Turnout fault detection method, device, electronic equipment and storage medium

Abstract

The present invention provides a turnout fault detection method, device, electronic device and storage medium, the method comprising: obtaining source monitoring data of a turnout to be detected; extracting key point data of the source monitoring data; reconstructing the source monitoring data and the key point data based on a fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result based on the reconstruction error between the reconstructed monitoring data and the source monitoring data; the fault detection model is obtained by training based on positive sample source monitoring data. The method provided by the present invention obtains a fault detection model by training with positive sample source monitoring data, extracts key point data, reconstructs the source monitoring data and the key point data based on the fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data, thereby realizing accurate, reliable and comprehensive turnout fault detection and enhancing the safety of train operation.

Description

Technical Field

[0001] The present invention relates to the field of rail transit safety technology, and in particular to a turnout fault detection method, device, electronic equipment and storage medium. Background Art

[0002] With the rapid development of rail transit systems, ensuring the safe and stable operation of rail equipment has become the core task of rail transit operations. As a key component of rail signal equipment, turnouts are responsible for track switching when trains change tracks, and their normal operation is crucial to the safety of train operation. At present, turnout fault detection is mainly achieved through statistical anomaly detection methods. Most existing statistical anomaly detection methods rely on a large amount of abnormal data for model training.

[0003] However, in practical applications, the abnormal data of the turnout system is relatively scarce and difficult to collect, and the different types of abnormal patterns are complex and diverse, which makes comprehensive modeling and detection based on limited abnormal samples more difficult. This makes the detection method based on abnormal samples limited in practical applications, and the detection accuracy is not high. Summary of the invention

[0004] The present invention provides a turnout fault detection method, device, electronic device and storage medium, which are used to solve the defects in the prior art that the training of the turnout fault detection model is based on abnormal data, resulting in limitations in fault detection and low accuracy of detection results.

[0005] The present invention provides a turnout fault detection method, comprising: Obtain source monitoring data of the turnout to be inspected; Extracting key point data of the source monitoring data; Reconstructing the source monitoring data and the key point data based on the fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data; The fault detection model is trained based on positive sample source monitoring data.

[0006] According to a turnout fault detection method provided by the present invention, the key point data includes at least one of: source monitoring data start and end values, source monitoring data extreme values, and conversion cycle duration.

[0007] According to a turnout fault detection method provided by the present invention, the source monitoring data and the key point data are reconstructed based on a fault detection model to obtain reconstructed monitoring data, including: Reconstructing a data curve of the source monitoring data based on the fault detection model to obtain reconstructed curve data; Reconstructing the key point data individually based on the fault detection model to obtain reconstructed key data; The reconstructed monitoring data is obtained based on the reconstructed curve data and / or the reconstructed key data.

[0008] According to a turnout fault detection method provided by the present invention, the fault detection result of the turnout to be detected is determined based on the reconstruction error between the reconstructed monitoring data and the source monitoring data, comprising: constructing a curve reconstruction error based on the reconstructed curve data and the source monitoring data; constructing a key point reconstruction error based on the reconstructed key data and key point data in the source monitoring data; The curve reconstruction error and the key point reconstruction error are weightedly calculated to obtain the reconstruction error, so as to determine the fault detection result based on the reconstruction error.

[0009] According to a turnout fault detection method provided by the present invention, the source monitoring data is reconstructed based on a fault detection model to obtain reconstructed monitoring data, and the fault detection result of the turnout to be detected is determined based on a reconstruction error between the reconstructed monitoring data and the source monitoring data, and further includes: Performing threshold detection on the key point data to obtain a key point detection result; In the case where there is a fault in the key point detection result, obtaining the fault detection result based on the key point detection result; When there is no fault in the key point detection result, the source monitoring data and the key point data are reconstructed based on the fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data.

[0010] According to a turnout fault detection method provided by the present invention, the source monitoring data is obtained based on acquiring the conversion force exerted on the turnout to be detected.

[0011] The present invention also provides a turnout fault detection device, comprising: An acquisition unit, for acquiring source monitoring data of the turnout to be detected; An extraction unit, extracting key point data of the source monitoring data; A fault detection unit, which reconstructs the source monitoring data and the key point data based on a fault detection model to obtain reconstructed monitoring data, so as to determine a fault detection result of the turnout to be detected based on a reconstruction error between the reconstructed monitoring data and the source monitoring data; The fault detection model is trained based on positive sample source monitoring data.

[0012] The present invention also 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, any of the above-mentioned turnout fault detection methods is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the turnout fault detection method described in any one of the above methods is implemented.

[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the turnout fault detection method described above is implemented.

[0015] The turnout fault detection method, device, electronic device and storage medium provided by the present invention obtain a fault detection model through forward sample source monitoring data training, reconstruct the source monitoring data based on the fault detection model to obtain reconstructed monitoring data, and determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data, thereby realizing accurate, reliable and comprehensive turnout fault detection and enhancing the safety of train operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces 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 creative work.

[0017] Figure 1 It is one of the flow charts of the turnout fault detection method provided by the present invention; Figure 2 is a schematic diagram of a conversion force curve provided by the present invention; Figure 3 This is the second flow chart of the turnout fault detection method provided by the present invention; Figure 4 This is the third flow chart of the turnout fault detection method provided by the present invention; Figure 5 It is a structural schematic diagram of a turnout fault detection device provided by the present invention; Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] It should be noted that due to the complex mechanical structure, high switching frequency and long-term outdoor operation of the turnout system, it is susceptible to the combined effects of external environmental influences and internal mechanical wear, which in turn leads to hidden faults. If these abnormal conditions cannot be discovered and handled in a timely manner, they may cause train operation failures and even endanger personal safety. Therefore, achieving real-time monitoring of turnout status and efficient fault diagnosis has become an important task in the maintenance of rail transit systems. At present, the fault detection technology of rail turnouts mainly relies on manual inspections and electrical signal monitoring methods. For example, current curve monitoring judges the status of the turnout by analyzing the current changes during the turnout switching process, which has been applied to a certain extent in the existing technology.

[0020] However, the existing detection methods have the following major shortcomings: First, manual inspections have certain limitations, as follows: Manual inspections are usually carried out regularly, and real-time monitoring of turnouts cannot be achieved. In addition, since manual inspections are time-consuming and costly, the long inspection cycle may result in early turnout failures not being discovered in time, posing potential safety hazards. Therefore, the current automated detection method based on the current curve is usually used to detect turnout faults. However, most of the current automated detection methods rely on a large amount of abnormal data for model training. In actual applications, the abnormal data of the turnout system is relatively scarce and difficult to collect, and the different types of abnormal patterns are complex and diverse, which makes comprehensive modeling and detection based on limited abnormal samples more difficult, which in turn makes the traditional detection method based on abnormal samples limited in actual applications.

[0021] In view of the above problems, the present invention provides a turnout fault detection method to achieve comprehensive and accurate turnout fault detection. Figure 1 It is one of the flow charts of the turnout fault detection method provided by the present invention, such as Figure 1 As shown, the method includes: Step 110, obtaining source monitoring data of the turnout to be detected; Here, source monitoring data refers to data used to reflect the actual operating state of mechanical components in the turnout, which can be current data or conversion force data when the mechanical components in the turnout to be detected are working, or other data that directly or indirectly reflects the working state of the turnout components. In the embodiment of the present invention, conversion force data that can directly and accurately reflect the mechanical state of the turnout to be detected is preferably used as source monitoring data.

[0022] Specifically, a conversion resistance sensor can be set on the components of the turnout to be tested, and the conversion force data of the turnout to be tested in the conversion process can be collected in real time through the sensor. The entire conversion process of the turnout to be tested from the initial position, such as the normally closed or normally open position, to the final position is recorded to generate a complete conversion force curve.

[0023] It should be noted that the conversion force on the turnout to be tested corresponding to different times can be obtained by analyzing the collected conversion force curve, thereby reflecting the mechanical operating state of the turnout, and then the conversion force curve can be used as an analysis basis for analyzing whether the turnout to be tested has a fault. Figure 2 Schematic diagram of the conversion force curve provided by the present invention, such as Figure 2 As shown, the horizontal axis represents the conversion force in the x-axis direction that the switch to be detected is subjected to, and the vertical axis represents the conversion force in the y-axis direction that the switch to be detected is subjected to. Figure 2 The curve in can clearly reflect the real-time force data of the turnout to be tested during the switching process. For example, the starting value of the conversion force can be obtained through the point where the conversion force starts to change, and the final value of the conversion force can be obtained through the point where the conversion force remains unchanged. The time spent on the conversion of the turnout to be tested can be obtained through the start and end time, as well as the changing characteristics of the conversion force during the turnout switching process.

[0024] It should also be noted that after obtaining the source monitoring data collected by the sensor, in order to ensure the accuracy of fault detection analysis based on the source monitoring data, the initial source monitoring data can be preprocessed to ensure the accuracy and stability of subsequent fault detection. Specifically, since the sensor may be interfered by external factors during the data collection process, such as equipment failure, sensor accuracy problems or network transmission failures, missing values ​​may appear in the turnout conversion force time series data. If these missing values ​​are not properly handled, it will have an adverse effect on the fault detection results. Therefore, the sliding window method can be used to preprocess the initial source monitoring data, and the missing points can be filled by calculating the average value of a period of time near the missing value. That is, select several data points before and after the missing point, and take their average value to replace the missing value. In addition, the source monitoring data after the missing value processing can also be normalized to scale the source monitoring data of different ranges to a unified range. For example, the minimum and maximum normalization can be used to scale the data to the [0,1] interval, which can be achieved by the following formula, as shown in the following formula: In the formula, represents the normalized source monitoring data; represents the source monitoring data to be normalized; represents the minimum source monitoring data to be normalized; Indicates the maximum source monitoring data to be normalized.

[0025] Therefore, the source monitoring data obtained after missing value supplementation and normalization processing has higher data accuracy, and the fault analysis based on this source monitoring data can obtain higher accuracy of the fault detection results.

[0026] Step 120, extracting key point data of the source monitoring data; Here, the key point data is used to directly reflect whether the mechanical state of the turnout is normal, that is, whether the turnout has a fault. The key point data may include at least one of the start and end values ​​of the source monitoring data, the extreme value of the source monitoring data, and the duration of the conversion cycle.

[0027] Specifically, at least one of the start and end values ​​of the source monitoring data, the extreme value of the source monitoring data, and the duration of the conversion cycle can be extracted as key point data. It should be noted that compared with the prior art that only performs data analysis on the complete source monitoring data, the method provided in the embodiment of the present invention further improves the comprehensiveness and accuracy of the fault analysis by extracting the key point data from the source monitoring data and performing fault analysis on the complete source monitoring data and the key point data.

[0028] Step 130, reconstructing the source monitoring data based on the fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on a reconstruction error between the reconstructed monitoring data and the source monitoring data; The fault detection model is trained based on positive sample source monitoring data.

[0029] Here, the fault detection model refers to a neural network model that can deeply understand the key point features and overall curve features in the positive sample source monitoring data, learn the potential features of the positive sample source monitoring data, and then reconstruct the collected source monitoring data. The fault detection model can be constructed based on a machine learning algorithm, for example, it can be constructed based on an autoencoder and a single-classification support vector machine. And it is obtained through unsupervised training with pre-constructed positive sample source monitoring data.

[0030] The positive sample source monitoring data here can be regarded as the source monitoring data of each component in the turnout under normal operation. Considering that in actual applications, the operating state of the components of the turnout is normal most of the time, the daily source monitoring data obtained by daily monitoring of the turnout can be used as the positive sample source monitoring data. For example, the conversion force curve data of several times during the turnout conversion process is collected by sensors, and the data includes the time series of mechanical changes of the turnout from the initial position to the final position. The positive sample source monitoring data here can also be obtained after missing value supplementation and normalization.

[0031] In addition, the reconstructed monitoring data here can be used to reflect the normal data predicted by the fault detection model based on the source monitoring data and the key point data, that is, the normal source monitoring data and normal key point data predicted based on the source monitoring data.

[0032] Specifically, in the application stage of the fault detection model, first, the source monitoring data and the key point data can be input into the fault detection model respectively, and the source monitoring data and the key point data can be reconstructed respectively through the fault detection model. For example, the source monitoring data can be encoded by the encoder in the fault detection model to obtain the encoding features. Then, the decoder in the fault detection model predicts based on the encoding features to realize the reconstruction of the source monitoring data and obtain the reconstructed curve data.

[0033] In addition, the key point data can be reconstructed by a single-classification support vector machine in the fault detection model to obtain reconstructed key point data. Thus, the reconstructed curve data and the reconstructed key point data can be used as reconstructed monitoring data.

[0034] Furthermore, the deviation between the reconstructed monitoring data and the source monitoring data can be calculated to obtain a reconstruction error between the source monitoring data and the reconstructed monitoring data. It can be understood that the reconstruction error can reflect the deviation between the source monitoring data and the reconstructed monitoring data representing a normal value. When the reconstruction error is greater than a preset threshold, it means that the source monitoring data differs greatly from the reconstructed monitoring data representing a normal value, and the fault detection result is that a fault exists; when the reconstruction error is less than the preset threshold, it means that the source monitoring data differs little from the reconstructed monitoring data representing a normal value, and it belongs to the normal deviation range, and the fault detection result is that there is no fault.

[0035] It should be noted that in practical applications, data reconstruction of source monitoring data is performed based on the fault detection model to achieve high-accuracy fault detection without the need for a large amount of abnormal sample source monitoring data, which greatly improves the accuracy and comprehensiveness of anomaly detection, especially for minor or potential fault situations.

[0036] In the training phase of the fault detection model, first, the daily source monitoring data obtained through daily monitoring can be used as positive sample source monitoring data. The initial fault detection model is unsupervisedly trained through the positive sample source monitoring data, so that the initial fault model learns the potential connections in the positive sample source monitoring data and adaptively captures subtle anomalies, so that the fault detection model finally trained can identify deviations from the normal state even in the absence of a large number of abnormal sample data.

[0037] It should be noted that compared with the prior art, most fault detection models are trained based on abnormal data of turnouts. Considering that there is less abnormal source monitoring data of turnouts, it is more difficult to collect abnormal source monitoring data, and it is even more difficult to obtain abnormal source monitoring data under a comprehensive fault state, which leads to the low monitoring accuracy and comprehensiveness of the fault detection model trained in the prior art. The method provided in the embodiment of the present invention obtains a fault detection model by training the forward sample source monitoring data, which greatly enriches the sample data used to train the fault detection model, improves the training effect of the fault detection model, and thus improves the accuracy, reliability and comprehensiveness of the fault detection results.

[0038] The method provided by the embodiment of the present invention obtains a fault detection model through forward sample source monitoring data training, extracts key point data of the source monitoring data, reconstructs the source monitoring data and the key point data through the fault detection model, and obtains reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data, thereby realizing accurate, reliable and comprehensive turnout fault detection and enhancing the safety of train operation.

[0039] Based on any of the above embodiments, the key point data includes: at least one of the start and end values ​​of the source monitoring data, the extreme value of the source monitoring data, and the duration of the conversion cycle.

[0040] Specifically, it is assumed that the source monitoring data is a conversion force curve, wherein the starting value of the source monitoring data can be used to reflect the mechanical state of the turnout at the beginning of the conversion. It should be noted that if the starting conversion force is too high or too low, it indicates that there may be a problem with the drive device of the turnout. The final value of the source monitoring data can be used to reflect whether the turnout encounters resistance or incomplete conversion when completing the conversion. The maximum value of the source monitoring data can be used to reflect whether there is excessive friction or jamming of the mechanical parts of the turnout. The duration of the conversion cycle can be used to reflect whether there is an abnormality in the overall market for the turnout conversion. It should be noted that the turnout conversion time has a fixed range under normal working conditions, and a conversion cycle that is too long or too short may mean that there is a mechanical failure in the equipment.

[0041] It should be noted that the present invention improves the accuracy and reliability of turnout fault detection by extracting key point data including but not limited to the start and end values ​​of monitoring data, extreme values ​​of source monitoring data and duration of conversion cycle, and performing fault analysis based on key point data that is strongly correlated with the fault.

[0042] Based on any of the above embodiments, in step 130, reconstructing the source monitoring data and the key point data based on the fault detection model to obtain reconstructed monitoring data includes: Reconstructing a data curve of the source monitoring data based on the fault detection model to obtain reconstructed curve data; Reconstructing the key point data individually based on the fault detection model to obtain reconstructed key data; The reconstructed monitoring data is obtained based on the reconstructed curve data and / or the reconstructed key data.

[0043] Specifically, the fault detection model can be constructed based on an autoencoder and a single-classification support vector machine. Among them, the source monitoring data can be encoded by the autoencoder to obtain the encoding features. Then, the encoding features are decoded based on the decoder in the fault detection model to obtain the reconstructed curve data of the turnout switching process. The reconstructed curve data here can be used to reflect the normal conversion force value of the entire turnout switching process. In addition, the key point data can be reconstructed individually by a single-classification support vector machine to obtain the normal value data corresponding to the key point data, that is, the reconstructed key data. Finally, the reconstructed curve data and / or the reconstructed key data can be used as reconstructed monitoring data.

[0044] It is understandable that in the training phase, the fault detection model constructed based on the autoencoder and the single-classification support vector machine can reconstruct data based on the real-time source monitoring data and key point data by learning the potential features in the conversion force curve and each key point data in the forward sample source monitoring data, and then in the application phase, reconstruct the data to obtain reconstructed monitoring data that can represent normal values. By reconstructing both the overall source monitoring data and the key point data in the source monitoring data, the sensitivity of fault detection is greatly improved, and the reliability, accuracy and comprehensiveness of fault detection based on the fault detection model are further improved.

[0045] Based on any of the above embodiments, the determining the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data includes: constructing a curve reconstruction error based on the reconstructed curve data and the source monitoring data; constructing a key point reconstruction error based on the reconstructed key data and key point data in the source monitoring data; The curve reconstruction error and the key point reconstruction error are weightedly calculated to obtain the reconstruction error, so as to determine the fault detection result based on the reconstruction error.

[0046] Specifically, the curve reconstruction error can be constructed by reconstructing the deviation between the curve data and the source monitoring data. The key point reconstruction error can be constructed by reconstructing the deviation between the key data and the key point data in the source monitoring data. Then, the curve reconstruction error and the key point reconstruction error can be weighted and added by the preset weighting coefficient to obtain the reconstruction error. Then, the final fault detection result can be determined based on the reconstruction error and the preset threshold. The reconstruction error here can be calculated by the following formula, as shown in the following formula: In the formula, represents the reconstruction error obtained by weighted calculation; , represents the weighting coefficient, which can be set by statistical analysis or through expert experience. , Both can be 0; Represents the curve reconstruction error calculated based on the autoencoder; Represents the key point reconstruction error calculated based on the single-class support vector machine.

[0047] In addition, when only curve reconstruction error or key point reconstruction error exists, the calculated curve reconstruction error or key point reconstruction error can be directly used as the final reconstruction error. Alternatively, the curve reconstruction error or key point reconstruction error can be multiplied by the corresponding weight to calculate the final reconstruction error.

[0048] The method provided by the embodiment of the present invention improves the accuracy and sensitivity of fault detection by reconstructing the overall source monitoring data and the key point data in the source monitoring data based on an autoencoder and a single-classification support vector machine, and performing weighted addition on the reconstruction errors obtained.

[0049] Based on any of the above embodiments, step 120 includes: Performing threshold detection on the key point data to obtain a key point detection result; In the case where there is a fault in the key point detection result, obtaining the fault detection result based on the key point detection result; When there is no fault in the key point detection result, the source monitoring data and the key point data are reconstructed based on the fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data.

[0050] Specifically, a threshold value detection can be performed on the key point data to obtain a key point detection result. For example, a normal value range corresponding to each key point data can be pre-set, and the extracted key point data can be compared with the normal value range by threshold value, and the key point detection result of the key point data exceeding the threshold value can be regarded as the presence of a fault; and the key point detection result of the key point data exceeding the threshold value can be regarded as the absence of a fault. The normal value range here can be set by statistical analysis of a large number of normal conversion force curves or by expert experience.

[0051] Furthermore, in the case where there is a fault in the key point detection result, the key point detection result can be directly used as the fault detection result of the turnout to be detected. In the case where there is no fault in the key point detection result, it means that there is no abnormality in the single key point data. However, in order to further improve the accuracy and comprehensiveness of fault detection and avoid missing detection of situations that may not exceed the preset threshold range but still contain abnormal risks, the source monitoring data and key point data can be reconstructed through the fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data, thereby realizing the detection of subtle abnormalities or potential faults.

[0052] The method provided by the embodiment of the present invention realizes efficient fault detection of prominent abnormal situations by extracting key point data of source monitoring data and performing threshold detection on the key point data. In addition, by combining data reconstruction of key point data and source monitoring data, the detection of subtle abnormalities or potential faults is realized. The combination of the two detection methods greatly improves the accuracy and comprehensiveness of fault detection of turnouts.

[0053] Based on any of the above embodiments, Figure 3 FIG. 2 is a flow chart of the turnout fault detection method provided by the present invention. Figure 3 As shown, first, data collection is performed. Then, data preprocessing is performed to obtain smooth source monitoring data. Next, key point data can be obtained by extracting features from the source monitoring data. Then, fault detection is performed on the key point data through an anomaly detection algorithm, and an abnormal alarm is issued based on the fault detection result.

[0054] Based on any of the above embodiments, Figure 4 FIG. 3 is a flow chart of the turnout fault detection method provided by the present invention. Figure 4 As shown, the method includes: first, obtaining the collected conversion force curve. Then, feature extraction is performed on the conversion force curve to obtain a curve graph feature set, that is, key point data. For example, the start and end values ​​of the source monitoring data, the extreme values ​​of the source monitoring data, and the duration of the conversion cycle. Then, each key point data in the curve graph feature set is compared with the corresponding preset threshold value to determine whether each key point data is greater than the preset threshold value. When there is key point data greater than the preset threshold, an abnormal alarm is issued, and the fault detection result is that a fault exists.

[0055] When there is no key point data greater than the preset threshold, it means that no obvious fault is found based on the threshold detection of the key point data, and fault detection can be achieved by reconstructing the data of the conversion force curve and the curve feature set.

[0056] In detail, the curve feature set can be input into a single-classification support vector machine, and the single-classification support vector machine is used to reconstruct the data of each key point in the curve feature set to obtain the reconstructed key data. Then, based on the deviation between the reconstructed key data and the key point data in the curve feature set, the abnormal score is calculated. In addition, the conversion force curve can be input into the autoencoder, and the conversion force curve can be encoded by the autoencoder to obtain the encoding feature. Then, the encoding feature is decoded based on the decoder to obtain the reconstructed curve data. Next, the abnormal score can be calculated based on the deviation between the reconstructed curve data and the initial conversion force curve. Finally, the abnormal score calculated based on the single-classification support vector machine and the autoencoder can be weighted and added to obtain the weighted abnormal score. Finally, the weighted abnormal score can be judged with the preset threshold. When the weighted abnormal score is not greater than the preset threshold, it means that there is no abnormal alarm, and the abnormal score is output; when the weighted abnormal score is greater than the preset threshold, it means that there is a minor fault in the turnout to be detected, that is, an abnormal alarm is issued.

[0057] Based on any of the above embodiments, the source monitoring data is obtained based on acquiring the conversion force exerted on the turnout to be detected.

[0058] It should be noted that the conversion force curve can directly reflect the force conditions during the turnout conversion process and more accurately capture mechanical abnormalities of the turnout, such as mechanical jamming, component wear or aging, etc. Therefore, compared with the data obtained by other sensors, selecting the conversion force of the turnout to be tested as the source monitoring data for turnout fault analysis can improve the accuracy of fault detection from the source.

[0059] Based on any of the above embodiments, Figure 5 Schematic diagram of the structure of the turnout fault detection device provided by the present invention. Figure 5 As shown, the device comprises: An acquisition unit 510 acquires source monitoring data of the turnout to be detected; An extraction unit 520 extracts key point data of the source monitoring data; A fault detection unit 530 reconstructs the source monitoring data and the key point data based on a fault detection model to obtain reconstructed monitoring data, so as to determine a fault detection result of the turnout to be detected based on a reconstruction error between the reconstructed monitoring data and the source monitoring data; The fault detection model is trained based on positive sample source monitoring data.

[0060] The device provided by the embodiment of the present invention obtains a fault detection model through forward sample source monitoring data training, extracts key point data of the source monitoring data, reconstructs the source monitoring data and the key point data based on the fault detection model, and obtains reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data, thereby realizing accurate, reliable and comprehensive turnout fault detection and enhancing the safety of train operation.

[0061] Based on any of the above embodiments, the key point data includes: at least one of the start and end values ​​of the source monitoring data, the extreme value of the source monitoring data, and the duration of the conversion cycle.

[0062] Based on any of the above embodiments, the fault detection unit is further specifically used for: Reconstructing a data curve of the source monitoring data based on the fault detection model to obtain reconstructed curve data; Reconstructing the key point data individually based on the fault detection model to obtain reconstructed key data; The reconstructed monitoring data is obtained based on the reconstructed curve data and / or the reconstructed key data.

[0063] Based on any of the above embodiments, the fault detection unit is further specifically used for: constructing a curve reconstruction error based on the reconstructed curve data and the source monitoring data; constructing a key point reconstruction error based on the reconstructed key data and key point data in the source monitoring data; The curve reconstruction error and the key point reconstruction error are weightedly calculated to obtain the reconstruction error, so as to determine the fault detection result based on the reconstruction error.

[0064] Based on any of the above embodiments, the fault detection unit is further specifically used for: Performing threshold detection on the key point data to obtain a key point detection result; In the case where there is a fault in the key point detection result, obtaining the fault detection result based on the key point detection result; When there is no fault in the key point detection result, the source monitoring data and the key point data are reconstructed based on the fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data.

[0065] Based on any of the above embodiments, the source monitoring data is obtained based on acquiring the conversion force exerted on the turnout to be detected.

[0066] Figure 6An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the turnout fault detection method, which includes: obtaining source monitoring data of the turnout to be detected; extracting key point data of the source monitoring data; reconstructing the source monitoring data and the key point data based on the fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data; the fault detection model is obtained by training based on the positive sample source monitoring data.

[0067] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, 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, and the computer software product is stored in a storage medium, including a number of instructions for 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: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0068] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the turnout fault detection method provided by the above-mentioned methods, which includes: obtaining source monitoring data of the turnout to be detected; extracting key point data of the source monitoring data; reconstructing the source monitoring data and the key point data based on a fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data; the fault detection model is obtained by training based on positive sample source monitoring data.

[0069] 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 turnout fault detection method provided by the above-mentioned methods, the method comprising: obtaining source monitoring data of the turnout to be detected; extracting key point data of the source monitoring data; reconstructing the source monitoring data and the key point data based on a fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data; the fault detection model is obtained by training based on positive sample source monitoring data.

[0070] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0071] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for 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 some parts of the embodiments.

[0072] 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 embodiments of the present invention.

Claims

1. A turnout fault detection method, characterized in that: include: Obtain source monitoring data of the turnout to be inspected; Extracting key point data of the source monitoring data; Reconstructing the source monitoring data and the key point data based on the fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data; The fault detection model is trained based on positive sample source monitoring data.

2. The turnout fault detection method according to claim 1, characterized in that: The key point data includes at least one of: the start and end values ​​of the source monitoring data, the extreme value of the source monitoring data, and the duration of the conversion cycle.

3. The turnout fault detection method according to claim 1, characterized in that: The reconstructing the source monitoring data and the key point data based on the fault detection model to obtain the reconstructed monitoring data includes: Reconstructing a data curve of the source monitoring data based on the fault detection model to obtain reconstructed curve data; Reconstructing the key point data individually based on the fault detection model to obtain reconstructed key data; The reconstructed monitoring data is obtained based on the reconstructed curve data and / or the reconstructed key data.

4. The turnout fault detection method according to claim 3, characterized in that: The step of determining the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data comprises: constructing a curve reconstruction error based on the reconstructed curve data and the source monitoring data; constructing a key point reconstruction error based on the reconstructed key data and key point data in the source monitoring data; The curve reconstruction error and the key point reconstruction error are weightedly calculated to obtain the reconstruction error, so as to determine the fault detection result based on the reconstruction error.

5. The turnout fault detection method according to any one of claims 1 to 4, characterized in that: The method of reconstructing the source monitoring data based on the fault detection model to obtain reconstructed monitoring data, and determining the fault detection result of the turnout to be detected based on a reconstruction error between the reconstructed monitoring data and the source monitoring data, further includes: Performing threshold detection on the key point data to obtain a key point detection result; In the case where there is a fault in the key point detection result, obtaining the fault detection result based on the key point detection result; When there is no fault in the key point detection result, the source monitoring data and the key point data are reconstructed based on the fault detection model to obtain reconstructed monitoring data, so as to determine the fault detection result of the turnout to be detected based on the reconstruction error between the reconstructed monitoring data and the source monitoring data.

6. The turnout fault detection method according to any one of claims 1 to 4, characterized in that: The source monitoring data is obtained based on acquiring the conversion force exerted on the turnout to be detected.

7. A turnout fault detection device, characterized in that: include: An acquisition unit, for acquiring source monitoring data of the turnout to be detected; An extraction unit, extracting key point data of the source monitoring data; A fault detection unit, which reconstructs the source monitoring data and the key point data based on a fault detection model to obtain reconstructed monitoring data, so as to determine a fault detection result of the turnout to be detected based on a reconstruction error between the reconstructed monitoring data and the source monitoring data; The fault detection model is trained based on positive sample source monitoring data.

8. 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 turnout fault detection method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the turnout fault detection method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the turnout fault detection method according to any one of claims 1 to 6 is implemented.

Images