A method and device for identifying potential risks of contact network based on dynamic detection data

Abstract

The present invention discloses a method and device for identifying potential risks of contact network based on dynamic detection data, including: obtaining historical contact network dynamic detection data, and performing data preprocessing on the historical contact network dynamic detection data to obtain preprocessed historical contact network dynamic detection data; using singular spectrum analysis to perform singular value decomposition of trajectory matrix on the preprocessed historical contact network dynamic detection data to obtain reconstruction components; selecting corresponding reconstruction components and the contribution of reconstruction components to reconstruct the trajectory matrix to obtain a reconstructed data sequence; obtaining real-time contact network dynamic detection data as measured values, performing residual calculation based on the reconstructed data sequence and the measured values ​​to obtain a residual sequence; identifying hidden danger points based on the residual sequence; and identifying potential risks of contact network for a single anchor section based on the hidden danger points. The identification method of the present invention has high accuracy, greatly improves the identification efficiency, and provides effective early warning information for railway operating units.

Description

Technical Field

[0001] The present invention relates to the field of contact network defect prevention and data mining technology, and in particular to a contact network potential risk identification method and device based on dynamic detection data. Background Art

[0002] With the rapid development of my country's railway construction, the service performance of the overhead contact system is the basis for ensuring the safe operation of high-speed railways. In order to accurately grasp the service performance of the overhead contact system, the manpower and material resources invested by each power supply section have increased year by year. The bow-net comprehensive detection device is now widely used in the daily operation and maintenance of the overhead contact system. According to the "High-speed Railway Overhead Contact Operation and Maintenance Rules", the dynamic detection data is currently limited to the binary judgment of "normal" and "overlimit", and the relationship between dynamic detection data and potential risks of the overhead contact network has not been fully explored.

[0003] However, the prior art does not consider the relationship between the contact network dynamic detection data and the contact network potential risk, which results in low accuracy and efficiency in the contact network potential risk identification. That is, the prior art lacks an effective contact network potential risk identification method based on dynamic detection data. Summary of the invention

[0004] The technical problem to be solved by the present invention is that the identification of potential risks of the contact network in the prior art does not take into account the relationship between the dynamic detection data of the contact network and the potential risks of the contact network, which results in the problem of low accuracy and low efficiency in the identification of potential risks of the contact network. The purpose of the present invention is to provide a method and device for identifying potential risks of the contact network based on dynamic detection data, and for the first time apply the singular spectrum analysis method to the mining of dynamic detection data of the contact network, obtain the standard state area of ​​the dynamic detection data of the contact network and identify potential risks; and then, by formulating preventive maintenance strategies, repair the contact network sections with hidden dangers, so as to avoid the occurrence of faults; the present invention has high recognition accuracy, greatly improves the recognition efficiency, and provides effective early warning information for railway operating units.

[0005] The present invention is achieved through the following technical solutions:

[0006] In a first aspect, the present invention provides a method for identifying potential risks of a contact network based on dynamic detection data, the method comprising:

[0007] Acquiring historical contact network dynamic detection data, and performing data preprocessing on the historical contact network dynamic detection data to obtain preprocessed historical contact network dynamic detection data;

[0008] According to the pre-processed historical contact network dynamic detection data, singular value decomposition of the trajectory matrix of the pre-processed historical contact network dynamic detection data is performed using a singular spectrum analysis method to obtain reconstruction components; corresponding reconstruction components and contributions of the reconstruction components are selected to reconstruct the trajectory matrix to obtain a reconstructed data sequence;

[0009] Real-time contact network dynamic detection data is obtained as measured values, and residual operation is performed on the reconstructed data sequence and the measured values ​​to obtain a residual sequence; based on the residual sequence, hidden danger points are identified; and potential risks of the contact network are identified for a single anchor section based on the hidden danger points.

[0010] The working principle is: the potential risk identification of the contact network in the prior art does not consider the relationship between the dynamic detection data of the contact network and the potential risk of the contact network, which leads to the problem of low accuracy and low efficiency in the potential risk identification of the contact network. The present invention designs a potential risk identification method for the contact network based on dynamic detection data, and for the first time applies the singular spectrum analysis method to the mining of dynamic detection data of the contact network, obtains the standard state area of ​​the dynamic detection data of the contact network and identifies the potential risks. Specifically, the present invention first takes the historical dynamic detection data of the contact network as input, and uses the singular spectrum analysis method to perform singular value decomposition and reconstruction of the trajectory matrix of the input historical dynamic detection data of the contact network to obtain a reconstructed data sequence; then, residual calculation is performed based on the reconstructed data sequence and the measured value to obtain a residual sequence, thereby identifying the hidden danger points; finally, the potential risk of the contact network is identified for each single anchor section according to the hidden danger points; thereby, the contact network section with hidden dangers is inspected and repaired by formulating a preventive maintenance strategy, thereby avoiding the occurrence of failures.

[0011] The recognition method of the present invention has high accuracy, greatly improves recognition efficiency, and provides effective early warning information for railway operating units.

[0012] Furthermore, the contact network dynamic detection data includes contact network geometric parameters and bow-cat contact force detection data, and the contact network geometric parameters include contact line height and contact line pull-out value.

[0013] Furthermore, the data preprocessing of the historical contact network dynamic detection data is to use an abnormal point search method to identify abnormal points in the historical contact network dynamic detection data, eliminate the abnormal points, and perform interpolation replacement processing on the abnormal points.

[0014] The interpolation replacement process for the abnormal point usually uses the mean of the detection data at the previous and next points to replace the data at the abnormal point.

[0015] Furthermore, the outlier finding method includes an outlier finding method based on an inverse complementary error function and an outlier finding method based on a dynamic change rate;

[0016] The abnormal point search method is used to count the number of abnormal points per kilometer of contact line height / contact line pull-out value / bow-net contact force in multiple manually confirmed correct detection data, recorded as N / km; when a new contact network dynamic detection data is input, the number of abnormal points per kilometer of contact line height / contact line pull-out value / bow-net contact force is counted M / km; referring to the discrete data statistical method, it is defaulted that when M is greater than three times N*c, the current data is invalid, and c is the abnormal point detection constant.

[0017] Furthermore, the singular spectrum analysis method is used to perform singular value decomposition of the trajectory matrix on the pre-processed historical contact network dynamic detection data to obtain reconstruction components; the corresponding reconstruction components and the contribution of the reconstruction components are selected to reconstruct the trajectory matrix to obtain a reconstructed data sequence, specifically including:

[0018] Step A, arranging multiple detection data of the historical contact network dynamic detection data after preprocessing of a single anchor segment in time series to obtain a trajectory matrix; the historical contact network dynamic detection data after preprocessing of the single anchor segment is recorded as a sequence of length n f0, f1, f2, ..., f n-1 ; The trajectory matrix X is expressed as follows:

[0019]

[0020] Step B, performing singular value decomposition on the trajectory matrix to obtain a reconstructed component of each eigenvector;

[0021]

[0022] In the formula, X i is the i-th reconstruction component; i is the i-th eigenvalue; U i The standard orthogonalization of the eigenvalue corresponding to the i-th eigenvalue gives the eigenvector; V i T V i The transposed matrix of ; 1≤i≤n;

[0023] Step C: Select corresponding reconstruction components and their contributions according to requirements to reconstruct the trajectory matrix and obtain a reconstructed data sequence.

[0024] Furthermore, in step C, the trajectory matrix is ​​reconstructed by selecting corresponding reconstruction components and their contributions according to the requirements, which specifically includes:

[0025] When performing noise reduction processing, the components with larger eigenvalues ​​in the singular value decomposition results are selected for reconstruction;

[0026] When extracting periodic information, select the periodic components for reconstruction;

[0027] The contribution calculation formula of the reconstruction component is: CR k is the contribution rate of the kth reconstruction component.

[0028] Furthermore, the step of identifying the potential risk points according to the residual sequence specifically includes:

[0029] Arrange the residual values ​​in the residual sequence from small to large, and calculate the quartiles Q1, Q2, and Q3;

[0030] For each residual value, the IQR criterion is used to test and calculate its robust ratio quantile statistic Z: IQR' = 0.7413*(Q3-Q1);

[0031] According to the robust ratio quantile statistic, hidden danger points are identified; when the absolute value of the robust ratio quantile statistic is greater than or equal to 3, the detected data point is a hidden danger point; when the robust ratio quantile statistic is less than 3, the detected data point is a non-hidden danger point, that is, the range of the detected data point is the standard state area.

[0032] The residual sequence is a sequence formed by subtracting the measured value from each reconstructed data in the reconstructed data sequence to obtain the absolute value results.

[0033] Furthermore, the step of identifying the potential risks of the contact network for a single anchor section according to the potential risk point includes:

[0034] If the probability of a hidden danger point in a single anchor section is greater than 2%, the anchor section is considered to have potential risks and an alarm message is pushed to the anchor section;

[0035] If the probability of a hidden danger point in a single anchor section is less than or equal to 2%, it is considered that the anchor section has no potential risk and no alarm information is pushed to the anchor section.

[0036] In a second aspect, the present invention further provides a device for identifying potential risks of a contact network based on dynamic detection data, the device supporting the method for identifying potential risks of a contact network based on dynamic detection data, the device comprising:

[0037] An acquisition unit is used to acquire historical contact network dynamic detection data; the contact network dynamic detection data includes contact network geometric parameters and bow-net contact force detection data, and the contact network geometric parameters include contact line height and contact line pull-out value;

[0038] A preprocessing unit, used for performing data preprocessing on the historical contact network dynamic detection data to obtain preprocessed historical contact network dynamic detection data;

[0039] A singular spectrum analysis unit is used to perform singular value decomposition of a trajectory matrix on the preprocessed historical contact network dynamic detection data using a singular spectrum analysis method to obtain reconstruction components; and to reconstruct the trajectory matrix by selecting corresponding reconstruction components and their contributions to obtain a reconstructed data sequence;

[0040] A residual sequence calculation unit, used for obtaining real-time contact network dynamic detection data as measured values, performing residual calculation on the reconstructed data sequence and the measured values ​​to obtain a residual sequence;

[0041] A potential danger point identification unit, used for identifying potential danger points according to the residual sequence;

[0042] The contact network potential risk identification unit is used to identify the contact network potential risks of a single anchor section according to the hidden danger point.

[0043] Further, the singular spectrum analysis unit includes a trajectory matrix forming subunit, a singular value decomposition subunit and a reconstruction subunit;

[0044] The trajectory matrix forming subunit is used to arrange multiple detection data of the historical contact network dynamic detection data after preprocessing of a single anchor segment in a time series to obtain a trajectory matrix; the historical contact network dynamic detection data after preprocessing of a single anchor segment is recorded as a sequence of length n f0, f1, f2, ..., f n-1 ; The trajectory matrix X is expressed as follows:

[0045]

[0046] The singular value decomposition subunit is used to perform singular value decomposition on the trajectory matrix to obtain a reconstructed component of each eigenvector;

[0047]

[0048] In the formula, X i is the i-th reconstruction component; i is the i-th eigenvalue; U i The standard orthogonalization of the eigenvalue corresponding to the i-th eigenvalue gives the eigenvector; V i T V i The transposed matrix of ; 1≤i≤n;

[0049] The reconstruction subunit is used to select corresponding reconstruction components and the contribution of the reconstruction components according to the requirements to reconstruct the trajectory matrix to obtain a reconstructed data sequence; the reconstruction includes:

[0050] When performing noise reduction processing, the components with larger eigenvalues ​​in the singular value decomposition results are selected for reconstruction;

[0051] When extracting periodic information, select the periodic components for reconstruction;

[0052] The contribution calculation formula of the reconstruction component is: CR k is the contribution rate of the kth reconstruction component.

[0053] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0054] 1. The present invention discloses a method and device for identifying potential risks of a contact network based on dynamic detection data. The present invention applies the singular spectrum analysis method to mining dynamic detection data of the contact network for the first time, obtains the standard state area of ​​the dynamic detection data of the contact network, and identifies potential risks. Specifically, the present invention first uses the historical dynamic detection data of the contact network as input, and uses the singular spectrum analysis method to perform singular value decomposition and reconstruction of the trajectory matrix of the input historical dynamic detection data of the contact network, thereby obtaining a reconstructed data sequence; then, residual calculation is performed based on the reconstructed data sequence and the measured value to obtain a residual sequence, thereby identifying hidden danger points; finally, the potential risks of the contact network are identified for each single anchor section based on the hidden danger points; thereby, preventive maintenance strategies are formulated to inspect and repair the contact network sections with hidden dangers, thereby avoiding the occurrence of faults. The identification method of the present invention has high accuracy, greatly improves the identification efficiency, and provides effective early warning information for railway operating units.

[0055] 2. Compared with the outlier recognition technology based on regression model, deterministic and hybrid model, the technology based on singular spectrum analysis of the present invention does not require the artificial establishment of mathematical models in advance, and has great application potential in the field of intelligent diagnosis of detection big data with a large number of measurement points and the need for timely feedback.

[0056] 3. The contact network defect prevention method proposed in the present invention is closely aligned with the contact network business and is applicable to the operation and maintenance of subway contact networks and large railway contact networks. The present invention is concise in calculation and can not only identify hidden dangers for a single detection value, but also realize hidden danger warnings for spatial units of anchor sections, intervals, etc., which has obvious application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0058] Figure 1 The present invention is a flow chart of a method for identifying potential risks of a contact network based on dynamic detection data.

[0059] Figure 2 This is the outlier query effect diagram of the present invention.

[0060] Figure 3 This is an abnormal point search diagram based on the dynamic change rate of the present invention.

[0061] Figure 4 This is the guide height residual curve diagram of the present invention.

[0062] Figure 5 It is a schematic diagram of the dynamic contact line height detection data and the standard state area of ​​the present invention.

[0063] Figure 6 For the present invention Figure 5 Enlarged detail image of .

[0064] Figure 7 The present invention is a schematic structural diagram of a contact network potential risk identification device based on dynamic detection data. DETAILED DESCRIPTION

[0065] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary implementation modes of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0066] Example 1

[0067] Based on the existing technology, the potential risk identification of the contact network does not consider the relationship between the dynamic detection data of the contact network and the potential risk of the contact network, which leads to the problem of low accuracy and low efficiency in the potential risk identification of the contact network. The present invention designs a potential risk identification method for the contact network based on dynamic detection data, and applies the singular spectrum analysis method to the mining of dynamic detection data of the contact network for the first time, obtains the standard state area of ​​the dynamic detection data of the contact network and identifies the potential risks. Specifically, firstly, the present invention takes the historical dynamic detection data of the contact network as input, and uses the singular spectrum analysis method to perform singular value decomposition and reconstruction of the trajectory matrix of the input historical dynamic detection data of the contact network to obtain a reconstructed data sequence; then, residual calculation is performed based on the reconstructed data sequence and the measured value to obtain a residual sequence, thereby identifying the hidden danger points; finally, the potential risk of the contact network is identified for each single anchor section according to the hidden danger points; thereby, the contact network section with hidden dangers is inspected and repaired by formulating a preventive maintenance strategy, thereby avoiding the occurrence of failures.

[0068] like Figure 1 As shown, the present invention provides a method for identifying potential risks of a contact network based on dynamic detection data, the method comprising:

[0069] Step 1, obtaining historical contact network dynamic detection data, and performing data preprocessing on the historical contact network dynamic detection data to obtain preprocessed historical contact network dynamic detection data; wherein the contact network dynamic detection data includes contact network geometric parameters and bow-net contact force detection data, and the contact network geometric parameters include contact line height and contact line pull-out value.

[0070] Considering that the contact network geometric parameters (contact line height and contact line pull-out value) and the bow-net contact force test data are relatively continuous under normal circumstances, there will be no obvious mutation points. When there are a large number of "discrete" mutation points in such test data, it means that the test may be interfered by external factors, resulting in abnormal data and no subsequent reference value.

[0071] Therefore, the present invention adopts an abnormal point search method to identify abnormal points in the historical contact network dynamic detection data, removes the abnormal points, and performs interpolation replacement processing on the abnormal points. Among them, the interpolation replacement processing on the abnormal points usually uses the average of the detection data at the previous and next points to replace the data at the abnormal points.

[0072] Specifically, the outlier point search method includes an outlier point search method based on an inverse complementary error function and an outlier point search method based on a dynamic change rate;

[0073] (1) Outlier search method based on inverse complementary error function

[0074] In statistics, values ​​that differ from the local median by more than three times the converted median absolute deviation (MAD) are called outliers. The MAD calculation formula is as follows:

[0075] MAD=c*median(abs(A-median(A)))

[0076]

[0077] Where A is the data point in the window, c is the outlier detection constant, median is the median, and abs() is the absolute value function;

[0078] erfcinv stands for the inverse complementary error function and is defined as follows:

[0079]

[0080] The outlier query effect diagram is as follows Figure 2 As shown, Figure 2 The horizontal axis represents the sampling point number, the vertical axis represents the detection value, the line is the detection data waveform, and the outlier is the abnormal point.

[0081] (2) Outlier point search method based on dynamic change rate

[0082] The dynamic detection data of the contact network changes with the detection position. When there is a phenomenon that the continuous detection data is almost unchanged, it is usually caused by abnormal detection sensors. This type of data is meaningless for the state analysis of the contact network system. The dynamic change rate of the i-th point in the dynamic detection data is defined as follows:

[0083]

[0084] Among them, v i represents the i-th dynamic detection data value, v i-1 represents the i-1th dynamic detection data value, and the dynamic change rate of the first point is set to 0. i The point where |<ε is recorded as an abnormal point. For different contact network detection parameters, the value of ε is shown in Table 1:

[0085] Table 1 Detection parameter abnormal point threshold table

[0086] Detection parameters ε Contact line height 1.0005 Contact wire pull-out value 1.002 Pantograph-catenary contact force 1.007

[0087] The abnormal point search diagram based on dynamic change rate is as follows Figure 3 As shown, Figure 3 The horizontal axis represents the kilometer mark, the vertical axis represents the detection value, Data is the detection data, and Outlier is the outlier.

[0088] According to the above method, the abnormal point search method is used to count the number of abnormal points per kilometer of contact line height / contact line pull-out value / bow-net contact force in multiple manually confirmed correct detection data, recorded as N / km; when a new contact network dynamic detection data is input, the number of abnormal points per kilometer of contact line height / contact line pull-out value / bow-net contact force is counted as M / km; referring to the discrete data statistical method, it is defaulted that when M is greater than three times N*c, the current data is invalid, and c is the abnormal point detection constant.

[0089] Step 2, according to the pre-processed historical contact network dynamic detection data, the singular value decomposition of the trajectory matrix of the pre-processed historical contact network dynamic detection data is performed by using the singular spectrum analysis method to obtain reconstruction components; the corresponding reconstruction components and the contribution of the reconstruction components are selected to reconstruct the trajectory matrix to obtain a reconstructed data sequence; specifically including:

[0090] Step A: The historical contact network dynamic detection data after preprocessing of a single anchor segment is recorded as a sequence of length n: f0, f1, f2, ..., f n-1 ; In order to grasp the time evolution structure implied by the detection parameters, the multiple detection data of the historical contact network dynamic detection data after preprocessing of a single anchor section are arranged in time series to obtain the trajectory matrix; the trajectory matrix X is expressed as follows:

[0091]

[0092] Step B, perform singular value decomposition on the trajectory matrix to obtain the reconstruction component of each eigenvector; let S = XX T , whose eigenvalues ​​are λ1,λ2,...,λ n , the corresponding standard orthogonalized eigenvectors are U1,U2,...,U n ,make Then the singular value decomposition of the trajectory matrix X is:

[0093]

[0094] In the formula, X i is the i-th reconstruction component; i is the i-th eigenvalue; U i The standard orthogonalization of the eigenvalue corresponding to the i-th eigenvalue gives the eigenvector; V i T V i The transposed matrix of ; 1≤i≤n;

[0095] Step C, reconstruct the trajectory matrix by selecting the corresponding reconstruction components and their contribution according to the requirements. For example, when performing noise reduction, select the component with the larger eigenvalue in the singular value decomposition result, and when extracting periodic information, select the reconstruction with periodic components. For the trajectory matrix X, the contribution rate CR of the kth reconstruction component is k for:

[0096]

[0097] The main components with significantly larger contribution rates can be selected to reconstruct the data sequence. Generally, the first k components with cumulative contribution rates of about 95% can be selected to reconstruct the data sequence.

[0098] Step 3, based on the reconstructed data sequence obtained in step 2, obtain real-time contact network dynamic detection data as measured values, perform residual calculation based on the reconstructed data sequence and the measured values ​​to obtain a residual sequence; identify potential hazards based on the residual sequence; and identify potential risks of the contact network for a single anchor section based on the potential hazards. The analysis steps are as follows:

[0099] (1) Arrange the residual values ​​in the residual sequence from small to large and calculate the quartiles Q1, Q2, and Q3; wherein the residual sequence is a sequence formed by subtracting each reconstructed data in the reconstructed data sequence from the measured value and obtaining the absolute value results.

[0100] (2) For each residual value, the IQR criterion is used to test and calculate its robust ratio quantile statistic Z:

[0101]

[0102] IQR' = 0.7413*(Q3-Q1);

[0103] According to the robust ratio quantile statistic, hidden danger points are identified; when the absolute value of the robust ratio quantile statistic is greater than or equal to 3 (|Z|≥3), the corresponding confidence level is 99%, and the detected data point is a hidden danger point; when the robust ratio quantile statistic is less than 3 (|Z|<3), the detected data point is not a hidden danger point, that is, the range of the detected data point is the standard state area.

[0104] (3) Considering the continuity of the catenary detection parameters, if the probability of a potential risk point in a single anchor section is greater than 2%, the anchor section is considered to have potential risks, and an alarm message is pushed to the anchor section;

[0105] If the probability of a hidden danger point in a single anchor section is less than or equal to 2%, it is considered that the anchor section has no potential risk and no alarm information is pushed to the anchor section.

[0106] This embodiment takes the dynamic detection data of a subway line as an example, selects 5 to calculate the mean of the dynamic contact line height data and absolute error in the db file of other trips, and includes them in the following Table 2:

[0107] Table 2 Error mean table

[0108] Trips 1 2 3 4 Mean error(mm) 0.358 0.3322 0.2897 0.2117

[0109] It can be concluded that the device detection error (err d ) is approximately 0.3mm.

[0110] Typical anchor section (Y061) height guide residual curve is as follows Figure 4 As shown, Figure 4 The horizontal axis represents the kilometer mark, the vertical axis represents the height guide residual value, and the residual curve is the height guide residual curve.

[0111] The residual distribution information table is shown in Table 3:

[0112] Table 3 Residual distribution information table

[0113] Maximum value / mm Mean / mm IQR / mm 0.5839 0.08 0.08

[0114] The allowable error of each dynamic detection point should not be less than the detection equipment error, so the standard state range of each single point is [reValue-max(IQR,err d ),reValue+max(IQR,err d )], where reValue is the reconstructed value.

[0115] The dynamic contact line height detection data of a typical anchor section (Y061) and the schematic diagram of the standard state area are shown in the figure. Figure 5 As shown, Figure 5 The horizontal axis represents the kilometer mark, and the vertical axis represents the contact line height value. Figure 6 for Figure 5 Enlarged detail image of .

[0116] According to statistics, the number of abnormal points in the contact line height of the anchor section accounts for about 2.7%.

[0117] The innovative features of the present invention are as follows:

[0118] 1. Innovative method for identifying potential risks of overhead contact network

[0119] The present invention applies the singular spectrum analysis method to the mining of contact network dynamic detection data for the first time, obtains the standard state area of ​​the contact network dynamic detection data and identifies potential risks. Specifically, the present invention first uses the historical contact network dynamic detection data as input, and uses the singular spectrum analysis method to perform singular value decomposition of the trajectory matrix of the input historical contact network dynamic detection data and reconstruct it to obtain a reconstructed data sequence; then, residual calculation is performed based on the reconstructed data sequence and the measured value to obtain a residual sequence, thereby identifying the hidden danger points; finally, the potential risks of the contact network are identified for each single anchor section according to the hidden danger points; thereby, the contact network sections with hidden dangers are inspected and repaired by formulating preventive maintenance strategies, thereby avoiding the occurrence of faults. The identification method of the present invention has high accuracy, greatly improves the identification efficiency, and provides effective early warning information for railway operating units.

[0120] 2. Real-time and effectiveness

[0121] Compared with outlier identification technologies based on regression models, deterministic and hybrid models, technologies based on singular spectrum analysis do not require the artificial establishment of mathematical models in advance, and have great application potential in the field of intelligent diagnosis of detection big data with a large number of measurement points and the need for timely feedback.

[0122] 3. Universality and convenience of railway system applications

[0123] The method for preventing defects in the contact network proposed in the present invention is closely related to the contact network business and is applicable to the operation and maintenance of subway contact networks and large railway contact networks. The present invention has concise calculations and can not only identify hidden dangers for a single detection value, but also realize hidden danger warnings for spatial units of anchor sections, intervals, etc., which has obvious application value.

[0124] Example 2

[0125] like Figure 7As shown, the difference between this embodiment and embodiment 1 is that this embodiment provides a contact network potential risk identification device based on dynamic detection data, the device supports a contact network potential risk identification method based on dynamic detection data described in embodiment 1, and the device includes:

[0126] An acquisition unit is used to acquire historical contact network dynamic detection data; the contact network dynamic detection data includes contact network geometric parameters and bow-net contact force detection data, and the contact network geometric parameters include contact line height and contact line pull-out value;

[0127] A preprocessing unit, used for performing data preprocessing on the historical contact network dynamic detection data to obtain preprocessed historical contact network dynamic detection data;

[0128] A singular spectrum analysis unit is used to perform singular value decomposition of a trajectory matrix on the preprocessed historical contact network dynamic detection data using a singular spectrum analysis method to obtain reconstruction components; and to reconstruct the trajectory matrix by selecting corresponding reconstruction components and their contributions to obtain a reconstructed data sequence;

[0129] A residual sequence calculation unit, used for obtaining real-time contact network dynamic detection data as measured values, performing residual calculation on the reconstructed data sequence and the measured values ​​to obtain a residual sequence;

[0130] A potential danger point identification unit, used for identifying potential danger points according to the residual sequence;

[0131] The contact network potential risk identification unit is used to identify the contact network potential risks of a single anchor section according to the hidden danger point.

[0132] In this embodiment, the singular spectrum analysis unit includes a trajectory matrix forming subunit, a singular value decomposition subunit and a reconstruction subunit;

[0133] The trajectory matrix forming subunit is used to arrange multiple detection data of the historical contact network dynamic detection data after preprocessing of a single anchor segment in a time series to obtain a trajectory matrix; the historical contact network dynamic detection data after preprocessing of a single anchor segment is recorded as a sequence of length n f0, f1, f2, ..., f n-1 ; The trajectory matrix X is expressed as follows:

[0134]

[0135] The singular value decomposition subunit is used to perform singular value decomposition on the trajectory matrix to obtain a reconstructed component of each eigenvector;

[0136]

[0137] In the formula, X i is the i-th reconstruction component; i is the i-th eigenvalue; U i The standard orthogonalization of the eigenvalue corresponding to the i-th eigenvalue gives the eigenvector; V i T V i The transposed matrix of ; 1≤i≤n;

[0138] The reconstruction subunit is used to select corresponding reconstruction components and the contribution of the reconstruction components according to the requirements to reconstruct the trajectory matrix to obtain a reconstructed data sequence; the reconstruction includes:

[0139] When performing noise reduction processing, the components with larger eigenvalues ​​in the singular value decomposition results are selected for reconstruction;

[0140] When extracting periodic information, select the periodic components for reconstruction;

[0141] The contribution calculation formula of the reconstruction component is: CR k is the contribution rate of the kth reconstruction component.

[0142] Among them, the execution process of each unit can be executed according to the process steps of a contact network potential risk identification method based on dynamic detection data described in Example 1, and will not be repeated one by one in this embodiment.

[0143] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0144] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0145] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0146] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0147] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for identifying potential risks of overhead contact network based on dynamic detection data, characterized in that: The method includes: Acquiring historical contact network dynamic detection data, and performing data preprocessing on the historical contact network dynamic detection data to obtain preprocessed historical contact network dynamic detection data; According to the pre-processed historical contact network dynamic detection data, singular value decomposition of the trajectory matrix of the pre-processed historical contact network dynamic detection data is performed using a singular spectrum analysis method to obtain reconstruction components; corresponding reconstruction components and contributions of the reconstruction components are selected to reconstruct the trajectory matrix to obtain a reconstructed data sequence; Acquire real-time contact network dynamic detection data as measured values, perform residual operation on the reconstructed data sequence and the measured values ​​to obtain a residual sequence; identify potential risk points according to the residual sequence; and identify potential risks of the contact network for a single anchor section according to the potential risk points; The contact network dynamic detection data includes contact network geometric parameters and pantograph-catenary contact force detection data, wherein the contact network geometric parameters include contact line height and contact line pull-out value; The step of identifying the potential risk points according to the residual sequence specifically includes: Arrange the residual values ​​in the residual sequence from small to large, and calculate the quartiles Q1, Q2, and Q3; For each residual value, the IQR criterion is used to test and calculate its robust ratio quantile statistic Z: IQR' = 0.7413*(Q3-Q1); According to the robust ratio quantile statistic, hidden danger points are identified; when the absolute value of the robust ratio quantile statistic is greater than or equal to 3, the detected data point is a hidden danger point; when the robust ratio quantile statistic is less than 3, the detected data point is a non-hidden danger point, that is, the range of the detected data point is the standard state area.

2. The method for identifying potential risks of a contact network based on dynamic detection data according to claim 1 is characterized in that: The data preprocessing of the historical contact network dynamic detection data is to use an abnormal point search method to identify abnormal points in the historical contact network dynamic detection data, remove the abnormal points, and perform interpolation replacement processing on the abnormal points.

3. The method for identifying potential risks of overhead line based on dynamic detection data according to claim 2 is characterized in that: The outlier finding method includes an outlier finding method based on an inverse complementary error function and an outlier finding method based on a dynamic change rate; The abnormal point search method is used to count the number of abnormal points per kilometer of contact line height / contact line pull-out value / bow-net contact force in multiple manually confirmed correct detection data, recorded as N / km; when a new contact network dynamic detection data is input, the number of abnormal points per kilometer of contact line height / contact line pull-out value / bow-net contact force is counted M / km; referring to the discrete data statistical method, when M is greater than three times N*c, the current data is invalid, and c is the abnormal point detection constant.

4. The method for identifying potential risks of overhead line based on dynamic detection data according to claim 1 is characterized in that: The singular spectrum analysis method is used to perform singular value decomposition of the trajectory matrix on the pre-processed historical contact network dynamic detection data to obtain the reconstructed components; Selecting corresponding reconstruction components and their contributions to reconstruct the trajectory matrix to obtain a reconstructed data sequence specifically includes: Step A, arranging multiple detection data of the historical contact network dynamic detection data after preprocessing of a single anchor segment in time series to obtain a trajectory matrix; the historical contact network dynamic detection data after preprocessing of the single anchor segment is recorded as a sequence of length n f0, f1, f2, ..., f n-1 ; The trajectory matrix X is expressed as follows: Step B, performing singular value decomposition on the trajectory matrix to obtain a reconstructed component of each eigenvector; Where, X i is the i-th reconstruction component; i is the i-th eigenvalue; U i The standard orthogonalization of the eigenvalue corresponding to the i-th eigenvalue gives the eigenvector; V i T V i The transposed matrix of ; 1≤i≤n; Step C: Select corresponding reconstruction components and their contributions according to requirements to reconstruct the trajectory matrix and obtain a reconstructed data sequence.

5. A method for identifying potential risks of overhead line based on dynamic detection data according to claim 4, characterized in that: In step C, the trajectory matrix is ​​reconstructed by selecting corresponding reconstruction components and their contributions according to the requirements, which specifically includes: When performing noise reduction processing, the components with larger eigenvalues ​​in the singular value decomposition results are selected for reconstruction; When extracting periodic information, select the periodic components for reconstruction; The contribution calculation formula of the reconstruction component is: CR k is the contribution rate of the kth reconstruction component.

6. The method for identifying potential risks of overhead line based on dynamic detection data according to claim 1 is characterized in that: The identifying of potential risks of the contact network for a single anchor section according to the potential risk point includes: If the probability of a hidden danger point in a single anchor section is greater than 2%, the anchor section is considered to have potential risks and an alarm message is pushed to the anchor section; If the probability of a hidden danger point in a single anchor section is less than or equal to 2%, it is considered that the anchor section has no potential risk and no alarm information is pushed to the anchor section.

7. A device for identifying potential risks of contact network based on dynamic detection data, characterized in that: The device supports a method for identifying potential risks of a contact network based on dynamic detection data as described in any one of claims 1 to 6, and the device comprises: An acquisition unit is used to acquire historical contact network dynamic detection data; the contact network dynamic detection data includes contact network geometric parameters and bow-net contact force detection data, and the contact network geometric parameters include contact line height and contact line pull-out value; A preprocessing unit, used for performing data preprocessing on the historical contact network dynamic detection data to obtain preprocessed historical contact network dynamic detection data; A singular spectrum analysis unit is used to perform singular value decomposition of a trajectory matrix on the preprocessed historical contact network dynamic detection data using a singular spectrum analysis method to obtain reconstruction components; and to reconstruct the trajectory matrix by selecting corresponding reconstruction components and their contributions to obtain a reconstructed data sequence; A residual sequence calculation unit, used for obtaining real-time contact network dynamic detection data as measured values, performing residual calculation on the reconstructed data sequence and the measured values ​​to obtain a residual sequence; A potential danger point identification unit, used for identifying potential danger points according to the residual sequence; A contact network potential risk identification unit, used for identifying the contact network potential risk of a single anchor section according to the hidden danger point; The contact network dynamic detection data includes contact network geometric parameters and pantograph-catenary contact force detection data, wherein the contact network geometric parameters include contact line height and contact line pull-out value; The step of identifying the potential risk points according to the residual sequence specifically includes: Arrange the residual values ​​in the residual sequence from small to large, and calculate the quartiles Q1, Q2, and Q3; For each residual value, the IQR criterion is used to test and calculate its robust ratio quantile statistic Z: IQR' = 0.7413*(Q3-Q1); According to the robust ratio quantile statistic, hidden danger points are identified; when the absolute value of the robust ratio quantile statistic is greater than or equal to 3, the detected data point is a hidden danger point; when the robust ratio quantile statistic is less than 3, the detected data point is a non-hidden danger point, that is, the range of the detected data point is the standard state area.

8. The device for identifying potential risks of overhead line based on dynamic detection data according to claim 7 is characterized in that: The singular spectrum analysis unit includes a trajectory matrix forming subunit, a singular value decomposition subunit and a reconstruction subunit; The trajectory matrix forming subunit is used to arrange multiple detection data of historical contact network dynamic detection data after preprocessing of a single anchor section in time series to obtain a trajectory matrix; The singular value decomposition subunit is used to perform singular value decomposition on the trajectory matrix to obtain a reconstructed component of each eigenvector; The reconstruction subunit is used to select corresponding reconstruction components and the contribution of the reconstruction components according to the requirements to reconstruct the trajectory matrix to obtain a reconstructed data sequence; the reconstruction includes: When performing noise reduction processing, the components with larger eigenvalues ​​in the singular value decomposition results are selected for reconstruction; When extracting periodic information, the periodic components are selected for reconstruction.

Images