Sleeper health diagnosis method, device and electronic equipment

Abstract

The present invention provides a sleeper health diagnosis method, device and electronic equipment, which belongs to the field of fault diagnosis technology. The method includes: using multiple sensors pre-set on the sleeper to obtain monitoring data of the sleeper to construct a complete monitoring data matrix; performing a random missing operation on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix; inputting the missing monitoring data matrix into a neural network regression model to output a complete monitoring data estimation matrix; and performing health diagnosis on the sleeper based on the difference between the complete monitoring data estimation matrix and each row of monitoring data of the complete monitoring data matrix. The present invention does not require feature selection, extraction and feature preprocessing, and only uses the correlation between multi-channel monitoring data to diagnose sleeper health. In addition, the method can flexibly diagnose all channel monitoring data at the same time, and has the advantages of flexibility, reliability and high diagnostic efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to a sleeper health diagnosis method, device and electronic equipment. Background Art

[0002] As railways age, under the influence of long-term vehicle loads and environmental factors, sleeper structures inevitably develop various types of defects such as cracks and breakage, affecting track smoothness and even threatening driving safety. Therefore, to ensure the normal and safe operation of the track, accurate safety assessment of the sleeper structure is crucial.

[0003] At present, the commonly used method for sleeper structural health diagnosis is statistical pattern recognition method. The evaluation effect of this type of method depends largely on the quality of feature extraction, that is, whether the extracted structural service performance characteristics can maximize the representation of the information contained in the structural health monitoring data. Feature extraction often requires professional knowledge and complex and sophisticated data preprocessing process.

[0004] Therefore, the existing sleeper structure health diagnosis scheme relies too much on feature extraction and preprocessing, which makes the health diagnosis process too complicated and the diagnosis efficiency low. Summary of the Invention

[0005] The present invention provides a sleeper health diagnosis method, device and electronic equipment to solve the defects of the prior art that the health diagnosis process is too complicated and the diagnosis efficiency is low.

[0006] In a first aspect, the present invention provides a method for diagnosing the health of a sleeper, comprising: obtaining monitoring data of the sleeper using a plurality of sensors pre-set on the sleeper to construct a complete monitoring data matrix; each row of monitoring data in the complete monitoring data matrix corresponds one-to-one to the sensor; performing a random missing operation on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix; inputting the missing monitoring data matrix into a neural network regression model to output a complete monitoring data estimation matrix; the neural network regression model is obtained by training a complete monitoring data matrix training sample set and a missing monitoring data matrix training sample set of the sleeper in a healthy state; and performing a health diagnosis on the sleeper based on the difference between the complete monitoring data estimation matrix and the monitoring data in each row of the complete monitoring data matrix.

[0007] According to a rail sleeper health diagnosis method provided by the present invention, a random missing operation is performed on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix, including: randomly selecting at least one row of the complete monitoring data matrix; replacing all the monitoring data of the selected at least one row with 0 to construct the missing monitoring data matrix.

[0008] According to a rail sleeper health diagnosis method provided by the present invention, the health diagnosis of the rail sleeper is performed based on the difference between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix, including: calculating the error vector between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix; determining the health index corresponding to each sensor based on the error vector of each row of monitoring data; and performing health diagnosis on the rail sleeper based on the health index corresponding to each sensor and the preset health index threshold corresponding to each sensor.

[0009] According to a sleeper health diagnosis method provided by the present invention, the health indicator corresponding to each sensor is determined based on the error vector of each row of monitoring data, specifically:

[0010]

[0011] Among them, e i represents the health indicator corresponding to the i-th sensor, represents the i-th row of the complete monitoring data estimation matrix, Z[i,:] represents the i-th row of the complete monitoring data matrix, L is the number of columns of the complete monitoring data matrix, || || 2 It means to find the two norm.

[0012] According to a rail sleeper health diagnosis method provided by the present invention, a health diagnosis of the rail sleeper is performed based on the health indicator corresponding to each sensor and the preset health indicator threshold corresponding to each sensor, including: when the health indicator corresponding to each sensor is less than or equal to the preset health indicator threshold corresponding to each sensor, determining that the rail sleeper is in a healthy state; otherwise, determining that the rail sleeper is in an abnormal state.

[0013] According to a rail sleeper health diagnosis method provided by the present invention, after determining that the rail sleeper is in an abnormal state, it also includes: determining a target sensor whose health index is greater than a preset health index threshold; and locating the abnormal position of the rail sleeper according to the target sensor.

[0014] According to a rail sleeper health diagnosis method provided by the present invention, before performing health diagnosis on the rail sleeper based on the health indicator corresponding to each sensor and the preset health indicator threshold corresponding to each sensor, the method further includes: obtaining a complete monitoring data matrix verification sample set corresponding to each sensor; using the complete monitoring data matrix verification sample set to obtain a health indicator output set corresponding to each sensor according to the neural network regression model; and using the maximum health indicator in the health indicator output set as the preset health indicator threshold.

[0015] According to a rail sleeper health diagnosis method provided by the present invention, the sensor is a strain sensor; and the neural network regression model is a long short-term memory artificial neural network model.

[0016] In a second aspect, the present invention further provides a sleeper health diagnosis device, comprising: an acquisition module, a missing matrix construction module, a regression prediction module and a health diagnosis module.

[0017] The acquisition module is used to acquire monitoring data of the sleeper using a plurality of sensors pre-installed on the sleeper to construct a complete monitoring data matrix; each row of monitoring data in the complete monitoring data matrix corresponds one-to-one to the sensor;

[0018] A missing matrix construction module is used to perform a random missing operation on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix;

[0019] A regression prediction module is used to input the missing monitoring data matrix into a neural network regression model and output a complete monitoring data estimation matrix; the neural network regression model is trained using a complete monitoring data matrix training sample set and a missing monitoring data matrix training sample set of the sleeper in a healthy state;

[0020] A health diagnosis module is used to perform health diagnosis on the sleeper according to the difference between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix.

[0021] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the above-described sleeper health diagnosis methods are implemented.

[0022] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described sleeper health diagnosis methods.

[0023] The sleeper health diagnosis method, device and electronic equipment provided by the present invention do not require feature selection, extraction and preprocessing (for example, dimensionality reduction processing), and only use the correlation between multi-channel monitoring data to diagnose the sleeper health. In addition, this method can flexibly diagnose all channel monitoring data at the same time, and has the advantages of flexibility, reliability and high diagnostic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 This is one of the flow charts of the sleeper health diagnosis method provided by the present invention;

[0026] Figure 2 This is the second flow chart of the sleeper health diagnosis method provided by the present invention;

[0027] Figure 3 It is a structural schematic diagram of the sleeper health diagnosis device provided by the present invention;

[0028] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0029] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0030] It should be noted that, in the description of the embodiments of the present invention, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the presence of other identical elements in the process, method, article or device comprising the elements. The orientation or positional relationship indicated by the terms "upper", "lower", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0031] The following combination Figures 1-4 The sleeper health diagnosis method and device provided in the embodiments of the present invention are described.

[0032] Figure 1 This is one of the flow charts of the sleeper health diagnosis method provided by the present invention, such as Figure 1 As shown, including but not limited to the following steps:

[0033] Step 101: Utilize a plurality of sensors pre-installed on the sleepers to obtain monitoring data of the sleepers to construct a complete monitoring data matrix.

[0034] The sensors may be strain sensors. Each sensor can record sleeper monitoring data when a train passes over the sleepers. Optionally, each sensor has the same sampling frequency, and each row of monitoring data in the complete monitoring data matrix corresponds one-to-one with the sensor. The number of rows in the complete monitoring data matrix corresponds to the number of sensors, and the number of columns in the complete monitoring data matrix corresponds to the number of monitoring data points.

[0035] Step 102: Perform a random missing operation on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix.

[0036] Optionally, the random missing operation may be to cause partial data in the complete monitoring data matrix to be missing (channel missing) to simulate a data missing situation.

[0037] The matrix generated after the random missing operation can be referred to as the missing monitoring data matrix in the present invention.

[0038] Step 103: Input the missing monitoring data matrix into a neural network regression model and output a complete monitoring data estimation matrix.

[0039] The neural network regression model is obtained by training using a complete monitoring data matrix training sample set and a missing monitoring data matrix training sample set of the sleeper in a healthy state.

[0040] It can be understood that the complete monitoring data matrix training sample set is a set of complete monitoring data matrix samples; the missing monitoring data matrix training sample set is a set of missing monitoring data matrix samples; wherein the missing monitoring data matrix samples correspond one-to-one to the complete monitoring data matrix samples.

[0041] It should be noted that the specific type and structure of the neural network regression model can be selected as needed with the goal of achieving the technical solution of the present invention, and there are no excessive restrictions on it.

[0042] Optionally, the present invention adopts a long short-term memory artificial neural network model (Long Short-Term Memory, LSTM) as the neural network regression model.

[0043] Furthermore, the present invention can employ unsupervised training to train the LSTM network model. The loss function can employ a mean square error function, and the parameter optimization method can employ the Adam optimization algorithm. Convergence of the LSTM network model training indicates that the network has the ability to recover the remaining channel monitoring data based on partial channel monitoring data under the sleeper health condition, i.e., regression prediction capability.

[0044] Optionally, the time expansion step of the LSTM network model can be automatically determined based on the length of the input sequence. In addition, to enhance the approximation (regression prediction) capability of the LSTM network model, the number of network layers can be appropriately increased.

[0045] Step 104: Perform health diagnosis on the sleeper according to the difference between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix.

[0046] Optionally, the present invention can characterize the differences in the monitoring data in each row through a certain statistical parameter (such as residual or mean square error, etc., which can be selected according to actual needs). When the above statistical parameter is greater than a preset threshold, it is determined that the sleeper is in a fault state.

[0047] The sleeper health diagnosis method provided by the present invention does not require feature selection, extraction and preprocessing (for example, dimensionality reduction processing), and only uses the correlation between multi-channel monitoring data to diagnose the sleeper health. In addition, this method can flexibly diagnose all channel monitoring data at the same time, and has the advantages of flexibility, reliability and high diagnostic efficiency.

[0048] Based on the contents of the above embodiments, as an optional embodiment, the sleeper health diagnosis method provided by the present invention performs a random missing operation on the complete monitoring data matrix to obtain the corresponding missing monitoring data matrix, including: randomly selecting at least one row of the complete monitoring data matrix; replacing all the monitoring data of the selected at least one row with 0 to construct the missing monitoring data matrix.

[0049] Specifically, let the complete monitoring data matrix be Z, and the number of rows of matrix Z be M; the random missing operation can be:

[0050] (1) Generate a random integer k between 1 and M-1;

[0051] (2) Randomly select k rows of elements (i.e., monitoring data) from Z and set them to 0 to form a missing monitoring data matrix, which is set as Z.

[0052] The random missing operation in the present invention can set a row or multiple rows in the complete monitoring data matrix to zero, so that the channel is missing to generate a missing monitoring data matrix.

[0053] Based on the contents of the above embodiments, as an optional embodiment, the present invention provides a sleeper health diagnosis method, which performs health diagnosis on the sleeper based on the difference between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix, including: calculating the error vector between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix; determining the health indicator corresponding to each sensor based on the error vector of each row of monitoring data; and performing health diagnosis on the sleeper based on the health indicator corresponding to each sensor and the preset health indicator threshold corresponding to each sensor.

[0054] Figure 2 This is the second flow chart of the sleeper health diagnosis method provided by the present invention. Figure 2 As well as the above embodiments, the implementation process of the present invention is further explained.

[0055] Optionally, the health indicator corresponding to each sensor is determined based on the error vector of each row of monitoring data, specifically:

[0056]

[0057] Among them, e i represents the health indicator corresponding to the i-th sensor, represents the i-th row of the complete monitoring data estimation matrix, Z[i,:] represents the i-th row of the complete monitoring data matrix, L is the number of columns of the complete monitoring data matrix, || || 2 It means to find the two norm.

[0058] It is understandable that That is the error vector of the monitoring data corresponding to sensor i.

[0059] As an optional embodiment, a health diagnosis of the sleeper is performed based on the health indicator corresponding to each sensor and the preset health indicator threshold corresponding to each sensor, including: when the health indicator corresponding to each sensor is less than or equal to the preset health indicator threshold corresponding to each sensor, determining that the sleeper is in a healthy state; otherwise, determining that the sleeper is in an abnormal state.

[0060] Specifically, if This means that the local position monitored by sensor i is normal. If all sensors monitor the local position normally, then If the sleeper is in good condition, then the sleeper is in good condition. Otherwise, it indicates that the sleeper is in abnormal condition and needs to be repaired in time. is the preset health indicator threshold of sensor i.

[0061] Furthermore, after determining that the sleeper is in an abnormal state, the sleeper health diagnosis method provided by the present invention also includes: determining a target sensor whose health index is greater than a preset health index threshold; and locating the abnormal position of the sleeper based on the target sensor.

[0062] It can be understood that the sleeper health diagnosis method provided by the present invention can not only determine whether the sleeper is in a healthy state, but also further locate the fault position of the sleeper according to the installation position of the target sensor, so as to facilitate timely inspection and maintenance.

[0063] Based on the contents of the above embodiments, as an optional embodiment, the sleeper health diagnosis method provided by the present invention, before performing health diagnosis on the sleeper according to the health indicator corresponding to each sensor and the preset health indicator threshold corresponding to each sensor, further includes: obtaining a complete monitoring data matrix verification sample set corresponding to each sensor; using the complete monitoring data matrix verification sample set, according to the neural network regression model, obtaining a health indicator output set corresponding to each sensor; and using the maximum health indicator in the health indicator output set as the preset health indicator threshold.

[0064] Specifically, the implementation process of the embodiment of the present invention is as follows:

[0065] (1) Collect sufficient monitoring data from each strain sensor when a train passes through the steel tube concrete sleeper in a healthy state, and construct multiple complete monitoring data matrices corresponding to each strain sensor to form a complete monitoring data matrix verification sample set.

[0066] (2) The present invention can use the complete monitoring data matrix to verify the complete monitoring data matrix in the sample set, and obtain the corresponding complete monitoring data estimation matrix based on the neural network regression model; further, obtain the health indicator output set corresponding to each sensor.

[0067] The implementation method is the same as the process of obtaining health indicators in the above embodiment. The specific process can be referred to the content of the above embodiment and will not be repeated here. The generated health indicator output set is in the following form:

[0068]

[0069] Among them, E i represents the health indicator output set corresponding to sensor i, Represents the n health indicators corresponding to the n complete monitoring data matrices of sensor i.

[0070] (3) Optionally, the maximum value of the health indicator in the health indicator output set is taken as the preset health indicator threshold. The specific form is as follows:

[0071]

[0072] in, Indicates the preset health indicator threshold corresponding to sensor i.

[0073] In summary, the present invention performs sequence-to-sequence regression analysis based on the LSTM network; utilizes the correlation and information redundancy between the monitoring data of each strain sensor to restore the remaining sensor monitoring data through partial sensor monitoring data; finally, the health status of the sleeper is judged by the mean square error of the restored residual, which has the advantages of flexibility, reliability and high diagnostic efficiency.

[0074] Figure 3 This is a schematic diagram of the structure of the sleeper health diagnosis device provided by the present invention. Figure 3 As shown, the device includes: an acquisition module 301, a missing matrix construction module 302, a regression prediction module 303 and a health diagnosis module 304.

[0075] The acquisition module 301 is used to acquire monitoring data of the sleeper using multiple sensors pre-installed on the sleeper to construct a complete monitoring data matrix;

[0076] Each row of monitoring data in the complete monitoring data matrix corresponds one-to-one to the sensor.

[0077] The missing matrix construction module 302 is used to perform a random missing operation on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix.

[0078] The regression prediction module 303 is used to input the missing monitoring data matrix into the neural network regression model and output the complete monitoring data estimation matrix.

[0079] The neural network regression model is obtained by using a complete monitoring data matrix training sample set and a missing monitoring data matrix training sample set of the sleeper in a healthy state.

[0080] The health diagnosis module 304 is configured to perform health diagnosis on the sleeper according to the difference between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix.

[0081] It should be noted that the sleeper health diagnosis device provided in the embodiment of the present invention can execute the sleeper health diagnosis method described in any of the above embodiments during specific operation, which will not be described in detail in this embodiment.

[0082] Figure 4 Schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 4As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other via the communications bus 440. The processor 410 may call logic instructions in the memory 430 to execute a sleeper health diagnosis method, which includes: using multiple sensors pre-installed on the sleeper to obtain monitoring data of the sleeper to construct a complete monitoring data matrix; each row of monitoring data in the complete monitoring data matrix corresponds one-to-one with the sensor; performing a random missing operation on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix; inputting the missing monitoring data matrix into a neural network regression model to output a complete monitoring data estimation matrix; the neural network regression model is trained using a complete monitoring data matrix training sample set and a missing monitoring data matrix training sample set of the sleeper in a healthy state; and performing a health diagnosis on the sleeper based on the difference between the complete monitoring data estimation matrix and the monitoring data in each row of the complete monitoring data matrix.

[0083] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0084] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the rail sleeper health diagnosis method provided by the above-mentioned embodiments, the method including: using multiple sensors pre-set on the rail sleeper to obtain monitoring data of the rail sleeper to construct a complete monitoring data matrix; each row of monitoring data in the complete monitoring data matrix corresponds one-to-one with the sensor; performing a random missing operation on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix; inputting the missing monitoring data matrix into a neural network regression model to output a complete monitoring data estimation matrix; the neural network regression model is obtained by training using a complete monitoring data matrix training sample set and a missing monitoring data matrix training sample set of the rail sleeper in a healthy state; performing a health diagnosis on the rail sleeper based on the difference between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix.

[0085] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the sleeper health diagnosis method provided in the above-mentioned embodiments, the method comprising: using a plurality of sensors pre-arranged on the sleeper to obtain monitoring data of the sleeper to construct a complete monitoring data matrix; each row of monitoring data in the complete monitoring data matrix corresponds one-to-one with the sensor; performing a random missing operation on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix; inputting the missing monitoring data matrix into a neural network regression model to output a complete monitoring data estimation matrix; the neural network regression model is obtained by training using a complete monitoring data matrix training sample set and a missing monitoring data matrix training sample set of the sleeper in a healthy state; performing a health diagnosis on the sleeper based on the difference between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix.

[0086] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0087] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A sleeper health diagnosis method, characterized in that: include: Using multiple sensors pre-installed on the sleepers, monitoring data of the sleepers is acquired to construct a complete monitoring data matrix; Each row of monitoring data in the complete monitoring data matrix corresponds one-to-one to the sensor; Performing a random missing operation on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix; Inputting the missing monitoring data matrix into a neural network regression model and outputting a complete monitoring data estimation matrix; the neural network regression model is trained using a complete monitoring data matrix training sample set and a missing monitoring data matrix training sample set of the sleeper in a healthy state; Performing health diagnosis on the sleeper according to the difference between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix; The step of performing health diagnosis on the sleeper according to the difference between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix comprises: Calculating an error vector between the complete monitoring data estimation matrix and each row of monitoring data of the complete monitoring data matrix; Determine the health indicator corresponding to each sensor based on the error vector of each row of monitoring data; Performing health diagnosis on the sleeper according to the health indicator corresponding to each sensor and the preset health indicator threshold value corresponding to each sensor; The health diagnosis of the sleeper according to the health indicator corresponding to each sensor and the preset health indicator threshold corresponding to each sensor includes: If the health indicator corresponding to each sensor is less than or equal to the preset health indicator threshold corresponding to each sensor, it is determined that the sleeper is in a healthy state; otherwise, determining that the sleeper is in an abnormal state; After determining that the sleeper is in an abnormal state, the method further includes: Determine a target sensor whose health indicator is greater than a preset health indicator threshold; The abnormal position of the sleeper is located according to the target sensor.

2. The sleeper health diagnosis method according to claim 1, characterized in that: Performing a random missing operation on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix includes: Randomly selecting at least one row of the complete monitoring data matrix; The monitoring data of at least one selected row are all replaced with 0 to construct the missing monitoring data matrix.

3. The sleeper health diagnosis method according to claim 1, characterized in that: The health indicator corresponding to each sensor is determined based on the error vector of each row of monitoring data, specifically: Among them, e i represents the health indicator corresponding to the i-th sensor, represents the i-th row of the complete monitoring data estimation matrix, Z[i,:] represents the i-th row of the complete monitoring data matrix, L is the number of columns of the complete monitoring data matrix, ‖‖ 2 It means to find the two norm.

4. The sleeper health diagnosis method according to claim 1, characterized in that: Before performing health diagnosis on the sleeper according to the health indicator corresponding to each sensor and the preset health indicator threshold corresponding to each sensor, the method further includes: Obtain the complete monitoring data matrix verification sample set corresponding to each sensor; Using the complete monitoring data matrix to verify the sample set, and according to the neural network regression model, obtaining the health indicator output set corresponding to each sensor; The maximum health indicator in the health indicator output set is used as the preset health indicator threshold.

5. The sleeper health diagnosis method according to claim 1, characterized in that: The sensor is a strain sensor; the neural network regression model is a long short-term memory artificial neural network model.

6. A sleeper health diagnosis device, characterized in that: A method for diagnosing sleeper health according to any one of claims 1 to 5, comprising: An acquisition module is used to acquire monitoring data of the sleeper using a plurality of sensors pre-installed on the sleeper to construct a complete monitoring data matrix; each row of monitoring data in the complete monitoring data matrix corresponds one-to-one to the sensor; A missing matrix construction module is used to perform a random missing operation on the complete monitoring data matrix to obtain a corresponding missing monitoring data matrix; A regression prediction module is used to input the missing monitoring data matrix into a neural network regression model and output a complete monitoring data estimation matrix; the neural network regression model is trained using a complete monitoring data matrix training sample set and a missing monitoring data matrix training sample set of the sleeper in a healthy state; A health diagnosis module is used to perform health diagnosis on the sleeper according to the difference between the complete monitoring data estimation matrix and the monitoring data of each row of the complete monitoring data matrix.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the sleeper health diagnosis method according to any one of claims 1 to 5 are implemented.

Images