:

[0028] The data preprocessing module is used to preprocess the initial ultrasonic echo signal of the track to be tested to obtain the ultrasonic echo signal to be tested after the initial ultrasonic echo signal is acquired.

[0029] The model processing module is used to input the ultrasonic echo signal to be tested into a pre-trained track damage recognition model to obtain the damage recognition result output by the track damage recognition model; the damage recognition result includes the damage location in the track to be tested, at least one damage type corresponding to each damage location, and the damage probability corresponding to each damage type;

[0030] The damage warning module is used to determine the damage warning information of each of the damage locations based on the damage identification results, and send the damage warning information to the track damage warning device, so as to provide audio and visual warnings on the damage status of each of the damage locations in the track under test based on the damage warning device.

[0031] The audio-visual warning includes the following displayed in the track damage warning device, which displays a simulated image of the track to be tested: the target damage identifier corresponding to each damage location in the simulated image, the target damage type corresponding to each damage location, the damage probability corresponding to each target damage type, and the target damage warning audio played by the track damage warning device.

[0032] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the first aspect.

[0033] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the first aspect.

[0034] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps in the first aspect.

[0035] The track damage early warning method, device, computer equipment, computer-readable storage medium, and computer program product provided in this application, wherein the track damage early warning method preprocesses the initial ultrasonic echo signal of the track to be tested after acquiring the initial ultrasonic echo signal, thereby obtaining the ultrasonic echo signal to be tested, which helps to ensure the data stability of the ultrasonic echo signal to be tested after preprocessing, and also helps to speed up the subsequent data processing of the track damage identification model; the ultrasonic echo signal to be tested is input into a pre-trained track damage identification model to obtain the damage identification result output by the track damage identification model; the damage identification result includes the damage location in the track to be tested, at least one damage type corresponding to each damage location, and the damage probability corresponding to each damage type; the pre-trained track damage identification model determines the damage identification result of the track to be tested based on the ultrasonic echo signal to be tested, which helps to improve the accuracy and efficiency of obtaining the damage identification result; based on damage identification... The results determine the damage warning information for each damaged location and send the warning information to the track damage warning device. Based on this, the device provides audio-visual warnings regarding the damage at each location on the track under test. The audio-visual warnings include: the target damage identifier corresponding to each damaged location in the simulated image of the track under test, the target damage type for each damaged location, the damage probability for each target damage type, and the target damage warning audio played by the track damage warning device. In other words, the damage warning device can display the damage status of each damaged location on the track under test to the user, while simultaneously providing audio warnings. This allows the user to be informed of the damage on the track under test through the device and to promptly identify and address any damage issues, thereby improving track safety. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is an application environment diagram of a track damage early warning method in one embodiment;

[0038] Figure 2 This is a flowchart illustrating a track damage early warning method in one embodiment;

[0039] Figure 3This is a flowchart illustrating the sub-steps of a track damage early warning method in one embodiment;

[0040] Figure 4 This is a structural block diagram of a track damage early warning device in one embodiment;

[0041] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0043] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0044] As described in the background section, related technologies often rely on human experience to identify the B-mode ultrasound images of the track in order to determine the extent of damage in the track. However, personnel with different levels of experience and skill may make mistakes or omissions when identifying damage, which may lead to the failure to detect some internal defects of the rails in a timely manner, posing potential risks to railway transportation safety.

[0045] With the development of computer technology, advanced technologies such as deep learning have demonstrated powerful data analysis and recognition capabilities in many fields. However, in the field of rail flaw detection, there is currently no complete and efficient method that can detect and analyze damage and defects in real time and issue timely alarms during the flaw detection process.

[0046] Therefore, this application provides a method for early warning of track damage, aiming to provide an online identification method for rail damage defects through simultaneous detection and analysis. Specifically, after acquiring the initial ultrasonic echo signal of the track to be tested, the initial ultrasonic echo signal is preprocessed to obtain the ultrasonic echo signal to be tested; the ultrasonic echo signal to be tested is input into a pre-trained track damage identification model to obtain the damage identification result output by the track damage identification model; the damage identification result includes the location of the damage in the track to be tested, at least one damage type corresponding to each damage location, and the damage type corresponding to each damage type. Damage probability; based on the damage identification results, damage warning information for each damage location is determined and sent to the track damage warning equipment. The equipment then provides audio-visual warnings based on the damage location at each damage location on the track under test. These audio-visual warnings include: the target damage identifier corresponding to each damage location in the simulated image of the track under test, the target damage type corresponding to each damage location, the damage probability corresponding to each target damage type, and the target damage warning audio played by the track damage warning equipment. This method aims to solve the problems of missed or misjudged damage in existing manual rail flaw detection methods, as well as the inability to analyze defects online in real time. By combining a rail flaw detector with edge computing technology, it enables real-time defect analysis of rail flaw detection data during the flaw detection process and timely alarms to the flaw detector operator when defects are detected, improving the accuracy and efficiency of rail flaw detection and ensuring railway traffic safety.

[0047] The track damage early warning method provided in this application embodiment can be applied to, for example, Figure 1 In the application environment shown, both the rail flaw detector 101 and the terminal 102 communicate with the server 103 via a network. The data storage system can store the data that the server 103 needs to process. The data storage system can be integrated onto the server 103 or placed on the cloud or other network servers. For example, the data storage system can store the initial ultrasonic echo signals collected, processed, and uploaded by the rail flaw detector 101, pre-trained track damage identification models, etc. The rail flaw detector 101 is an instrument that can move on the track and perform ultrasound imaging of the track; it can also have a display screen for touch operation by the flaw detector operator and display of ultrasound images. The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc. The server 103 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0048] In one exemplary embodiment, such as Figure 2 As shown, a method for early warning of track damage is provided, which can be applied to... Figure 1Taking server 103 as an example, the explanation includes the following steps 201 to 203. Wherein:

[0049] Step 201: After acquiring the initial ultrasonic echo signal of the track to be tested, preprocess the initial ultrasonic echo signal to obtain the ultrasonic echo signal to be tested.

[0050] The initial ultrasonic echo signal can originate from a rail flaw detector, specifically, it can be collected by a small, slip-on rail flaw detector performing flaw detection operations along the rail. The movement of the slip-on rail flaw detector on the rail can be computer-controlled or manually controlled by the flaw detector operator; this application does not impose specific limitations on either method. The slip-on rail flaw detector may be equipped with a display screen, which is used at least for the real-time display of the signal data collected during the rail flaw detection operation.

[0051] The initial ultrasonic echo signal is, for example, a digital signal (digital ultrasonic echo signal); the purpose of preprocessing may be to distribute the data related to the initial ultrasonic echo signal in a specific range, so as to ensure the effectiveness and efficiency of subsequent processing of the initial ultrasonic echo signal using a relevant track damage identification model.

[0052] For example, when the server acquires the initial ultrasonic echo signal transmitted by the rail flaw detector running on the track to be tested, it can further preprocess the initial ultrasonic echo signal so that the data related to the preprocessed initial ultrasonic echo signal can be distributed in a specific range, so as to obtain the test ultrasonic echo signal corresponding to the preprocessed initial ultrasonic echo signal.

[0053] Step 202: Input the ultrasonic echo signal to be tested into the pre-trained track damage identification model to obtain the damage identification result output by the track damage identification model; the damage identification result includes the damage location in the track to be tested, at least one damage type corresponding to each damage location, and the damage probability corresponding to each damage type.

[0054] Among them, the pre-trained track damage recognition model is the track damage recognition model that has been trained in advance; the track damage recognition model can receive the ultrasonic echo signal corresponding to the track under test and output the damage recognition result of the track under test corresponding to the ultrasonic echo signal under test.

[0055] For example, when the server obtains the ultrasonic echo signal corresponding to the track to be tested, it can input the ultrasonic echo signal to be tested into a pre-trained track damage identification model, so as to calculate and output the damage identification result of the track to be tested corresponding to the ultrasonic echo signal through the track damage identification model.

[0056] The damage identification results include at least the location of the damage in the track under test, at least one type of damage corresponding to each damage location, and the probability of damage corresponding to each type of damage. The damage location refers to the specific position of the damage in the track under test, facilitating subsequent confirmation and repair of the damage by the flaw detector. Damage types may include, for example, crack types, inclusion types, porosity types, etc. This application does not limit the specific sub-categories of damage types and can adjust the range of damage types based on requirements. A damage location may include one or more damage types. The damage type is not limited to a single damage location being identifiable as only one type of damage. The damage probability is the probability of each damage type at a given damage location. For example, (damage location: point A; damage type: crack; damage probability: 80%) indicates that the probability of a crack type damage at point A in the track under test is 80%. Similarly, (damage location: point B; damage type: crack, damage probability: 40%; damage type: porosity, damage probability: 80%) indicates that the probability of a crack type damage at point B in the track under test is 40%, and the probability of a porosity type damage is 80%.

[0057] Step 203: Based on the damage identification results, determine the damage warning information for each damage location and send the damage warning information to the track damage warning device. The damage warning device will then provide audio-visual warnings based on the damage status of each damage location in the track under test. The audio-visual warnings include the following displayed on the track damage warning device, which shows a simulated image of the track under test: the target damage identifier corresponding to each damage location in the simulated image, the target damage type corresponding to each damage location, the damage probability corresponding to each target damage type, and the target damage warning audio played by the track damage warning device.

[0058] Among them, track damage early warning equipment includes portable devices such as mobile phones and tablets used by railway flaw detectors, which can have the functions of displaying images and playing sound.

[0059] The simulated image of the track to be tested can be a three-dimensional simulation image. Correspondingly, the coordinate information of each damaged location can also be set as three-dimensional information to ensure the accuracy of the position display of the damaged location in the simulated image of the track to be tested. In addition, if the simulated image of the track to be tested is a two-dimensional image, the coordinate information of the damaged location can also be set as two-dimensional information. This application does not make specific limitations on this.

[0060] The damage warning information for each damage location can be determined based on the damage identification results, and the damage warning information can be obtained after data compression processing of the damage identification results. For example, if the damage warning result is "Damage location: Point B; Damage type: Crack type, Damage probability: 40%; Damage type: Porosity type, Damage probability: 80%", and the preset setting is that subsequent damage warnings will only be issued when the damage probability exceeds 60%, then the damage warning information corresponding to the damage warning result "Damage location: Point B; Damage type: Crack type, Damage probability: 40%; Damage type: Porosity type, Damage probability: 80%" will be determined as "Damage location: Point B; Damage type: Porosity type, Damage probability: 80%". This ensures that when issuing a damage warning for point B, no data, images, or sounds related to the damage type being crack will be displayed. Therefore, damage warning information can be understood as the determination of relevant damage information that is more suitable for subsequent warning in the damage identification results based on a pre-set warning determination threshold.

[0061] Another alternative implementation is that the damage warning information can also be set to further filter the damage identification results to obtain the filtering results (such as "damage location: point B; damage type: pore type, damage probability: 80%" as mentioned in the previous paragraph). This involves further determining the display method and warning audio of the damage situation at point B in the track damage warning device. For example, it would determine how point B is marked in the simulated image of the track under test—whether to use a square or a circle, the color of the selection, or a specific pattern. It would also determine how the pore type is displayed in the simulated image of the track under test—whether using text or combining the marking style or color of point B. Furthermore, it would determine how the damage probability is displayed in the simulated image of the track under test—whether using text or combining the marking style or color of point B. Finally, it would also determine the sound and volume of the warning audio for point B.

[0062] For example, after obtaining the damage identification results corresponding to the track to be tested, the server will further determine the damage warning information for each damage location based on the damage identification results, and then send the determined damage warning information to the track damage warning device. The track damage warning device will then display the damage situation of the track to be tested in the simulated screen, based on the simulated screen of the track to be tested. Specifically, the screen can display the target damage identifier corresponding to each damage location in the simulated screen, the target damage type corresponding to each damage location, the damage probability corresponding to each target damage type, and other information. At the same time, the track damage warning device will play the target damage warning audio.

[0063] The track damage early warning method provided in this application preprocesses the initial ultrasonic echo signal of the track under test after acquisition, thereby obtaining the ultrasonic echo signal to be tested. This helps ensure the data stability of the preprocessed ultrasonic echo signal and accelerates the subsequent data processing using a track damage identification model. The ultrasonic echo signal to be tested is then input into a pre-trained track damage identification model to obtain the damage identification result output by the model. The damage identification result includes the location of the damage in the track under test, at least one damage type corresponding to each damage location, and the damage probability corresponding to each damage type. The pre-trained track damage identification model determines the damage identification result of the track under test based on the ultrasonic echo signal, improving the accuracy and efficiency of the damage identification result. Based on the damage identification result, damage early warning information for each damage location is determined, and... Damage warning information is sent to the track damage warning device, which then provides audio and visual warnings based on the damage status at each damaged location on the track under test. The audio and visual warnings include: the target damage identifier corresponding to each damaged location in the simulated image of the track under test, the target damage type corresponding to each damaged location, the damage probability corresponding to each target damage type, and the target damage warning audio played by the track damage warning device. In other words, the damage warning device can display the damage status at each damaged location on the track under test to the user, while simultaneously providing audio warnings. This allows the user to be informed of the damage on the track under test through the damage warning device and to promptly identify and address any damage issues, thereby improving track safety.

[0064] In an exemplary embodiment, after obtaining the damage identification result output by the track damage identification model, the track damage early warning method provided in this application further includes steps 301 and 302, wherein:

[0065] Step 301: For each injury location, compare the injury probability corresponding to each injury type at that injury location with the preset injury probability threshold for each injury type.

[0066] Step 302: If the damage probability corresponding to the damage type characterized by the comparison result is equal to or greater than the corresponding damage probability threshold, the damage type is determined as the target damage type of the relevant damage location.

[0067] Each damage type can have a pre-set damage probability threshold. For example, the damage probability threshold for cracks is 70%, for inclusions it is 50%, and for pores it is 60%. The pre-set damage probability threshold for each damage type can be set based on the hazard level of each damage type to the track under test. The hazard level of each damage type to the track under test can be determined by technical experts in the relevant field based on historical damage data. Correspondingly, a warning will only be issued for crack damage at a certain damage location if the damage probability of cracks at that location is equal to or greater than 70%; a warning will only be issued for inclusion damage at a certain damage location if the damage probability of inclusions at that location is equal to or greater than 50%; and a warning will only be issued for pore damage at a certain damage location if the damage probability of pores at that location is equal to or greater than 60%. Alternatively, you can choose to have the same injury probability threshold for all injury types, for example, set to 60%. In this way, regardless of the type of injury, if the corresponding injury probability is equal to or greater than 60%, a warning will be issued for the relevant injury type at that injury location.

[0068] For example, after obtaining the damage identification results of the track to be tested, the server can further compare the damage probability corresponding to each damage type at each damage location with the preset damage probability threshold for each damage type to obtain the probability comparison result corresponding to each damage type. If the comparison result indicates that the damage probability corresponding to any damage type is equal to or greater than the corresponding damage probability threshold, the relevant damage type is determined as the target damage type of the relevant damage location. The target damage type of a damage location can be one or more, and this application does not specifically limit this.

[0069] When there is only one target damage type at the damage location, the simulation screen of the track under test of the track damage early warning device will only display data related to one damage type at that damage location. When there are multiple target damage types at the damage location, the simulation screen of the track under test of the track damage early warning device can choose to display all data related to all damage types, or it can choose to display at least two data related to the damage types with the highest damage probability values ​​corresponding to the target damage types.

[0070] In this embodiment, for each damaged location, the probability of the damage type included in each damaged location is compared with a pre-set damage probability threshold for the relevant damage type. Based on the comparison result, it is determined whether the damage type included in each damaged location should be displayed in a subsequent warning. Specifically, damage types with a damage probability equal to or greater than the corresponding damage probability threshold are identified as the target damage type for the relevant damaged location. In this way, the target damage type to be warned can be further determined, which can improve the accuracy of the warning and avoid the waste of human and maintenance resources caused by false alarms.

[0071] In an exemplary embodiment, the track damage early warning method provided in this application further includes: for each target damage type at each damage location, comparing the damage probability of each target damage type with the damage probability of each preset early warning level, determining the early warning intensity of each target damage type based on the comparison result, and determining the highest early warning intensity as the first early warning intensity of the damage location; for each damage location, comparing the number of target damage types with the number range of each preset early warning level, and determining the second early warning intensity of the damage location based on the comparison result; and determining the target early warning intensity of the damage location based on the highest intensity of the first and second early warning intensities.

[0072] For example, after determining the target damage type for each damage location based on steps 301 and 302, the server can further compare the damage probability of each target damage type with the damage probability of each preset warning level to determine the warning intensity for each target damage type. For example, the preset warning levels include Level 1, Level 2, and Level 3. It can be preset that when the damage probability is equal to or greater than 90%, the corresponding warning level is Level 1 (highest level); when the damage probability is equal to or greater than 80% and less than 90%, the corresponding warning level is Level 2 (intermediate level); and when the damage probability is equal to or greater than 50% and less than 80%, the corresponding warning level is Level 3 (lowest level). Correspondingly, when the damage probability is less than 50%, no further warning display is provided for the relevant damage type.

[0073] If a damage location has only one target damage type and its corresponding warning level is determined, then the warning intensity corresponding to that warning level is the same as the warning intensity corresponding to that damage location.

[0074] In a damage location, if at least two target damage types have been identified and their corresponding warning levels have been determined, the warning intensities corresponding to these at least two warning levels can be compared, and the one with the highest warning intensity can be determined as the first warning intensity for that damage location.

[0075] Since the number of target damage types at a single damage location may not be limited to just one, the server can determine the number of target damage types at each damage location and then compare this number with the preset range of each warning level to obtain the corresponding warning level. Based on this warning level, a second warning intensity for each damage location can then be determined. For example, if the preset warning levels include Level 1, Level 2, and Level 3, the warning level can be preset as follows: Level 1 (highest level) if the number of target damage types at a damage location is equal to or greater than 3; Level 2 (intermediate level) if the number of target damage types at a damage location is equal to or greater than 2 but less than 3; and Level 3 (lowest level) if the number of target damage types at a damage location is equal to or greater than 1 but less than 2.

[0076] Furthermore, if the first warning intensity and the second warning intensity are determined simultaneously at a damaged location using the two methods described above, the highest of the two warning intensities can be used to determine the target warning intensity for the relevant damaged location.

[0077] It should be added that there can be a preset correlation between the warning level and the warning intensity, and this application does not make specific limitations on this. For example, it can be set that the warning intensity is the highest intensity when the warning level is level 1, the warning intensity is the medium intensity when the warning level is level 2, and the warning intensity is the lowest intensity when the warning level is level 3; for example, it can be set that the warning intensity is high intensity when the warning level is level 1, and the warning intensity is normal intensity when the warning level is level 2 or 3.

[0078] The display of different warning intensities in the virtual image of the track under test can vary. For example, the highest warning intensity is indicated by red, the medium intensity by yellow, and the lowest intensity by green; conversely, a high intensity warning is indicated by red, and a normal intensity warning by yellow. This example uses color to differentiate warning intensities, but it is not limited to this. For instance, different styles of indicator frames or different types of warning audio can also be used to differentiate warning intensities.

[0079] In this embodiment, the first warning intensity of each damaged location is determined by the damage probability of each target damage type at each damaged location, and / or the second warning intensity is determined by the number of target damage types at each damaged location. Then, the target warning intensity of the damaged location is further determined based on the highest of the first and second warning intensities. This improves the accuracy of the determined target warning intensity for each damaged location and enhances the warning effect on each damaged location in the subsequent virtual image of the track under test. It also allows flaw detectors to quickly obtain the warning intensity of each damaged location, enabling them to quickly determine the urgency of repairs at multiple damaged locations, thus improving their work efficiency and effectiveness.

[0080] In an exemplary embodiment, the track damage warning method provided in this application further includes: determining a target damage warning audio suitable for each damage location based on the target warning intensity of each damage location; wherein, different target warning intensities correspond to different intensity ranges, and different target damage warning audios are pre-set to match them; the different target damage warning audios can be set to have different audio types and / or different audio volumes.

[0081] For example, after determining the target warning intensity at each damage location in the track under test, the server can further determine the corresponding target damage warning audio based on the target warning intensity at each damage location. For instance, when the target warning intensity is the first warning intensity, a more urgent audio type can be selected as the target damage warning audio; when the target warning intensity is the second warning intensity, a more soothing audio type can be selected as the target damage warning audio. For example, when the target warning intensity is the first warning intensity, the audio volume can be set to 30, and when the target warning intensity is the second warning intensity, the audio volume can be set to 10. Furthermore, different target warning intensities can also result in different audio volumes based on the different audio types.

[0082] In this embodiment, the target damage warning audio of different target warning intensities is determined by audio type and / or audio volume. This helps flaw detectors to know the severity of the damage when they hear the relevant warning audio. In the case of multiple damage locations, the priority of repairs at each damage location can be quickly determined, which helps to improve the work efficiency and effectiveness of flaw detectors.

[0083] In one exemplary embodiment, for multiple damage locations within a preset range, an audio signal corresponding to the highest warning intensity can be determined as the target damage warning audio signal based on the target warning intensity corresponding to each damage location. This avoids the impact of playing multiple audio signals within a small range on the damage assessment by the flaw detector. For example, along the track extension direction, only one target damage warning audio signal can be determined for each 100 meters.

[0084] In an exemplary embodiment, the track damage early warning method provided in this application further includes: determining a target damage identifier suitable for each damage location based on the target early warning intensity of each damage location; wherein, different target early warning intensities correspond to different intensity ranges, and different target damage identifiers are pre-set to match them; different target damage identifiers can be set to have different identifier types and / or different identifier colors.

[0085] For example, after determining the target warning intensity for each damaged location, the server can further determine the target damage marker for each damaged location based on the target warning intensity, so as to obtain the final display method of each damaged location in the virtual screen; for example, for different target warning intensities, at least one of different marker types (marker styles) and marker colors can be set; so that flaw detectors can determine the priority of the repair status of each damaged location by looking at the target damage markers of each damaged location displayed in the track damage warning device, which is conducive to improving the work efficiency and effectiveness of flaw detectors.

[0086] In an exemplary embodiment, the preprocessing of the initial ultrasonic echo signal to obtain the ultrasonic echo signal to be tested in the above steps can be specifically performed as follows: linear normalization processing is performed on the initial ultrasonic echo signal to obtain the ultrasonic echo signal to be tested; wherein, the initial ultrasonic echo signal is obtained by analog-to-digital conversion based on a preset quantization accuracy of the acquired analog ultrasonic echo signal of the track to be tested, and the initial ultrasonic echo signal is a digital ultrasonic echo signal.

[0087] The preset quantization precision can be set in the range of 1-1.176V / BIT.

[0088] In an exemplary embodiment, the track damage early warning method provided in this application further includes: the server can train an initial track damage identification model based on multiple initial ultrasonic echo signals from a training sample feature set of the track, to obtain a pre-trained track damage identification model. The training sample feature set may include training samples and validation samples.

[0089] Regarding the track damage early warning method provided in this application, an alternative implementation method is also provided, including the following data acquisition steps, data transmission steps, edge computing analysis steps, and alarm prompting steps.

[0090] (1) Data acquisition steps:

[0091] A small, slip-on rail flaw detector is used to perform flaw detection along the rail. During the flaw detection process, the small rail flaw detector performs ultrasonic testing on the interior of the rail at a fixed sampling frequency (f) (unit: Hz), acquiring raw ultrasonic echo signal data (S(t)) containing information about the internal structure of the rail, where (t) represents the time variable (unit: s). This raw ultrasonic echo signal data (S(t)) is a continuously changing analog signal over time, and its mathematical expression can be expressed as: ;

[0092] Where (A(t)) is the amplitude function that varies with time, reflecting the intensity change of the ultrasonic echo; w is the angular frequency (w=2πf), a fixed parameter determined by the sampling frequency; and (φ(t)) is the phase function that varies with time, reflecting the phase characteristics of the ultrasonic echo signal. The analog signal (S(t)) is converted into a digital signal (D[n]) according to the preset quantization accuracy (q) (unit: V / bit) using the built-in analog-to-digital converter (ADC) of the flaw detection instrument, where (n) represents the number of discrete sampling points. The conversion relationship can be expressed by the following formula:

[0093] ;

[0094] in, , The sampling interval is denoted by 'round'. The 'round' function represents the rounding operation, which rounds the quantized value to the nearest integer, thereby obtaining discrete digital ultrasound echo signal data (D[n]) for subsequent analysis and processing.

[0095] (2) Data transmission steps:

[0096] The acquired digital ultrasound echo signal data (D[n]), i.e., the initial ultrasound echo signal, is transmitted in real time to the edge computing box via a wireless communication module (such as Wi-Fi, Bluetooth, etc.). Let the baud rate of the wireless transmission be (B) (unit: bit / s), and the time required to transmit the data (D[n]) of one sampling point be... It can be calculated using the following formula:

[0097] ;

[0098] Here, bitwidth(D[n]) represents the number of bits occupied by the digital signal (D[n]), which depends on factors such as quantization precision (q) and the bit depth of the analog-to-digital converter. To ensure the real-time performance and integrity of data transmission, a check and retransmission mechanism, such as Cyclic Redundancy Check (CRC), is used during transmission. If the received data is found to be erroneous after check, a retransmission operation is triggered until the correct data is received, thus ensuring the accuracy of the data in subsequent edge computing analysis.

[0099] (3) Edge computing analysis steps:

[0100] A rail damage and defect identification model is pre-built and trained in an edge computing box. This model is based on deep learning algorithms, such as a convolutional neural network (CNN) structure, which includes an input layer, multiple hidden layers (convolutional layers, pooling layers, etc.), and an output layer. The received digital ultrasonic echo signal data (D[n]) is preprocessed. First, it is normalized (preprocessing) to distribute the data within a specific interval, for example, by using linear normalization to convert the data (D[n]) into (D... norm [n]), its calculation formula is as follows: ;

[0101] Where D min and D max These are the minimum and maximum values ​​in the data (D[n]), respectively.

[0102] Then the normalized data (D) norm [n]), that is, the ultrasonic echo signal to be measured, is input into the rail damage identification model (rail damage identification model). The model extracts features and classifies the data through internal parameters such as convolution kernels, and outputs a damage identification result (R), that is, the damage identification result, which can be represented as a probability vector. For example, if there are (m) types of damage (such as cracks, inclusions, pores, etc.), then (R=[r1,r2,…,r... m ]), where (r i ) represents the probability of belonging to the (i)th disease (0≤r). i ≤1), and satisfy Among them, the disease recognition result can be optionally set to at least reflect the disease category and the probability of the disease category.

[0103] By setting a disease determination threshold T (0 < T < 1), when there is a certain (r j ≥ T), it is determined that the (j)-th type of disease exists at the current detection position of the rail, realizing the identification and judgment of rail damage diseases.

[0104] (4) Alarm prompt step:

[0105] When the edge computing analysis concludes that there is a rail damage disease, that is, when the disease determination condition is met, the edge computing box sends an alarm signal through the alarm device (such as an audible and visual alarm) connected to it. The alarm intensity (I) of the alarm device can be set in levels according to the severity of the disease. For example, it is set that the severity of the disease is measured by an index (S) ((S) can be comprehensively calculated based on factors such as disease type, probability, etc.). There is a mapping relationship between the alarm intensity (I) and the severity index (S), such as using a linear mapping relationship: ; where (k) and (b) are pre-set constants. In this way, the flaw detector can intuitively understand the approximate severity of the disease according to the alarm intensity and promptly confirm and process the disease location. Throughout the process, the above steps are carried out synchronously and in real time. That is, as the flaw detection instrument moves along the rail for flaw detection (flaw detection while moving), operations such as data collection, transmission, edge computing analysis, and alarm prompt are carried out sequentially and continuously, realizing the online identification function of rail damage diseases while detecting and analyzing. Once a disease is found, it can immediately remind the flaw detector, effectively improving the timeliness and accuracy of rail flaw detection. Exemplarily, the alarm method can be optionally selected to adopt the suspected disease box selection method, that is, when a suspected disease is found, it is boxed, the disease category is displayed, graded (minor injury, severe injury), the probability is shown, and an alarm sound is emitted. The graded alarm can specifically be a combination of text display and different voice prompts.

[0106] The rail damage early warning method provided by this application is equivalent to providing a method for online identification of rail damage diseases with on-the-go detection and analysis, aiming to solve the problems of easy missed judgment, misjudgment, and inability to perform real-time online analysis of diseases in existing rail flaw detection. By combining the rail flaw detector with edge computing technology, it realizes real-time disease analysis of rail flaw detection data during the flaw detection process, and promptly alarms and reminds the flaw detector when a disease is found, improving the accuracy and efficiency of rail flaw detection and ensuring railway operation safety.

[0107] Furthermore, the beneficial effects of the rail damage early warning method provided by this application include:

[0108] Improved flaw detection accuracy: By using a damage and defect identification model built with deep learning technology for automatic analysis, the omissions and misjudgments caused by differences in the skill level of flaw detectors during manual identification are avoided, which significantly improves the accuracy of identifying rail damage and defects.

[0109] Real-time online analysis and alarm: During the flaw detection process, data can be collected, transmitted and analyzed in real time. Once a defect is detected, an alarm will be triggered immediately, so that the flaw detection worker can know the defect situation in time and take corresponding measures in time, thus ensuring the safety of railway operation.

[0110] Highly efficient and convenient: The entire process is highly automated. The flaw detector operator only needs to push the small flaw detector instrument normally to carry out the flaw detection work. No additional complicated operations are required. The mode of simultaneous detection and analysis improves the efficiency of flaw detection work and reduces labor and time costs.

[0111] Furthermore, this application also provides a specific embodiment of a track damage early warning method, including the following five stages: equipment preparation, data acquisition, data transmission, edge computing analysis, and alarm notification.

[0112] Equipment preparation:

[0113] Select a suitable slip-on rail flaw detector, with its ultrasonic probe frequency set to 5MHz, i.e., the sampling frequency (f=5×10). 6 The instrument features a built-in 12-bit analog-to-digital converter (ADC) with a quantization accuracy of q=0.1V / bit. It is equipped with an edge computing box, which houses a rail damage identification model based on a convolutional neural network structure built using the TensorFlow framework. This model has been thoroughly trained using a large amount of labeled rail flaw detection data. Three types of defects are defined (cracks, inclusions, and porosity), with a defect judgment threshold of T=0.8. Simultaneously, a Wi-Fi-enabled wireless module with a baud rate of B=115200bit / s is connected between the flaw detection instrument and the edge computing box for data transmission. The alarm device uses an audible and visual alarm, with a linear mapping relationship between alarm intensity and defect severity (k=10, b=5).

[0114] Data collection:

[0115] The flaw detector operator pushes a small rail flaw detector along a section of rail to be inspected at a constant speed, following the normal flaw detection procedure. The ultrasonic probe of the flaw detector continuously emits ultrasonic waves into the rail and receives the ultrasonic echoes, collecting raw ultrasonic echo signal data (S(t)) according to the aforementioned sampling frequency (f). For example, at a certain moment (t=0.001s), the amplitude of the collected raw ultrasonic echo signal is (A(0.001)=5V), and the phase is (φ(0.001)=π / 4). Then, according to the expression for (S(t)): The original ultrasonic echo signal value at this time is calculated. Then, it is converted into a digital signal (D[n]) by an analog-to-digital converter, and the corresponding discrete sampling point values ​​are calculated according to the quantization formula.

[0116] Data transmission:

[0117] The acquired digital ultrasound echo signal data (D[n]) is transmitted to the edge computing box via a Wi-Fi module. Assuming that the data (D[n]) from one sampling point occupies 12 bits, the transmission time is calculated using the formula... During transmission, a CRC check is performed on each set of data transmitted. If an error is found during the check, the data is immediately retransmitted to ensure that the data is transmitted to the edge computing box accurately.

[0118] Edge computing analysis:

[0119] After receiving the digital ultrasonic echo signal data (D[n]), the edge computing box first performs normalization processing on it to find the minimum value (D) in this batch of data. min =0) (assuming the minimum value after quantization is 0), maximum value (D) max =2047) (For 12-bit data, the maximum value is (2 12 -1), and the data of each sampling point is transformed according to the normalization formula to obtain (D) norm [n]) is then input into the rail damage and defect identification model. The model extracts features through internal convolution, pooling and other operations and outputs a probability vector (R=[r1,r2,r3]), for example, the output result is (R=[0.1,0.85,0.05]). Since (r2=0.85≥0.8) (the defect judgment threshold (T)), it is determined that there is a mixed defect at the current rail detection position.

[0120] Alarm notification:

[0121] After the edge computing box determines the presence of defects, it calculates the alarm intensity based on the severity index (severity index set to 0.8 here, considering defect type and probability) using a linear mapping relationship (I=10×0.8+5=13). The audible and visual alarm then emits an alarm signal of corresponding intensity (I=13) to alert the flaw detector operator that there are embedded defects in the rail. The operator can then mark the location and conduct further inspection. Through this process of simultaneous detection and analysis, the above steps are repeated throughout the rail flaw detection operation to achieve comprehensive, accurate, and real-time online identification and alarm for rail damage and defects.

[0122] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0123] Based on the same inventive concept, this application also provides a track damage early warning device for implementing the track damage early warning method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more track damage early warning device embodiments provided below can be found in the limitations of the track damage early warning method described above, and will not be repeated here.

[0124] In one exemplary embodiment, such as Figure 4 As shown, a track damage early warning device 400 is provided, including: a data preprocessing module 41, a model processing module 42, and a damage early warning module 43, wherein:

[0125] The data preprocessing module 41 is used to preprocess the initial ultrasonic echo signal of the track to be tested after the initial ultrasonic echo signal is acquired, so as to obtain the ultrasonic echo signal to be tested.

[0126] The model processing module 42 is used to input the ultrasonic echo signal to be tested into the pre-trained track damage recognition model to obtain the damage recognition result output by the track damage recognition model; the damage recognition result includes the damage location in the track to be tested, at least one damage type corresponding to each damage location, and the damage probability corresponding to each damage type;

[0127] The damage warning module 43 is used to determine the damage warning information of each damage location based on the damage identification result, and send the damage warning information to the track damage warning device so as to provide audio and visual warnings based on the damage warning device for each damage location in the track under test.

[0128] The audio-visual warning includes the following displayed in the track damage warning device, which shows a simulated image of the track to be tested: the target damage identifier corresponding to each damage location in the simulated image, the target damage type corresponding to each damage location, the damage probability corresponding to each target damage type, and the target damage warning audio played by the track damage warning device.

[0129] In an exemplary embodiment, after the model processing module 42 obtains the damage identification result output by the track damage identification model, the damage warning module 43 can also be used to compare the damage probability corresponding to each damage type at each damage location with the damage probability threshold preset for each damage type. If the comparison result indicates that the damage probability corresponding to the damage type is equal to or greater than the corresponding damage probability threshold, the damage type is determined as the target damage type of the relevant damage location.

[0130] In an exemplary embodiment, the damage warning module 43 is further configured to: for each target damage type at each damage location, compare the damage probability of each target damage type with the damage probability of each preset warning level, determine the warning intensity of each target damage type based on the comparison result, and determine the highest warning intensity as the first warning intensity of the damage location; for each damage location, compare the number of target damage types with the number range of each preset warning level, and determine the second warning intensity of the damage location based on the comparison result; and determine the target warning intensity of the damage location based on the highest intensity of the first and second warning intensities.

[0131] In an exemplary embodiment, the damage warning module 43 is further configured to: determine the target damage warning audio appropriate for each damage location based on the target warning intensity of each damage location; wherein, the intensity range corresponding to different target warning intensities is pre-set with different matching target damage warning audios; the different target damage warning audios can be set to have different audio types and / or different audio volumes.

[0132] In an exemplary embodiment, the damage warning module 43 is further configured to: determine the target damage identifier suitable for each damage location based on the target warning intensity of each damage location; wherein, the intensity range corresponding to different target warning intensities is pre-set with different matching target damage identifiers; the different target damage identifiers can be set to have different identifier types and / or different identifier colors.

[0133] In an exemplary embodiment, the data preprocessing module 41 is used to preprocess the initial ultrasonic echo signal to obtain the ultrasonic echo signal to be tested. Specifically, it is used to: perform linear normalization processing on the initial ultrasonic echo signal to obtain the ultrasonic echo signal to be tested; wherein, the initial ultrasonic echo signal is obtained by analog-to-digital conversion based on a preset quantization accuracy of the acquired analog ultrasonic echo signal of the track to be tested, and the initial ultrasonic echo signal is a digital ultrasonic echo signal.

[0134] In an exemplary embodiment, the track damage early warning device 400 further includes a model training module for training an initial track damage identification model based on multiple initial ultrasonic echo signals from a training sample feature set of the track, thereby obtaining a pre-trained track damage identification model.

[0135] The various modules in the aforementioned track damage early warning device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0136] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and databases. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for early warning of track damage.

[0137] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0138] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a track damage early warning method.

[0139] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program being executed by a processor to implement the steps of a track damage early warning method.

[0140] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of a track damage early warning method.

[0141] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0142] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0143] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0144] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for early warning of track damage, characterized in that, The method includes: If the initial ultrasonic echo signal of the track to be tested is acquired, the initial ultrasonic echo signal is preprocessed to obtain the ultrasonic echo signal to be tested. The ultrasonic echo signal to be tested is input into a pre-trained track damage identification model to obtain the damage identification result output by the track damage identification model; the damage identification result includes the damage location in the track to be tested, at least one damage type corresponding to each damage location, and the damage probability corresponding to each damage type; Based on the damage identification results, damage warning information for each of the damage locations is determined, and the damage warning information is sent to the track damage warning device so as to provide audio-visual warnings on the damage status of each of the damage locations in the track under test based on the damage warning device. The audio-visual warning includes the following displayed in the track damage warning device, which displays a simulated image of the track to be tested: the target damage identifier corresponding to each damage location in the simulated image, the target damage type corresponding to each damage location, the damage probability corresponding to each target damage type, and the target damage warning audio played by the track damage warning device.

2. The method according to claim 1, characterized in that, After obtaining the damage identification result output by the track damage identification model, the method further includes: For each of the damage locations, the damage probability corresponding to each damage type at that damage location is compared with a preset damage probability threshold for each damage type; If the comparison result indicates that the probability of damage corresponding to the damage type is equal to or greater than the corresponding damage probability threshold, the damage type is determined as the target damage type for the relevant damage location.

3. The method according to claim 2, characterized in that, The method further includes: For each target damage type at each damage location, the damage probability of each target damage type is compared with the damage probability of each preset warning level. Based on the comparison results, the warning intensity of each target damage type is determined, and the warning intensity with the highest intensity is determined as the first warning intensity of the damage location. For each of the aforementioned damage locations, the number of the target damage types is compared with the preset range of each warning level, and a second warning intensity for the damage location is determined based on the comparison result; The target warning intensity for the damaged location is determined based on the highest of the first and second warning intensities at the damaged location.

4. The method according to claim 3, characterized in that, The method further includes: Based on the target warning intensity of each damage location, a target damage warning audio is determined for each damage location; wherein, for different target warning intensities, different target damage warning audios are pre-set to match the intensity ranges; different target damage warning audios can be set to have different audio types and / or different audio volumes.

5. The method according to claim 3, characterized in that, The method further includes: Based on the target warning intensity at each of the damage locations, a target damage identifier suitable for each of the damage locations is determined; wherein, for different target warning intensities, different target damage identifiers are pre-set to match different intensity ranges; different target damage identifiers can be set to have different identifier types and / or different identifier colors.

6. The method according to claim 1, characterized in that, The preprocessing of the initial ultrasonic echo signal to obtain the ultrasonic echo signal to be measured includes: The initial ultrasonic echo signal is linearly normalized to obtain the ultrasonic echo signal to be measured. The initial ultrasonic echo signal is obtained by analog-to-digital conversion of the acquired analog ultrasonic echo signal of the track to be tested based on a preset quantization accuracy. The initial ultrasonic echo signal is a digital ultrasonic echo signal.

7. The method according to any one of claims 1-6, characterized in that, The method further includes: The initial ultrasonic echo signals of multiple samples in the training sample feature set of the track are used to train the initial track damage identification model, thus obtaining a pre-trained track damage identification model.

8. A track damage early warning device, characterized in that, The device includes: The data preprocessing module is used to preprocess the initial ultrasonic echo signal of the track to be tested to obtain the ultrasonic echo signal to be tested after the initial ultrasonic echo signal is acquired. The model processing module is used to input the ultrasonic echo signal to be tested into a pre-trained track damage recognition model to obtain the damage recognition result output by the track damage recognition model; the damage recognition result includes the damage location in the track to be tested, at least one damage type corresponding to each damage location, and the damage probability corresponding to each damage type; The damage warning module is used to determine the damage warning information of each of the damage locations based on the damage identification results, and send the damage warning information to the track damage warning device, so as to provide audio and visual warnings on the damage status of each of the damage locations in the track under test based on the damage warning device. The audio-visual warning includes the following displayed in the track damage warning device, which displays a simulated image of the track to be tested: the target damage identifier corresponding to each damage location in the simulated image, the target damage type corresponding to each damage location, the damage probability corresponding to each target damage type, and the target damage warning audio played by the track damage warning device.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1-7.

Images