A comprehensive diagnosis method for overhead line based on multi-source detection parameters

Abstract

The present invention discloses a method for comprehensive diagnosis of contact network based on multi-source detection parameters, comprising the following steps: S1, automatic alignment of detection data; S2, standard state generation of detection data after automatic alignment; S3, potential defect analysis of detection data in standard state, including time domain analysis and frequency domain analysis, to obtain potential defects of contact network. The method for comprehensive diagnosis of contact network based on multi-source detection parameters mentioned in this aspect mainly utilizes multiple multi-dimensional detection data to conduct comprehensive diagnosis and evaluation of contact network. Multi-dimensional detection data include but are not limited to pull-out value, guide height, bow-net contact force, and hard points.

Description

Technical Field

[0001] The invention relates to the field of contact network detection, and in particular to a contact network comprehensive diagnosis method based on multi-source detection parameters. Background Art

[0002] With the rapid development of urban rail transit, the contact network detection system has become more accurate and comprehensive, and can comprehensively monitor the contact network status from multiple aspects. At present, the use of detection data mainly stays on the judgment of whether it is over-limit, and the threshold setting of over-limit is generally large, and few people pay attention to the data that is not over-limit. The existing methods for using detection data are: 1. Use fault tree analysis, event number analysis, and credibility theory to establish a reliability model of the entire contact network system; 2. Analyze the contact network detection data based on Hilbert-Huang transform to evaluate the entire operating status of the contact network; 3. Use fuzzy comprehensive evaluation method, gray clustering, cloud model, etc. to evaluate the health status of the contact network; 4. Use set pair analysis method and evidence theory to evaluate the health status of the contact network, and integrate each evaluation index in the contact network health status evaluation index system layer by layer. Among the above methods, method 1 only uses defect data for fault analysis and does not diagnose non-overlimit data; method 2 analyzes the overall line data and cannot judge the specific location; method 3 does not adequately analyze the uncertainty of information processing and information fusion, and directly weights and fuses the various evaluation indicators of the contact network status, which can easily misjudge the contact network status; method 4 can accurately complete the health status evaluation of the contact network, but cannot accurately locate the specific defect location and only evaluates the health status of the entire line. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention provides a comprehensive contact network diagnosis method based on multi-source detection parameters to solve the problem of incomplete contact network detection data.

[0004] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a method for comprehensive diagnosis of contact network based on multi-source detection parameters, comprising the following steps:

[0005] S1. Automatically align the detection data;

[0006] S2, generating a standard state for the detection data after automatic alignment;

[0007] S3. Perform potential defect analysis on the standard state detection data, including time domain analysis and frequency domain analysis, to obtain the potential defects of the contact network.

[0008] Further: the specific steps of step S1 are:

[0009] S11, standard data establishment and test data preprocessing;

[0010] S12, positioning the anchor section area through data matching and calculation;

[0011] S13, performing corresponding difference sampling operations on the data according to the matching results, and aligning the detection data waveform with the design curve.

[0012] Further: the establishment of the standard data includes:

[0013] Using the design values ​​of the contact network's positioning points and anchor sections and the corresponding mileage information, standard data are established through the linear difference method.

[0014] Further: the detection data preprocessing includes:

[0015] Starting from the first detected data, the difference is made with the subsequent data in sequence. When the difference value is greater than the set threshold T, the previous and next data are divided into two point sets. Finally, the points with the number of point sets less than the threshold M are removed to filter the abnormal jump points.

[0016] Further: the step S12 is specifically as follows:

[0017] S121, dividing the test data and the standard data into multiple parts based on the pull-out value mutation;

[0018] S122, using the correlation coefficient, matching each anchor segment of the test data with the anchor segment of the standard data, and selecting the longest ordered subsequence from the matching results as the optimal solution;

[0019] S123, by calculating the stretching scale of the optimal solution for matching with respect to the standard data, so as to obtain a stretching scale less than the threshold value L max result.

[0020] Further: the step S2 comprises:

[0021] According to the measured value of each detection point and the confidence interval statistics, the measurement standard range corresponding to each detection point is generated to generate the standard state area of ​​the detection data.

[0022] Furthermore: the time domain analysis in step S3 includes change trend diagnosis and / or variance analysis.

[0023] Further: the change trend diagnosis is specifically:

[0024] Calculate the maximum difference between each acquisition point of the current detection data and the standard state area data i , if it is in the standard state area, it is 0, and the maximum difference and ∑diff between each positioning point are obtained by span i, using the sum of the maximum difference and the center value of the standard state area, calculate the change rate Δm of the test data compared with the standard state area;

[0025] The change rates of the detection data of different points are analyzed synchronously. If the change trends of multiple detection data are the same and the detection data gradually exceed the standard state range, it is regarded as a sub-healthy state of the contact network.

[0026] Further: the variance analysis includes:

[0027] S311, respectively obtaining the variance value of the conductor height pull-out value of the entire line;

[0028] S312, setting the conductor height weight value to twice the pull-out value weight; respectively adding the variance and the assigned value of the conductor height pull-out value to obtain the sum of the variances based on the conductor height pull-out value;

[0029] S313, multiple detection data are used to calculate the variance of the detection data at different times. When the variance of the values ​​of the continuous high-conductor pull-out gradually increases, there is a hidden danger in the contact network.

[0030] Further: the frequency domain analysis in step S3 includes:

[0031] S321, de-averaging the test data, and subtracting the overall mean from each test data, so that the overall mean of the data is 0;

[0032] S322, using Hilbert-Huang transformation to realize time-frequency conversion of detection data, calculating the instantaneous energy of the signal through its time-frequency information, setting an energy threshold, and if the instantaneous energy exceeds the energy threshold, it is determined to be a potential defect area.

[0033] The beneficial effects of the present invention are:

[0034] 1. The present invention relates to a method for comprehensive diagnosis of the contact network based on multi-source detection parameters, which mainly utilizes multiple multi-dimensional detection data to conduct comprehensive diagnosis and evaluation of the contact network. The multi-dimensional detection data includes but is not limited to pull-out value, guide height, bow-net contact force, and hard points.

[0035] 2. The present invention automatically realizes the alignment of multiple detection data to avoid the failure of subsequent analysis caused by the mileage error of the detection data.

[0036] 3. The present invention selects a specific analysis method according to the characteristics of different source detection points, and amplifies the characteristics of the contact network to achieve effective diagnosis of the contact network detection data. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of the present invention;

[0038] Figure 2Schematic diagram of standard pull-out value data and actual detected pull-out value data in an embodiment of the present invention;

[0039] Figure 3 It is a schematic diagram of the alignment result of the detection data waveform and the design value curve in an embodiment of the present invention;

[0040] Figure 4 Schematic diagram of the relationship between variance and time in an embodiment of the present invention;

[0041] Figure 5 It is a schematic diagram of pressure data of a certain line in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0043] like Figure 1 As shown, a comprehensive contact network diagnosis method based on multi-source detection parameters includes the following steps:

[0044] S1. Automatically align the detection data;

[0045] The location information of the detection data usually comes from the vehicle speed sensor, GPS, etc., but the cumulative mileage error is inevitable due to the relative sliding of the wheel and rail, the shortening of the wheel diameter caused by wheel wear, etc. The mileage error of the detection data will directly affect the subsequent evaluation and analysis of the contact network status, the accurate positioning of defects and the judgment of their causes. Figure 2 Represents the standard pull-out value data and the actual detected pull-out value data. Due to the gradually accumulated mileage error, the actual detected pull-out value data has a more obvious mileage deviation.

[0046] S11, standard data establishment and test data preprocessing;

[0047] Collect the design values ​​of the positioning points and anchor sections of the overhead line, use the design data of the positioning points and anchor section joints and the corresponding mileage information, and use the linear difference method to establish the standard data of the entire line based on the sampling interval of 0.25m per point. Let the mileage sequence be (x1, x2, … x m ), the sequence of values ​​pulled out is (y1,y2,…y m );

[0048]

[0049] In the above formula, y iis the i-th pull-out value, y0 is the design value of the pull-out value at the positioning point, y m is the mth pulled value, x m is the mth mileage, x0 is the mileage information at the positioning point, x i is the i-th mileage;

[0050] As for the contact network detection data itself, it is usually affected by data transmission, electromagnetic interference, etc., and may produce many abnormal points. Before performing subsequent alignment operations, the abnormal points need to be excluded, and the abnormal jump points are filtered by data clustering. Starting from the first detection data, it is subtracted from the next data in turn. When the difference is greater than the set threshold T, the previous and next data are divided into two point sets, and so on. Finally, the points whose number of point sets is less than the threshold M are removed.

[0051] S12, positioning the anchor section area;

[0052] The main feature of the contact network pull-out value curve in the anchor section area is the pull-out value mutation. The test data and the standard data are divided into multiple parts based on the pull-out value mutation, which are the standard data s md1 ,s md2 …s mdn , detection data A md1 ,A md2 …A mdm ;

[0053] Using the correlation coefficient formula, each anchor segment of the test data is matched with the anchor segment of the standard data. The correlation coefficient formula is:

[0054]

[0055] In the above formula, r is the correlation coefficient, x i With x i ' is the i-th standard data and test data, m is the number of standard data and test data;

[0056] Record the matching pairs whose correlation coefficient exceeds the threshold T, select the longest optimal order matching pairs, record the positions of the corresponding matching detection data in the order of the standard data, select the longest ordered subsequence as the optimal solution in the matching results, calculate the stretching scale ΔL of the matching optimal solution compared to the standard data, and delete the stretching scale greater than the threshold L max The matching results of the retained stretching scale are less than the threshold L max The matching results.

[0057] S13, perform the corresponding difference sampling operation on the data according to the matching result, align the detection data waveform with the design curve, and the alignment result is as follows: Figure 3 shown.

[0058] S2, generating a standard state for the detection data after automatic alignment;

[0059] After the detection data is aligned, the mileage error of the detection data is eliminated. Due to certain system errors in the detection equipment itself, the state judgment of the detection parameters cannot rely solely on the comparison between the design value and the measured value. The present invention provides a method for generating a standard state of detection data. The method relies on multiple detection data. After the detection data is aligned, the measured value of each detection point is counted, and the measurement standard range corresponding to each point is counted according to the 95% confidence interval. in is the mean, σ is the variance, n is the number of data, and the standard state range of the detection data is generated according to the measurement standard range.

[0060] S3. Perform potential defect analysis on the standard state detection data, including time domain analysis and frequency domain analysis, to obtain the potential defects of the contact network.

[0061] The static parameters of the contact network (conductor height pull-out value) are relatively stable and less affected by external factors, so the relatively direct time domain analysis of the static parameters of the contact network is mainly used for comprehensive diagnosis. The time domain analysis includes change trend diagnosis and variance analysis.

[0062] The specific change trend diagnosis is as follows:

[0063] Calculate the maximum difference between each acquisition point of the current detection data and the standard state area data i , if it is in the middle of the standard state, it is 0, and the maximum difference and ∑diff between each positioning point are obtained by span i , the sum of the maximum difference and the center value of the corresponding standard state area is regarded as the change rate Δm of the current detection data compared with the standard state area;

[0064] The calculation formula is:

[0065]

[0066]

[0067] In the above formula, [X imin ,X imax ] is the standard state area range;

[0068] This change rate can correctly reflect the trend change of the test data, and the standard state area can effectively filter out the changes caused by the normal fluctuations of the test itself. The change rate of the test data of multiple different points is analyzed synchronously. If the change trend of multiple test data is the same, the test data gradually exceeds the standard state range, which is regarded as the sub-health state of the contact network.

[0069] The overall variance of the detection data is an important element in evaluating whether the contact network is smooth. Based on the multiple detection data of the line, the before and after comparison diagnosis of the contact network can be realized. Due to the uncertainty of dynamic detection parameters, the present invention uses the static detection parameter high-conducting pull-out value to comprehensively evaluate the contact network.

[0070] The variance analysis is specifically as follows: respectively obtain the variance of the conductor height pull-out value of the entire line; according to the weighted consideration of the static quality calculation of the contact network in the "CQI contact network static detection and evaluation method", the conductor height weight value is set to 2 times the pull-out value weight, and the conductor height weight is 0.67 after normalization, and the pull-out value weight is 0.33; respectively add the variance and assignment of the conductor height pull-out value, and obtain the variance of the detection data at different times based on the variance and of the conductor height pull-out value, and calculate the variance of the detection data at different times based on multiple detection data, such as Figure 4 As shown in the figure, when the variance of the pull-out values ​​increases gradually for multiple consecutive times, there is a hidden danger in the contact network.

[0071] The dynamic parameters of the overhead contact network (pressure hard points, etc.) are greatly affected by the outside world. Different pantographs and different vehicles will lead to large differences in the detection data. Therefore, the frequency domain analysis method is used to extract the corresponding frequency components from the detection data for fault identification and judgment.

[0072] The frequency domain analysis is as follows:

[0073] De-mean the test data, subtract the overall mean from each test data, so that the overall mean of the data is 0;

[0074] The Hilbert-Huang transform is used to realize the time-frequency conversion of the detection data. The instantaneous energy of the signal is calculated through its time-frequency information, and the energy threshold is set. If the instantaneous energy exceeds the energy threshold, it is determined to be a potential defect area.

[0075] Figure 5 Taking the pressure data of a certain line as an example, it can be seen that when the instantaneous energy exceeds the threshold line, the detection data above it also obviously fluctuates the most at the same time. For dynamic detection parameters, its instantaneous energy can sensitively reflect the fluctuation of dynamic detection parameters, reflecting the smoothness of the contact network at various locations, and can effectively avoid the influence of different detection equipment.

Claims

1. A method for comprehensive diagnosis of contact network based on multi-source detection parameters, characterized in that: The following steps are involved: S1. Automatically align the detection data; S2, generating a standard state for the detection data after automatic alignment; S3. Perform potential defect analysis on the standard state detection data, including time domain analysis and frequency domain analysis, to obtain potential defects of the contact network; The time domain analysis in step S3 includes change trend diagnosis and / or variance analysis; The change trend diagnosis is specifically as follows: Calculate the maximum difference between each acquisition point of the current detection data and the standard state area data i , if it is in the standard state area, it is 0, and the maximum difference and ∑diff between each positioning point are obtained by span i , using the sum of the maximum difference and the center value of the standard state area, calculate the change rate Δm of the test data compared with the standard state area; Synchronously analyze the change rate of the test data of different points for multiple times. If the change trend of the test data is the same for multiple times, and the test data gradually exceeds the standard state range, it is regarded as a sub-healthy state of the contact network. The variance analysis comprises: S311, respectively obtaining the variance value of the conductor height pull-out value of the entire line; S312, setting the conductor height weight value to twice the pull-out value weight; respectively adding the variance and the assigned value of the conductor height pull-out value to obtain the sum of the variances based on the conductor height pull-out value; S313, the variance of the detection data at different times is calculated from the multiple detection data. When the variance of the values ​​of the continuous high-conductor pull-out gradually increases, there is a hidden danger in the contact network; The frequency domain analysis in step S3 includes: S321, de-averaging the test data, and subtracting the overall mean from each test data, so that the overall mean of the data is 0; S322, using Hilbert-Huang transformation to realize time-frequency conversion of detection data, calculating the instantaneous energy of the signal through its time-frequency information, setting an energy threshold, and if the instantaneous energy exceeds the energy threshold, it is determined to be a potential defect area.

2. The method for comprehensive diagnosis of contact network based on multi-source detection parameters according to claim 1 is characterized in that: The specific steps of step S1 are: S11, standard data establishment and test data preprocessing; S12, positioning the anchor section area through data matching and calculation; S13, performing corresponding difference sampling operations on the data according to the matching results, and aligning the detection data waveform with the design curve.

3. The method for comprehensive diagnosis of overhead line based on multi-source detection parameters according to claim 2 is characterized in that: The standard data establishment includes: Using the design values ​​of the contact network's positioning points and anchor sections and the corresponding mileage information, standard data are established through the linear difference method.

4. The method for comprehensive diagnosis of overhead line based on multi-source detection parameters according to claim 2 is characterized in that: The detection data preprocessing includes: Starting from the first detected data, the difference is made with the subsequent data in sequence. When the difference value is greater than the set threshold T, the previous and next data are divided into two point sets. Finally, the points with the number of point sets less than the threshold M are removed to filter the abnormal jump points.

5. The method for comprehensive diagnosis of contact network with multi-source detection parameters according to claim 2, characterized in that: The step S12 is specifically as follows: S121, dividing the test data and the standard data into multiple parts based on the pull-out value mutation; S122, using the correlation coefficient, matching each anchor segment of the test data with the anchor segment of the standard data, and selecting the longest ordered subsequence from the matching results as the optimal solution; S123, by calculating the stretching scale of the optimal solution for matching with respect to the standard data, so as to obtain a stretching scale less than the threshold value L max result.

6. The method for comprehensive diagnosis of overhead line based on multi-source detection parameters according to claim 1, characterized in that: The step S2 comprises: According to the measured value of each detection point and the confidence interval statistics, the measurement standard range corresponding to each detection point is generated to generate the standard state area of ​​the detection data.

Images