Urban rail vehicle trend data anomaly identification method, device and equipment

Abstract

The invention relates to the technical field of data processing, in particular to an anomaly identification method, device and equipment for urban rail vehicle trend data, and the method comprises the steps: determining a difference value between time adjacent data in a trend data set of a target category; determining whether single-point mutation exists or not according to the obtained difference value sequence; performing feature calculation on multiple pieces of data in each sliding window of the trend data set to obtain a first feature set; determining whether continuous mutation exists or not according to the first feature set corresponding to the adjacent sliding windows; performing feature calculation on multiple pieces of data of the target sliding windows of the trend data sets corresponding to the multiple positions to obtain second feature sets corresponding to the multiple positions; according to the corresponding data of the second feature set and the trend data set, whether the same-condition comparison significant difference exists or not is determined; and generating an abnormal result according to the three results. According to the method, various abnormal conditions in the data are considered, category division is carried out, and the analysis mode is simple and reliable.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and in particular to a method, device and equipment for identifying anomalies in trend data of urban rail vehicles. Background Art

[0002] With the continuous advancement of urban rail transit system construction and the development of digital information technology, the accuracy and generalizability of the application of urban rail transit intelligent operation and maintenance platforms in urban traffic management has become one of the current research focuses in the field of rail transit. Therefore, the quality of the platform's collection and storage of real-time status information of urban rail vehicles will directly affect the operation and maintenance platform's assessment of vehicle operation status and vehicle maintenance and repair recommendations.

[0003] The on-board fault diagnosis technology of the running gear is an important means to obtain the operating status of urban rail vehicles. The status characteristic quantity information of each component of the vehicle is collected and transmitted by the detection device to form time series trend data, but due to the influence of factors such as manual operation and actual environment, anomalies are inevitable in the detection data. In order to ensure the data quality of vehicle operating status evaluation, data anomaly identification will play a key role.

[0004] At present, due to the complex working conditions of urban rail vehicles, according to the operating characteristics of vehicle running parts and monitoring equipment, trend data anomalies usually include multiple situations. For example, the sensor device or communication system is affected by the operating environment interference, recording errors and other factors, resulting in data loss, noise samples deviating from the normal data distribution expectation, sample values ​​remain unchanged, improper transmission processing leads to data disorder and loss, etc. In order to ensure the data quality of subsequent vehicle health status assessment work, such anomalies unrelated to vehicle equipment can be screened out by data cleaning; however, long-term trend changes caused by vehicle equipment anomalies, mutation data values ​​in correlation sequence pairs or correlation sequence sets at the same time, data migration in correlation sequences, high noise or high variance data, etc., are the true manifestation of the equipment operation status, which does not belong to the scope of data cleaning and needs to be identified by anomaly detection algorithm methods.

[0005] At present, there are many methods for identifying anomalies in trend data in the rail transit industry, including statistical methods, traditional machine learning methods, and deep learning methods. However, they can only identify anomalies in short-term stable data, and cannot use long-term non-stationary historical data for analysis and mining. Moreover, this identification method is relatively complex and requires a lot of manpower and material resources. Summary of the invention

[0006] The purpose of the present application is to provide a method, device and equipment for identifying abnormality in trend data of urban rail vehicles, which can identify abnormal data.

[0007] Acquire a trend data set of a target category, the trend data set of the target category including a plurality of data within a preset time period, wherein the target category is any one of a plurality of categories monitored by an onboard system of an urban rail vehicle;

[0008] Determine the difference between the temporally adjacent data in the trend data set to obtain a difference sequence; and determine whether there is a first situation of a single point mutation according to the difference sequence;

[0009] Performing feature calculation on multiple data in each sliding window of the trend data set to obtain a first feature set, wherein the first feature set includes at least one of a first window mean, a first window variance, a first window maximum value, and a first window minimum value; and determining whether there is a second situation of continuous mutation based on the first feature sets corresponding to adjacent sliding windows;

[0010] Perform feature calculation on multiple data of the target sliding window of the trend data set corresponding to each of the multiple positions to obtain a second feature set corresponding to each of the multiple positions; determine whether there is a third situation in which there is a significant difference when compared under the same conditions according to the second feature sets of each of the multiple positions and the multiple data of the target sliding window of the trend data set corresponding to each of the multiple positions, where the target sliding window is any window of the multiple sliding windows of the trend data set;

[0011] An abnormal result is generated according to the first situation, the second situation, and the third situation.

[0012] In a preferred example, the present application can be further configured as follows: determining whether there is a first case of a single point mutation according to the difference sequence, including:

[0013] Determining whether there are outliers in the difference sequence;

[0014] If there is no outlier, it is determined that the first case is that there is no single point mutation;

[0015] If there is an outlier, and the number of outliers is 1, then when the outlier and the first single-point mean are not equal, and the outlier and the second single-point mean are not equal, it is determined that the first situation is that there is a single-point mutation; otherwise, it is determined that the first situation is that there is no single-point mutation, wherein the first single-point mean is determined by the mean of all differences before the outlier in the difference sequence and the identification weight, and the second single-point mean is determined by the mean of all differences after the outlier in the difference sequence and the identification weight;

[0016] If there are outliers and the number of outliers is greater than 1, then when there are no consecutive outliers, and, for each outlier, the outlier and the third single-point mean are not equal, and the outlier and the fourth single-point mean are not equal, the first situation is determined to be the presence of a single-point mutation, otherwise, the first situation is determined to be the absence of a single-point mutation, wherein the third single-point mean is the mean of all differences between the outlier and the previous outlier in the difference sequence, and the fourth single-point mean is the mean of all differences between the outlier and the next outlier in the difference sequence.

[0017] In a preferred example, the present application may be further configured as follows: when the outlier value is not equal to the first single point mean, and the outlier value is not equal to the second single point mean, determining that the first situation is that a single point mutation exists includes:

[0018] When the outlier value is greater than the first single point mean, it is determined that the first situation is a single point mutation increase; when the outlier value is less than the first single point mean, it is determined that the first situation is a single point mutation decrease;

[0019] When the outlier value is smaller than the second single point mean, the first situation is determined to be a single point mutation increase; when the outlier value is larger than the second single point mean, the first situation is determined to be a single point mutation decrease;

[0020] Alternatively, for each outlier, when the outlier and the third single point mean are not equal, and the outlier and the fourth single point mean are not equal, determining that the first situation is that a single point mutation exists includes:

[0021] For each outlier, if the outlier is greater than the third single-point mean, it is determined that the first situation includes the existence of a single-point mutation increase; if the outlier is less than the third single-point mean, it is determined that the first situation includes the existence of a single-point mutation decrease; if the outlier is less than the fourth single-point mean, it is determined that the first situation includes the existence of a single-point mutation increase; if the outlier is greater than the fourth single-point mean, it is determined that the first situation includes the existence of a single-point mutation decrease.

[0022] In a preferred example, the present application may be further configured as follows: determining whether there is a second situation of continuous mutation according to the first feature set corresponding to the adjacent sliding windows, including:

[0023] Calculate the linear correlation value of the first feature set of adjacent sliding windows;

[0024] When the linear correlation value is not less than the preset correlation value threshold, it is determined that the second situation is that there is no continuous mutation;

[0025] When the linear correlation value is less than the preset correlation value threshold, it is determined that the second situation is that there is a continuous mutation.

[0026] In a preferred example, the present application may be further configured as follows: when the linear correlation value is less than a preset correlation value threshold, determining that the second situation is a continuous mutation, includes:

[0027] When the linear correlation value is less than the preset correlation value threshold, the first target trend is determined according to the size relationship of the maximum values ​​of adjacent sliding windows, and the second situation is determined as the existence of a continuous mutation of the first target trend.

[0028] In a preferred example, the present application may be further configured as follows: the second feature set includes: a second window maximum value, a window maximum adjacent difference value, and a window minimum adjacent difference value;

[0029] Determining whether there is a third situation with significant difference in the same condition comparison according to the second feature sets of the multiple positions and the multiple data of the target sliding window includes:

[0030] For the target second feature corresponding to the target sliding window, determine the maximum target second feature according to the target second features corresponding to the plurality of positions, where the target second feature is any feature in the second feature set;

[0031] According to the remaining target second characteristics except the maximum target second characteristic, using the 3σ method, determine whether the maximum target second characteristic exceeds the range;

[0032] If so, it is determined that the third situation is that there is a difference in the comparison under the same conditions; the target second feature is any feature in the second feature set corresponding to each sliding window.

[0033] In a preferred example, the present application can be further configured as follows: If yes, determining that the third situation is a difference in the comparison under the same conditions, including:

[0034] If yes, determining the second target trend according to the second window maximum value, the second window minimum value and the second window mean value corresponding to each sliding window;

[0035] The third situation is determined as the same-condition comparison difference with the second target trend.

[0036] In a preferred example, the present application may be further configured as follows: the step of acquiring a trend data set of a target category includes:

[0037] Acquire an initial trend data set of a target category, where the trend data set of the target category includes a plurality of initial data within a preset time period;

[0038] For each initial data, when the initial data is not within the set data range, the current data value is processed by replacement;

[0039] The replacement method includes any one of mean replacement, median replacement, and regression replacement.

[0040] In a second aspect, a device for identifying abnormalities in trend data of urban rail vehicles is provided, comprising:

[0041] An acquisition module is used to acquire a trend data set of a target category, the trend data set of the target category includes a plurality of data within a preset time period, and the target category is any one of a plurality of categories monitored by an on-board system of an urban rail vehicle;

[0042] A first determination module is used to determine the difference between the temporally adjacent data in the trend data set to obtain a difference sequence; and determine whether there is a first situation of a single point mutation according to the difference sequence;

[0043] A second determination module is configured to perform feature calculation on multiple data in each sliding window of the trend data set to obtain a first feature set, wherein the first feature set includes at least one of a first window mean, a first window variance, a first window maximum value, and a first window minimum value; and determine whether there is a second situation of continuous mutation based on the first feature sets corresponding to adjacent sliding windows;

[0044] A third determination module is used to perform feature calculation on multiple data of a target sliding window of a trend data set corresponding to each of the multiple positions to obtain a second feature set corresponding to each of the multiple positions; and determine whether there is a third situation of significant difference in the same condition comparison according to the second feature sets of each of the multiple positions and the multiple data of the target sliding window of the trend data set corresponding to each of the multiple positions, where the target sliding window is any window of the multiple sliding windows of the trend data set;

[0045] A generation module is used to generate an abnormal result according to the first situation, the second situation, and the third situation.

[0046] In a third aspect, an electronic device is provided, including:

[0047] one or more processors;

[0048] Memory;

[0049] One or more applications, wherein one or more applications are stored in a memory and configured to be executed by one or more processors, and one or more programs are configured to: perform operations corresponding to the method shown in any possible implementation of the first aspect.

[0050] In a fourth aspect, a computer-readable storage medium is provided, wherein the storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, at least one program, a code set or an instruction set is loaded by a processor and executes the steps of the method shown in any possible implementation of the first aspect.

[0051] In a fifth aspect, a computer program product is provided, comprising a computer program, wherein when the computer program is executed by a processor, operations corresponding to the method shown in any possible implementation manner in the first aspect are implemented.

[0052] In summary, the method provided by this application includes the following beneficial technical effects:

[0053] A trend data set of a target category is obtained, which covers multiple data points within a preset time period, so that the changes in the data time series can be analyzed; after determining the difference between the temporally adjacent data in the trend data set and generating a difference sequence, the first situation of a single point mutation that has undergone a significant change at a certain moment can be quickly identified; feature calculation is performed on the data in each sliding window of the trend data set, and a first feature set including window mean, variance, maximum value, minimum value, etc. is obtained, which can reflect the overall trend and fluctuation of the data over a period of time. Then, by comparing the first feature sets corresponding to adjacent sliding windows, a second situation of continuous mutation in which the data continuously deviates from the normal fluctuation range over a period of time is identified, so that the long-term abnormal trend in the data can be captured more accurately, providing strong support for fault warning and early intervention; feature calculation is performed on the data of the target sliding windows of the trend data set corresponding to multiple locations, and a second feature set of multiple locations is obtained; by combining these feature sets and the original data, a third situation is evaluated whether the data in the same time window at different locations have significant differences compared under the same conditions. This enables data anomalies to be identified from a spatial dimension. For situations involving multiple monitoring points, it is possible to have a more comprehensive understanding of the operating status and timely discover and locate potential fault points. Combining the judgment results of the first, second, and third situations, it is possible to generate anomaly results that reflect the abnormal characteristics of the data in time and space. This application considers a variety of abnormal situations in trend data, summarizes and divides the abnormal categories, and the analysis method is simple and reliable.

[0054] In addition, the present application also provides an abnormality identification device, equipment and medium for urban rail vehicle trend data, all of which have the above-mentioned beneficial technical effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present application or the technical solutions of 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 some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 This is a flow chart of a method for identifying abnormalities in trend data of urban rail vehicles provided in an embodiment of the present application;

[0057] Figure 2 It is a schematic diagram of the definition of three identification types provided in an embodiment of the present application;

[0058] Figure 3 It is a flow chart of a trend identification method provided in an embodiment of the present application;

[0059] Figure 4 It is a structural schematic diagram of an abnormality identification device for urban rail vehicle trend data provided by an embodiment of the present application;

[0060] Figure 5 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0061] This specific embodiment is merely an explanation of the present application and is not a limitation of the present application. After reading this specification, a person skilled in the art may make non-creative modifications to the present embodiment as needed, but such modifications are protected by the patent law as long as they are within the scope of the present application.

[0062] It should be noted that in the optional embodiments of the present application, the object information and other related data involved, when the embodiments in the present application are applied to specific products or technologies, need to obtain the permission or consent of the object, and the collection, use and processing of the relevant data need to comply with the relevant laws, regulations and standards of the relevant countries and regions. In other words, if the embodiments of the present application involve data related to the object, it needs to be obtained with the authorization and consent of the object, the authorization and consent of the relevant departments, and in compliance with the relevant laws, regulations and standards of the country and region. If personal information is involved in the embodiments, the acquisition of all personal information needs to obtain the consent of the individual. If sensitive information is involved, the separate consent of the information subject needs to be obtained. The embodiments also need to be implemented with the authorization and consent of the object.

[0063] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0064] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally means that the associated objects before and after are in an "or" relationship.

[0065] The current trend data anomaly identification methods for the rail transit industry include:

[0066] Trend anomaly identification method based on statistical methods. This is a method that uses the probability part of trend data to determine whether it is abnormal. For example, the Grubbs hypothesis test method is used to obtain abnormal measurement point temperature samples that have undergone data histogram distribution statistics, the ARIMA method is used to capture abnormal vehicle speed information under a stable state, and the combination of month-on-month and year-on-year data and the heuristic segmentation and fusion method are used to identify abnormal discrete samples of axle temperature data of co-located related measurement points of locomotive multiple axles.

[0067] Trend anomaly identification method based on traditional machine learning. This method is based on traditional statistical methods and classic machine learning algorithms. It learns rules and patterns from data in order to perform anomaly detection related tasks. For example, it uses K-nearest neighbor algorithm, local anomaly factor (LOF), K-means clustering, density-based spatial clustering with noise (DBSCAN), isolation forest and other algorithm strategies to calculate sample space distances and build anomaly detection decisions based on the overall spatial distribution of samples.

[0068] A trend anomaly recognition method based on deep learning. This method captures the complex contextual feature associations in trend data by learning high-level representations of data, so as to better identify trend anomalies in the data. For example, the constructed convolutional neural network combined with long short-term memory network model (CNN-LSTM) is used to predict the axle temperature range of the vehicle under different operating scenarios using quantile regression combined with gated recurrent unit (QR-GRU) method, so as to achieve efficient recognition of trend temperature anomalies. Li proposed a combination of axle temperature prediction model based on particle swarm optimization support vector machine (PSO-SVM) and a bidirectional long short-term memory network axle temperature prediction model (Bi-LSTM). The Kalman filter algorithm is used to update the time and observation quantity, so as to avoid the error accumulation in the long time series data prediction process and improve the accuracy of axle temperature trend data anomaly recognition.

[0069] Statistical methods require data to meet specific assumptions, such as normal distribution, linearity, and stationarity. Actual shaft temperature trend data often does not meet these assumptions and needs to be processed and converted in advance, resulting in information loss. In addition, when processing a large amount of redundant data, statistical methods will face performance bottlenecks and cannot be easily expanded.

[0070] Compared with statistics, traditional machine learning methods have significantly improved performance and can reduce computing costs to a certain extent. However, the application effect of this type of method depends entirely on the quality of training data and is easily interfered by noise data. Therefore, it is necessary to set parameters and thresholds in advance. If the parameters and thresholds are set differently, the detection results will be significantly different, resulting in unstable model performance.

[0071] In most cases, deep learning-based methods can well capture the intrinsic connection of trend data, and can also model multivariate / high-dimensional data features to meet the application scenarios responsible for trend abnormal conditions. However, due to the complexity of deep learning models, high requirements for hardware equipment conditions, and a long training process, they do not meet the low-latency detection requirements of on-board real-time computing in trains. When the data scale is not large, the model may be overfitted and the performance may be unstable.

[0072] Traditional trend anomaly identification methods have the following defects in the rail transit industry:

[0073] Trend anomaly identification methods often require manual analysis experience, which is time-consuming and labor-intensive, and prone to subjective bias. It is impossible to accurately quantify the current trend anomaly and then apply it to other trend data anomaly analysis;

[0074] Anomaly identification is often based on some simple, single statistical methods, lacking some intelligent means to obtain the correlation and difference characteristics between normal trend data and abnormal trend data, which limits the efficiency and accuracy of anomaly identification;

[0075] Lack of analysis and mining of historical trend data: Traditional trend anomaly identification methods can often only identify anomalies in short-term stable data, and cannot use long-term non-stationary historical data for analysis and mining, making it difficult to discover some potential anomalies.

[0076] In order to solve the abnormal problem of urban rail vehicle monitoring data, a trend anomaly identification method is proposed, which aims to identify different abnormal situations of various trend data. By statistically analyzing the various types of trend data monitored, the embodiment of the present application provides an abnormal identification method for urban rail vehicle trend data, such as Figure 1 As shown, the method provided in the embodiment of the present application can be executed by an electronic device, and the electronic device is a server. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. The terminal device can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc., but is not limited to this. The terminal device and the electronic device can be directly or indirectly connected via wired or wireless communication, which is not limited in the embodiment of the present application. The method includes:

[0077] S101, obtaining a trend data set of a target category, where the trend data set of the target category includes a plurality of data within a preset period, and the target category is any one of a plurality of categories monitored by an onboard system of an urban rail vehicle;

[0078] In the embodiment of the present application, the multi-type trend data monitored by the on-board system of the urban rail vehicle may be the multi-type trend data of the vehicle running gear bearings, gears and other components, or the multi-type trend data of multiple components of the transmission part, which is not limited in the embodiment of the present application. For the running gear, the multi-type trend data includes but is not limited to rotation speed, temperature, vibration, and impact, and the electronic device can directly obtain the trend data of a certain period of time.

[0079] In the embodiment of the present application, abnormal data identification can be performed in combination with time-continuous data, and the continuity of data can be fully considered. The preset time period can be within one month or two months, and the specific time period can be set by the user according to actual needs.

[0080] S102, determining the difference between the time-adjacent data in the trend data set to obtain a difference sequence; and determining whether there is a first situation of a single point mutation according to the difference sequence;

[0081] Conventionally, the data distribution is determined directly based on the original data to determine the single point mutation. However, in the embodiment of the present application, the change trend or difference between data points is used to determine the single point mutation, which is particularly effective for time series data and can capture the change trend of data points. It can be applied to the scenario in the present application where sudden changes or abnormal fluctuations between data points need to be detected, so as to identify isolated points or mutation points that do not conform to the overall trend.

[0082] Among them, the original trend data is differentiated and the difference between adjacent data is calculated , and obtain the difference sequence ;

[0083] In one achievable method of determining whether a single point mutation exists, the difference sequence is traversed and each difference is compared with a preset difference threshold; if the difference is greater than an upper threshold or less than a lower threshold, it is determined that a first case of a single point mutation exists.

[0084] In another achievable method of determining whether there is a single point mutation, the first case of determining whether there is a single point mutation according to the difference sequence includes: Sa1-Sa4, wherein:

[0085] Sa1. Determine whether there are outliers in the difference sequence;

[0086] Sa2. If there is no outlier, it is determined that the first situation is that there is no single point mutation;

[0087] Sa3. If there are outliers, and the number of outliers is 1, then when the outlier and the first single-point mean are not equal, and the outlier and the second single-point mean are not equal, it is determined that the first case is that there is a single-point mutation; otherwise, it is determined that the first case is that there is no single-point mutation, wherein the first single-point mean is determined by the mean of all differences before the outlier in the difference sequence and the recognition weight, and the second single-point mean is determined by the mean of all differences after the outlier in the difference sequence and the recognition weight;

[0088] Among them, the single point mutation recognition weight value is set to be the recognition weight (The weight range can be set between 1 and 3, and the current default is 1.0). The first single point mean is , the second single point mean is .

[0089] In one possible case, when the outlier value is not equal to the first single point mean, and the outlier value is not equal to the second single point mean, determining that the first case is that a single point mutation exists includes:

[0090] When the outlier value is greater than the first single point mean, it is determined that the first situation is a single point mutation increase; when the outlier value is less than the first single point mean, it is determined that the first situation is a single point mutation decrease;

[0091] When the outlier value is smaller than the second single point mean, the first situation is determined to be a single point mutation increase; when the outlier value is larger than the second single point mean, the first situation is determined to be a single point mutation decrease;

[0092] Among them, if the outlier value and the single point mean are equal, it means that there is no significant mutation between the outliers, and the whole is a relatively uniform set. At this time, it is determined that there is no single point mutation.

[0093] Specifically, when outliers When the number is 0, there is no single point mutation phenomenon;

[0094] When outliers When the number of is 1, the outlier value is :

[0095] If the current , it is considered that the trend data in this time period has a single point mutation increase;

[0096] If the current , it is considered that there is no single point mutation phenomenon;

[0097] If the current , it is considered that the trend data in this time period has a single-point mutation decline;

[0098] like , it is considered that the trend data in this time period has a single point mutation increase;

[0099] like , it is considered that there is no single point mutation phenomenon;

[0100] like , it is considered that there is a single-point mutation decline in the trend data within this time period.

[0101] Sa4. If there are outliers and the number of outliers is greater than 1, then when there are no continuous outliers, and, for each outlier, the outlier and the third single-point mean are not equal, and the outlier and the fourth single-point mean are not equal, it is determined that the first case is that there is a single-point mutation, otherwise, it is determined that the first case is that there is no single-point mutation, wherein the third single-point mean is determined by the mean of all differences between the outlier and the previous outlier in the difference sequence and the recognition weight, and the fourth single-point mean is determined by the mean of all differences between the outlier and the next outlier in the difference sequence and the recognition weight.

[0102] For each outlier For , the third single point mean is: ;in, is the previous difference of the current outlier value, is the previous abnormal value adjacent to the current abnormal value, is the next difference between the previous abnormal value and the current abnormal value. The fourth single point mean is: ;in, is the next difference of the current outlier value, The next abnormal value adjacent to the current abnormal value.

[0103] In one possible case, for each outlier, when the outlier and the third single point mean are not equal, and the outlier and the fourth single point mean are not equal, determining that the first case is that a single point mutation exists includes:

[0104] For each outlier, if the outlier is greater than the third single-point mean, it is determined that the first situation includes the existence of a single-point mutation increase; if the outlier is less than the third single-point mean, it is determined that the first situation includes the existence of a single-point mutation decrease; if the outlier is less than the fourth single-point mean, it is determined that the first situation includes the existence of a single-point mutation increase; if the outlier is greater than the fourth single-point mean, it is determined that the first situation includes the existence of a single-point mutation decrease.

[0105] Specifically, when outliers When the number is greater than 1, the difference sequence data There are k (k>1) outliers ,in . Traverse the entire difference sequence data and query whether it exists in the sequence (Indicates that the abnormal values ​​are continuous). If they are continuous, it is considered that there is no single-point mutation phenomenon;

[0106] If it is not continuous, the difference trend data is segmented according to the outlier value as the dividing point, and each outlier value For example, if the current abnormal value is greater than the third single-point mean, it is considered that the trend data in this time period has a single-point mutation increase;

[0107] If the current outlier value is equal to the third single point mean, it is considered that there is no single point mutation phenomenon;

[0108] If the current outlier value is smaller than the third single-point mean, it is considered that there is a single-point mutation decline in the trend data within this time period.

[0109] If the current outlier value is less than the fourth single-point mean, it is considered that the trend data within this time period has a single-point mutation increase; if the current outlier value is equal to the fourth single-point mean, it is considered that there is no single-point mutation phenomenon; if the current outlier value is greater than the fourth single-point mean, it is considered that the trend data within this time period has a single-point mutation decrease.

[0110] S103, performing feature calculation on multiple data in each sliding window of the trend data set to obtain a first feature set, where the first feature set includes: at least one of a first window mean, a first window variance, a first window maximum value, and a first window minimum value; and determining whether there is a second situation of continuous mutation based on the first feature sets corresponding to adjacent sliding windows;

[0111] In the embodiment of the present application, the size of the sliding window can be set by the user according to actual needs. The mean, variance, maximum value, and minimum value are calculated based on multiple data in each sliding window, and a feature vector w, i.e., a first feature set, is formed based on at least one of these calculation results. Preferably, the first feature set includes: a first window mean, a first window variance, a first window maximum value, and a first window minimum value.

[0112] Furthermore, determining whether there is a second situation of continuous mutation according to the first feature set corresponding to the adjacent sliding windows includes:

[0113] Calculate the linear correlation value of the first feature set of adjacent sliding windows;

[0114] When the linear correlation value is not less than the preset correlation value threshold, it is determined that the second situation is that there is no continuous mutation;

[0115] When the linear correlation value is less than the preset correlation value threshold, it is determined that the second situation is that there is a continuous mutation.

[0116] The method of calculating the linear correlation value is not limited in the present embodiment, and the Pearson correlation method can be used. For the preset correlation value threshold, the user can set it to a fixed value according to actual needs, such as 0.2, or can determine a dynamic threshold according to actual conditions.

[0117] Furthermore, in order to facilitate subsequent maintenance by technical personnel, it is necessary not only to determine whether there is a continuous mutation, but also to determine the trend of the continuous mutation. Then, when the linear correlation value is less than the preset correlation value threshold, the second situation is determined to be the existence of a continuous mutation, including:

[0118] When the linear correlation value is less than the preset correlation value threshold, the first target trend is determined according to the size relationship of the maximum values ​​of adjacent sliding windows, and the second situation is determined as the existence of a continuous mutation of the first target trend.

[0119] Among them, when the maximum value of adjacent sliding windows and , and exists , then it is considered that the first target trend within this time period is rising, otherwise, the first target trend is falling.

[0120] S104, performing feature calculation on multiple data of the target sliding window of the trend data set corresponding to each of the multiple positions to obtain a second feature set corresponding to each of the multiple positions; determining whether there is a third situation in which there is a significant difference when compared under the same conditions according to the second feature sets of each of the multiple positions and the multiple data of the target sliding window of the trend data set corresponding to each of the multiple positions, where the target sliding window is any window of the multiple sliding windows of the trend data set;

[0121] In the embodiment of the present application, the second feature set includes but is not limited to: the second window maximum value, the window maximum adjacent difference value, and the window minimum adjacent difference value.

[0122] In the embodiment of the present application, for a target category, there may be multiple sensors set at different positions to collect trend data corresponding to multiple positions. Therefore, there may be at least one trend data set corresponding to the target category. Therefore, under the same conditions, trend data sets corresponding to multiple positions for the target category are obtained. Then, the trend data set at each position is grouped according to the sliding window, and feature calculation is performed after grouping to obtain a second feature set for each sliding window corresponding to each of the multiple positions.

[0123] Exemplarily, when the target category is temperature, trend data set 1 is collected at position 1, and trend data set 2 is collected at position 2; then, the data are grouped according to the sliding window, and feature calculation is performed based on the grouped data to obtain the second feature set 11 for the first sliding window, the second feature set 12 for the first sliding window, and so on at position 1, and the second feature set 21 for the first sliding window, the second feature set 22 for the first sliding window, and so on at position 2.

[0124] Specifically, feature calculation is performed on multiple data of the target sliding window of the trend data set corresponding to each of the multiple positions to obtain a second feature set corresponding to each of the multiple positions, including:

[0125] For the data of the target sliding window corresponding to a certain position, calculate the second window maximum value of the data in the target sliding window and the difference between the adjacent data;

[0126] From the differences of adjacent data, determine the maximum adjacent difference of the window and the minimum adjacent difference of the window;

[0127] A second feature set corresponding to a certain position for a target sliding window is obtained, where the second feature set includes: a second window maximum value, a window maximum adjacent difference value, and a window minimum adjacent difference value corresponding to the target sliding window.

[0128] Specifically, calculate the maximum value of the trend data at different positions under the same conditions in the target sliding window and adjacent differences , Indicates the mark value of different positions, and the adjacent difference sequence is expressed as , get the maximum value of the adjacent difference sequence , minimum .

[0129] Accordingly, determining whether there is a third situation in which there is a significant difference in the same condition comparison based on the second feature sets of the multiple locations and the multiple data of the target sliding window of the trend data set corresponding to the multiple locations includes:

[0130] For the target second feature corresponding to the target sliding window, determine the maximum target second feature according to the target second features corresponding to the multiple positions, where the target second feature is any feature in the second feature set;

[0131] According to the remaining target second characteristics except the maximum target second characteristic, using the 3σ method, determine whether the maximum target second characteristic exceeds the range;

[0132] If so, it is determined that the third situation is that there is a difference in the same condition comparison; the target second feature is any feature in the second feature set corresponding to each sliding window. Otherwise, it is determined that the third situation is that there is no difference in the same condition comparison.

[0133] If it does not exceed the range, it is determined that the third situation is that there is no difference in the comparison under the same conditions.

[0134] For the target sliding window, the sequence To judge abnormality, For example, calculate the maximum value in the sequence , using the remaining values ​​to form a set of data ,according to Rule, judgment Is it out of range? If so, it is considered The trend data of this position have significant differences in the same position, otherwise it is considered that there is no significant difference in the same position. Make a judgment.

[0135] Furthermore, in order to facilitate subsequent maintenance by technical personnel, it is necessary not only to determine whether there is a continuous mutation, but also to determine the trend of the continuous mutation. If so, the third situation is determined to be a difference in the comparison under the same conditions, including:

[0136] If yes, determining the second target trend according to the second window maximum value, the second window minimum value and the second window mean value corresponding to each sliding window;

[0137] The third situation is determined as the same-condition comparison difference with the second target trend.

[0138] Specifically, if the maximum value of the second window = the minimum value of the second window = the mean value of the second window, then it is considered that the trend data has no trend when compared under the same conditions, that is, the third situation is determined to be a difference compared under the same conditions with a trend; if they are not the same, then the third situation is determined to be a difference compared under the same conditions without a trend.

[0139] S105. Generate an abnormal result according to the first situation, the second situation, and the third situation.

[0140] In the embodiment of the present application, various abnormal situations in the trend data are considered, the abnormal categories are summarized and divided, and the trend data is judged abnormally using the corresponding abnormal identification method for the listed abnormal situations to improve the reliability of data analysis. It is understandable that there may be one or more abnormal situations for the trend data.

[0141] It can be seen that in the embodiment of the present application, a trend data set of a target category is obtained, and the data set covers multiple data points within a preset time period, so that the changes in the data time series can be analyzed; after determining the difference between the temporally adjacent data in the trend data set and generating a difference sequence, the first situation of a single point mutation that has undergone a significant change at a certain moment can be quickly identified; feature calculation is performed on the data in each sliding window of the trend data set, and a first feature set including window mean, variance, maximum value, minimum value, etc. is obtained, which can reflect the overall trend and fluctuation of the data over a period of time. Then, by comparing the first feature sets corresponding to adjacent sliding windows, the second situation of continuous mutations in which the data continuously deviates from the normal fluctuation range over a period of time is identified, so that the long-term abnormal trend in the data can be captured more accurately, providing strong support for fault warning and early intervention; feature calculation is performed on the data of the target sliding windows of the trend data set corresponding to each of the multiple positions, and the second feature sets of the multiple positions are obtained; by combining these feature sets and the original data, the third situation of evaluating whether the data in the same time window at different positions have significant differences compared under the same conditions is evaluated. This enables data anomalies to be identified from a spatial dimension. For situations involving multiple monitoring points, it is possible to have a more comprehensive understanding of the operating status and timely discover and locate potential fault points. Combining the judgment results of the first, second, and third situations, it is possible to generate anomaly results that reflect the abnormal characteristics of the data in time and space. This application considers a variety of abnormal situations in trend data, summarizes and divides the abnormal categories, and the analysis method is simple and reliable.

[0142] It is understandable that abnormal results can also provide a basis for subsequent troubleshooting, maintenance decisions, etc., improve the stability and reliability of the system, and reduce the risk of failure and maintenance costs.

[0143] Furthermore, the validity of the trend data can also be judged to provide a fourth abnormal situation. Specifically, a trend data set of a target category is obtained, including: obtaining an initial trend data set of the target category, the trend data set of the target category includes multiple initial data within a preset time period; for each initial data, when the initial data is not within the set data range, the current data value is processed by replacement; the replacement method includes any one of mean replacement, median replacement, and regression replacement.

[0144] In some feasible methods, based on the result of whether the initial data is within the set data range, it is determined whether there is a fourth situation of sensor out of limit; if so (the initial data is not within the set data range), the current data value is processed by replacement, and subsequent S102-S104 are executed. Accordingly, S105 specifically includes: generating abnormal results according to the first situation, the second situation, the third situation and the fourth situation.

[0145] In the embodiment of the present application, the situation of sensor over-limit is also taken into consideration, wherein the data range is set to be beyond the range of the sensor acquisition range. By replacing the data in the early stage, the cross-correlation between various types of anomalies is reduced.

[0146] Furthermore, the impact of differences in vehicle operating conditions on different urban routes on trend changes in trend data (such as temperature, speed, and mileage) can also be considered, and the trend abnormal mutation threshold can be reasonably corrected, which has better adaptability in practical applications.

[0147] In one possible scenario, when the working condition of the city is good, 3σ can be reduced by a preset value, which is user-defined; when the working condition of the city is poor, 3σ can be increased by a preset value to increase the abnormal fluctuation range. In one possible scenario, when the working condition of the city is good, the preset correlation value threshold can be increased, otherwise, the preset correlation value threshold can be reduced, and the specific adjustment size can be user-defined. For the good or bad working conditions, users can set it according to the actual situation.

[0148] Based on the above embodiments, the embodiments of the present application provide a specific solution to the abnormal problem of urban rail vehicle monitoring data and propose a trend anomaly identification method, which aims to identify different abnormal conditions of each trend data. Through statistical analysis of the various types of trend data monitored, the current trend data abnormal conditions can be divided into four categories, namely single point mutation, continuous mutation, sensor over-limit and significant difference in isotopic comparison. Among them, single point mutation, continuous mutation and significant difference in isotopic comparison are defined as follows: Figure 2 shown.

[0149] Specifically, the trend identification method process is as follows: Figure 3 As shown, including:

[0150] Step 1: Get trend data within a certain period of time;

[0151] The trend data currently monitored by the on-board system of urban rail vehicles include multiple types of trend data such as rotation speed, temperature, vibration, and impact, and the original trend data of a certain period of time can be directly obtained.

[0152] Step 2: Data validity judgment and processing;

[0153] Data validity judgment and processing. Set the data range threshold. If the current data value is not within the threshold range, the current data value is processed by replacement, including mean replacement, median replacement, regression replacement, etc. Otherwise, go to step three.

[0154] Step 3: Identify single point mutation anomalies in trend data;

[0155] The single point mutation anomaly in the trend data is identified by the following method:

[0156] ①Differentiate the original trend data and calculate the difference between adjacent data , and obtain the difference sequence , set the single point mutation recognition weight value (The weight range can be set between 1 and 3, and the current default is 1.0);

[0157] ②Use Hampel filter to extract the difference sequence data Get outliers in ;

[0158] ③ When the outlier When the number is 0, there is no single point mutation phenomenon;

[0159] ④ When the outlier When the number is 1,

[0160] If the current , it is considered that the trend data in this time period has a single point mutation increase; if the current , it is considered that there is no single point mutation phenomenon; if the current , it is considered that the trend data in this time period has a single-point mutation decline; if , it is considered that the trend data in this time period has a single-point mutation increase; if , it is considered that there is no single point mutation phenomenon; if , it is considered that there is a single-point mutation decline in the trend data within this time period.

[0161] ⑤ When the outlier When the number is greater than 1, the difference sequence data There are k (k>1) outliers ,in . Traverse the entire difference sequence data and query whether it exists in the sequence (continuous outliers) If they are continuous, it is considered that there is no single-point mutation phenomenon;

[0162] If it is not continuous, the difference trend data is segmented according to the outlier value as the dividing point, and each outlier value For example, if the current , it is considered that the trend data in this time period has a single point mutation increase; if the current , it is considered that there is no single point mutation phenomenon; if the current , it is considered that there is a single-point mutation decline in the trend data within this time period.

[0163] like , it is considered that the trend data in this time period has a single point mutation increase; if the current , it is considered that there is no single point mutation phenomenon; if the current , it is considered that there is a single-point mutation decline in the trend data within this time period.

[0164] Step 4: Identify the abnormal continuous mutations in the trend data;

[0165] Identify the abnormal situation of continuous mutation in trend data. The identification method is as follows:

[0166] ① Set the sliding window size, use the sliding window to divide the trend data, calculate the corresponding mean, variance, maximum value, and minimum value of the time series data in each window, and form a feature vector from these four statistical features ;

[0167] ② Calculate the feature vectors of adjacent sliding windows through Pearson correlation analysis , Is there a stable linear relationship? If the Pearson correlation coefficient When , the linear correlation is weak, then there are abnormal sample points;

[0168] ③ When there are outliers, if the maximum value of the adjacent sliding windows and ,exist , it is considered that the trend data within this time period has a continuous mutation increase, otherwise it is a continuous mutation decrease.

[0169] Step 5: Perform a peer comparison of the trend data to determine if the differences are significant.

[0170] The significant difference of the same position comparison in the trend data is judged as follows:

[0171] ① Set the sliding window size and use the sliding window to divide the trend data;

[0172] ② Calculate the maximum value of the trend data at different positions under the same conditions in the current window and adjacent differences Indicates the mark value of different positions, and the adjacent difference sequence is expressed as , get the maximum value of the adjacent difference sequence , minimum ;

[0173] ② Sequence To judge abnormality, For example, calculate the maximum value in the sequence , using the remaining values ​​to form a set of data ,according to Rule, judgment Is it out of range? If so, it is considered The trend data of this position have significant differences in the same position, otherwise it is considered that there is no significant difference in the same position. Make a judgment.

[0174] ③ If the trend data in the current sliding window has significant differences in the same position comparison, calculate the minimum, average and maximum values ​​of the time series data in the current sliding window. If the minimum value = average value = maximum value, it is considered that the trend data has no trend in the same position comparison.

[0175] Normally, we can assume that a rail transit vehicle can be composed of 6 or 8 carriages, each carriage has a fixed number and contains 2 bogies, each bogie has two axles, and a carriage contains 4 axles. According to the driving direction of the vehicle, the axles in a carriage are numbered from front to back, namely axle 1, axle 2, axle 3, and axle 4. There will be multiple measuring point positions on each axle. Therefore, in this embodiment, "comparison of trend data in the same position" is also "determining whether there is a third situation with significant differences in the same condition comparison", where "comparison under the same condition" is also "comparison of the same position", which means: the data obtained from the same measuring point position of different axles of any carriage of the rail transit vehicle are compared.

[0176] It can be seen that the trend anomaly identification method proposed in the application embodiment covers abnormal operating conditions at the root of three types of trend anomaly problems, and realizes the identification of abnormal point mutations, trend mutations and co-location comparison anomalies through codes. Taking into account the possible interference between different abnormal operating conditions, invalid data is replaced. The cross-correlation between various abnormal operating conditions is reduced. At the same time, considering the impact of vehicle operating conditions differences on temperature, speed and mileage trend changes on different urban lines, the trend abnormal mutation threshold is reasonably corrected, which has good adaptability in practical applications.

[0177] The following is an introduction to a device provided in an embodiment of the present application. The device described below and the method described above can be referred to each other. The device 200 of this embodiment is set in an electronic device. Figure 5 , Figure 5 It is a structural block diagram of a device of one embodiment of the present application, comprising:

[0178] An acquisition module 210 is used to acquire a trend data set of a target category, where the trend data set of the target category includes a plurality of data within a preset period, and the target category is any one of a plurality of categories monitored by the on-board system of an urban rail vehicle;

[0179] The first determination module 220 is used to determine the difference between the time-adjacent data in the trend data set to obtain a difference sequence; and determine whether there is a first situation of a single point mutation according to the difference sequence;

[0180] The second determination module 230 is used to perform feature calculation on multiple data in each sliding window of the trend data set to obtain a first feature set, where the first feature set includes at least one of a first window mean, a first window variance, a first window maximum value, and a first window minimum value; and determine whether there is a second situation of continuous mutation based on the first feature sets corresponding to adjacent sliding windows;

[0181] The third determination module 240 is used to perform feature calculation on multiple data of the target sliding window of the trend data set corresponding to each of the multiple positions to obtain a second feature set corresponding to each of the multiple positions; determine whether there is a third situation of significant difference in the same condition comparison according to the second feature sets of each of the multiple positions and the multiple data of the target sliding window of the trend data set corresponding to each of the multiple positions, where the target sliding window is any window of the multiple sliding windows of the trend data set;

[0182] The generating module 250 is used to generate an abnormal result according to the first situation, the second situation and the third situation.

[0183] In an implementable manner, the first determining module 220 is further configured to:

[0184] The first case of determining whether there is a single point mutation based on the difference sequence includes:

[0185] Determine if there are outliers in the difference series;

[0186] If there is no outlier, it is determined that the first case is that there is no single point mutation;

[0187] If there is an outlier, and the number of outliers is 1, then when the outlier and the first single point mean are not equal, and the outlier and the second single point mean are not equal, it is determined that the first case is that there is a single point mutation; otherwise, it is determined that the first case is that there is no single point mutation, wherein the first single point mean is determined by the mean of all differences before the outlier in the difference sequence and the recognition weight, and the second single point mean is determined by the mean of all differences after the outlier in the difference sequence and the recognition weight;

[0188] If there are outliers and the number of outliers is greater than 1, then when there are no consecutive outliers, and, for each outlier, the outlier and the third single-point mean are not equal, and the outlier and the fourth single-point mean are not equal, it is determined that the first case is that there is a single-point mutation, otherwise, it is determined that the first case is that there is no single-point mutation, wherein the third single-point mean is the mean of all differences between the outlier and the previous outlier in the difference sequence, and the fourth single-point mean is the mean of all differences between the outlier and the next outlier in the difference sequence.

[0189] In an implementable manner, the first determining module 220 is further configured to:

[0190] When the outlier value is not equal to the first single point mean, and the outlier value is not equal to the second single point mean, determining that the first situation is that a single point mutation exists includes:

[0191] When the outlier value is greater than the first single point mean, it is determined that the first situation is a single point mutation increase; when the outlier value is less than the first single point mean, it is determined that the first situation is a single point mutation decrease;

[0192] When the outlier value is smaller than the second single point mean, the first situation is determined to be a single point mutation increase; when the outlier value is larger than the second single point mean, the first situation is determined to be a single point mutation decrease;

[0193] Alternatively, for each outlier, when the outlier and the third single point mean are not equal, and the outlier and the fourth single point mean are not equal, determining that the first situation is that a single point mutation exists includes:

[0194] For each outlier, if the outlier is greater than the third single-point mean, it is determined that the first situation includes the existence of a single-point mutation increase; if the outlier is less than the third single-point mean, it is determined that the first situation includes the existence of a single-point mutation decrease; if the outlier is less than the fourth single-point mean, it is determined that the first situation includes the existence of a single-point mutation increase; if the outlier is greater than the fourth single-point mean, it is determined that the first situation includes the existence of a single-point mutation decrease.

[0195] In an implementable manner, the second determining module 230 is further configured to:

[0196] When the linear correlation value is less than the preset correlation value threshold, the first target trend is determined according to the size relationship of the maximum values ​​of adjacent sliding windows, and the second situation is determined as the existence of a continuous mutation of the first target trend.

[0197] In an implementable manner, the second feature set includes: a second window maximum value, a window maximum adjacent difference value, and a window minimum adjacent difference value;

[0198] The third determining module 240 is further configured to:

[0199] For the target second feature corresponding to the target sliding window, determine the maximum target second feature according to the target second features corresponding to the multiple positions, where the target second feature is any feature in the second feature set;

[0200] According to the remaining target second characteristics except the maximum target second characteristic, using the 3σ method, determine whether the maximum target second characteristic exceeds the range;

[0201] If so, it is determined that the third situation is that there is a difference in the same condition comparison; the target second feature is any feature in the second feature set corresponding to each sliding window.

[0202] In an implementable manner, the third determining module 240 is further configured to:

[0203] If yes, determining the second target trend according to the second window maximum value, the second window minimum value and the second window mean value corresponding to each sliding window;

[0204] The third situation is determined as the same-condition comparison difference with the second target trend.

[0205] In one achievable manner, the module is obtained, specifically for:

[0206] Acquire an initial trend data set of a target category, where the trend data set of the target category includes a plurality of initial data within a preset time period;

[0207] For each initial data, when the initial data is not within the set data range, the current data value is processed by replacement;

[0208] The replacement method includes any one of mean replacement, median replacement, and regression replacement.

[0209] An electronic device is provided in an embodiment of the present application, such as Figure 5 As shown, Figure 5 The electronic device 300 shown includes: a processor 301 and a memory 303. The processor 301 and the memory 303 are connected, such as through a bus 302. Optionally, the electronic device 300 may also include a transceiver 304. It should be noted that in actual applications, the transceiver 304 is not limited to one, and the structure of the electronic device 300 does not constitute a limitation on the embodiments of the present application.

[0210] The processor 301 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. The processor 301 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0211] The bus 302 may include a path to transmit information between the above components. The bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 302 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0212] The memory 303 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0213] The memory 303 is used to store the application code for executing the solution of the present application, and the execution is controlled by the processor 301. The processor 301 is used to execute the application code stored in the memory 303 to implement the contents shown in the above method embodiment.

[0214] Figure 5 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0215] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer can execute the corresponding content in the aforementioned method embodiment.

[0216] An embodiment of the present application provides a computer program product, including a computer program, which implements the corresponding contents of the aforementioned method embodiment when the computer program is executed by a processor.

[0217] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0218] The above are only some implementation methods of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for identifying anomalies in trend data of urban rail vehicles, characterized in that: include: Acquire a trend data set of a target category, the trend data set of the target category including a plurality of data within a preset time period, wherein the target category is any one of a plurality of categories monitored by an onboard system of an urban rail vehicle; Determine the difference between the temporally adjacent data in the trend data set to obtain a difference sequence; and determine whether there is a first situation of a single point mutation according to the difference sequence; Performing feature calculation on multiple data in each sliding window of the trend data set to obtain a first feature set, wherein the first feature set includes: at least one of a first window mean, a first window variance, a first window maximum value, and a first window minimum value; and determining whether there is a second situation of continuous mutation according to the first feature sets corresponding to adjacent sliding windows; Perform feature calculation on multiple data of the target sliding window of the trend data set corresponding to each of the multiple positions to obtain a second feature set corresponding to each of the multiple positions; determine whether there is a third situation in which there is a significant difference when compared under the same conditions according to the second feature sets of each of the multiple positions and the multiple data of the target sliding window of the trend data set corresponding to each of the multiple positions, where the target sliding window is any window of the multiple sliding windows of the trend data set; An abnormal result is generated according to the first situation, the second situation, and the third situation.

2. The method according to claim 1, characterized in that: The first case of determining whether there is a single point mutation according to the difference sequence includes: Determining whether there are outliers in the difference sequence; If there is no outlier, it is determined that the first case is that there is no single point mutation; If there is an outlier, and the number of outliers is 1, then when the outlier and the first single-point mean are not equal, and the outlier and the second single-point mean are not equal, it is determined that the first situation is that there is a single-point mutation; otherwise, it is determined that the first situation is that there is no single-point mutation, wherein the first single-point mean is determined by the mean of all differences before the outlier in the difference sequence and the identification weight, and the second single-point mean is determined by the mean of all differences after the outlier in the difference sequence and the identification weight; If there are outliers and the number of outliers is greater than 1, then when there are no consecutive outliers, and, for each outlier, the outlier and the third single-point mean are not equal, and the outlier and the fourth single-point mean are not equal, the first situation is determined to be the presence of a single-point mutation, otherwise, the first situation is determined to be the absence of a single-point mutation, wherein the third single-point mean is the mean of all differences between the outlier and the previous outlier in the difference sequence, and the fourth single-point mean is the mean of all differences between the outlier and the next outlier in the difference sequence.

3. The method according to claim 2, characterized in that When the outlier value is not equal to the first single point mean, and the outlier value is not equal to the second single point mean, determining that the first situation is that a single point mutation exists includes: When the outlier value is greater than the first single point mean, it is determined that the first situation is a single point mutation increase; when the outlier value is less than the first single point mean, it is determined that the first situation is a single point mutation decrease; When the outlier value is smaller than the second single point mean, the first situation is determined to be a single point mutation increase; when the outlier value is larger than the second single point mean, the first situation is determined to be a single point mutation decrease; Alternatively, for each outlier, when the outlier and the third single point mean are not equal, and the outlier and the fourth single point mean are not equal, determining that the first situation is that a single point mutation exists includes: For each outlier, if the outlier is greater than the third single-point mean, it is determined that the first situation includes the existence of a single-point mutation increase; if the outlier is less than the third single-point mean, it is determined that the first situation includes the existence of a single-point mutation decrease; if the outlier is less than the fourth single-point mean, it is determined that the first situation includes the existence of a single-point mutation increase; if the outlier is greater than the fourth single-point mean, it is determined that the first situation includes the existence of a single-point mutation decrease.

4. The method according to claim 1, characterized in that Determining whether there is a second situation of continuous mutation according to the first feature set corresponding to the adjacent sliding windows includes: Calculate the linear correlation value of the first feature set of adjacent sliding windows; When the linear correlation value is not less than the preset correlation value threshold, it is determined that the second situation is that there is no continuous mutation; When the linear correlation value is less than the preset correlation value threshold, it is determined that the second situation is that there is a continuous mutation.

5. The method according to claim 4, characterized in that When the linear correlation value is less than the preset correlation value threshold, the second situation is determined as the existence of a continuous mutation, including: When the linear correlation value is less than the preset correlation value threshold, the first target trend is determined according to the size relationship of the maximum values ​​of adjacent sliding windows, and the second situation is determined as the existence of a continuous mutation of the first target trend.

6. The method according to claim 1, characterized in that The second feature set includes: a second window maximum value, a window maximum adjacent difference value, and a window minimum adjacent difference value; Determining whether there is a third situation with significant difference in the same condition comparison according to the second feature sets of the multiple positions and the multiple data of the target sliding window includes: For the target second feature corresponding to the target sliding window, determine the maximum target second feature according to the target second features corresponding to the plurality of positions, where the target second feature is any feature in the second feature set; According to the remaining target second characteristics except the maximum target second characteristic, using the 3σ method, determine whether the maximum target second characteristic exceeds the range; If so, it is determined that the third situation is that there is a difference in the comparison under the same conditions; the target second feature is any feature in the second feature set corresponding to each sliding window.

7. The method according to claim 6, characterized in that If so, the third situation is determined to be a difference in the comparison under the same conditions, including: If yes, determining the second target trend according to the second window maximum value, the second window minimum value and the second window mean value corresponding to each sliding window; The third situation is determined as the same-condition comparison difference with the second target trend.

8. The method according to any one of claims 1 to 7, characterized in that: The step of obtaining a trend data set of a target category includes: Acquire an initial trend data set of a target category, where the trend data set of the target category includes a plurality of initial data within a preset time period; For each initial data, when the initial data is not within the set data range, the current data value is processed by replacement; The replacement method includes any one of mean replacement, median replacement, and regression replacement.

9. An abnormality identification device for trend data of urban rail vehicles, characterized in that: include: An acquisition module is used to acquire a trend data set of a target category, the trend data set of the target category includes a plurality of data within a preset time period, and the target category is any one of a plurality of categories monitored by an on-board system of an urban rail vehicle; A first determination module is used to determine the difference between the temporally adjacent data in the trend data set to obtain a difference sequence; and determine whether there is a first situation of a single point mutation according to the difference sequence; A second determination module is configured to perform feature calculation on multiple data in each sliding window of the trend data set to obtain a first feature set, wherein the first feature set includes at least one of a first window mean, a first window variance, a first window maximum value, and a first window minimum value; and determine whether there is a second situation of continuous mutation based on the first feature sets corresponding to adjacent sliding windows; A third determination module is used to perform feature calculation on multiple data of a target sliding window of a trend data set corresponding to each of the multiple positions to obtain a second feature set corresponding to each of the multiple positions; and determine whether there is a third situation of significant difference in the same condition comparison according to the second feature sets of each of the multiple positions and the multiple data of the target sliding window of the trend data set corresponding to each of the multiple positions, where the target sliding window is any window of the multiple sliding windows of the trend data set; A generation module is used to generate an abnormal result according to the first situation, the second situation, and the third situation.

10. An electronic device, characterized in that: include: one or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more applications are configured to: execute the steps of the method for identifying anomalies in trend data of urban rail vehicles according to any one of claims 1 to 8.

Images