Method and device for identifying abnormal change of spatial position of overhead line system

Abstract

The invention discloses an overhead line system spatial position abnormal change identification method and device. The method comprises the following steps: acquiring detection data of overhead line system geometric parameters acquired at different dates; calculating the variable quantity of the geometric parameters of the overhead line system at each sampling position along with time, and constructing a variable quantity sequence of the geometric parameters of the overhead line system at each sampling position; constructing a spatial position change feature vector of each span by taking the geometric parameter variable quantity sequence of the overhead line system as a vector component; and outlier diagnosis is carried out on each feature vector, and an overhead line system section with abnormal spatial position change is determined. The method can automatically identify the place where the spatial position of the overhead line system changes abnormally, improves the detection efficiency of the overhead line system, and reduces the detection cost of the overhead line system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of catenary detection, in particular to a catenary spatial position abnormal change identification method and device. BACKGROUND

[0002] This section is intended to provide background information to facilitate an understanding of embodiments of the application as set forth in the claims. The description herein does not constitute admission of prior art.

[0003] In an electrified railway system, as the operation time of high-speed railway continues to increase, the catenary components are subjected to continuous random vibration and frequent load, and the spatial position of the catenary may deviate from the initial state, thereby affecting the normal contact between the train pantograph and the catenary, and even causing drilling and punching of the pantograph, which seriously threatens the safety of train operation.

[0004] For the identification of abnormal changes in the spatial position of the catenary, existing solutions mainly include manual inspection and the use of some traditional monitoring equipment. Manual inspection relies on professional maintenance personnel to periodically view the catenary along the railway line, and to determine whether the spatial position of the catenary is abnormal based on experience. Traditional monitoring equipment, such as simple displacement sensors installed on catenary supports, can only monitor the small displacement changes in one direction of the support; some monitoring systems based on video images capture catenary images through fixed cameras, and then analyze the position of the catenary in the images by manual or simple image recognition algorithms.

[0005] However, these existing solutions have many shortcomings. Manual inspection is inefficient, requires a large amount of manpower, material resources and time cost, and due to the long railway line, the manual inspection interval is long, making it difficult to discover abnormal changes in the spatial position of the catenary in real time, and the detection results are largely dependent on the professional level and working state of the inspection personnel, which is subjective and has the risk of missed detection. In terms of traditional monitoring equipment, simple displacement sensors have single function and can only monitor limited parameters, and cannot fully reflect the complex spatial position information of the catenary; the monitoring system based on video images is greatly affected by environmental factors, such as light changes and bad weather (rain, snow, fog, etc.), which reduces the image quality and leads to reduced recognition accuracy, and simple image recognition algorithms cannot accurately and quickly identify subtle abnormal changes in the spatial position of the catenary. Therefore, there is an urgent need for a more efficient, accurate and reliable catenary spatial position abnormal change identification method to ensure the safe and stable operation of the electrified railway catenary. SUMMARY

[0006] The embodiments of the present application provide a catenary spatial position abnormal change identification method for automatically identifying places where the spatial position of the catenary has abnormally changed, improving the detection efficiency of the catenary and reducing the detection cost of the catenary, which comprises:

[0007] obtain detection data of the catenary geometric parameters collected on different dates; wherein the detection data comprises a plurality of data points; each data point has a unique sampling serial number within the same collection date, which is used to identify the collection sequence and the sampling position;

[0008] constitute a detection data set; in each detection data set, the data points are operated as follows in the order of increasing collection dates to obtain the catenary geometric parameter variation sequence corresponding to each sampling serial number: calculate the difference between the data point collected on the current date and the data point collected on the next date, and add the difference to the end of the catenary geometric parameter variation sequence; wherein the catenary geometric parameter variation sequence is initially empty;

[0009] arrange the catenary geometric parameter variation sequences corresponding to the sampling serial numbers in the order of increasing sampling serial numbers to construct the spatial position variation feature vectors of the spans; wherein the catenary section between the adjacent two catenary supports is a span;

[0010] calculate the local outlier factors of the spatial position variation feature vectors of all spans, and determine the span corresponding to the spatial position variation feature vector whose local outlier factor is greater than a preset threshold as the catenary section with abnormal spatial position variation.

[0011] The embodiment of the application also provides a catenary spatial position abnormal variation recognition device for automatically identifying the place where the catenary spatial position has abnormal variation, improving the catenary detection efficiency and reducing the catenary detection cost, and the device comprises:

[0012] a data acquisition module configured to obtain detection data of the catenary geometric parameters collected on different dates; wherein the detection data comprises a plurality of data points; each data point has a unique sampling serial number within the same collection date, which is used to identify the collection sequence and the sampling position;

[0013] a data analysis module configured to constitute a detection data set by using the data points with the same sampling serial number in the detection data collected on different dates; in each detection data set, the data points are operated as follows in the order of increasing collection dates to obtain the catenary geometric parameter variation sequence corresponding to each sampling serial number: calculate the difference between the data point collected on the current date and the data point collected on the next date, and add the difference to the end of the catenary geometric parameter variation sequence; wherein the catenary geometric parameter variation sequence is initially empty;

[0014] The feature vector construction module is configured to arrange the contact net geometry parameter variation sequence corresponding to each sampling number in ascending order of the sampling number, and construct a spatial position variation feature vector of each span, wherein the contact net section between the adjacent two contact net pillars is a span.

[0015] The abnormal variation identification module is configured to calculate a local outlier factor of the spatial position variation feature vector of all spans, and determine the span corresponding to the spatial position variation feature vector with the local outlier factor greater than a preset threshold as a contact net section with abnormal spatial position variation.

[0016] The embodiment of the present application also provides a computer device, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, and the processor implements the contact net spatial position abnormal variation identification method when executing the computer program.

[0017] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the contact net spatial position abnormal variation identification method.

[0018] The embodiment of the present application also provides a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the contact net spatial position abnormal variation identification method.

[0019] Compared with the prior art of manual inspection combined with traditional monitoring equipment, the traditional monitoring equipment combined with a simple image recognition algorithm, in the embodiment of the present application, the detection data of the contact net geometry parameter is analyzed, the variation of the contact net geometry parameter is analyzed from the time angle, the spatial variation feature vector of the contact net geometry parameter is constructed in units of spans, the calculation amount is reduced and the identification efficiency is improved; by using the characteristics that the spatial position of the contact wire will produce a specific mode response after the equipment state is abnormal, the outlier point diagnosis is performed on the constructed spatial variation feature vector, the local outlier factor of each spatial variation feature vector is calculated, and the automatic identification of the contact net spatial position abnormal variation section is realized. Different from the prior art of simple image recognition of the detection data of the traditional monitoring equipment, the embodiment of the present application relies on the detection data collected by the contact net detection vehicle, uses the data mining technology to obtain the useful information hidden in the data, and uses the useful information to represent the running state of the equipment, thereby improving the accuracy of the contact net detection, and achieving the purpose of accurately identifying the contact net section with abnormal spatial position variation, and avoiding the problems of low efficiency, high cost of manpower, material resources and time in the prior art which depends on manual inspection. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to make the technical solutions in the embodiments of the present application or the prior art clearer, the accompanying drawings needed in the embodiments or prior art description will be briefly introduced. Obviously, the accompanying drawings in the following description only constitute some embodiments of the present application, and other accompanying drawings can be obtained by those skilled in the art without any creative effort on the basis of these accompanying drawings. In the drawings:

[0021] Figure 1 A flow chart of the contact net spatial position abnormal change recognition method in the embodiment of the present application;

[0022] Figure 2 A schematic diagram of the system state value prediction principle in the embodiment of the present application;

[0023] Figure 3 A flow chart of the mileage correction method of the contact net geometric parameter detection data in the embodiment of the present application;

[0024] Figure 4 A schematic diagram of the contact net anchor section in the embodiment of the present application;

[0025] Figure 5 An example diagram of the pull-out value change curve in the embodiment of the present application;

[0026] Figure 6 A schematic diagram of the CCR operator construction principle in the embodiment of the present application;

[0027] Figure 7 An example diagram of the contact net diagnostic index of each span in the embodiment of the present application;

[0028] Figure 8 A schematic diagram of the contact net spatial position abnormal change recognition device in the embodiment of the present application. DETAILED DESCRIPTION

[0029] In order to make the technical solutions in the embodiments of the present application or the prior art clearer, the accompanying drawings needed in the embodiments or prior art description will be briefly introduced. Obviously, the accompanying drawings in the following description only constitute some embodiments of the present application, and other accompanying drawings can be obtained by those skilled in the art without any creative effort on the basis of these accompanying drawings. In the drawings:

[0030] The spatial position of the contact net may deviate from the initial state due to continuous random vibration and frequent load, thereby affecting the normal contact between the train pantograph and the contact net, and even causing the train to drill or hit the pantograph, which seriously threatens the safety of train operation. In order to avoid such failures, railway operating units usually invest a large amount of manpower and material resources to carry out comprehensive equipment state inspection and rectification, which has a significant effect, but the overall efficiency is low.

[0031] With the maintenance and repair mode of railway power supply specialty changing from "planned repair" to "state repair", it is urgent to serve production through strong scientific and technological means, reduce equipment maintenance and repair cost, and improve production efficiency. Therefore, the embodiment of the present application proposes a catenary spatial position abnormal change identification method, which has solved the above problems and improved the maintenance efficiency. Figure 1 The flow chart of the catenary spatial position abnormal change identification method in the embodiment of the present application is shown in the figure. Figure 1 As shown in the figure, the method comprises the following steps:

[0032] Step 101, acquiring detection data of catenary geometric parameters collected on different dates; wherein the detection data contains a plurality of data points; each data point has a unique sampling serial number within the same collection date, which is used to identify the collection sequence and sampling position;

[0033] Step 102, constructing a detection data set from data points with the same sampling serial number in detection data collected on different dates; in each detection data set, the data points are operated as follows in the order of increasing collection date: calculating the difference between the data point collected on the current date and the data point collected on the next date, and adding the difference to the end of the catenary geometric parameter change sequence; wherein the catenary geometric parameter change sequence is initially empty;

[0034] Step 103, arranging the catenary geometric parameter change sequence corresponding to each sampling serial number in the order of increasing sampling serial number to construct the spatial position change feature vector of each span; wherein the catenary section between adjacent two catenary supports is a span;

[0035] Step 104, calculating the local outlier factor of the spatial position change feature vector of all spans, and determining the span corresponding to the spatial position change feature vector with a local outlier factor greater than a preset threshold as the catenary section with abnormal spatial position change.

[0036] In the embodiment of the present application, the detection data of catenary geometric parameters is analyzed, the change amount of catenary geometric parameters is analyzed from the time angle, the spatial change feature vector of catenary geometric parameters is constructed in units of spans, the calculation amount is reduced and the identification efficiency is improved; by using the characteristics that the spatial position of the contact wire will produce a specific mode response after the device state appears abnormal, the outlier point diagnosis is performed on the constructed spatial change feature vector, and the automatic identification of the catenary section with abnormal spatial position change is realized. Relying on the detection data collected by the catenary detection vehicle, the data mining technology is used to obtain the useful information hidden in the data, so as to represent the running state of the equipment, improve the accuracy of catenary detection, and realize the purpose of accurately identifying the catenary section with abnormal spatial position change.

[0037] In order to realize accurate analysis on the changing state of the overhead contact system spatial position, first, detection data of the overhead contact system geometric parameters collected on different dates is acquired, and through in-depth analysis of the data, scientific decision basis is provided for railway maintenance.

[0038] In the embodiment of the application, detection data of the overhead contact system geometric parameters collected on different dates is acquired, wherein the detection data comprises a plurality of data points; each data point has a unique sampling serial number in the same collection date, which is used for identifying the collection sequence and the sampling position.

[0039] The overhead contact system geometric parameter detection device is installed on the roof of a vehicle, and is affected by strong light, rain and snow and other factors, so that the geometric detection data often contains interference data. In order to reduce the influence of the interference data on the subsequent steps, the detection data needs to be filtered. In one embodiment, H is the detection data of the overhead contact system geometric parameters, and specifically: H = h(i) is the data point at the sampling serial number i, i = 1, 2, …, k (k is the total number of sampling serial numbers); the detection data and the rectangular window function are used to define the system state value at each sampling serial number, as shown in formula (1).

[0040]

[0041] In the formula, is the system state value at the sampling serial number t, f δ (t) is the rectangular window function, and δ is the window length.

[0042] Then, the system state value at the sampling serial number t-1 is relied on to predict the system state value at the sampling serial number t. is compared with the true system state value at the sampling serial number t. If the difference between and exceeds a preset threshold value ε, it is considered that the coarse error point is contained in , and is corrected to Otherwise, it is considered that does not contain coarse error, and is discarded.

[0043] Figure 2 is the schematic diagram of the system state value prediction principle in the embodiment of the application. As shown in Figure 2 , taking the sampling serial number as the horizontal axis and the detection data of the geometric parameters as the vertical axis, the circular data points are the true measurement values, the square data points are the predicted values, and the data points in the gray rectangular frame are used to calculate the system state value at the sampling serial number t-1. The data points in the blue rectangular frame are used to calculate the system state value at the sampling serial number t. The data points in the red rectangular frame are used to predict the system state value at the sampling serial number t. ​

[0044] Specifically, for example, the rectangular window function truncates the data points h(t-1), h(t),..., h(t+delta-2) at the sampling sequence number t-1; using h(t-1), h(t),..., h(t+delta-2), the predicted value of the next data point h(t+delta-1) is calculated by a regression algorithm According to h(t),..., h(t+delta-2), The predicted value of the system state value at the sampling sequence number t is calculated

[0045] In the data acquisition process of the overhead contact line geometric parameter detection device, due to factors such as speed fluctuation and positioning error, there is a common mileage deviation problem in the detection data collected multiple times. In order to facilitate the analysis of multiple detection data at the same position, the mileage deviation existing in the overhead contact line geometric parameter detection data needs to be corrected. Figure 3 The flowchart of the mileage correction method of the overhead contact line geometric parameter detection data in the embodiment of the present application. In one embodiment, as shown in Figure 3 The mileage correction of the overhead contact line geometric parameter detection data can be realized according to the following steps:

[0046] Step 301, divide the reference data of the overhead contact line geometric parameter into multiple reference data segments according to the overhead contact line anchor segment; divide the detection data of the overhead contact line geometric parameter collected on different dates into multiple detection data segments according to the overhead contact line anchor segment;

[0047] Step 302, use the dynamic time warping algorithm to match each detection data segment with each reference data segment, determine the reference data segment with the highest similarity to each detection data segment; adjust the length of each detection data segment to be consistent with the length of the reference data segment with the highest similarity by using the linear interpolation method; use the reference data segment with the highest similarity to each detection data segment to correct the mileage of each detection data segment;

[0048] Step 303, according to the sequence of the overhead contact line anchor segment, splice multiple detection data segments after mileage correction to obtain the detection data of the overhead contact line geometric parameter collected on different dates after mileage correction.

[0049] In order to meet the needs of power supply and mechanical stress, the overhead contact line is divided into several independent anchor segments. In the connection section of two adjacent anchor segments, there are two contact lines at the same time, and the one in contact with the pantograph is the working branch, and the one away from the pantograph is the non-working branch. The conversion position of the working branch and the non-working branch is called the equal height point. Figure 4 The schematic diagram of the overhead contact line anchor segment in the embodiment of the present application. As shown in Figure 4As shown, the contour point is the boundary of two adjacent anchor sections, and the contour point of the catenary can be taken as the section point in step 301, and then the section of the catenary detection data is completed.

[0050] In order to focus on the change characteristics of the detection data in the time dimension, the embodiment of the application highlights the fluctuation and change trend of the data by the difference between adjacent data, so that the slight change is clear, and the calculation efficiency and the analysis accuracy are improved.

[0051] In the embodiment of the application, the data points with the same sampling sequence number in the detection data collected on different dates form a detection data set; in each detection data set, the data points are operated as follows in the order of increasing collection date, to obtain the catenary geometric parameter change sequence corresponding to each sampling sequence number: the difference between the data points collected on the current date and the data points collected on the next date is calculated, and the difference is added to the end of the catenary geometric parameter change sequence; wherein the catenary geometric parameter change sequence is initially empty.

[0052] For example, suppose that the data point at the collection sequence number i collected on date d is s d (i), the difference between the data points at the collection sequence number i collected on date d1 and date d2 is defined as shown in formula (2).

[0053]

[0054] In the formula: is the difference between the data points at the collection sequence number i collected on date d1 and date d2, is the data point at the collection sequence number i collected on date d1, is the data point at the collection sequence number i collected on date d2.

[0055] If the total number of collection dates is m, the catenary geometric parameter change sequence at the collection sequence number i can be expressed as

[0056] The section of the catenary between two adjacent catenary supports is called a span, and the span is the basic structural unit of the catenary. The entire catenary is divided into multiple relatively independent sections, and the data characteristics in each section are relatively consistent. This not only facilitates targeted analysis of each section, but also significantly reduces the overall calculation amount and improves the efficiency and accuracy of data processing.

[0057] Outlier diagnosis is a data analysis method for identifying abnormal patterns and special events, and the analysis effect is limited by the complexity of the data. Therefore, the spatial position change feature vector is designed, which can not only accurately capture the intrinsic characteristics of the data, but also reduce the calculation amount and improve the detection efficiency.

[0058] In the embodiment of the present application, the contact network geometric parameter variation sequence corresponding to each sampling number is arranged in the order of increasing sampling number to construct the spatial position variation feature vector of each span, wherein the contact network section between the adjacent two contact network pillars is a span.

[0059] The contact network pillar is not only a key component for supporting the contact network structure, but also provides a basis for the zoning and anchoring of the contact network. The contact network section between the adjacent two contact network pillars is called a span, and the span is a basic unit for the design, construction and maintenance of the contact network. By accurately positioning the contact network pillar, the boundary of each span can be determined, thereby providing an accurate spatial reference for the geometric parameter measurement, state evaluation and fault maintenance of the contact network.

[0060] For example, if the collection number of the first data point in the lth span is i, the spatial position variation feature vector of the lth span is represented as n is the total amount of data points contained in the lth span.

[0061] In one embodiment, the contact network geometric parameters include a pull-out value, and the method further comprises obtaining the position of the contact network pillar by using the pull-out value and the sampling number to form a coordinate point, connecting the coordinate points in the order of increasing sampling number to obtain a pull-out value variation curve, performing corner point detection on the pull-out value variation curve to obtain a plurality of corner points of the pull-out value variation curve, and determining the position identified by the collection number of the corner point as the position of the contact network pillar.

[0062] Figure 5 FIG. 1 is an example diagram of the pull-out value variation curve in the embodiment of the present application. For example, as shown in FIG. 1, the dashed line represents the reference data before the local variation of the pull-out value, and the solid line represents the pull-out value variation curve after the local variation of the pull-out value, wherein the overlapping part of the dashed line and the solid line is represented by the solid line. Figure 5

[0063] ​In one embodiment, the coordinate points on the pull-out value change curve are traversed in the following manner to obtain multiple corner point feature quantities: the Nth coordinate point before the current coordinate point is determined as the first endpoint, and the Nth coordinate point after the current coordinate point is determined as the second endpoint, where N is a preset positive integer; if the current coordinate point does not have a corresponding first endpoint and second endpoint, then the current coordinate point has no corner point feature quantity; the arc length of the pull-out value change curve between the first endpoint and the current coordinate point is determined as the first arc length, and the arc length of the pull-out value change curve between the current coordinate point and the second endpoint is determined as the second arc length; the sum of the first arc length and the second arc length, as well as the chord length of the pull-out value change curve between the first endpoint and the second endpoint are determined; and the ratio of the chord length to the sum of the arc lengths is determined as the corner point feature quantity of the current coordinate point. Then, the corner point feature quantities are connected in ascending order of the acquisition serial numbers to obtain the corner point feature curve; a non-maximum suppression algorithm is used to obtain multiple local maxima of the corner point feature curve, and the coordinate points of the pull-out value change curve corresponding to the multiple local maxima are determined as multiple corner points of the pull-out value change curve; the positions identified by the acquisition serial numbers corresponding to the multiple corner points are determined as the positions of the contact network pillars.

[0064] For example, the pull-out value detection data is regarded as a two-dimensional contour curve consisting of the acquisition sequence number and the pull-out value, and the curve to chord ratio (CCR) operator is used to construct the corner feature, as shown in formula (3).

[0065]

[0066] Where u is the sampling number, C(u) is the characteristic value of the corner point at sampling number u, d1 is the first arc length, d2 is the second arc length, and d3 is the chord length of the pull-out value change curve between the first endpoint and the second endpoint.

[0067] Figure 6 This is a schematic diagram of the construction principle of the CCR operator in an embodiment of the present invention. CCR uses a simple triangle side length principle to achieve corner point detection. The construction principle is as follows: Figure 6 As shown: the pull-out value change curve is represented by s(u); taking u as the starting point, extrapolate δ points forward (δ is a positive integer greater than 1) to obtain the point s(u-δ) on the contour; similarly, taking u as the starting point, extrapolate δ points backward to obtain the point s(u+δ) on the contour; connect the points s(u), s(u-δ), and s(u+δ) with straight lines to obtain three chords, which together form a triangle; according to the triangle principle, the sum of two sides must be greater than the third side, and the arc lengths d1 and d2 are both greater than the corresponding chord lengths, so the sum of the arc lengths d1 and d2 must be greater than the chord length d3.

[0068] After obtaining the positioning result of the position of the overhead line support, in order to further improve the positioning accuracy, in an embodiment, the positioning result can also be evaluated as follows: the reliability of the initial positioning point is evaluated, the closer the distance between two adjacent initial positioning points, the greater the penalty coefficient value applied to the angular point feature quantity; if the angular point feature quantity after the penalty is less than a preset threshold, the initial positioning point is removed; if the distance between two adjacent initial positioning points is significantly greater than the distance between any two adjacent positioning points in the vicinity, then a positioning point is supplemented at the maximum value in the interval with the two adjacent initial positioning points (angular points) as the endpoints.

[0069] For example, let o = {1, -1}, o = 1 represents the next positioning point, and o = -1 represents the previous positioning point; when 1≤i+o≤M (M is the total number of positioning points), the initial positioning point k j The distance between k j+o , that is, the adjacent span, is denoted as The reliability evaluation index of the initial positioning point is shown in formula (4).

[0070]

[0071] In the formula, C(k ) is the reliability evaluation index of the initial positioning point k j , j represents the serial number of the initial positioning point, P j is the penalty coefficient, C(k j ) is the angular point feature quantity of the initial positioning point k j , and ω is a constant term; θ min is the actual minimum allowable span.

[0072] As can be seen from the above formula, the smaller the minimum value of the adjacent span i and of the initial positioning point k i , the smaller the value of i , that is, the closer the distance between the initial positioning point k and other initial positioning points, and the greater the penalty coefficient value P

[0001] applied to the angular point feature quantity.

[0073] In the embodiment of the application, the local outlier factor of the spatial position change feature vector of all spans is calculated, and the span corresponding to the spatial position change feature vector with a local outlier factor greater than a preset threshold is determined as the overhead line section with abnormal spatial position change.

[0074] The main reasons for changes in the contact network geometry detection data are differences in operating conditions and abnormal equipment status. Among them, the difference in operating conditions refers to the inconsistency of the operating conditions of the detection device when collecting the contact network geometry parameter detection data, such as inconsistent pantograph models, inconsistent detection speeds, etc., which leads to changes in the dynamic matching relationship between the pantograph and the network, and then causes overall differences in the contact network geometry parameter detection data. Abnormal equipment status refers to abnormalities in the state of the contact network equipment, such as pillar tilt, pillar foundation settlement, positioning support slippage, etc., which lead to local changes in the spatial position of the contact line. Differences in operating conditions are normal. Therefore, how to avoid the impact of operating conditions and accurately identify changes in geometric parameter detection data caused by abnormal equipment status is the main technical difficulty in identifying abnormal changes in the spatial position of the contact network.

[0075] In order to solve this problem, the spans are classified based on the pantograph model, detection speed and other operating conditions. The places where the contact network geometric parameter detection data changes in each category are regarded as outliers. The problem of identifying abnormal changes in the spatial position of the contact network can be transformed into a situational outlier diagnosis problem.

[0076] In one embodiment, a classification vector of each span is constructed with operating conditions as components, and spans with the same classification vector are recorded as the same span category, thereby obtaining multiple span categories reflecting the operating conditions; the operating conditions include: detection speed, pantograph model; in each span category, the local outlier factor of the spatial position change characteristic vector of all spans is calculated, and the span corresponding to the spatial position change characteristic vector with a local outlier factor greater than a preset threshold is determined as: under the operating conditions represented by each span category, the contact network section with abnormal spatial position changes.

[0077] In one embodiment, before calculating the local outlier factors of the spatial position change feature vectors of all spans, it also includes: recording the span categories with fewer spans than a preset number as the first span category; calculating the similarity between the classification vectors of other span categories and the classification vector of the first span category, and recording the span category with the highest similarity to the classification vector of the first span category as the second span category; adding the spans in the first span category to the second span category, and deleting the first span category.

[0078] For example, scenario outlier diagnosis is implemented as follows:

[0079] 1. Let the number of eigenvectors contained in a certain cross-category be m, and let the set composed of the eigenvectors in the cross-category be Calculate any object in P K-distance This distance represents With another object The distance between satisfy:

[0080] 1) At least K objects such that

[0081] 2) At least K-1 objects such that

[0082] 2, Calculate the K-distance neighborhood of object , which is all objects in the data set P with the distance to not greater than , recorded as:

[0083] 3, Calculate the reachable distance from another object τ to object , recorded as:

[0084] 4, Calculate the local reachable density of object , recorded as:

[0085] 5, Define the local outlier factor of object , recorded as:

[0086] 6, Take as the diagnostic index of abnormal changes of the spatial position of the contact network, and output the span whose diagnostic index is greater than the preset threshold as the section of abnormal changes of the spatial position of the contact network.

[0087] Figure 7 is an example diagram of the diagnostic index of the contact network of each span in the embodiment of the application. For example, as shown in Figure 7 , K takes the value of 5, wherein the span with larger spatial position changes corresponds to the diagnostic index LOF5 which is significantly greater than other spans, indicating that the diagnostic index constructed in the embodiment of the application has a significant effect in distinguishing local changes of the geometric parameters of the contact network.

[0088] The embodiment of the application also provides a contact network spatial position abnormal change recognition device, as described in the following embodiment. Since the principle of solving the problem of the device is similar to that of the contact network spatial position abnormal change recognition method, the implementation of the device can be referred to the implementation of the contact network spatial position abnormal change recognition method, and the repeated parts will not be described again.

[0089] Figure 8 is a schematic diagram of the contact network spatial position abnormal change recognition device in the embodiment of the application. As shown in Figure 8 , the device comprises:

[0090] The data acquisition module 801 is configured to acquire detection data of the overhead line geometric parameters collected on different dates, wherein the detection data comprises a plurality of data points, each data point has a unique sampling serial number in the same collection date, and the unique sampling serial number is used to identify a collection sequence and a sampling position.

[0091] The data analysis module 802 is configured to construct a detection data set by using data points with the same sampling serial number in the detection data collected on different dates, and perform the following operation on the data points in each detection data set in an ascending order of collection dates to obtain a variation sequence of the overhead line geometric parameters corresponding to each sampling serial number: calculate a difference value between a data point collected on a current date and a data point collected on a next date, and add the difference value to an end of the variation sequence of the overhead line geometric parameters, wherein the variation sequence of the overhead line geometric parameters is initially empty.

[0092] The feature vector construction module 803 is configured to arrange the variation sequence of the overhead line geometric parameters corresponding to each sampling serial number in an ascending order of the sampling serial number, and construct a spatial position variation feature vector of each span, wherein a section of the overhead line between two adjacent overhead line supports is a span.

[0093] The abnormal variation identification module 804 is configured to calculate a local outlier factor of the spatial position variation feature vector of all spans, and determine a span corresponding to a spatial position variation feature vector with a local outlier factor greater than a preset threshold as an overhead line section with abnormal spatial position variation.

[0094] In an embodiment, the data acquisition module 801 is further configured to, after acquiring the detection data of the overhead line geometric parameters collected on different dates:

[0095] divide the reference data of the overhead line geometric parameters into a plurality of reference data segments according to the overhead line anchor sections, and divide the detection data of the overhead line geometric parameters collected on different dates into a plurality of detection data segments according to the overhead line anchor sections.

[0096] match each detection data segment with each reference data segment by using a dynamic time warping algorithm to determine a reference data segment with the highest similarity to each detection data segment, adjust a length of each detection data segment to be consistent with a length of the reference data segment with the highest similarity by using a linear interpolation method, and perform mile correction on each detection data segment by using the reference data segment with the highest similarity to each detection data segment.

[0097] splice a plurality of mile-corrected detection data segments in an order of the overhead line anchor sections to obtain the detection data of the mile-corrected overhead line geometric parameters collected on different dates.

[0098] In an embodiment, the overhead line geometric parameter comprises a pull-out value.

[0099] The apparatus further includes a support positioning module 805, configured to:

[0100] The position of the catenary support is obtained in the following manner:

[0101] The coordinate points are composed of the pulling-out value and the sampling serial number, and the coordinate points are connected in ascending order of the sampling serial number to obtain a pulling-out value change curve;

[0102] The pulling-out value change curve is subjected to corner point detection to obtain a plurality of corner points of the pulling-out value change curve; and the position indicated by the sampling serial number corresponding to the corner points is determined as the position of the catenary support.

[0103] In an embodiment, the support positioning module 805 is specifically configured to:

[0104] The coordinate points on the pulling-out value change curve are traversed in the following manner to obtain a plurality of corner point feature quantities: the Nth coordinate point before the current coordinate point is determined as a first end point, and the Nth coordinate point after the current coordinate point is determined as a second end point, N being a preset positive integer; if the current coordinate point has no first end point and second end point, the current coordinate point has no corner point feature quantity; the arc length of the pulling-out value change curve between the first end point and the current coordinate point is determined as a first arc length, and the arc length of the pulling-out value change curve between the current coordinate point and the second end point is determined as a second arc length; the sum of the first arc length and the second arc length, and the chord length of the pulling-out value change curve between the first end point and the second end point are determined; and the ratio of the chord length to the sum of the arc lengths is determined as the corner point feature quantity of the current coordinate point;

[0105] The corner point feature quantities are connected in ascending order of the sampling serial number to obtain a corner point feature curve; a non-maximum suppression algorithm is adopted to obtain a plurality of local maximum values of the corner point feature curve, and the coordinate points of the pulling-out value change curve corresponding to the plurality of local maximum values are determined as the plurality of corner points of the pulling-out value change curve;

[0106] The positions indicated by the sampling serial numbers corresponding to the plurality of corner points are determined as the positions of the catenary supports.

[0107] In an embodiment, the abnormal change identification module 804 is specifically configured to:

[0108] The classification vectors of the spans are constructed with the running conditions as components, the spans with the same classification vectors are recorded as the same span category, and a plurality of span categories reflecting the running conditions are obtained; the running conditions include: detection speed, pantograph model;

[0109] In each span category, the local outlier factors of the spatial position change feature vectors of all spans are calculated, and the span corresponding to the spatial position change feature vector with a local outlier factor greater than a preset threshold is determined as a catenary section with abnormal spatial position change under the running condition represented by the span category.

[0110] In one embodiment, the abnormal change identification module 804 is further configured to, before calculating the local outlier factors of all the spatial position change feature vectors of the spans:

[0111] record a span category with a number of spans less than the preset number of spans as a first span category;

[0112] calculate the similarity between the classification vector of the first span category and the classification vectors of other span categories, and record a span category with the highest similarity to the classification vector of the first span category as a second span category;

[0113] add the spans in the first span category to the second span category, and delete the first span category.

[0114] The embodiment of the present application also provides a computer device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the above-mentioned catenary spatial position abnormal change identification method when executing the computer program.

[0115] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the computer program implements the above-mentioned catenary spatial position abnormal change identification method when executed by a processor.

[0116] The embodiment of the present application also provides a computer program product, which comprises a computer program, and the computer program implements the above-mentioned catenary spatial position abnormal change identification method when executed by a processor.

[0117] Compared with the prior art of manual inspection combined with traditional monitoring equipment and the prior art of traditional monitoring equipment combined with a simple image recognition algorithm, in the embodiment of the present application, the detection data of the catenary geometric parameters is analyzed, the variation of the catenary geometric parameters is analyzed from the time angle, the spatial change feature vector of the catenary geometric parameters is constructed in units of spans, the calculation amount is reduced, and the identification efficiency is improved; the spatial position of the contact line will produce a specific mode response after the device state is abnormal, and the spatial change feature vector constructed is subjected to outlier diagnosis; each span is classified according to the operating conditions, the local outlier factors of each spatial change feature vector in different span categories are calculated, the automatic identification of the catenary spatial position abnormal change section is realized, the influence of the operating conditions can be avoided, the change of the geometric parameter detection data caused by the abnormal device state can be accurately identified, the accuracy of the catenary detection is improved, and the purpose of accurately identifying the catenary section with abnormal spatial position change is achieved, thereby avoiding the problems of low efficiency, high labor cost, material cost and time cost in the prior art of relying on manual inspection.

[0118] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0119] The present application is described in reference to the flowchart and / or block diagrams of the method, apparatus (system) and computer program product according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing device or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for carrying out each of the one or more functions specified in the flowchart and / or block diagram block or blocks.

[0120] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for carrying out each of the one or more functions specified in the flowchart and / or block diagram block or blocks.

[0121] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for carrying out each of the one or more functions specified in the flowchart and / or block diagram block or blocks.

[0122] The specific embodiments described above are illustrative for purposes of the present application. The particular implementations are not intended to limit the scope of the present application, which is defined by the appended claims. Numerous variations, changes, and substitutions can be made without departing from the application. It is the intention, therefore, to limit the application only as indicated by the scope of the claims.

Claims

1. A method for identifying abnormal changes in the spatial position of a contact network, characterized in that: include: Acquire detection data of contact network geometric parameters collected on different dates; wherein the detection data includes multiple data points; within the same collection date, each data point has a unique sampling sequence number for identifying the collection order and sampling location; The data points with the same sampling number in the test data collected on different dates are combined into a test data set; in each test data set, the data points are subjected to the following operations in ascending order of collection date to obtain a catenary geometric parameter variation sequence corresponding to each sampling number: the difference between the data point collected on the current date and the data point collected on the next date is calculated, and the difference is added to the end of the catenary geometric parameter variation sequence; wherein the catenary geometric parameter variation sequence is initially empty; The contact network geometric parameter variation sequence corresponding to each sampling number is arranged in ascending order of the sampling number to construct the spatial position variation feature vector of each span; wherein the contact network section between two adjacent contact network pillars is a span; The local outlier factors of the spatial position change characteristic vectors of all spans are calculated, and the spans corresponding to the spatial position change characteristic vectors whose local outlier factors are greater than a preset threshold are determined as the contact network sections with abnormal spatial position changes.

2. The method according to claim 1, wherein After obtaining the detection data of the contact network geometric parameters collected on different dates, it also includes: Based on the catenary anchor section, the reference data of the catenary geometric parameters are divided into a plurality of reference data segments; based on the catenary anchor section, the detection data of the catenary geometric parameters collected on different dates are divided into a plurality of detection data segments; Using a dynamic time warping algorithm, each test data segment is matched with each reference data segment to determine the reference data segment with the highest similarity to each test data segment. Using linear interpolation, the length of each test data segment is adjusted to match the length of the reference data segment with the highest similarity. Mileage correction is performed on each test data segment using the reference data segment with the highest similarity to each test data segment. According to the order of the catenary anchor sections, multiple mileage-corrected detection data segments are spliced ​​together to obtain the mileage-corrected catenary geometric parameter detection data collected on different dates.

3. The method according to claim 1, wherein The contact network geometric parameters include: pull-out value; The method further comprises obtaining the position of the contact network support by: Use the pull-out value and the sampling sequence number to form the coordinate points, and connect the coordinate points in the order of increasing sampling sequence numbers to obtain the pull-out value change curve; Corner point detection is performed on the pull-out value variation curve to obtain multiple corner points of the pull-out value variation curve; the position identified by the acquisition serial number corresponding to the corner point is determined as the position of the contact network pillar.

4. The method according to claim 3, wherein Performing corner point detection on the pull-out value variation curve to obtain multiple corner points of the pull-out value variation curve; determining the position identified by the acquisition serial number corresponding to the corner point as the position of the contact network support, including: The coordinate points on the pull-out value change curve are traversed in the following manner to obtain multiple corner point feature quantities: the Nth coordinate point before the current coordinate point is determined as the first endpoint, and the Nth coordinate point after the current coordinate point is determined as the second endpoint, where N is a preset positive integer; if the current coordinate point does not have a corresponding first endpoint and second endpoint, then the current coordinate point has no corner point feature quantity; the arc length of the pull-out value change curve between the first endpoint and the current coordinate point is determined as the first arc length, and the arc length of the pull-out value change curve between the current coordinate point and the second endpoint is determined as the second arc length; the sum of the first arc length and the second arc length, as well as the chord length of the pull-out value change curve between the first endpoint and the second endpoint, is determined; and the ratio of the chord length to the sum of the arc lengths is determined as the corner point feature quantity of the current coordinate point; Connecting corner feature quantities in ascending order of acquisition sequence numbers to obtain a corner feature curve; using a non-maximum suppression algorithm to obtain multiple local maxima of the corner feature curve, and determining the coordinate points of the pull-out value change curve corresponding to the multiple local maxima as multiple corner points of the pull-out value change curve; The positions identified by the collection serial numbers corresponding to the multiple corner points are determined as the positions of the contact network pillars.

5. The method according to claim 1, wherein Calculate the local outlier factors of the spatial position change feature vectors of all spans, and identify the spans corresponding to the spatial position change feature vectors whose local outlier factors are greater than the preset threshold as the contact network sections with abnormal spatial position changes, including: Using the operating conditions as components, a classification vector for each span is constructed. Spans with the same classification vector are recorded as the same span category, resulting in multiple span categories reflecting the operating conditions. The operating conditions include: detection speed, pantograph model; In each span category, the local outlier factors of the spatial position change characteristic vectors of all spans are calculated, and the spans corresponding to the spatial position change characteristic vectors whose local outlier factors are greater than the preset threshold are determined as: the contact network sections with abnormal spatial position changes under the operating conditions represented by each span category.

6. The method according to claim 5, wherein Before calculating the local outlier factors of all span spatial position variation eigenvectors, also include: The span categories with fewer spans than the preset number are recorded as the first span category; Calculate the similarity between the classification vectors of other cross-categories and the classification vector of the first cross-category, and record the cross-category with the highest similarity to the classification vector of the first cross-category as the second cross-category; Add the spans in the first span category to the second span category and delete the first span category.

7. A device for identifying abnormal changes in the spatial position of a contact network, characterized in that: include: A data acquisition module is used to acquire detection data of contact network geometric parameters collected on different dates; wherein the detection data includes multiple data points; within the same collection date, each data point has a unique sampling sequence number for identifying the collection order and sampling location; The data analysis module is used to form a detection data set from data points with the same sampling sequence number in the detection data collected on different dates; in each detection data set, the following operations are performed on the data points in ascending order of collection date to obtain a contact network geometric parameter change sequence corresponding to each sampling sequence number: the difference between the data point collected on the current date and the data point collected on the next date is calculated, and the difference is added to the end of the contact network geometric parameter change sequence; wherein the contact network geometric parameter change sequence is initially empty; The feature vector construction module is used to arrange the contact network geometric parameter variation sequence corresponding to each sampling number in the order of increasing sampling number to construct the spatial position variation feature vector of each span; wherein the contact network section between two adjacent contact network pillars is a span; The abnormal change identification module is used to calculate the local outlier factors of the spatial position change feature vectors of all spans, and determine the spans corresponding to the spatial position change feature vectors whose local outlier factors are greater than a preset threshold as the contact network sections with abnormal spatial position changes.

8. The device according to claim 7, wherein The data acquisition module is also used to obtain the detection data of the contact network geometric parameters collected on different dates: Based on the catenary anchor section, the reference data of the catenary geometric parameters are divided into multiple reference data segments; Based on the catenary anchor section, the detection data of the catenary geometric parameters collected on different dates are divided into multiple detection data segments; Using a dynamic time warping algorithm, each test data segment is matched with each reference data segment to determine the reference data segment with the highest similarity to each test data segment; using linear interpolation, the length of each test data segment is adjusted to be consistent with the length of the reference data segment with the highest similarity; Perform mileage correction on each detection data segment using the reference data segment with the highest similarity to each detection data segment; According to the order of the catenary anchor sections, multiple mileage-corrected detection data segments are spliced ​​together to obtain the mileage-corrected catenary geometric parameter detection data collected on different dates.

9. The device according to claim 7, wherein The contact network geometric parameters include: pull-out value; The device also includes a pillar positioning module for: The position of the catenary support is obtained as follows: Use the pull-out value and the sampling sequence number to form the coordinate points, and connect the coordinate points in the order of increasing sampling sequence numbers to obtain the pull-out value change curve; Corner point detection is performed on the pull-out value variation curve to obtain multiple corner points of the pull-out value variation curve; the position identified by the acquisition serial number corresponding to the corner point is determined as the position of the contact network pillar.

10. The device according to claim 9, wherein Pillar positioning module, specifically used for: The coordinate points on the pull-out value change curve are traversed in the following manner to obtain multiple corner point features: the Nth coordinate point before the current coordinate point is determined as the first endpoint, and the Nth coordinate point after the current coordinate point is determined as the second endpoint, where N is a preset positive integer; if the current coordinate point does not have a corresponding first endpoint and second endpoint, then the current coordinate point has no corner point feature; Determine the arc length of the value change curve between the first endpoint and the current coordinate point as the first arc length, and determine the arc length of the value change curve between the current coordinate point and the second endpoint as the second arc length; determine the sum of the first arc length and the second arc length, and the chord length of the value change curve between the first endpoint and the second endpoint; and determine the ratio of the chord length to the sum of the arc lengths as the corner point feature quantity of the current coordinate point; Connecting corner feature quantities in ascending order of acquisition sequence numbers to obtain a corner feature curve; using a non-maximum suppression algorithm to obtain multiple local maxima of the corner feature curve, and determining the coordinate points of the pull-out value change curve corresponding to the multiple local maxima as multiple corner points of the pull-out value change curve; The positions identified by the collection serial numbers corresponding to the multiple corner points are determined as the positions of the contact network pillars.

11. The device according to claim 7, wherein The abnormal change identification module is specifically used to: Using the operating conditions as components, a classification vector for each span is constructed. Spans with the same classification vector are recorded as the same span category, resulting in multiple span categories reflecting the operating conditions. The operating conditions include: detection speed, pantograph model; In each span category, the local outlier factors of the spatial position change characteristic vectors of all spans are calculated, and the spans corresponding to the spatial position change characteristic vectors whose local outlier factors are greater than the preset threshold are determined as: the contact network sections with abnormal spatial position changes under the operating conditions represented by each span category.

12. The device according to claim 11, wherein The abnormal change identification module is also used before calculating the local outlier factors of all span spatial position change feature vectors: The span categories with fewer spans than the preset number are recorded as the first span category; Calculate the similarity between the classification vectors of other cross-categories and the classification vector of the first cross-category, and record the cross-category with the highest similarity to the classification vector of the first cross-category as the second cross-category; Add the spans in the first span category to the second span category and delete the first span category.

13. A computer 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 method according to any one of claims 1 to 6 is implemented.

14. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

15. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images