Multi-dimensional track data association method, device, equipment and medium

Abstract

The invention discloses a multi-dimensional orbit data association method and device, equipment and a medium, and relates to the technical field of data mining, and the method comprises the steps: carrying out the trend analysis of the extreme value change direction of each piece of orbit data in a multi-source data subset within a preset time range, and obtaining a first analysis result; performing trend analysis on the data difference value change among the running gear data in the multi-source data subset to obtain a second analysis result; integrating and marking the first analysis result, the second analysis result, the track data in the multi-source data subset and the running gear data to obtain a marked data set; performing layer-by-layer searching and pruning processing on the marked data set to obtain a frequent item set; and generating a plurality of association rules based on each piece of non-empty subset information of the frequent item set, calculating confidence information of each association rule, and taking the association rule meeting a preset confidence threshold condition as an effective track data association rule. The internal relation between the track data and the running gear data can be obtained.

Description

Technical Field

[0001] The present invention relates to the field of data mining technology, and in particular to a multi-dimensional track data association method, device, equipment and medium. Background Art

[0002] With the development of rail transit, the amount of data generated by operations is huge and the types are complex and diverse, such as line conditions, train speed, passenger capacity, climate change, random distribution of track unevenness, uncertainty of passenger behavior, etc., which brings great challenges to processing and analysis. Track correlation analysis requires the integration of multi-source data, but the data such as running gear trend data, track vibration impact trend data and track design parameters are very different in format, semantics, and spatiotemporal scale, making them difficult to integrate.

[0003] In summary, how to accurately mine the correlation relationship of rail transit data to obtain accurate correlation analysis results is a technical problem to be solved in this field. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a multi-dimensional rail data association method, device, equipment and medium, which can accurately mine the association relationship of rail transit data to obtain accurate association analysis results. The specific scheme is as follows:

[0005] In a first aspect, the present application discloses a multi-dimensional track data association method, comprising:

[0006] Performing trend analysis of extreme value change directions on each track data in the multi-source data subset within a preset time range to obtain a first analysis result;

[0007] Performing trend analysis on the data difference changes between each running part data in the multi-source data subset to obtain a second analysis result;

[0008] Integrate and label the first analysis result, the second analysis result, the track data in the multi-source data subset, and the running gear data according to the common dimension to obtain a labeled data set;

[0009] Performing layer-by-layer search and pruning processing on the labeled data set to obtain a frequent itemset consisting of labeled data items that meet a preset support threshold condition;

[0010] A plurality of association rules are generated based on the non-empty subset information of the frequent item sets, and the confidence information of each association rule is calculated, and the association rules whose confidence information meets the preset confidence threshold condition are taken as valid track data association rules.

[0011] Optionally, before performing trend analysis of extreme value change direction on each track data in the multi-source data subset within the preset time range, the method further includes:

[0012] A multi-source data subset related to track-related targets is selected from a multi-source data set including track vibration and impact data, running gear data, track design parameters, and track operation and maintenance data.

[0013] Optionally, after selecting a multi-source data subset related to the track-related target from the multi-source data set including track vibration and impact data, running gear data, track design parameters, and track operation and maintenance data, the method further includes:

[0014] Deleting abnormal data in the multi-source data subset to obtain a cleaned multi-source data subset;

[0015] Performing data filling processing on the cleaned multi-source data subset according to preset data filling rules to obtain a preprocessed multi-source data subset; wherein the preset data filling rules include using the median method to fill in the missing numerical data in the cleaned multi-source data subset, and / or using the mode method to fill in the missing discrete data in the cleaned multi-source data subset;

[0016] Accordingly, the trend analysis of the extreme value change direction of each track data in the multi-source data subset within the preset time range is performed to obtain a first analysis result; and the trend analysis of the data difference change between each running part data in the multi-source data subset is performed to obtain a second analysis result, including:

[0017] Performing trend analysis of extreme value change directions on each track data in the preprocessed multi-source data subset within a preset time range to obtain a first analysis result;

[0018] A trend analysis is performed on the data difference changes between the running part data in the preprocessed multi-source data subset to obtain a second analysis result.

[0019] Optionally, performing trend analysis of extreme value change directions on each track data in the multi-source data subset within a preset time range to obtain a first analysis result includes:

[0020] The orbital vibration value of each orbital data is decomposed to obtain the decomposed target vibration trend, target seasonal component and target residual information that vary with time as the first analysis result.

[0021] Optionally, the orbital vibration value of each orbital data is decomposed to obtain the decomposed target vibration trend, target seasonal component and target residual information that vary with time as the first analysis result, including:

[0022] The period length of the seasonal component is set according to the periodic characteristics of the orbital vibration values ​​of each orbital data;

[0023] Initializing the vibration trend and seasonal components to be empty to set the orbital vibration value as residual information;

[0024] Performing LOESS smoothing on the residual information according to a preset smoothing window size, so as to use the smoothed result as the current vibration trend;

[0025] removing the current vibration trend from the track vibration value to obtain detrended data;

[0026] Segmenting the detrended data according to the cycle length to obtain detrended data segments, and performing LOESS smoothing on each detrended data segment to obtain a preliminary seasonal component;

[0027] Performing intra-period average processing on the preliminary seasonal component to obtain a revised seasonal component;

[0028] removing the corrected seasonal component from the orbital vibration value to obtain deseasonalized data;

[0029] Performing LOESS smoothing processing on the deseasonalized data according to a preset smoothing window size to update the current vibration trend, obtain a new current vibration trend, and jump to execute the step of removing the current vibration trend from the orbital vibration value until the change information of the vibration trend and the seasonal component is less than a preset change threshold, and outputting the current vibration trend and the corrected seasonal component as the target vibration trend and the target seasonal component;

[0030] Target residual information is determined using the orbital vibration value, the target vibration trend, and the target seasonal component to obtain a first analysis result.

[0031] Optionally, the determining target residual information by using the orbit vibration value, the target vibration trend and the target seasonal component includes:

[0032] pass Determine target residual information;

[0033] in, express The target vibration trend at each moment; express The target seasonal component of the moment, express The target residual information at the moment, express The orbital vibration value at the moment.

[0034] Optionally, performing trend analysis on the data difference changes between each running part data in the multi-source data subset to obtain a second analysis result includes:

[0035] Pairing the data of each running part to obtain a number of running part data pairs;

[0036] Performing difference processing on each pair of the running part data to obtain a difference result of each pair of the running part data; the difference result includes a positive difference result and a negative difference result;

[0037] A second analysis result of the running gear data is determined based on the quantity information of the positive difference result and the quantity information of the negative difference result.

[0038] Optionally, determining the second analysis result of the running gear data based on the quantity information of the positive difference result and the quantity information of the negative difference result includes:

[0039] Based on the sum of the number of the positive difference results and the number of the negative difference results as the total number of data pairs;

[0040] Perform binomial distribution processing using the total number of data pairs, the success probability of the positive difference result / negative difference result of the running part data pair, and the number of successes to obtain a success probability value, and calculate significance level information based on the success probability value;

[0041] The trend of the data difference change of the running data is analyzed according to the size result of the quantity information of the positive difference result and the quantity information of the negative difference result and the significance level information to obtain a second analysis result.

[0042] Optionally, analyzing the trend of the data difference change of the running data according to the quantity information of the positive difference result and the quantity information of the negative difference result and the significance level information to obtain a second analysis result includes:

[0043] When the number of the positive difference results is greater than the number of the negative difference results, and the significant level information is less than a preset significant level threshold, it is determined that the data difference change of the running part data is an upward trend, so as to obtain a second analysis result in which the trend analysis result is an upward trend;

[0044] When the number of the positive difference results is less than the number of the negative difference results, and the significance level information is less than the preset significance level threshold, it is determined that the data difference change of the running data is a downward trend, so as to obtain a second analysis result in which the trend analysis result is a downward trend.

[0045] Optionally, the integrating and labeling the first analysis result, the second analysis result, the track data in the multi-source data subset, and the running gear data according to the common dimension to obtain a labeled data set includes:

[0046] Integrate the track data and the running gear data according to the common dimension to obtain a process integrated data set;

[0047] Inserting the first analysis result and the second analysis result into the process integration data set in the form of fields to obtain an integrated data set;

[0048] The integrated data of the numerical type in the integrated data set is divided into intervals, and the integrated data after the interval division is discretely marked in a discrete marking manner to obtain a marked data set.

[0049] Optionally, the step of performing layer-by-layer searching and pruning on the labeled data set to obtain a frequent itemset consisting of labeled data items that meet a preset support threshold condition includes:

[0050] Scanning the labeled data set to count the number of times each labeled data item appears in the labeled data set;

[0051] Calculate the support of the frequency information of each marked data item to the number of data records in the marked data set;

[0052] Prune the labeled data items whose support is less than a preset support threshold condition to obtain updated labeled data items;

[0053] The updated marked data items in each data record are connected according to the preset connection rules, and the connected marked data items are used as new marked data items, and the step of counting the number of times each marked data item appears in the marked data set is jumped to execute until the pruning is completed and the frequent item set is output.

[0054] In a second aspect, the present application discloses a multi-dimensional track data association device, comprising:

[0055] A first trend analysis module, used to perform trend analysis of extreme value change direction on each track data in the multi-source data subset within a preset time range to obtain a first analysis result;

[0056] A second trend analysis module is used to perform trend analysis on the data difference changes between each running part data in the multi-source data subset to obtain a second analysis result;

[0057] An integration and labeling module, configured to integrate and label the first analysis result, the second analysis result, the track data in the multi-source data subset, and the running gear data according to a common dimension to obtain a labeled data set;

[0058] A search processing module, used for performing layer-by-layer search and pruning processing on the labeled data set to obtain a frequent itemset consisting of labeled data items that meet a preset support threshold condition;

[0059] The association generation module is used to generate a plurality of association rules based on the non-empty subset information of the frequent item sets, calculate the confidence information of each association rule, and take the association rules whose confidence information meets the preset confidence threshold condition as the effective track data association rules.

[0060] In a third aspect, the present application discloses an electronic device, comprising:

[0061] Memory, used to store computer programs;

[0062] The processor is used to execute the computer program to implement the steps of the multi-dimensional track data association method disclosed above.

[0063] In a fourth aspect, the present application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of the multi-dimensional track data association method disclosed above are implemented.

[0064] It can be seen that the present application discloses a multi-dimensional track data association method, including: performing a trend analysis on the direction of extreme value changes of each track data in a multi-source data subset within a preset time range to obtain a first analysis result; performing a trend analysis on the data difference changes between each running part data in the multi-source data subset to obtain a second analysis result; integrating and marking the first analysis result, the second analysis result, the track data in the multi-source data subset, and the running part data according to a common dimension to obtain a marked data set; performing layer-by-layer search and pruning on the marked data set to obtain a frequent item set composed of marked data items that meet a preset support threshold condition; generating a number of association rules based on each non-empty subset information of the frequent item set, and calculating the confidence information of each of the association rules, and taking the association rules whose confidence information meets the preset confidence threshold condition as valid track data association rules. It can be seen that by analyzing the trend of the extreme value change direction of the track data and the difference change trend analysis of the running gear data, the first analysis result and the second analysis result are obtained, and the changing trend of the track and running gear operation status is deeply analyzed, and the abnormal increase of track vibration or abnormal fluctuation of running gear data is discovered in time, which provides a basis for predicting potential faults. The above data are integrated according to the common dimensions, breaking the data barriers and forming a comprehensive and systematic labeled data set. The problem of multi-source data fusion is solved, and a rich data foundation is provided for mining multi-dimensional association relationships, ensuring that the analysis results are more accurate and comprehensive. The layer-by-layer search and pruning technology is used to process the labeled data set, which can quickly and accurately obtain frequent item sets that meet the preset support threshold, greatly improve the mining efficiency, accurately locate the key labeled data items, and reduce the consumption of computing resources. Finally, the association rules are generated based on the non-empty subsets of the frequent item sets, and the rules with confidence that meet the threshold are strictly calculated and screened to ensure that the effective track data association rules are true and reliable, and the intrinsic connection between the track data and the running gear data can be obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0066] Figure 1 A flow chart of a multi-dimensional track data association method disclosed in this application;

[0067] Figure 2 A flowchart of a frequent item set generation method disclosed in this application;

[0068] Figure 3 This is a schematic diagram of the structure of a multi-dimensional track data association device disclosed in this application;

[0069] Figure 4 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0071] With the continuous advancement of urbanization, the rail transit industry has developed rapidly. As an efficient, safe and environmentally friendly mode of transportation, rail transit occupies an important position in the urban transportation system. Track monitoring data and vehicle running gear trend data are crucial to the safe operation of rail transit. The interaction between tracks directly affects the running smoothness, safety and comfort of trains. By analyzing track trend data such as track vibration, wavelength, and wave depth, we can understand the dynamic characteristics of the track and discover potential problems in a timely manner. Trend data such as impact and vibration of vehicle running gear reflects the dynamic response of the vehicle during operation, which is of great significance for evaluating the operating status and safety of the vehicle.

[0072] Track vibration impact trend data is closely related to vehicle running gear trend data. The intensification of track vibration often leads to the increase of impact and vibration of vehicle running gear, and this positive correlation can be reduced to a certain extent by optimizing track design parameters. Track trend data and track design parameters influence each other. A smaller curve radius will increase the contact force between the rails, which will lead to the intensification of track vibration, and the wavelength and wave depth may also increase. Reasonable selection of track design parameters can reduce the adverse changes in track trend data. Vehicle running gear trend data is closely related to track design parameters. Unreasonable track design parameters may lead to the deterioration of vehicle running gear trend data, and optimizing track design parameters can improve the operating status of vehicle running gear.

[0073] At present, track correlation analysis needs to integrate multiple data such as running gear trend data, track vibration and impact trend data, and track design parameters. However, these data have differences in format, semantics, and spatiotemporal scale, which brings great technical challenges to the integration work. Track vibration and impact trend data and vehicle running gear trend data are characterized by massive, multi-source heterogeneity, and high noise. Extracting effective information from them and conducting correlation analysis requires complex data processing and analysis methods. Traditional data processing technology is inefficient in processing big data, and it is difficult to meet real-time requirements, and it has limitations in mining deep-level characteristics and laws of data; different types of data require different processing methods and tools in the data processing process, and it is difficult to integrate and analyze structured and unstructured data. For correlation analysis and mining work, to accurately sort out valuable correlation relationships is like accurately "fishing for a needle" in the vast ocean of data, which requires a lot of computing resources and time costs. The operating conditions of rail transit are complex and diverse, including different line conditions, train speeds, passenger capacity, etc. These factors will affect track interactions and the performance of vehicle running gear. The current correlation analysis method is not adaptable enough under different working conditions, which affects the accuracy and reliability of the correlation analysis results.

[0074] Faced with massive amounts of track data, data sampling methods such as random sampling are currently used; in track correlation analysis and mining, the current practice for data cleaning and organization is often to first use manual screening, with professional data processing personnel conducting preliminary review and identification of track data based on experience, and then use statistical methods to mine track correlation relationships, such as correlation analysis (Pearson correlation coefficient, Spearman correlation coefficient, etc.). However, there are many uncertain factors in the operating environment and conditions of rail transit, such as line conditions, train speed, passenger capacity, climate change, random distribution of track unevenness, uncertainty in passenger behavior, etc. These factors will affect track interactions and vehicle running gear performance, increase the difficulty of correlation analysis, and affect the accuracy and reliability of correlation analysis results.

[0075] To this end, the present invention proposes a multi-dimensional track data association scheme, which can process and analyze complex nonlinear problems and multi-factor coupling problems, accurately reveal the intrinsic relationship between track vibration impact trend data, running gear data, track design parameters and other data, and mine the association relationship of rail transit data to obtain accurate association analysis results.

[0076] Reference Figure 1 As shown, an embodiment of the present invention discloses a multi-dimensional track data association method, including:

[0077] Step S11: performing trend analysis of extreme value change directions on each track data in the multi-source data subset within a preset time range to obtain a first analysis result.

[0078] In this embodiment, before the trend analysis of the extreme value change direction of each track data in the multi-source data subset within the preset time range is performed, it also includes: screening the multi-source data subset related to the track-related target from the multi-source data set including track vibration impact data, running gear data, track design parameters, and track operation and maintenance data. It can be understood that the data from multiple sources is obtained, specifically, the multi-source data sets such as track vibration impact data, running gear data, track design parameters, track operation and maintenance data (track status data, track maintenance data); wherein the track vibration impact data include: time, city, line, train number, vehicle number, station, up and down, departure station, destination station, left and right track, axle number, position number, starting kilometer mark, ending kilometer mark, sampling start time, sampling end time, starting speed, ending speed, vibration peak-to-peak value, vibration average value, wavelength, wave depth, vibration maximum value, vibration effective value, wave grinding mark, impact SV effective value, etc.; running gear Track data include: time, city, line, train number, vehicle number, axle number, position number, impact dB, impact SV, temperature, tread out-of-roundness, etc.; track design parameters include: city, line, up and down lines, stations, left and right tracks, kilometer mark, curve, fastener, weld, turnout, slope, roadbed, sleeper, curve radius, etc.; track status data include: time, city, line, up and down lines, station name, left and right tracks, kilometer mark, wave depth, wavelength measurement value, damage, unevenness, etc.; track maintenance data include: city, line, up and down lines, station name, left and right tracks, track grinding time, grinding method, grinding kilometer mark information, etc.;

[0079] Due to the wide variety of data in the multi-source data set, it is necessary to screen out the target data based on the track-related targets of the track-related relationship to construct a multi-source data subset. Specifically, the track-related targets generally include three different mining targets, namely: 1. Mining targets of association rules between track inspection results and track design parameters, speed, etc.; 2. Mining targets of correlation analysis between trend changes in track vibration values, running gear data changes, and track inspection results; 3. Mining targets of association rules between large track vibration values ​​and track design parameters, speed, etc.

[0080] When the mining target is the association rules between track inspection results and track design parameters, speed, etc., the screened target data should be track vibration impact trend data (time, city, line, train number, vehicle number, station, up and down lines, left and right tracks, axle number, position number, departure station, destination station, kilometer mark, speed, etc.), track design parameters (city, line, up and down lines, station, left and right tracks, kilometer mark, curve, fasteners, welds, turnouts, slope, roadbed, sleepers, curve radius, etc.), track status data (time, city, line, up and down lines, station name, left and right tracks, kilometer mark, wave depth, wavelength measurement value, damage, unevenness, etc.) and track maintenance data (city, line, up and down lines, station, left and right tracks, track grinding time, grinding method, grinding kilometer mark information, etc.).

[0081] When the mining target is the correlation analysis between the trend change of track vibration value, the change of running gear data and the track inspection results, the screened target data are track trend data (time, city, line, train number, vehicle number, station, up and down line, departure station, destination station, left and right tracks, axle number, position number, kilometer mark, vibration effective value, vibration peak-to-peak value, etc.), running gear data (time, city, line, train number, vehicle number, axle number, position number, time, tread dB, tread vibration effective value) and track maintenance data (city, line, up and down line, station, left and right tracks, track grinding time, grinding method, grinding kilometer mark information, etc.).

[0082] When the mining target is the association rules between excessively large track vibration values ​​and track design parameters, speed, etc., the target data to be screened should be track trend data (time, city, line, train number, vehicle number, station, up and down lines, departure station, destination station, left and right tracks, axle number, position number, kilometer mark, vibration effective value, speed, etc.) and track design parameters (city, line, station, up and down lines, left and right tracks, kilometer mark, curves, fasteners, welds, switches, slope, ballast, sleepers, curve radius, etc.).

[0083] In this way, according to the actual orbital correlation target, the target data corresponding to the current orbital correlation target can be screened from a large number of multi-source data sets, and the key data can be accurately located according to different mining targets, and irrelevant information interference can be eliminated, so that the analysis direction is clearer and focused on the core mining target. In addition, if all data in the multi-source data set are fully analyzed, the computing resources and time costs are high. Screening the target data greatly reduces the amount of data and simplifies the processing flow; finally, screening the target data can ensure that the data involved in the analysis are highly correlated, reduce the influence of noise data and irrelevant factors, make the correlation analysis more targeted, and the mined correlation relationship is more accurate, and the analysis results are more reliable.

[0084] In this embodiment, after selecting a multi-source data subset related to the track-related target from a multi-source data set including track vibration and impact data, running gear data, track design parameters, and track operation and maintenance data, it also includes: deleting abnormal data in the multi-source data subset to obtain a cleaned multi-source data subset; performing data filling processing on the cleaned multi-source data subset according to a preset data filling rule to obtain a pre-processed multi-source data subset; wherein the preset data filling rule includes using the median method to fill in the numerical missing data in the cleaned multi-source data subset, and / or using the mode method to fill in the discrete type missing data in the cleaned multi-source data subset. It can be understood that after determining the multi-source data subset related to the track-related target, the abnormal values ​​in the current multi-source data subset are deleted to obtain the cleaned multi-source data subset; the missing values ​​of the numerical type in the cleaned multi-source data subset are filled by the median method, and the missing values ​​of the discrete variables are filled by the mode filling method to ensure the integrity of the data, thereby obtaining the pre-processed multi-source data subset.

[0085] In a specific implementation, the steps of filling in missing numerical data using the median method are as follows:

[0086] The median refers to the data in the middle of a set of data. It divides the cleaned multi-source data subset into two equal parts, half of which is greater than the median and half is less than the median. The basic principle of using the median to fill missing values ​​is to replace missing values ​​with the median of the attribute to reduce the deviation caused by missing values. First, sort the data by size, and set the sorting result to:

[0087] ;

[0088] in, Represents the first The value of the data, , is a positive integer, the median calculation formula is as follows:

[0089] ;

[0090] In another specific implementation, the mode method is used to fill missing values ​​in discrete data. Using the mode to fill missing values ​​can preserve the distribution characteristics of the data, which is particularly suitable for nominal variables. The mode filling method is simple and does not require a complex statistical model.

[0091] In this embodiment, a trend analysis of the direction of extreme value changes is performed on each orbital data in the preprocessed multi-source data subset within a preset time range to obtain a first analysis result; it can be understood that performing a trend analysis of the direction of extreme value changes on each orbital data (discrete orbital data after filling, numerical orbital data after filling) in the preprocessed multi-source data subset within the preset time range to obtain the first analysis result can ensure the accuracy, continuity and effectiveness of the orbital data and avoid errors in the trend analysis of the direction of extreme value changes of the orbital data.

[0092] In this embodiment, the orbital vibration value of each orbital data is decomposed to obtain the decomposed target vibration trend, target seasonal component and target residual information that change with time as the first analysis result. Specifically, the period length of the seasonal component is set according to the periodic characteristics of the orbital vibration value of each orbital data; the vibration trend and seasonal component are initialized to empty to set the orbital vibration value as residual information; the residual information is subjected to LOESS smoothing processing according to a preset smoothing window size to use the smoothed result as the current vibration trend; the current vibration trend is removed from the orbital vibration value to obtain detrended data; the detrended data is segmented according to the period length to obtain each detrended data segment, and each detrended data segment is subjected to LOESS smoothing processing to obtain a preliminary seasonal component; the preliminary seasonal component is subjected to intra-period average processing, to obtain a corrected seasonal component; to remove the corrected seasonal component from the orbital vibration value to obtain de-seasoned data: to perform LOESS smoothing on the de-seasoned data according to a preset smoothing window size to update the current vibration trend to obtain a new current vibration trend, and to jump to the step of removing the current vibration trend from the orbital vibration value until the change information of the vibration trend and the seasonal component is less than a preset change threshold, and to output the current vibration trend and the corrected seasonal component as the target vibration trend and the target seasonal component; to determine the target residual information using the orbital vibration value, the target vibration trend and the target seasonal component to obtain a first analysis result.

[0093] It can be understood that the orbital vibration value (vibration effective value, vibration peak-to-peak value) of the orbital data is decomposed into three parts: vibration trend (Trend), seasonal component (Seasonal), and residual information (Residual) using STL (Seasonal-Trend Decomposition using LOESS). The decomposition formula is as follows:

[0094] ;

[0095] in, express The target vibration trend at each moment indicates the long-term change trend of the data; express The target seasonal component at time represents periodic fluctuations with a fixed period length, express The target residual information at the moment represents the random fluctuation or noise part.

[0096] First, the trend components after STL decomposition are averaged on a daily basis to form new daily average vibration effective value trend data, recorded as X. Then, the extreme value change of X is calculated as the standard value. The formula is:

[0097] ;

[0098] Among them, max and min represent the maximum and minimum values ​​in the effective trend data of the entire daily average vibration, respectively. argmax and argmin represent the sequence numbers of the maximum and minimum values ​​in the entire data. The change value of each time period (for example, data for 2 consecutive days) is defined as follows:

[0099] ;

[0100] in, For the starting point of each time period, For the end point of each time period, a total of three discrete changes are defined: decline, stability, and rise. If the value is less than or equal to the threshold (for example, 0.05), the change is defined as stable. If the value is greater than (for example, 0.05), it is defined as a decrease or increase. The decrease and increase are determined by the size of the starting point and the end point of the time period. If , it is decreasing.

[0101] In this way, the trend components decomposed by the daily average STL are used as the daily trend data, and then trend analysis is performed based on the extreme value changes of the trend data to obtain three analysis results: trend rising / falling / stable.

[0102] Step S12: performing trend analysis on the data difference changes between the running part data in the multi-source data subset to obtain a second analysis result.

[0103] In this embodiment, a trend analysis is performed on the data difference change between each running part data in the preprocessed multi-source data subset to obtain a second analysis result. It can be understood that the trend analysis of the data difference change direction between each running part data (discrete running part data after filling, numerical running part data after filling) in the preprocessed multi-source data subset within the preset time range to obtain the second analysis result can ensure the accuracy, continuity and validity of the running part data and avoid errors in the trend analysis of the data difference change direction of the running part data.

[0104] In this embodiment, each running part data is paired to obtain a number of running part data pairs; each running part data pair is subjected to difference processing to obtain the difference result of each running part data pair; the difference result includes positive difference result and negative difference result; the second analysis result of the running part data is determined based on the quantity information of the positive difference result and the quantity information of the negative difference result. Specifically, the total number of data pairs is taken as the sum of the number of positive difference results and the number of negative difference results; the total number of data pairs, the success probability of the positive difference result / negative difference result of the running part data pair, and the number of successes are used to perform binomial distribution processing to obtain the success probability value, and the significance level information is calculated based on the success probability value; the trend of the data difference change of the running part data is analyzed according to the size result of the quantity information of the positive difference result and the quantity information of the negative difference result, and the significance level information, to obtain the second analysis result. Among them, when the number of positive difference results is greater than the number of negative difference results, and the significance level information is less than the preset significance level threshold, it is determined that the data difference change of the running data is an upward trend, so as to obtain a second analysis result in which the trend analysis result is an upward trend; when the number of positive difference results is less than the number of negative difference results, and the significance level information is less than the preset significance level threshold, it is determined that the data difference change of the running data is a downward trend, so as to obtain a second analysis result in which the trend analysis result is a downward trend.

[0105] It can be understood that, firstly, based on the time range of the track vibration data, the running gear data in the corresponding time range is selected, and then the Cox-Stuart trend test is used to perform trend analysis on the running gear data (tread db, tread vibration effective value), and three analysis results of trend increase / decrease / no trend are obtained; the specific steps of running gear data trend analysis are as follows:

[0106] Select The running gear data of the data length forms a data set , .

[0107] Pick and Form a pair of data ( , ∈ ), which generates For the running gear data pair ( , ), , ), , ), etc., among which for:

[0108] ;

[0109] Use the difference between the two running gear data of each pair, recorded as ,set up Record is a positive quantity, that is, the number of positive difference results; Record The number of negative difference results.

[0110] Use binomial discrete distribution to test whether the trend is significant and the significance level The specific calculation formula of the value is as follows:

[0111] ;

[0112] in is a binomial distribution, expressed as follows:

[0113] ;

[0114] in, The number of successes, or test statistic, is the number of trials, that is, the total number of data pairs (excluding the number of pairs with a difference of 0), is the probability of success per trial (e.g. 0.5):

[0115] ;

[0116] when > When, and is less than the significant level threshold (0.05), there is an upward trend; when < When, and If it is less than the significant level threshold (0.05), there is a downward trend, otherwise there is no trend.

[0117] Step S13: Integrate and label the first analysis result, the second analysis result, the track data in the multi-source data subset, and the running gear data according to the common dimension to obtain a labeled data set.

[0118] In this embodiment, the track data and the running gear data are integrated according to the common dimensions to obtain a process integrated data set; the first analysis result and the second analysis result are inserted into the process integrated data set in the form of fields to obtain an integrated data set; the integrated data of the numerical type in the integrated data set is divided into intervals, and the integrated data after the interval division is discretely marked in a discrete marking manner to obtain a marked data set. It can be understood that data integration needs to be integrated according to the common dimensions of multi-source data and the mining targets of the association relationship. For example, if there are the following three different mining targets, the integrated data is as follows:

[0119] If the mining target is the association rules between track inspection results and track design parameters, speed, etc., the integrated data set should be time, city, line, train number, vehicle number, up and down lines, departure station, destination station, left and right tracks, axle number, position number, kilometer mark, speed, wavelength, wave depth, damage, unevenness, curves, fasteners, welds, switches, slopes, sleepers, and ballast, etc.

[0120] The mining target is the correlation analysis between the trend change of track vibration value, the trend data change of running gear, and the track inspection results. The integrated data should be time, city, line, train number, vehicle number, up and down line, departure station, destination station, left and right track, axle number, position number, kilometer mark, track vibration peak-to-peak value trend analysis results, track vibration effective value trend analysis results, running gear tread vibration effective value trend analysis results, running gear tread impact dB trend analysis results, whether it is polished, etc.

[0121] If the mining target is the association rules between excessive track vibration values ​​and track design parameters, speed, etc., the integrated data set should include time, city, line, train number, vehicle number, up and down lines, departure station, destination station, left and right tracks, axle number, position number, kilometer mark, track vibration effective value trend increase analysis results, speed, curves, fasteners, welds, switches, slope, sleepers, ballast, etc.

[0122] Furthermore, discrete data of track design parameters such as fastener type, sleeper type, and ballast bed type are marked by category (for example, if there are three types of fasteners, they are marked as fastener 1, fastener 2, and fastener 3 respectively); for numerical data such as curve radius data, wavelength, wave depth, and rotation speed, equal-width interval division is first used to convert continuous values ​​into discrete intervals, and then marked according to discrete intervals (for example, the curve radius is divided into two categories: 0~1000m and 1000~2000m, then the curve radius can be marked as curve radius 1 and curve radius 2).

[0123] In this way, the processed multi-source data are integrated and labeled into a unified dataset according to common dimensions (such as time, city, route, etc.) to form a labeled dataset.

[0124] Step S14: performing layer-by-layer search and pruning processing on the labeled data set to obtain frequent itemsets consisting of labeled data items that meet a preset support threshold condition.

[0125] In this embodiment, the labeled data set is scanned and a minimum support threshold (such as 0.2) and a minimum confidence threshold (such as 0.2) are set, and the relationship between multi-dimensional data items is repeatedly mined to obtain a frequent itemset consisting of data items that meet the above threshold conditions.

[0126] Step S15: generating a plurality of association rules based on the non-empty subset information of the frequent itemsets, calculating the confidence information of each association rule, and taking the association rules whose confidence information meets the preset confidence threshold condition as the valid track data association rules.

[0127] In this embodiment, each non-empty subset in the frequent item set is generated by a combination method, and then the corresponding association rules are generated by using the data item information in each non-empty subset, and the association rules are calculated according to the confidence. If the calculated confidence meets the preset confidence threshold condition, the corresponding association rule is used as the valid track data association rule.

[0128] It can be seen that the present application discloses a multi-dimensional track data association method, including: performing a trend analysis on the direction of extreme value changes of each track data in a multi-source data subset within a preset time range to obtain a first analysis result; performing a trend analysis on the data difference changes between each running part data in the multi-source data subset to obtain a second analysis result; integrating and marking the first analysis result, the second analysis result, the track data in the multi-source data subset, and the running part data according to a common dimension to obtain a marked data set; performing layer-by-layer search and pruning on the marked data set to obtain a frequent item set composed of marked data items that meet a preset support threshold condition; generating a number of association rules based on each non-empty subset information of the frequent item set, and calculating the confidence information of each of the association rules, and taking the association rules whose confidence information meets the preset confidence threshold condition as valid track data association rules. It can be seen that by analyzing the trend of the extreme value change direction of the track data and the difference change trend analysis of the running gear data, the first analysis result and the second analysis result are obtained, and the changing trend of the track and running gear operation status is deeply analyzed, and the abnormal increase of track vibration or abnormal fluctuation of running gear data is discovered in time, which provides a basis for predicting potential faults. The above data are integrated according to the common dimensions, breaking the data barriers and forming a comprehensive and systematic labeled data set. The problem of multi-source data fusion is solved, and a rich data foundation is provided for mining multi-dimensional association relationships, ensuring that the analysis results are more accurate and comprehensive. The layer-by-layer search and pruning technology is used to process the labeled data set, which can quickly and accurately obtain frequent item sets that meet the preset support threshold, greatly improve the mining efficiency, accurately locate the key labeled data items, and reduce the consumption of computing resources. Finally, the association rules are generated based on the non-empty subsets of the frequent item sets, and the rules with confidence that meet the threshold are strictly calculated and screened to ensure that the effective track data association rules are true and reliable, and the intrinsic connection between the track data and the running gear data can be obtained.

[0129] Reference Figure 2 As shown, the embodiment of the present invention specifically discloses step S14: performing layer-by-layer search and pruning processing on the labeled data set to obtain a frequent item set consisting of labeled data items that meet a preset support threshold condition. Compared with the previous embodiment, this embodiment further explains and optimizes the technical solution. Specifically:

[0130] Step S21: Scan the labeled data set to count the number of times each labeled data item appears in the labeled data set.

[0131] In this embodiment, an association mining model is established, such as an Apriori algorithm model, and the minimum support threshold (0.2) and the minimum confidence threshold (0.2) of the model are set. First, the Apriori algorithm performs a first scan on the labeled data set to count the number of times each single item (single labeled data item) appears in the labeled data set. For example, when the labeled database has Tid 1.ACD, Tid 2.BCE, Tid 3.ABCE, and Tid 4.BE, the first scan operation is performed to count the number of times the single labeled data items ABCDE appear respectively. The statistical results are: A appears 2 times, B appears 3 times, C appears 3 times, D appears 1 time, E appears 3 times, and the number of data records is 5.

[0132] Step S22: Calculate the support degree of the frequency information of each marked data item to the number of data records in the marked data set.

[0133] In this embodiment, the Apriori algorithm model is used to first calculate the support of all single items. According to the above statistical results and the number of data records, the support of ABCDE is calculated as: 0.4, 0.6, 0.6, 0.2, and 0.6 respectively.

[0134] Step S23: Pruning the labeled data items whose support is less than a preset support threshold condition to obtain updated labeled data items.

[0135] In this embodiment, according to the preset minimum support threshold of 0.2, the single items that meet the conditions are screened out as ABCE as the updated labeled data items, and the frequent itemset 1 formed by these single items is {A, B, C, E}.

[0136] Step S24: connect the updated marked data items in each data record according to the preset connection rules, and use the connected marked data items as new marked data items, and jump to execute the step of counting the number of times each marked data item appears in the marked data set until pruning is completed and the frequent item set is output.

[0137] In this embodiment, after obtaining the frequent item set 1, the candidate item set is generated by connecting the frequent item set 1 itself, wherein the connection rule is that if the first k-2 items of two frequent item sets 1 are the same and the k-1th items are different, then they can be connected to generate a candidate item set. Therefore, the self-connection result of the above frequent item set 1 is: {A, B}, {A, C}, {A, E}, {B, C}, {B, E}, {C, E}, and the number of occurrences of the marked data items after connection in the candidate item set is counted. The number of occurrences of the marked data items after connection {A, B}, {A, C}, {A, E}, {B, C}, {B, E}, {C, E} are respectively: 1, 2, 1, 2, 3, 2. Then the support of the newly generated candidate item set is calculated, and it is screened again according to the screening requirement based on the minimum support threshold, and the screening result is: {A, C}, {B, C}, {B, E}, {C, E}. For the generated candidate item sets, scan the data set again, count the number of times each candidate item set appears in the data set, and calculate its support. Then, according to the minimum support threshold, filter out the candidate item sets that meet the conditions. These candidate item sets constitute frequent item sets 2. Repeat the above steps until no new frequent item sets can be generated. Repeat the above process continuously to generate candidate k item sets, calculate their support and filter them until no new frequent item sets can be generated. The final frequent item set is {B, C, E}, and the number is 2.

[0138] In this way, the frequent itemset generation process uses layer-by-layer search and pruning technology, starting from frequent itemset 1 and gradually generating higher-order frequent itemsets. After each new candidate itemset is generated, support calculation and screening are performed to promptly exclude itemsets that do not meet the conditions. This reduces unnecessary calculations, improves the efficiency of the algorithm, and enables good performance to be maintained even when processing large-scale data sets.

[0139] In this embodiment, after the frequent itemsets are determined, for a frequent itemset, all its non-empty subsets are generated by combination. For example, for the frequent itemset {B, C, E}, its non-empty subsets are {B}, {C}, {E}, {B, C}, {B, E}, {C, E}.

[0140] For each generated rule , calculate its confidence, for each generated rule, according to the confidence formula Calculate its confidence.

[0141] The calculated confidence is compared with the preset minimum confidence threshold. If the calculated confidence is greater than the minimum confidence threshold 0.2, the rule is a valid association rule. Correspondingly, the rules corresponding to other non-empty subsets are as above. The judgment method is the same as that of , which will not be described in detail.

[0142] It can be seen that by calculating the support of a single item and a candidate item set, and filtering according to the minimum support threshold, it is possible to accurately find frequently occurring data item combinations, i.e., frequent item sets, from massive and complex data sets. These frequent item sets are the core parts of the data that are representative and relevant, which help focus on key information and avoid wasting computing resources and time on irrelevant data. In addition, the minimum support and minimum confidence thresholds are set to filter frequent item sets and association rules to ensure that the mined rules have a certain degree of universality and reliability. Only those rules that appear frequently in the data set and have a high degree of association will be retained as valid association rules. Avoid mining some false rules that appear by chance or have weak correlation, and improve the credibility and practicality of the analysis results.

[0143] Reference Figure 3 As shown, the present invention also discloses a multi-dimensional track data association device, including:

[0144] A first trend analysis module 11 is used to perform trend analysis of extreme value change directions on each track data in the multi-source data subset within a preset time range to obtain a first analysis result;

[0145] A second trend analysis module 12 is used to perform trend analysis on the data difference changes between the running part data in the multi-source data subset to obtain a second analysis result;

[0146] An integration and labeling module 13 is used to integrate and label the first analysis result, the second analysis result, the track data in the multi-source data subset, and the running gear data according to a common dimension to obtain a labeled data set;

[0147] A search processing module 14 is used to perform layer-by-layer search and pruning on the labeled data set to obtain a frequent item set consisting of labeled data items that meet a preset support threshold condition;

[0148] The association generating module 15 is used to generate a plurality of association rules based on the non-empty subset information of the frequent itemsets, calculate the confidence information of each association rule, and take the association rules whose confidence information meets the preset confidence threshold condition as the valid track data association rules.

[0149] It can be seen that the present application discloses performing a trend analysis on the direction of extreme value changes of each track data in a multi-source data subset within a preset time range to obtain a first analysis result; performing a trend analysis on the data difference changes between each running part data in the multi-source data subset to obtain a second analysis result; integrating and marking the first analysis result, the second analysis result, the track data in the multi-source data subset, and the running part data according to a common dimension to obtain a marked data set; performing a layer-by-layer search and pruning process on the marked data set to obtain a frequent item set composed of marked data items that meet a preset support threshold condition; generating a number of association rules based on each non-empty subset information of the frequent item set, and calculating the confidence information of each of the association rules, and taking the association rules whose confidence information meets the preset confidence threshold condition as valid track data association rules. It can be seen that by analyzing the trend of the extreme value change direction of the track data and the difference change trend analysis of the running gear data, the first analysis result and the second analysis result are obtained, and the changing trend of the track and running gear operation status is deeply analyzed, and the abnormal increase of track vibration or abnormal fluctuation of running gear data is discovered in time, which provides a basis for predicting potential faults. The above data are integrated according to the common dimensions, breaking the data barriers and forming a comprehensive and systematic labeled data set. The problem of multi-source data fusion is solved, and a rich data foundation is provided for mining multi-dimensional association relationships, ensuring that the analysis results are more accurate and comprehensive. The layer-by-layer search and pruning technology is used to process the labeled data set, which can quickly and accurately obtain frequent item sets that meet the preset support threshold, greatly improve the mining efficiency, accurately locate the key labeled data items, and reduce the consumption of computing resources. Finally, the association rules are generated based on the non-empty subsets of the frequent item sets, and the rules with confidence that meet the threshold are strictly calculated and screened to ensure that the effective track data association rules are true and reliable, and the intrinsic connection between the track data and the running gear data can be obtained.

[0150] Furthermore, the present application also discloses an electronic device. Figure 4 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram cannot be regarded as any limitation on the scope of use of the present application.

[0151] Figure 4 The structural diagram of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the multi-dimensional track data association method disclosed in any of the above embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0152] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0153] Among them, the processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 21 can be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 21 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 21 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 21 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0154] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0155] Among them, the operating system 221 is used to manage and control the hardware devices and computer programs 222 on the electronic device 20, so as to realize the operation and processing of the massive data 223 in the memory 22 by the processor 21, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the multi-dimensional track data association method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks. In addition to data transmitted from an external device received by the electronic device, the data 223 can also include data collected by its own input and output interface 25, etc.

[0156] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein the computer program, when executed by a processor, implements the multi-dimensional track data association method disclosed above. For the specific steps of the method, reference may be made to the corresponding contents disclosed in the above embodiments, and no further description will be given here.

[0157] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0158] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application. The steps of the method or algorithm described in conjunction with the embodiments disclosed herein can be implemented directly with hardware, a software module executed by a processor, or a combination of the two. The software module can be placed in a random access memory RAM (Random Access Memory), memory, read-only memory ROM (Read Only Memory), electrically programmable EPROM (Electrically Programmable Read Only Memory), electrically erasable programmable EEPROM (ElectricErasable Programmable Read Only Memory), register, hard disk, removable disk, CD-ROM (CompactDisc-Read Only Memory), or any other form of storage medium known in the technical field.

[0159] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0160] The scheme provided by the present invention is introduced in detail above. Specific examples are used in this article to illustrate the principle and implementation mode of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation mode and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A multi-dimensional track data association method, characterized in that: include: Performing trend analysis of extreme value change directions on each track data in the multi-source data subset within a preset time range to obtain a first analysis result; Performing trend analysis on the data difference changes between each running part data in the multi-source data subset to obtain a second analysis result; Integrate and label the first analysis result, the second analysis result, the track data in the multi-source data subset, and the running gear data according to the common dimension to obtain a labeled data set; Performing layer-by-layer search and pruning processing on the labeled data set to obtain a frequent itemset consisting of labeled data items that meet a preset support threshold condition; A plurality of association rules are generated based on the non-empty subset information of the frequent item sets, and the confidence information of each association rule is calculated, and the association rules whose confidence information meets the preset confidence threshold condition are taken as valid track data association rules.

2. The multi-dimensional track data association method according to claim 1, characterized in that: Before performing trend analysis on the extreme value change direction of each track data in the multi-source data subset within the preset time range, the method further includes: A multi-source data subset related to track-related targets is selected from a multi-source data set including track vibration and impact data, running gear data, track design parameters, and track operation and maintenance data.

3. The multi-dimensional track data association method according to claim 2, characterized in that: After selecting a multi-source data subset related to the track-related target from the multi-source data set including track vibration and impact data, running gear data, track design parameters, and track operation and maintenance data, the method further includes: Deleting abnormal data in the multi-source data subset to obtain a cleaned multi-source data subset; Performing data filling processing on the cleaned multi-source data subset according to preset data filling rules to obtain a preprocessed multi-source data subset; wherein the preset data filling rules include using the median method to fill in the missing numerical data in the cleaned multi-source data subset, and / or using the mode method to fill in the missing discrete data in the cleaned multi-source data subset; Accordingly, the trend analysis of the extreme value change direction of each track data in the multi-source data subset within the preset time range is performed to obtain a first analysis result; and the trend analysis of the data difference change between each running part data in the multi-source data subset is performed to obtain a second analysis result, including: Performing trend analysis of extreme value change directions on each track data in the preprocessed multi-source data subset within a preset time range to obtain a first analysis result; A trend analysis is performed on the data difference changes between the running part data in the preprocessed multi-source data subset to obtain a second analysis result.

4. The multi-dimensional track data association method according to claim 1, characterized in that: The step of performing trend analysis of extreme value change directions on each track data in the multi-source data subset within a preset time range to obtain a first analysis result includes: The orbital vibration value of each orbital data is decomposed to obtain the decomposed target vibration trend, target seasonal component and target residual information that vary with time as the first analysis result.

5. The multi-dimensional track data association method according to claim 4, characterized in that: The orbital vibration value of each orbital data is decomposed to obtain the decomposed target vibration trend, target seasonal component and target residual information that change over time as the first analysis result, including: The period length of the seasonal component is set according to the periodic characteristics of the orbital vibration values ​​of each orbital data; Initializing the vibration trend and seasonal components to be empty to set the orbital vibration value as residual information; Performing LOESS smoothing on the residual information according to a preset smoothing window size, so as to use the smoothed result as the current vibration trend; removing the current vibration trend from the track vibration value to obtain detrended data; Segmenting the detrended data according to the cycle length to obtain detrended data segments, and performing LOESS smoothing on each detrended data segment to obtain a preliminary seasonal component; Performing intra-period average processing on the preliminary seasonal component to obtain a revised seasonal component; removing the corrected seasonal component from the orbital vibration value to obtain deseasonalized data; Performing LOESS smoothing processing on the deseasonalized data according to a preset smoothing window size to update the current vibration trend, obtain a new current vibration trend, and jump to execute the step of removing the current vibration trend from the orbital vibration value until the change information of the vibration trend and the seasonal component is less than a preset change threshold, and outputting the current vibration trend and the corrected seasonal component as the target vibration trend and the target seasonal component; Target residual information is determined using the orbital vibration value, the target vibration trend, and the target seasonal component to obtain a first analysis result.

6. The multi-dimensional track data association method according to claim 5, characterized in that: The determining target residual information by using the orbit vibration value, the target vibration trend and the target seasonal component includes: pass Determine target residual information; in, express The target vibration trend at each moment; express The target seasonal component of the moment, express The target residual information at the moment, express The orbital vibration value at the moment.

7. The multi-dimensional track data association method according to claim 1, characterized in that: The performing trend analysis on the data difference changes between each running part data in the multi-source data subset to obtain a second analysis result includes: Pairing the data of each running part to obtain a number of running part data pairs; Performing difference processing on each pair of the running part data to obtain a difference result of each pair of the running part data; the difference result includes a positive difference result and a negative difference result; A second analysis result of the running gear data is determined based on the quantity information of the positive difference result and the quantity information of the negative difference result.

8. The multi-dimensional track data association method according to claim 7, characterized in that: The determining of the second analysis result of the running gear data based on the quantity information of the positive difference result and the quantity information of the negative difference result includes: Based on the sum of the number of the positive difference results and the number of the negative difference results as the total number of data pairs; Perform binomial distribution processing using the total number of data pairs, the success probability of the positive difference result / negative difference result of the running part data pair, and the number of successes to obtain a success probability value, and calculate significance level information based on the success probability value; The trend of the data difference change of the running data is analyzed according to the size result of the quantity information of the positive difference result and the quantity information of the negative difference result and the significance level information to obtain a second analysis result.

9. The multi-dimensional track data association method according to claim 8, characterized in that: The step of analyzing the trend of the data difference change of the running data according to the quantity information of the positive difference result and the quantity information of the negative difference result and the significance level information to obtain a second analysis result includes: When the number of the positive difference results is greater than the number of the negative difference results, and the significance level information is less than a preset significance level threshold, it is determined that the data difference change of the running part data is an upward trend, so as to obtain a second analysis result in which the trend analysis result is an upward trend; When the number of the positive difference results is less than the number of the negative difference results, and the significance level information is less than the preset significance level threshold, it is determined that the data difference change of the running data is a downward trend, so as to obtain a second analysis result in which the trend analysis result is a downward trend.

10. The multi-dimensional track data association method according to claim 1, characterized in that: The integrating and labeling the first analysis result, the second analysis result, the track data in the multi-source data subset, and the running gear data according to the common dimension to obtain a labeled data set includes: Integrate the track data and the running gear data according to the common dimension to obtain a process integrated data set; Inserting the first analysis result and the second analysis result into the process integration data set in the form of fields to obtain an integrated data set; The integrated data of the numerical type in the integrated data set is divided into intervals, and the integrated data after the interval division is discretely marked in a discrete marking manner to obtain a marked data set.

11. The multi-dimensional track data association method according to claim 1, characterized in that: The step of performing layer-by-layer searching and pruning on the labeled data set to obtain a frequent itemset consisting of labeled data items that meet a preset support threshold condition includes: Scanning the labeled data set to count the number of times each labeled data item appears in the labeled data set; Calculate the support of the frequency information of each marked data item to the number of data records in the marked data set; Prune the labeled data items whose support is less than a preset support threshold condition to obtain updated labeled data items; The updated marked data items in each data record are connected according to the preset connection rules, and the connected marked data items are used as new marked data items, and the step of counting the number of times each marked data item appears in the marked data set is jumped to execute until the pruning is completed and the frequent item set is output.

12. A multi-dimensional track data association device, characterized in that: include: A first trend analysis module, used to perform trend analysis of extreme value change direction on each track data in the multi-source data subset within a preset time range to obtain a first analysis result; A second trend analysis module is used to perform trend analysis on the data difference changes between each running part data in the multi-source data subset to obtain a second analysis result; An integration and labeling module, configured to integrate and label the first analysis result, the second analysis result, the track data in the multi-source data subset, and the running gear data according to a common dimension to obtain a labeled data set; A search processing module, used for performing layer-by-layer search and pruning processing on the labeled data set to obtain a frequent itemset consisting of labeled data items that meet a preset support threshold condition; The association generation module is used to generate a plurality of association rules based on the non-empty subset information of the frequent item sets, calculate the confidence information of each association rule, and take the association rules whose confidence information meets the preset confidence threshold condition as the effective track data association rules.

13. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the multi-dimensional track data association method according to any one of claims 1 to 10.

14. A computer-readable storage medium, characterized in that: Used to store a computer program; wherein, when the computer program is executed by a processor, the steps of the multi-dimensional track data association method according to any one of claims 1 to 10 are implemented.

Images