Abnormal event analysis and processing method for railway infrastructure

Abstract

The invention relates to the technical field of abnormal event analysis and processing, in particular to an abnormal event analysis and processing method for railway infrastructure. The method comprises the steps of obtaining credibility factors corresponding to similar windows in a similar window set corresponding to data to be filled; according to the credibility factor corresponding to the similar window, obtaining a filling value of the to-be-filled data, replacing the to-be-filled data with the filling value corresponding to the to-be-filled data, and recording the replaced target to-be-filled sequence as a second target sequence corresponding to the target rail section; and according to the standard deviation of the first target sequence and the second target sequence, obtaining a quality characterization value of the target rail section, and according to the quality characterization value, obtaining an analysis processing result of the abnormal event of the target rail section. According to the invention, the reliability of analyzing and processing the railway abnormal event can be ensured or improved.

Description

Technical Field

[0001] The present invention relates to the technical field of abnormal event analysis and processing, and in particular to an abnormal event analysis and processing method for railway infrastructure. Background Art

[0002] Railway abnormal events refer to abnormal situations that occur during railway transportation, and will have many impacts on railway transportation, such as cargo transportation delays, passenger travel obstructions, etc. Therefore, it is necessary to prevent the occurrence of railway abnormal events; and because the quality problems of railway infrastructure will also cause the occurrence of railway abnormal events, the quality of railway infrastructure will be monitored, especially railway tracks, such as monitoring the unevenness of railway tracks, and analyzing and evaluating the probability of abnormal events caused by the monitored tracks based on the quality monitoring results, and determining whether to check and repair the monitored tracks based on the evaluation results. In addition, when monitoring the quality of railway tracks, collection equipment is used to collect a variety of rail data, such as the left height, right height, left track direction, right track direction, track gauge, level and triangular pit of the rail, which are all rail data. Then, the collected data is analyzed and processed to obtain the quality monitoring results.

[0003] However, in the process of data collection, the collection equipment may cause the loss of transmitted data or the omission of certain data due to factors such as electromagnetic interference and communication problems between devices, that is, there are missing values ​​in the collected data sequence, and the existence of missing values ​​will directly affect the accuracy and reliability of the track quality monitoring results. Therefore, when there are missing values ​​in the collected data sequence, the traditional hot card filling method is generally used to fill the positions where data needs to be filled. However, the hot card filling method generally first constructs a local window at the location of the missing value in the acquired data sequence, and the data sequence includes but is not limited to the left high and low data sequence, the right high and low data sequence, the left track data sequence, etc. of the rail, and then obtains a local window that is most similar to the local window at the location of the missing value, and then determines the filling value of the missing value position based on the most similar local window found. However, this method of determining the filling value of the missing value position based only on a most similar local window will make the obtained filling value less reliable, and when the filling value is filled When it is relatively unreliable, the consistency and reliability of the filled data sequence will be poor. If the data in the most similar local window is seriously disturbed by external environmental factors during collection, that is, the data in the most similar local window is seriously affected by electromagnetic interference, then the most similar local window found at this time is not reliable, and the filling value of the missing value position obtained based on the most similar local window is also unreliable. That is, filling the missing values ​​according to the traditional filling method will make the consistency and reliability of the filled data sequence poor. When the consistency and reliability of the data sequence after the missing value filling are poor, the accuracy and reliability of the obtained track quality monitoring results will be low. When the accuracy and reliability of the obtained track quality monitoring results are low, the subsequent analysis and processing results of railway abnormal events will be relatively unreliable. Therefore, how to reliably fill the missing positions to ensure the reliability of the analysis and processing of railway abnormal events has become an urgent problem to be solved. Summary of the invention

[0004] In order to solve the above problems, the present invention provides an abnormal event analysis and processing method for railway infrastructure, and the technical solution adopted is as follows: An embodiment of the present invention provides an abnormal event analysis and processing method for railway infrastructure, comprising the following steps: Acquire a target sequence to be filled and a corresponding first target sequence corresponding to the target rail section, wherein the target sequence to be filled contains data to be filled; Taking the data to be filled in the target sequence to be filled as the center, constructing a target window corresponding to each data to be filled in the target sequence to be filled, and sliding the target window corresponding to the data to be filled on the target sequence to be filled to obtain a set of windows to be analyzed corresponding to the data to be filled; obtaining a set of similar windows corresponding to the data to be filled according to the target window corresponding to the data to be filled and the set of windows to be analyzed corresponding to the data to be filled; According to all similar windows in the similar window set, a credibility factor corresponding to each similar window in the similar window set is obtained; According to the credibility factor corresponding to the similarity window, a filling value of the data to be filled is obtained, and the data to be filled is replaced with the filling value corresponding to the data to be filled, and the target sequence to be filled after the replacement is recorded as a second target sequence corresponding to the target rail section; A quality characterization value of the target rail section is obtained according to the standard deviation of the first target sequence and the second target sequence, and an analysis and processing result of an abnormal event of the target rail section is obtained according to the quality characterization value.

[0005] Beneficial effects: The present invention first obtains a target sequence to be filled and a corresponding first target sequence corresponding to a target rail section; then, with the data to be filled in the target sequence to be filled as the center, a target window corresponding to each data to be filled in the target sequence to be filled is constructed, and the target window corresponding to the data to be filled is slid on the target sequence to be filled to obtain a set of windows to be analyzed corresponding to the data to be filled, and a set of similar windows corresponding to the data to be filled is obtained according to the target window corresponding to the data to be filled and the set of windows to be analyzed corresponding to the data to be filled; then, according to all similar windows in the similar window set, a credibility factor corresponding to each similar window in the similar window set is obtained; then, according to the credibility factor corresponding to the similar window, a filling value of the data to be filled is obtained, and the data to be filled is replaced with the filling value of the corresponding data to be filled, and the replaced target sequence to be filled is recorded as the second target sequence corresponding to the target rail section; finally, according to the standard deviation of the first target sequence and the second target sequence, a quality characterization value of the target rail section is obtained, and according to the quality characterization value, an analysis and processing result of abnormal events of the target rail section is obtained. Moreover, the filling value obtained by analyzing the credibility factors of multiple similar windows in the present invention is more reliable and accurate, and the consistency and reliability of the first target sequence determined based on the more reliable and accurate filling value are better. Therefore, the quality characterization value obtained based on the target sequence with better consistency and reliability is more reliable and accurate, and based on the more reliable and accurate quality characterization value, the reliability of analyzing and processing abnormal railway events can be guaranteed or improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0007] Figure 1 The present invention is a flow chart of an abnormal event analysis and processing method for railway infrastructure. DETAILED DESCRIPTION

[0008] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the embodiments of the present invention.

[0009] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0010] This embodiment provides an abnormal event analysis and processing method for railway infrastructure, which is described in detail as follows: like Figure 1 As shown, the abnormal event analysis and processing method for railway infrastructure includes the following steps: Step S001, obtaining a target sequence to be filled and a corresponding first target sequence corresponding to a target rail section, wherein the target sequence to be filled contains data to be filled.

[0011] During the data collection process, factors such as electromagnetic interference and communication problems between devices may lead to data loss, that is, electromagnetic interference, communication problems between devices and other factors may cause missing values ​​in the collected rail data sequence, and the presence of missing values ​​may affect the subsequent analysis and processing of abnormal railway events. In this embodiment, the reliability of the analysis and processing of abnormal railway events is ensured by reliably filling the missing value positions in the rail data sequence, that is, in this embodiment, the missing value positions in the rail data sequence are reliably filled, and then the quality inspection results of the rails are obtained based on the filled sequence, and then the abnormal railway events are analyzed and processed based on the quality inspection results of the rails. In addition, when the quality of the rails is currently monitored, it is generally monitored in sections. In order to facilitate understanding, in this embodiment, any rail section will be analyzed as an example, and the rail section will be recorded as the target rail section, that is, the rail sections that appear later in this embodiment are the same rail section. In addition, in a specific application, the implementer also needs to set the length of the rail section according to the actual situation. For example, in this embodiment, the length of the rail section can be set to 200 meters.

[0012] Since the rail data of the target rail section can reflect the quality problems of the rail section, this embodiment needs to use data acquisition equipment, that is, a track inspection vehicle to inspect the target rail section. After the inspection is completed, D different types of rail data sequences corresponding to the target rail section are obtained, and then the set formed by all the acquired rail data sequences corresponding to the target rail section is recorded as the rail data sequence set corresponding to the target rail section. The front or rear of the track inspection vehicle is usually equipped with a laser scanner and a laser rangefinder; and the process of data acquisition by the track inspection vehicle is as follows: the track inspection vehicle will start from the starting end of the target rail section and inspect along the target rail section. During the inspection process, each time it reaches a detection position, the track inspection vehicle needs to record the D different types of rail data monitored until the end of the target rail section is detected and stops.

[0013] And in specific applications, the implementer also needs to set the interval between adjacent detection positions according to actual conditions, such as it can be set to 0.1 meters; in addition, the implementer also needs to set the value of D and the type of rail data collected according to the quality problem of the rail to be detected. For example, the present embodiment focuses on detecting the unevenness of the rail, then the value of D can be set to 7 in this embodiment, and the type of rail data to be collected can be left high and low data, right high and low data, left track direction data, right track direction data, gauge data, horizontal data and triangular pit data, that is, the collected rail data sequence includes left high and low data sequence, right high and low data sequence, left track direction data sequence, right track direction data sequence, gauge data sequence, horizontal data sequence and triangular pit data sequence.

[0014] It should be noted that, during the data collection process, if it is detected that the data at a certain position is missing, the position will be marked as a missing value position; for example, if in the process of obtaining the left high-low data sequence corresponding to the target rail section, if the left high-low data collected by the detection vehicle at the third detection position on the target rail section is missing, and there is no missing at other detection positions, then there is a missing value position in the subsequently acquired left high-low data sequence, and this position is located behind and adjacent to the second data in the left high-low data sequence; in addition, during the data collection process, the method of detecting the missing value position is a well-known technology, so this embodiment will not be described in detail.

[0015] Therefore, this embodiment can obtain the set of rail data sequences corresponding to the target rail section through the above process; since only the rail data sequences with missing value positions need to be filled, after obtaining the rail data sequence set, all rail data sequences with missing value positions are obtained in the rail data sequence set, and are all recorded as initial sequences to be filled, and all rail data sequences without missing value positions in the rail data sequence set are obtained, and are all recorded as the first target sequence corresponding to the target rail section, that is, the first target sequence is a sequence that does not need to be filled.

[0016] To facilitate subsequent analysis, after obtaining the initial sequence to be filled, this embodiment sets a missing value marker symbol A, and then fills the missing value marker symbol A into all missing value positions in each initial sequence to be filled, and the filled sequences are recorded as the target sequence to be filled corresponding to the target rail section, and all missing value marker symbols A in the target sequence to be filled are recorded as data to be filled, and all data in the target sequence to be filled except the missing value marker symbol A are recorded as non-data to be filled.

[0017] Therefore, this embodiment can obtain the target sequence to be filled and the corresponding first target sequence corresponding to the target rail section through the above process, and there is data to be filled in the target sequence to be filled.

[0018] Step S002, taking the to-be-filled data in the target to-be-filled sequence as the center, constructing a target window corresponding to each to-be-filled data in the target to-be-filled sequence, and sliding the target window corresponding to the to-be-filled data on the target to-be-filled sequence to obtain a set of to-be-analyzed windows corresponding to the to-be-filled data; obtaining a set of similar windows corresponding to the to-be-filled data based on the target window corresponding to the to-be-filled data and the set of to-be-analyzed windows corresponding to the to-be-filled data.

[0019] Moreover, since reliably filling the missing value position is a key factor in ensuring the subsequent reliable analysis and processing of abnormal railway events, when determining the filling value of the missing value position, a local window that is most similar to the local window where the missing value position is located is generally obtained first, and then the filling value of the missing value position is determined based on the most similar local window found. However, this method of determining the filling value of the missing value position based only on a most similar local window will make the obtained filling value less reliable. When the filling value is unreliable, the consistency and reliability of the filled data sequence will be poor. When the consistency and reliability of the data sequence after the missing value filling are poor, it will directly affect the analysis of abnormal railway events. The reliability of event analysis and processing, if the data in the most similar local window found is seriously disturbed by external environmental factors during collection, that is, the data in the most similar local window is seriously affected by electromagnetic interference, which will lead to the filling value of the missing value position obtained based on the most similar local window is unreliable. In order to make the obtained filling value more reliable, this embodiment will first obtain multiple windows that are similar to the local window where the missing value position is located, that is, similar windows, and then analyze the credibility of all the similar windows obtained separately, and then finally determine the filling value based on the credibility of the multiple similar windows obtained, and this method of obtaining the filling value can make the obtained filling value more reliable.

[0020] Therefore, based on the above, we can know that we need to first build a local window where the missing value is located, that is, to build a target window where the data to be filled is located. The specific construction process is: For any target sequence to be filled W: first, with each data to be filled in the target sequence to be filled W as the center, a target window corresponding to each data to be filled in the target sequence to be filled W is constructed. The size of the target window is generally set to an empirical value. For example, the size of the target window can be set to , m1 is 11, that is, under normal circumstances, the number of data in the target window corresponding to the data to be filled is generally 11, but when constructing the target window corresponding to the data to be filled, the number of data on the left or right side of the data to be filled may be less than 5. When this happens, the amount of data in the constructed target window will be less than 11. However, no matter what the situation, the central data in the target window corresponding to a certain data to be filled is the data to be filled. If there are only 3 data on the left side of a certain data to be filled in the target sequence W to be filled, then the amount of data in the target window corresponding to the data to be filled is 9, that is, the length of the target window corresponding to the data to be filled is 9.

[0021] Therefore, through the above process, the target window corresponding to each data to be filled in each target sequence to be filled can be obtained. After the target window corresponding to the data to be filled is obtained, the target window corresponding to each data to be filled is slid on the target sequence to be filled to obtain a set of windows to be analyzed corresponding to each data to be filled. The set of windows to be analyzed is the basis for subsequently obtaining a set of similar windows. Therefore, the specific process of obtaining the set of windows to be analyzed corresponding to each data to be filled is as follows: For the b-th data to be filled in the target sequence W to be filled: on the target sequence W to be filled, the target window corresponding to the b-th data to be filled is slid leftward and rightward respectively with a preset sliding step length, and all sliding windows obtained in the sliding process are recorded as the windows to be analyzed corresponding to the b-th data to be filled, and the set constructed by all the windows to be analyzed corresponding to the b-th data to be filled is recorded as the set of windows to be analyzed corresponding to the b-th data to be filled, and the condition for the target window corresponding to the b-th data to be filled to stop sliding to the left is the first appearance of the first data in the target sequence W to be filled in the sliding window, and the condition for the target window corresponding to the b-th data to be filled to stop sliding to the right is the first appearance of the last data in the target sequence W to be filled in the sliding window; in addition, it should be noted that the position distance between the center position of any window to be analyzed corresponding to the b-th data to be filled and the center position of the target window corresponding to the b-th data to be filled is an integer multiple of the preset sliding step length.

[0022] In specific applications, the implementer needs to set the preset sliding step length according to the actual situation, and this embodiment requires setting the preset sliding step length according to the length of the target window corresponding to the data to be filled. For example, if the length of the target window corresponding to the data to be filled is 11, then the preset sliding step length of the target window corresponding to the data to be filled when sliding is .

[0023] Therefore, this embodiment can obtain the set of windows to be analyzed corresponding to each data to be filled in each target sequence to be filled through the above process. After obtaining the set of windows to be analyzed, this embodiment obtains a set of similar windows corresponding to the data to be filled according to the target window corresponding to the data to be filled and the set of windows to be analyzed corresponding to the data to be filled. The similar window set is an important basis for subsequently reliably and accurately determining the filling value. Therefore, the process of obtaining the similar window set corresponding to the data to be filled is as follows: For the b-th data to be filled in the target sequence W to be filled: first, obtain the structural similarity between the target window corresponding to the b-th data to be filled and each window to be analyzed in the set of windows to be analyzed corresponding to the b-th data to be filled, and record it as the similarity characterization value of the corresponding window to be analyzed; then, in the set of windows to be analyzed corresponding to the b-th data to be filled, record the set formed by all windows to be analyzed whose similarity characterization values ​​are greater than a preset similarity threshold as the first set; then, record the total number of data in the target window corresponding to the b-th data to be filled as the target length corresponding to the b-th data to be filled, and then obtain the total number of data in each window to be analyzed in the first set, and record it as the first length of the corresponding window to be analyzed; then, in the first set, obtain all windows to be analyzed whose first length is equal to the target length corresponding to the b-th data to be filled. The first set is a set of windows to be analyzed, and the set formed by all windows to be analyzed with the first length equal to the target length is recorded as the second set; then the position distance between each window to be analyzed in the second set and the target window corresponding to the b-th data to be filled is obtained, and recorded as the first distance of the corresponding window to be analyzed; then all windows to be analyzed in the second set are sorted in the order of the first distance from small to large, and the sorting result is recorded as the first sequence, and the length of the windows to be analyzed in the first sequence is consistent with the length of the target window corresponding to the b-th data to be filled; then the first preset number Q0 of windows to be analyzed located in the first sequence is obtained, and the set constructed by the first preset number Q0 of windows to be analyzed located in the first sequence is recorded as the similar window set corresponding to the b-th data to be filled, and the windows in the similar window set are all similar windows. And because the length of the target sequence to be filled W is generally long, in order to ensure that the number of windows to be analyzed in the first sequence is not less than Q0, the value of Q0 cannot be set too large, such as in this embodiment, the preset number Q0 can be set to 5.

[0024] In this embodiment, the specific process of obtaining the structural similarity between the target window corresponding to the b-th data to be filled in the target sequence W to be filled and each window to be analyzed in the set of windows to be analyzed corresponding to the b-th data to be filled is: For the target window corresponding to the b-th data to be filled in the target sequence W to be filled and the c-th window to be analyzed in the set of windows to be analyzed corresponding to the b-th data to be filled: First, a sequence formed by all non-to-be-filled data in the target window corresponding to the b-th data to be filled is obtained, and recorded as the first non-to-be-filled data sequence, and a sequence formed by all non-to-be-filled data in the c-th window to be analyzed is obtained, and recorded as the second non-to-be-filled data sequence; then, a first-order difference sequence of the first non-to-be-filled data sequence is obtained, and a linear regression function is used to perform a linear fit on all data in the first-order difference sequence of the first non-to-be-filled data sequence, and the slope of the straight line obtained by the fitting is recorded as the first fitting slope, and the first-order difference sequence of the second non-to-be-filled data sequence is obtained, and a linear regression function is used to perform a linear fit on all data in the first-order difference sequence of the second non-to-be-filled data sequence, and the slope of the straight line obtained by the fitting is recorded as the second fitting slope, and the t-th data in the first-order difference sequence of the non-to-be-filled data sequence is The value of the t+1th non-to-be-filled data in the corresponding non-to-be-filled data sequence is subtracted from the value of the tth non-to-be-filled data; then, the absolute value of the difference between the first fitting slope and the second fitting slope is obtained, and recorded as the first difference value, and the absolute value of the difference between the mean of the first non-to-be-filled data sequence and the mean of the second non-to-be-filled data sequence is obtained, and recorded as the second difference value; then, the first difference value, the second difference value and the preset first constant are added, and the inverse of the result obtained by the addition is recorded as the initial similarity value, and then the initial similarity value is normalized, and the value obtained after the normalization is recorded as the structural similarity between the target window corresponding to the bth data to be filled and the cth window to be analyzed; in addition, the above-mentioned method of obtaining the first-order difference sequence and the method of performing linear fitting are both well-known technologies, so this embodiment will not be described in detail.

[0025] The specific formula for calculating the structural similarity between the target window corresponding to the bth data to be filled and the cth window to be analyzed is: Among them, S0 is the structural similarity between the target window corresponding to the b-th data to be filled and the c-th window to be analyzed, norm() is the normalization function, D1 is the first difference value, D2 is the second difference value, c1 is the preset first constant, and c1 here is to prevent the denominator from being 0, and in this embodiment, the preset value of the first constant is 0.1.

[0026] In addition, the smaller the values ​​of D1 and D2 are, the larger the value of S0 is, and the larger the value of S0 is, the more similar the structures between the target window corresponding to the b-th data to be filled and the c-th window to be analyzed are, that is, the more similar the target window corresponding to the b-th data to be filled and the c-th window to be analyzed are, otherwise it means that the structures between the target window corresponding to the b-th data to be filled and the c-th window to be analyzed are less similar.

[0027] In this embodiment, the process of obtaining the position distance between each window to be analyzed in the second set and the target window corresponding to the bth data to be filled is: For any window b1 to be analyzed in the second set, if the first data in the window b1 to be analyzed is the Zth data in the target sequence W to be filled, and the first data in the target window corresponding to the bth data to be filled is the Z+10th data in the target sequence W to be filled, then the position distance between the window b1 to be analyzed and the target window corresponding to the bth data to be filled is .

[0028] In a specific application, the implementer needs to set the preset similarity threshold according to the actual situation. For example, in this embodiment, the preset similarity threshold can be set to 0.8.

[0029] Therefore, this embodiment can obtain the similar window set corresponding to each to-be-filled data in each target to-be-filled sequence through the above process.

[0030] Step S003: obtaining, according to all similar windows in the similar window set, a credibility factor corresponding to each similar window in the similar window set.

[0031] Based on step S002, the similar window set corresponding to each data to be filled has been obtained, but the data change trend in the similar window and the degree of data interference lead to differences in the reliability or credibility of different similar windows, and the reliability or credibility is directly related to the contribution or participation of the similar window in the subsequent calculation of the filling value. Therefore, this embodiment will then obtain the reliability or credibility of the similar window by analyzing the data change trend in the similar window and the degree of data interference. That is, this embodiment will then obtain the credibility factor corresponding to each similar window in the similar window set corresponding to the data to be filled according to all similar windows in the similar window set corresponding to the data to be filled, that is, the specific acquisition process of the credibility factor is as follows: For any similar window f in the similar window set corresponding to the b-th data to be filled in the target sequence W to be filled: First, the target window and the similar window f corresponding to the b-th data to be filled are enlarged according to the preset sizes to obtain the enlarged window of the target window and the enlarged window of the similar window f corresponding to the b-th data to be filled, and the enlarged window of the target window and the enlarged window of the similar window f corresponding to the b-th data to be filled are recorded as the first window and the second window respectively; and in a specific application, the implementer can set the preset size according to the actual situation, and the preset size must be greater than the length of the target window corresponding to the b-th data to be filled. If the length of the target window corresponding to the b-th data to be filled is m1, then the preset size can be set to , that is, the width of the window before and after expansion remains unchanged, while the length of the expanded window is ; Then, in this embodiment, the enlarged window of the target window corresponding to the b-th data to be filled refers to the window after the target window corresponding to the b-th data to be filled is enlarged to a preset size, and the enlarged window of the similar window f refers to the window after the similar window f is enlarged to a preset size; in addition, the enlargement rule can also be set according to actual conditions, such as requiring the window to be enlarged only to the right, or setting the window to be enlarged to both sides at the same time, but it is required that the enlargement rules selected when enlarging different windows are consistent.

[0032] Then, the first mark value of each data in the first window and the first mark value of each data in the second window are obtained, and in this embodiment, the first mark value of the g-th data in each window is g; then, the standard deviation of the sequence formed by all non-to-be-filled data in the second window is obtained and recorded as the first standard deviation, and then the reciprocal of the sum of the first standard deviation and the preset first constant is recorded as the first indicator value.

[0033] Then, according to the first mark value of each data in the first window and the second window, the first feature sequence corresponding to the first window and the second feature sequence corresponding to the second window are obtained, and the Pearson correlation coefficient between the first feature sequence and the second feature sequence is obtained, and the absolute value of the obtained Pearson correlation coefficient is recorded as the second indicator value.

[0034] Next, in the first window, obtain a data pair formed by two non-to-be-filled data located on the left side of the b-th data to be filled and closest to the b-th data to be filled, and record it as the first non-to-be-filled data pair; in the first window, obtain a data pair formed by two non-to-be-filled data located on the right side of the b-th data to be filled and closest to the b-th data to be filled, and record it as the second non-to-be-filled data pair.

[0035] Then, the data located at the center position of the similar window f is obtained and recorded as the first center data; since the second window is an enlarged window of the similar window f, the first center data also exists in the second window; therefore, next in the second window, a data pair formed by two non-to-be-filled data located on the left side of the first center data and closest to the first center data is obtained, and recorded as the third non-to-be-filled data pair; in the second window, a data pair formed by two non-to-be-filled data located on the right side of the first center data and closest to the first center data is obtained, and recorded as the fourth non-to-be-filled data pair.

[0036] Next, based on the acquired first non-to-be-filled data pair, second non-to-be-filled data pair, third non-to-be-filled data pair and fourth non-to-be-filled data pair, obtain the change rate characterization value of the first window and the change rate characterization value of the second window; then calculate the absolute value of the difference between the change rate characterization value of the first window and the change rate characterization value of the second window, and record it as the characteristic difference value, and then record the inverse of the addition of the characteristic difference value and the preset first constant as the third indicator value.

[0037] Finally, the first index value, the second index value and the third index value are added, and the result of the addition is normalized, and the value obtained after the normalization is recorded as the credibility factor corresponding to the similarity window f.

[0038] In this embodiment, the specific process of obtaining the first feature sequence corresponding to the first window and the second feature sequence corresponding to the second window according to the first mark value of each data in the first window and the second window is: First, the set formed by the first tag values ​​of all the data to be filled in the first window and the second window is recorded as the tag value set; then, in the first window, all the data whose first tag value does not belong to the tag value set are obtained, and the sequence formed by all the data whose first tag value does not belong to the tag value set in the first window is recorded as the first feature sequence corresponding to the first window; then, in the second window, all the data whose first tag value does not belong to the tag value set are obtained, and the sequence formed by all the data whose first tag value does not belong to the tag value set in the second window is recorded as the second feature sequence corresponding to the second window.

[0039] In this embodiment, the specific process of obtaining the change rate characterization value of the first window and the change rate characterization value of the second window according to the acquired first non-to-be-filled data pair, second non-to-be-filled data pair, third non-to-be-filled data pair and fourth non-to-be-filled data pair is: First, a mapping space corresponding to the target sequence to be filled W is constructed, and the data on the horizontal axis of the mapping space corresponding to the target sequence to be filled W is the first mark value, and the data on the vertical axis is the non-to-be-filled data in the target sequence to be filled W; then, a feature data point pair corresponding to the first non-to-be-filled data pair, a feature data point pair corresponding to the second non-to-be-filled data pair, a feature data point pair corresponding to the third non-to-be-filled data pair, and a feature data point pair corresponding to the fourth non-to-be-filled data pair are obtained; and the data points in the feature data point pair corresponding to the to-be-filled data pair refer to the non-to-be-filled data in the corresponding non-to-be-filled data pair. The first mark value of the data to be filled and the data not to be filled are mapped to the data points obtained in the mapping space corresponding to the target sequence to be filled W; that is, the first data point in the feature data point pair corresponding to the data pair not to be filled is the data point obtained by mapping the first non-data to be filled and the first mark value of the first non-data to be filled in the corresponding data pair not to be filled to the data point in the mapping space, and the second data point in the feature data point pair corresponding to the data pair not to be filled is the data point obtained by mapping the second non-data to be filled and the first mark value of the second non-data to be filled in the corresponding data pair not to be filled to the data point in the mapping space.

[0040] Next, obtain the slope of the straight line formed by the two data points in the characteristic data point pair corresponding to each non-to-be-filled data pair, and record it as the slope of the straight line corresponding to the corresponding non-to-be-filled data pair, that is, according to the above method, the slope of the straight line corresponding to the first non-to-be-filled data pair, the slope of the straight line corresponding to the second non-to-be-filled data pair, the slope of the straight line corresponding to the third non-to-be-filled data pair, and the slope of the straight line corresponding to the fourth non-to-be-filled data pair can be obtained; then obtain the ratio of the slope of the straight line corresponding to the first non-to-be-filled data pair to the slope of the straight line corresponding to the second non-to-be-filled data pair, and record it as the rate of change representation value of the first window, obtain the ratio of the slope of the straight line corresponding to the third non-to-be-filled data pair to the slope of the straight line corresponding to the fourth non-to-be-filled data pair, and record it as the rate of change representation value of the second window.

[0041] In addition, the specific expression for calculating the credibility factor corresponding to the similar window f in this embodiment is: in, is the credibility factor corresponding to the similar window f, is the standard deviation of the sequence formed by all non-to-be-filled data in the second window, P0 is the Pearson correlation coefficient between the first feature sequence and the second feature sequence, K1 is the change rate characterization value of the first window, K2 is the change rate characterization value of the second window, c1 is the preset first constant, and c1 here is also to prevent the denominator from being 0, is the first indicator value, is the second index value, is the third indicator value.

[0042] In addition, the difference between the change rate characterization value of the first window and the change rate characterization value of the second window mainly reflects the difference between the instantaneous change rates, that is, it reflects the difference in local change trends between the first window and the second window, and the Pearson correlation coefficient reflects the difference in overall trend changes between the first window and the second window, that is, the change rate characterization value and the Pearson correlation coefficient can reflect the trend change difference between the first window and the second window; and when acquiring data, if it is interfered by external factors, then the interference of external factors will cause part of the collected data to fluctuate, and the greater the degree of interference, the greater the fluctuation of the collected data, that is, when the standard deviation of the sequence formed by all non-to-be-filled data in the second window is larger, it indicates that the data in the second window is more interfered during collection and is less credible.

[0043] And if, when obtaining the filling value, a similar window with a large degree of interference is referred to to a large extent, the authenticity and coherence of the filled sequence will be lower, and a similar window with a large trend change difference from the first window is referred to to a large extent, the authenticity and coherence of the filled sequence will also be lower; therefore, in this embodiment, when The smaller the value, the larger the P0 and The smaller the time, the The larger the value of The larger the value of is, the more reliable the similarity window f is, and it also indicates that when calculating the filling value corresponding to the bth data to be filled, more reference should be made to the similarity window f. The larger the value, the smaller the P0 and The larger the The smaller the value of The smaller the value of is, the less reliable the similarity window f is, and it also indicates that the similarity window f should be referred to less frequently when calculating the filling value corresponding to the b-th data to be filled.

[0044] Therefore, this embodiment can obtain the credibility factor corresponding to each similar window in the similar window set corresponding to the data to be filled according to the above process.

[0045] Step S004, obtaining a filling value of the data to be filled according to the credibility factor corresponding to the similarity window, and replacing the data to be filled with the filling value corresponding to the data to be filled, and recording the replaced target sequence to be filled as a second target sequence corresponding to the target rail section.

[0046] After obtaining the credibility factors corresponding to each similar window in the similar window set corresponding to the data to be filled, the present embodiment will then obtain the filling values ​​of each data to be filled in the target sequence to be filled according to the credibility factors corresponding to the similar windows. After obtaining the filling values ​​of each data to be filled, the present embodiment will replace each data to be filled in the target sequence to be filled with the filling values ​​of the corresponding data to be filled, and record the replaced target sequence to be filled as the second target sequence corresponding to the target rail section, that is, one target sequence to be filled can obtain a second target sequence; for example, for the bth data to be filled in the target sequence W to be filled, the bth data to be filled will be replaced with the filling value of the bth data to be filled.

[0047] In this embodiment, according to the credibility factor corresponding to the similar window, the specific process of obtaining the filling value of each to-be-filled data in the target to-be-filled sequence is as follows: For the b-th data to be filled in the target sequence W to be filled: first, obtain the cumulative sum of the credibility factors corresponding to all similar windows in the similar window set corresponding to the b-th data to be filled, and record it as the comprehensive credibility factor; then obtain the ratio of the credibility factor corresponding to each similar window in the similar window set corresponding to the b-th data to be filled and the comprehensive credibility factor, and record it as the normalization factor of the corresponding similar window; then obtain the product of the central data of each similar window in the similar window set corresponding to the b-th data to be filled and the normalization factor of the corresponding similar window, and record it as the weighted eigenvalue of the corresponding similar window; and if the central data of a certain similar window is the data to be filled, then in the similar window, obtain a non-to-be-filled data closest to the central data of the similar window, and record the product of the non-to-be-filled data closest to the central data of the similar window and the normalization factor of the similar window as the weighted eigenvalue of the similar window; finally, the cumulative sum of the weighted eigenvalues ​​of all similar windows in the similar window set corresponding to the b-th data to be filled is recorded as the filling value of the b-th data to be filled.

[0048] Therefore, this embodiment can obtain the filling value of each data to be filled in each target sequence to be filled through the above process; in addition, when the first data or the last data in the target sequence to be filled is the data to be filled, the filling value cannot be obtained according to the above process. Therefore, if the first data or the last data in the target sequence to be filled is the data to be filled, this embodiment directly uses the mean of all non-data to be filled in the target sequence to be filled as the corresponding filling value.

[0049] Step S005: obtaining a quality characterization value of the target rail section according to the standard deviation of the first target sequence and the second target sequence, and obtaining an analysis and processing result of an abnormal event of the target rail section according to the quality characterization value.

[0050] Based on the above description, it can be known that the present embodiment can obtain multiple first target sequences and multiple second target sequences corresponding to the target rail section, and the first target sequence is a sequence that does not need to be filled, and the second target sequence refers to a sequence that has been filled, and then all the first target sequences and second target sequences corresponding to the target rail section are constructed into a set, which is recorded as a comprehensive set; then the standard deviation of each target sequence in the comprehensive set is obtained, and the cumulative sum of the standard deviations of all target sequences in the comprehensive set is recorded as the quality characterization value of the target rail section; and the larger the quality characterization value, the more serious the unevenness problem of the corresponding target rail section, and the greater the probability that the target rail section will occur or cause an abnormal railway event.

[0051] Finally, according to the quality characterization value of the target rail section, the analysis and processing results of the abnormal events of the target rail section are obtained; specifically: in this embodiment, a preset quality threshold is set to determine whether the normalized value of the quality characterization value of the target rail section is greater than the preset quality threshold. If so, it is determined that the target rail section will cause the occurrence of abnormal railway events, and it is necessary to immediately remind relevant staff to repair or investigate the target rail section. Otherwise, it is determined that the target rail section will not cause the occurrence of abnormal railway events, and it is not necessary to remind relevant staff to repair and maintain the target rail section. In addition, in this embodiment, it is also necessary to obtain the quality characterization value of each section on the rail to be detected according to the above process, and then sort all sections on the rail to be detected in order from large to small according to the quality characterization value, and record the sorting result as a section sequence, and then remind relevant staff to focus on the sections with higher rankings in the section sequence, and the sections with higher rankings are more likely to cause abnormal railway events.

[0052] In addition, the implementer needs to set the preset quality threshold according to the actual situation. For example, in this embodiment, the preset quality threshold can be set to 0.6.

[0053] So far, this embodiment has completed the reliable analysis and processing of abnormal railway events.

[0054] In summary, the present embodiment first obtains the target sequence to be filled and the corresponding first target sequence corresponding to the target rail section; then, with the data to be filled in the target sequence to be filled as the center, a target window corresponding to each data to be filled in the target sequence to be filled is constructed, and the target window corresponding to the data to be filled is slid on the target sequence to be filled to obtain a set of windows to be analyzed corresponding to the data to be filled, and a set of similar windows corresponding to the data to be filled is obtained according to the target window corresponding to the data to be filled and the set of windows to be analyzed corresponding to the data to be filled; then, according to all similar windows in the similar window set, a credibility factor corresponding to each similar window in the similar window set is obtained; then, according to the credibility factor corresponding to the similar window, a filling value of the data to be filled is obtained, and the data to be filled is replaced with the filling value of the corresponding data to be filled, and the replaced target sequence to be filled is recorded as the second target sequence corresponding to the target rail section; finally, according to the standard deviation of the first target sequence and the second target sequence, a quality characterization value of the target rail section is obtained, and according to the quality characterization value, an analysis and processing result of abnormal events of the target rail section is obtained. In addition, in this embodiment, the filling value obtained by analyzing the credibility factors of multiple similar windows is more reliable and accurate, and the consistency and reliability of the first target sequence determined based on the more reliable and accurate filling value are better. Therefore, the quality characterization value obtained based on the target sequence with better consistency and reliability is more reliable and accurate, and based on the more reliable and accurate quality characterization value, the reliability of analyzing and processing abnormal railway events can be guaranteed or improved.

[0055] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for analyzing and processing abnormal events for railway infrastructure, characterized in that: The method comprises the following steps: Acquire a target sequence to be filled and a corresponding first target sequence corresponding to the target rail section, wherein the target sequence to be filled contains data to be filled; Taking the data to be filled in the target sequence to be filled as the center, constructing a target window corresponding to each data to be filled in the target sequence to be filled, and sliding the target window corresponding to the data to be filled on the target sequence to be filled to obtain a set of windows to be analyzed corresponding to the data to be filled; Obtaining a similar window set corresponding to the data to be filled according to the target window corresponding to the data to be filled and the set of windows to be analyzed corresponding to the data to be filled; According to all similar windows in the similar window set, a credibility factor corresponding to each similar window in the similar window set is obtained; According to the credibility factor corresponding to the similarity window, a filling value of the data to be filled is obtained, and the data to be filled is replaced with the filling value corresponding to the data to be filled, and the target sequence to be filled after the replacement is recorded as a second target sequence corresponding to the target rail section; A quality characterization value of the target rail section is obtained according to the standard deviation of the first target sequence and the second target sequence, and an analysis and processing result of an abnormal event of the target rail section is obtained according to the quality characterization value.

2. The method for analyzing and processing abnormal events for railway infrastructure according to claim 1, characterized in that: The method for acquiring the target to-be-filled sequence and the corresponding first target sequence corresponding to the target rail section includes: Acquire a set of rail data sequences corresponding to the target rail section, wherein the set of rail data sequences consists of D different types of rail data sequences; Recording all rail data sequences with missing value positions in the rail data sequence set as initial sequences to be filled, and recording all rail data sequences without missing value positions in the rail data sequence set as first target sequences corresponding to the target rail segment; The missing value marker symbol A is filled into all the missing value positions in the initial sequence to be filled, and the filled sequence is recorded as the target sequence to be filled corresponding to the target rail segment, all the missing value marker symbols A in the target sequence to be filled are recorded as data to be filled, and all the data in the target sequence to be filled except the data to be filled are recorded as non-data to be filled.

3. The method for analyzing and processing abnormal events for railway infrastructure according to claim 1, characterized in that: The method for acquiring the set of windows to be analyzed corresponding to the data to be filled includes: For the b-th data to be filled in the target sequence to be filled, on the target sequence to be filled, the target window corresponding to the b-th data to be filled is slid to the left and right respectively, and the set constructed by all sliding windows obtained during the sliding process is recorded as the window to be analyzed corresponding to the b-th data to be filled, and the window in the set of windows to be analyzed corresponding to the b-th data to be filled is the window to be analyzed.

4. The method for analyzing and processing abnormal events for railway infrastructure according to claim 2, characterized in that: The method for acquiring the similar window set corresponding to the data to be filled includes: For the bth data to be filled in the target sequence to be filled: The structural similarity between the target window corresponding to the b-th data to be filled and each window to be analyzed in the set of windows to be analyzed corresponding to the b-th data to be filled is obtained, and recorded as the similarity characterization value of the corresponding window to be analyzed; in the set of windows to be analyzed corresponding to the b-th data to be filled, a set formed by all windows to be analyzed whose similarity characterization values ​​are greater than a preset similarity threshold is recorded as a first set; the total number of data in each window to be analyzed in the first set is recorded as the first length of the corresponding window to be analyzed, and the total number of data in the target window corresponding to the b-th data to be filled is recorded as the target length; in the first set, a set formed by all windows to be analyzed whose first length is equal to the target length is recorded as a second set; the position distance between each window to be analyzed in the second set and the target window corresponding to the b-th data to be filled is recorded as the first distance of the corresponding window to be analyzed; all windows to be analyzed in the second set are sorted in ascending order of the first distance, and the sorting result is recorded as a first sequence; a set constructed by the preset number Q0 windows to be analyzed located at the front of the first sequence is recorded as the similar window set corresponding to the b-th data to be filled, and all windows in the similar window set are recorded as similar windows.

5. The method for analyzing and processing abnormal events for railway infrastructure according to claim 4, characterized in that: The method for obtaining the structural similarity comprises: For the target window corresponding to the b-th data to be filled and the c-th window to be analyzed in the set of windows to be analyzed corresponding to the b-th data to be filled: Recording a sequence formed by all non-to-be-filled data in the target window corresponding to the b-th data to be filled as a first non-to-be-filled data sequence, and recording a sequence formed by all non-to-be-filled data in the c-th window to be analyzed as a second non-to-be-filled data sequence; performing linear fitting on the differential sequence of the first non-to-be-filled data sequence, and recording the slope of the straight line obtained by fitting as a first fitting slope; performing linear fitting on the differential sequence of the second non-to-be-filled data sequence, and recording the slope of the straight line obtained by fitting as a second fitting slope; The absolute value of the difference between the first fitting slope and the second fitting slope is recorded as the first difference value, and the absolute value of the difference between the mean of the first data sequence not to be filled and the mean of the second data sequence not to be filled is recorded as the second difference value; the reciprocal of the result obtained by adding the first difference value, the second difference value and the preset first constant is recorded as the initial similarity value; the normalized value of the initial similarity value is recorded as the structural similarity between the target window corresponding to the bth data to be filled and the cth window to be analyzed.

6. The method for analyzing and processing abnormal events for railway infrastructure according to claim 2, characterized in that: The method for obtaining the credibility factor corresponding to each similar window in the similar window set includes: For any similar window f in the similar window set corresponding to the b-th data to be filled in the target sequence to be filled: The target window corresponding to the b-th data to be filled and the similar window f are respectively enlarged according to preset sizes to obtain an enlarged window of the target window corresponding to the b-th data to be filled and an enlarged window of the similar window f, and the enlarged window of the target window corresponding to the b-th data to be filled and the enlarged window of the similar window f are respectively recorded as a first window and a second window; Obtaining the first mark value of each data in the first window and the first mark value of each data in the second window, wherein the first mark value of the g-th data in the window is g; recording the standard deviation of the sequence formed by all non-to-be-filled data in the second window as the first standard deviation, and recording the reciprocal of the sum of the first standard deviation and a preset first constant as the first indicator value; According to the first mark value of each data in the first window and the second window, a first feature sequence corresponding to the first window and a second feature sequence corresponding to the second window are obtained, and the absolute value of the Pearson correlation coefficient between the first feature sequence and the second feature sequence is recorded as a second indicator value; In the first window, a data pair formed by two non-to-be-filled data located on the left side of the b-th data to be filled and closest to the b-th data to be filled is recorded as a first non-to-be-filled data pair, and a data pair formed by two non-to-be-filled data located on the right side of the b-th data to be filled and closest to the b-th data to be filled is recorded as a second non-to-be-filled data pair; The data located at the center position of the similarity window f is recorded as the first center data, and in the second window, the data pair formed by the two non-to-be-filled data located on the left side of the first center data and closest to the first center data is recorded as the third non-to-be-filled data pair, and the data pair formed by the two non-to-be-filled data located on the right side of the first center data and closest to the first center data is recorded as the fourth non-to-be-filled data pair; According to the first non-to-be-filled data pair, the second non-to-be-filled data pair, the third non-to-be-filled data pair, and the fourth non-to-be-filled data pair, a change rate characterization value of the first window and a change rate characterization value of the second window are obtained; an absolute value of a difference between the change rate characterization value of the first window and the change rate characterization value of the second window is recorded as a characteristic difference value; and a reciprocal of the sum of the characteristic difference value and a preset first constant is recorded as a third indicator value; A normalized value of a result obtained by adding the first index value, the second index value, and the third index value is recorded as the credibility factor corresponding to the similarity window f.

7. The method for analyzing and processing abnormal events for railway infrastructure according to claim 6, characterized in that: The method for acquiring the first feature sequence corresponding to the first window and the second feature sequence corresponding to the second window includes: The set formed by the first tag values ​​of all the data to be filled in the first window and the second window is recorded as the tag value set; in the first window, the sequence composed of all data whose first tag values ​​do not belong to the tag value set is recorded as the first feature sequence corresponding to the first window; in the second window, the sequence composed of all data whose first tag values ​​do not belong to the tag value set is recorded as the second feature sequence corresponding to the second window.

8. The method for analyzing and processing abnormal events for railway infrastructure according to claim 6, characterized in that: The method for acquiring the change rate characterizing value of the first window and the change rate characterizing value of the second window includes: Constructing a mapping space, wherein the abscissa value of the mapping space is the first mark value, and the ordinate value is the non-to-be-filled data in the target to-be-filled sequence; Acquire a characteristic data point pair corresponding to the first non-to-be-filled data pair, a characteristic data point pair corresponding to the second non-to-be-filled data pair, a characteristic data point pair corresponding to the third non-to-be-filled data pair, and a characteristic data point pair corresponding to the fourth non-to-be-filled data pair; the data points in the characteristic data point pairs corresponding to the non-to-be-filled data pairs refer to the data points obtained by mapping the non-to-be-filled data and the first label values ​​of the non-to-be-filled data in the corresponding non-to-be-filled data pairs to the mapping space; The slope of the straight line formed by the two data points in the characteristic data point pair corresponding to the non-to-be-filled data pair is recorded as the slope of the straight line corresponding to the corresponding non-to-be-filled data pair; the ratio of the slope of the straight line corresponding to the first non-to-be-filled data pair to the slope of the straight line corresponding to the second non-to-be-filled data pair is recorded as the change rate characterization value of the first window; the ratio of the slope of the straight line corresponding to the third non-to-be-filled data pair to the slope of the straight line corresponding to the fourth non-to-be-filled data pair is recorded as the change rate characterization value of the second window.

9. The method for analyzing and processing abnormal events for railway infrastructure according to claim 1, characterized in that: The method for obtaining the filling value of the data to be filled includes: For the bth data to be filled in the target sequence to be filled: The cumulative sum of the credibility factors corresponding to all similar windows in the similar window set corresponding to the b-th data to be filled is recorded as the comprehensive credibility factor; the ratio of the credibility factor corresponding to each similar window in the similar window set corresponding to the b-th data to be filled to the comprehensive credibility factor is recorded as the normalization factor of the corresponding similar window; The product of the center data of each similar window in the similar window set corresponding to the b-th data to be filled and the normalization factor of the corresponding similar window is recorded as the weighted eigenvalue of the corresponding similar window; The cumulative sum of the weighted feature values ​​of all similar windows in the similar window set corresponding to the b-th data to be filled is recorded as the filling value of the b-th data to be filled.

10. The method for analyzing and processing abnormal events for railway infrastructure according to claim 1, characterized in that: The method for obtaining the quality characterization value of the target rail section includes: A set is constructed from all the first target sequences and the second target sequences, which is recorded as a comprehensive set; and the cumulative sum of the standard deviations of all the target sequences in the comprehensive set is recorded as the quality characterization value of the target rail section.

Images