Edge device monitoring frequency multistage control method based on abnormal value identification

Abstract

The invention relates to the technical field of railway engineering construction operation and maintenance, in particular to an edge equipment monitoring frequency multistage control method based on abnormal value identification, which comprises the following steps of: constructing a railway infrastructure monitoring knowledge graph; the method comprises the following steps: collecting railway infrastructure monitoring data, deploying a model at edge equipment, and performing data preprocessing on sensor data; performing anomaly judgment by adopting a sequence anomaly value judgment model considering time freshness and an anomaly value judgment model based on Fourier and Monte Carlo; searching an abnormal value discrimination threshold value for the current monitoring data needing to be discriminated by using the knowledge graph, and performing edge device monitoring frequency control according to a threshold value discrimination result; through the above mode, the problem of time delay caused by monitoring data transmission in a previous cloud computing method can be solved, abnormal value discrimination of multi-category data of the Internet of Things is realized, verification is carried out through a multi-level hierarchical discrimination method, the false abnormal alarm probability of a monitoring system is effectively reduced, and the monitoring efficiency and reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of railway engineering construction and operation and maintenance, and in particular to a multi-level control method for edge device monitoring frequency based on abnormal value identification. Background Art

[0002] Ensuring the healthy service status of railway infrastructure during railway engineering operation is the key to safe operation of the line. During railway operation and maintenance, infrastructure is often monitored with multiple indicators. Multiple monitoring devices constitute the Internet of Things, which can detect abnormalities in structural monitoring data in a timely manner to ensure the good operation status of the line.

[0003] The existing monitoring abnormal value discrimination adopts the cloud computing model, which transmits the monitoring data obtained by the front-end monitoring equipment to the cloud computing center, performs data calculation and analysis through the relevant algorithms deployed in the cloud, and then transmits the results to the user end to complete the information feedback. When the monitoring frequency is high, there is a delay in the transmission of large amounts of monitoring data. The processing of data information by the cloud computing center is limited by the network transmission speed, and the railway operation and maintenance management staff cannot obtain effective information feedback in time. From the perspective of IoT monitoring technology, the causes of abnormal values ​​in the monitoring process include: external environmental influences, poor transmission signals of monitoring equipment, abnormal influences of monitoring equipment, abnormalities in the basic structure, etc.

[0004] The existing data analysis algorithms deployed in the cloud do not consider the misjudgment of abnormal values ​​due to external factors, and cannot accurately eliminate "pseudo-anomalies", resulting in the system blindly alarming when encountering abnormal values, and cannot automatically adapt to adjust the monitoring equipment collection frequency to provide more available data information, which increases the workload of railway operation and maintenance personnel in identifying and preparing for abnormal situations, and cannot provide a better solution for the allocation of human and material resources. In addition, there are many types of sensing devices in the railway monitoring Internet of Things, and the existing abnormal value discrimination algorithms are mostly based on a single indicator. Due to the lack of adaptability to the monitoring indicators, the algorithm performs poorly in the discrimination of diversified indicators.

[0005] In summary, it is very necessary to propose a multi-level control method for edge device monitoring frequency based on outlier recognition, which can solve the problem of time delay in monitoring data transmission in previous cloud computing methods, effectively reduce the probability of false abnormal alarms in the monitoring system, and improve monitoring efficiency and reliability. Summary of the invention

[0006] The purpose of the present invention is to provide a multi-level control method for edge device monitoring frequency based on outlier recognition, aiming to solve the problem of time delay in monitoring data transmission in previous cloud computing methods, effectively reduce the probability of false anomaly alarms in the monitoring system, and improve monitoring efficiency and reliability.

[0007] To achieve the above object, the present invention adopts a multi-level control method for edge device monitoring frequency based on abnormal value identification, comprising the following steps:

[0008] Extract entities from railway infrastructure monitoring knowledge and build a railway infrastructure monitoring knowledge graph;

[0009] Collect railway infrastructure monitoring data, deploy models on edge devices, and perform data preprocessing on sensor data;

[0010] The outlier discrimination model of the sequence considering time freshness and the outlier discrimination model based on Fourier and Monte Carlo are used for anomaly discrimination respectively;

[0011] The knowledge graph is used to find the abnormal value discrimination threshold for the current monitoring data that needs to be judged, and the threshold judgment is performed. Based on the judgment result, the edge device monitoring frequency is controlled.

[0012] Among them, in the steps of respectively using the sequence outlier discrimination model considering time freshness and the outlier discrimination model based on Fourier and Monte Carlo to perform anomaly discrimination, the process of using the sequence outlier discrimination model to perform anomaly discrimination is:

[0013] Split the sequence data;

[0014] Establish a sequence outlier discrimination model that considers the time factor weighting;

[0015] Identify sequence outliers.

[0016] Among them, in the steps of respectively using the sequence outlier discrimination model considering time freshness and the outlier discrimination model based on Fourier and Monte Carlo to perform outlier discrimination, the process of using the outlier discrimination model based on Fourier and Monte Carlo to perform outlier discrimination is:

[0017] Construct multiple super Latin sampling datasets;

[0018] The constructed data sets were subjected to multi-level Fourier series expansion, and confidence intervals were formed using Monte Carlo simulation;

[0019] Use confidence intervals to identify outliers.

[0020] Among them, in the steps of using the sequence outlier discrimination model considering time freshness and the outlier discrimination model based on Fourier and Monte Carlo to perform outlier discrimination:

[0021] If the judgment shows an abnormality, the abnormal value judgment threshold is searched based on the knowledge graph;

[0022] If the display is judged to be normal, it is returned to the user end.

[0023] Among them, in the step of using the knowledge graph to find the abnormal value discrimination threshold for the current monitoring data to be judged, performing threshold judgment, and controlling the edge device monitoring frequency according to the judgment result:

[0024] If the threshold judgment shows an abnormality, abnormality identification is performed;

[0025] If the threshold value is normal, the monitoring frequency of the current monitoring data will be increased.

[0026] Wherein, in the step of performing abnormality identification when the threshold value judgment shows abnormality:

[0027] According to the railway infrastructure monitoring knowledge graph, highly correlated monitoring indicators are mobilized to perform multi-model anomaly identification, and based on the identification results, an anomaly identification risk alarm model is constructed to return the anomaly report and risk level to the user end.

[0028] Among them, in the step of adjusting the monitoring frequency of the current monitoring data when the threshold value shows normal:

[0029] The monitoring frequency of the current monitoring data is adjusted through the adaptive adjustment module, and the data preprocessing and abnormality identification are repeated for the current sensor data collected after the next round of monitoring frequency adjustment.

[0030] The present invention discloses a multi-level control method for edge device monitoring frequency based on outlier recognition, which extracts entities from railway infrastructure monitoring knowledge and constructs a railway infrastructure monitoring knowledge graph; collects railway infrastructure monitoring data, deploys models on edge devices, and performs data preprocessing on sensor data; respectively uses a sequence outlier discrimination model that considers time freshness and an outlier discrimination model based on Fourier and Monte Carlo to perform outlier discrimination; uses a knowledge graph to find an outlier discrimination threshold for the current monitoring data to be discriminated, performs threshold judgment, and controls the edge device monitoring frequency according to the judgment result. By using edge computing technology as a means at the sensor end and based on sensor monitoring data, local discrimination of monitoring outliers is achieved; the sequence outlier discrimination model that considers time freshness and the outlier discrimination model based on Fourier and Monte Carlo are used as the first-level discrimination method, and the railway-specific knowledge graph is used as the second-level discrimination basis and the interface for mobilizing relevant indicator discrimination to form a comprehensive discrimination and issue an early warning for the effective screening of outlier discriminations; hierarchical identification is performed for abnormal judgments, and the pseudo-anomalies generated by monitoring are effectively screened. For different abnormal situations, the acquisition frequency of edge monitoring equipment is adjusted in a targeted manner to improve the efficiency of monitoring and early warning, and ensure the reliability and accuracy of abnormal alarms of the monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] 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 some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 It is a flow chart of the multi-level control method of edge device monitoring frequency based on abnormal value identification of the present invention.

[0033] Figure 2 It is a flowchart of the steps of the multi-level control method of edge device monitoring frequency based on abnormal value identification of the present invention.

[0034] Figure 3 It is a schematic diagram of the network model structure of the multi-task learning objective of the present invention.

[0035] Figure 4 It is a visualization diagram of the railway infrastructure monitoring knowledge graph of the present invention.

[0036] Figure 5 It is a schematic diagram of the structure of the sequence outlier discrimination model considering the time factor of the present invention.

[0037] Figure 6 It is a schematic diagram of the structure of the outlier discrimination model based on Fourier and Monte Carlo of the present invention.

[0038] Figure 7 It is a flow chart of adaptive adjustment of the acquisition frequency of the sensor device of the present invention. DETAILED DESCRIPTION

[0039] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0040] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0041] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0042] See also Figure 1 to Figure 7 ,in Figure 1 This is a flow chart of a multi-level control method for edge device monitoring frequency based on abnormal value identification. Figure 2 This is a flowchart of the steps of the multi-level control method of edge device monitoring frequency based on abnormal value identification. Figure 3 It is a schematic diagram of the network model structure of the multi-task learning target. Figure 4 It is a visualization of the railway infrastructure monitoring knowledge graph. Figure 5 It is a schematic diagram of the structure of the sequence outlier discrimination model considering time factors. Figure 6 This is a schematic diagram of the outlier discrimination model structure based on Fourier and Monte Carlo. Figure 7 It is a flow chart of adaptive adjustment of the acquisition frequency of the sensor device.

[0043] The present invention provides a multi-level control method for edge device monitoring frequency based on abnormal value identification, comprising the following steps:

[0044] S100: Extract entities from railway infrastructure monitoring knowledge and build a railway infrastructure monitoring knowledge graph;

[0045] In this implementation, a network model of multi-task learning objectives is constructed based on deep learning, entities are extracted from railway infrastructure monitoring knowledge, and a railway infrastructure monitoring knowledge graph is constructed; the specific network structure is implemented as follows:

[0046] Input data—word embedding layer—shared encoding layer—[task branch 1 (entity recognition) + task branch 2 (attribute acquisition) + task branch 3 (relation acquisition)].

[0047] The network model of the multi-task learning target first captures the semantic information of the word in the word embedding layer to convert each word in the text into a vector representation of a fixed length. Then, the self-attention mechanism is used through the shared encoding layer to encode the contextual information of the text. The information is then passed to different task-specific high-level networks for specific task learning, including knowledge entity recognition, knowledge attribute acquisition, and knowledge relationship acquisition. In the knowledge entity recognition task, the sequence labeling method is used to assign a predefined entity label to each token in the text. In the knowledge attribute acquisition task, a fully connected layer classifier is used to realize attribute recognition, and then a regression model is used to extract specific attribute values ​​from the recognized entity context. In the knowledge relationship acquisition task, the two entity contexts are processed separately through a two-stream network, and then fused, merged, and classified to achieve the knowledge relationship acquisition task.

[0048] A multi-task joint loss function is designed for the network model of multi-task learning objectives, and dynamic weight adjustment is used to control the learning and training effect of the model. The specific loss function is defined as follows:

[0049] L total =λ1L ner +λ2L attr +λ3L re ;

[0050] L ner = -log P(y * |x;θ);

[0051]

[0052] Where: L ner is the knowledge entity recognition loss; L attr is the knowledge attribute acquisition loss; L re is the relationship acquisition loss; λ1, λ2, λ3 are the loss balance adjustment coefficients; x is the input sequence, y * is the correct label sequence; θ is the model parameter; P is the probability of a specific label sequence given the input and model parameters; N is the number of knowledge entities; M is the number of knowledge attributes; y ij is a binary indicator indicating whether entity i has attribute j; p ij is the probability that the model predicts that entity i has attribute j; v i is the correct attribute value of entity i; is the model predicted attribute value; R is the number of knowledge relationship categories.

[0053] S200: collects railway infrastructure monitoring data, deploys models on edge devices, and performs data preprocessing on sensor data;

[0054] In this embodiment, railway infrastructure monitoring data is collected, models are deployed on edge devices, and sensor data is preprocessed; wherein the outlier discrimination algorithm model is deployed on edge devices, including format conversion, model compression, and deployment integration. When preprocessing monitoring data, the key is to embed category labels into the data, which are used as query inputs in the constructed railway infrastructure monitoring knowledge graph to obtain the thresholds of the indicators corresponding to the categories. Data preprocessing includes interpolation of missing values ​​and data noise reduction for the monitoring data c1, c2, c3, etc. collected by the sensing devices.

[0055] Data preprocessing includes missing value interpolation and noise reduction. In this embodiment, missing value interpolation adopts K-nearest interpolation method, that is, a missing data interpolation algorithm based on K nearest neighbor algorithm. K samples closest to the missing sample are selected according to distance or correlation analysis, and the data of these K samples are weighted to estimate the missing data of the sample.

[0056] The above K-nearest interpolation method assumes that two samples y a and b The distance between them is D(y a ,y b ), using heterogeneous Euclidean distance metric to define D(y a ,y b )as follows:

[0057]

[0058] In the formula, Di(y ai ,y bi ) is the sample y a and b The distance between the i-th variables;

[0059] For sample y a and b If the ith variable has missing values, the maximum distance 1 is returned, which is calculated by the D0 function and is defined as:

[0060]

[0061] For the K nearest interpolation method, if the sample y a If there is a missing value in the i-th variable of , select the K samples closest to the sample, and the i-th variable of these K samples has no missing values, and place these K samples at the distance from sample y a From near to far, the collection is composed:

[0062] (m j is the distance sample y a recent samples);

[0063] By weighting the K nearest neighbor samples separately, adding the weights of each category, and selecting the category with the largest sum of weights as the interpolation value.

[0064] Data denoising uses wavelet transform to remove noise interference from monitoring data and obtain useful data information. First, the Daubechies wavelet basis function is selected, then the number of decomposition layers is determined for fast wavelet transform, and finally the data information is reconstructed. The formula involved is as follows:

[0065]

[0066] D j [n] = ∑ k f[n]·ψ j,k [n];

[0067] ψ j,k [n] = ψ[n]·2 -j ·ψ(2 -j nk);

[0068] Where: f[n] is the original monitoring data signal; k is the displacement parameter; C j [n] approximation coefficient; D j [n] detail coefficient; Scaling function; ψ[n] discrete function of the mother wavelet; j number of decomposition levels.

[0069] S300: Anomaly discrimination is performed by using a sequence outlier discrimination model that considers time freshness and an outlier discrimination model based on Fourier and Monte Carlo;

[0070] In this embodiment, a sequence outlier discrimination model considering time freshness and an outlier discrimination model based on Fourier and Monte Carlo are used to perform outlier discrimination; wherein:

[0071] The purpose of sequence outlier discrimination considering time factors is to filter out false alarms caused by single mutation anomalies due to external environmental influences. The process of using the sequence outlier discrimination model for anomaly discrimination is as follows:

[0072] S311: Segment the sequence data;

[0073] S312: Establishing a sequence outlier discrimination model that considers the weighting of time factors;

[0074] S313: Identify sequence outliers.

[0075] Furthermore, in S311: splitting the sequence data, taking the current monitoring data to be identified as the benchmark, the time series A containing 21 data points is intercepted forward along the time dimension and divided into 3 time step windows: T = 3;

[0076] Furthermore, in S312: establishing a sequence outlier discrimination model that considers the time factor weighting,

[0077] The structure of the sequence outlier discrimination model considering time factors is implemented as follows:

[0078] The time impact function is constructed to meet the principle of "near is large and far is small" in the time dimension, and the function satisfies non-negative single increasing. The constructed time impact function is as follows:

[0079]

[0080] Where: F(t) represents the weighting function considering the influence of time; t aq Indicates the starting time point of the previous sequence; t a Indicates the starting time point of the current sequence; e represents the base of the exponential function; S represents the error of the monitoring data.

[0081] The sequence outlier discrimination model considering time factors includes two parts: decoder and encoder, such as Figure 5 As shown in the figure, the specific structure is: encoder: input layer - LSTM layer - time factor weighted layer - re-parameterization layer - output layer; decoder: fully connected layer - LSTM layer - activation layer - output layer. The key steps of the above network implementation are as follows:

[0082] Time series datasets The feature vector observed by the LSTM layer at time step t is a t ,In the encoder, the time factor and weight are specifically implemented as follows:

[0083] w t =F(a t θ F );

[0084]

[0085] Among them, w t is the weight at time step t; θ F is the parameter of the time influence function; F(·) is the time influence function operation; is the weighted eigenvector; ⊙ represents element-wise multiplication; h t Represents the output of the time factor weighted layer.

[0086] Furthermore, in S313: determining sequence outliers, the threshold of the sequence outlier determination model considering the time factor is cross-validated to obtain the optimal threshold, and the reconstruction error is compared with the threshold to achieve anomaly determination. The determination result is fed back to the user end and the edge device acquisition frequency adaptive adjustment module.

[0087] The outlier discrimination model based on Fourier and Monte Carlo is to construct a discrimination standard based on the changing law of the monitoring data itself, and filter out false reports caused by ignoring the natural changes over time and being restricted by artificial fixed values; the process of using the outlier discrimination model based on Fourier and Monte Carlo for abnormal discrimination is as follows:

[0088] S321: Construct multiple super Latin sampling datasets;

[0089] S322: performing multi-level Fourier series expansion on the constructed data sets respectively, and forming confidence intervals using Monte Carlo simulation;

[0090] S323: Use confidence intervals to identify outliers.

[0091] Further, in S321: constructing multiple super Latin sampling data sets, constructing an outlier discrimination model based on Fourier and Monte Carlo requires first constructing a data set of current monitoring data through super Latin cubic sampling. Divide the monitoring data into N evenly spaced intervals and randomly select points, forming a new data set When the number of points m=1, the principle of data set composition can be expressed as:

[0092]

[0093] Further, in S322: the constructed data sets are respectively expanded by multi-level Fourier series, and the confidence interval is formed by Monte Carlo simulation. After the currently monitored data set is constructed by super Latin cube sampling, the new data set is Perform multi-level Fourier series expansion according to the Fourier series expansion principle:

[0094]

[0095] where a0, a n , b n are the coefficients of the multi-order Fourier series expansion.

[0096] After Fourier series expansion, a new data set is obtained Using the Monte Carlo simulation idea, S321 is executed multiple times to obtain k groups of data to form a data set group. Multi-level Fourier series expansion is performed to obtain the data matrix X M :

[0097]

[0098] Get Cap:

[0099]

[0100] Lower limit:

[0101]

[0102] The confidence interval is taken with the upper and lower limits of 95% as the basis for judging the current monitoring data. The confidence interval can be expressed as:

[0103] [X s -0.05(X s -X x ),X x +0.05(X s -X x )].

[0104] Furthermore, in S323: using confidence intervals to identify outliers, the confidence interval generated by the outlier identification model constructed based on the Fourier series expansion and Monte Carlo simulation concept is used to identify outliers on the current data.

[0105] The confidence interval generated by the outlier discrimination model constructed based on Fourier series expansion and Monte Carlo simulation ideas is used to discriminate outliers on the current data.

[0106] S400: Using the knowledge graph to find an abnormal value discrimination threshold for the current monitoring data to be discriminated, performing threshold judgment, and controlling the edge device monitoring frequency according to the judgment result;

[0107] In this embodiment, the category label of the current discriminant data object is read, and the category label is used as input. The threshold corresponding to the category indicator is queried in the railway infrastructure monitoring knowledge graph, and used as the judgment threshold of the current discriminant data. Taking the post-construction settlement monitoring of the roadbed as an example, when the sequence outlier discrimination model considering time freshness and the outlier discrimination model based on Fourier and Monte Carlo are detected to be abnormal, a search will be performed in the constructed knowledge graph. In the railway infrastructure monitoring knowledge graph, the threshold setting is stored according to the specification requirements: the post-construction settlement of the ballastless track should not exceed 15mm. If it exceeds the specification threshold, the abnormal information of the post-construction settlement of the ballastless track will be returned to the client.

[0108] Furthermore, if the threshold value judgment shows an abnormality, an abnormality identification is performed, and if the threshold value shows normal, the monitoring frequency of the current monitoring data is adjusted and strengthened.

[0109] Furthermore, in the step of anomaly identification, multi-model anomaly identification is performed by mobilizing highly correlated monitoring indicators according to the railway infrastructure monitoring knowledge graph, and based on the identification results, an anomaly identification risk alarm model is constructed to return the anomaly report and risk level to the user end.

[0110] Among them, according to the knowledge graph, other monitoring indicator data with high correlation with the current monitoring data indicators are mobilized, and the multi-model abnormal identification module is used for identification, and the abnormal identification risk alarm model is constructed. The monitoring indicators with a correlation of more than 60% are used as auxiliary. Assume that there are m relevant monitoring indicators that meet the requirements, and the correlation coefficient B = {b1, b2, …, b m}, and use (0,1) distribution to represent the abnormal judgment results of each related indicator and the current monitoring data indicator, and obtain the risk value Y = {y1,y2,…,y m ,y x}, then the risk alarm model is composed of:

[0111]

[0112] The corresponding risk alarm levels are shown in Table 1:

[0113] Table 1 Risk Alarm Level Table

[0114] Risk Level RiskSignal yellow RiskSignal<0.5 orange color 0.5≤RiskSignal<0.7 red 0.7≤Risk Signal

[0115] Taking the post-construction settlement monitoring of the roadbed as an example, the monitoring indicators with a correlation of more than 60% at the same time are used as auxiliary indicators. For example, the current abnormal judgment value of the post-construction settlement monitoring of the roadbed is y x , the current abnormal identification value of the monitoring indicator that meets the requirements is water content y1, horizontal displacement y2, and their correlation coefficients are b1, b2 (b1, b2>0.6), then the risk alarm model is RiskSignal=0.8y x +0.2×(b1y1+b2y2) / 2. Alarm is issued according to the Risk Signal value correspondence table.

[0116] Furthermore, in the step of adjusting the monitoring frequency of the current monitoring data, the monitoring frequency of the current monitoring data is adjusted through the adaptive adjustment module, and the data preprocessing and abnormality judgment are repeated for the current sensor data collected after the next round of adjustment of the monitoring frequency, that is, S200 to S300 are repeatedly executed, and for the next round of collected data that is judged to be normal, the original collection frequency is restored through the adaptive adjustment module.

[0117] The sensor edge device receives the signal and performs adaptive adjustment on the acquisition frequency. If an abnormal result is received, the sensor edge device automatically increases the acquisition frequency to collect data, providing data guarantee for repeated execution of S200 to S300. If the received result is normal, the sensor maintains the original acquisition frequency to work normally.

[0118] like Figure 7 As shown, the steps of adaptively adjusting the acquisition frequency of the sensor device include: receiving signals, analyzing signal properties, adjusting the real-time acquisition frequency, and collecting monitoring data.

[0119] The signal is received, and the judgment result obtained by edge computing is input into the adaptive adjustment module as a signal.

[0120] The abnormality nature is analyzed, and the input judgment result is analyzed to determine whether it is abnormal or normal.

[0121] The real-time acquisition frequency adjustment, if it is an abnormal signal, the system will increase the frequency of data acquisition in order to intensively monitor the development of this abnormal situation to ensure that all key data can be captured. If it is a normal signal, the current acquisition frequency is maintained or adjusted back to the normal level to avoid unnecessary data accumulation and resource waste.

[0122] The monitoring data is collected, and the adaptive adjustment module controls the working mode of the sensor to achieve real-time adjustment of the collection frequency. The operation feedback makes the adjusted collection frequency take effect immediately. The system continues to collect data according to the new collection frequency to provide data support for the abnormal value identification module. The yellow alarm of abnormal identification adjusts the equipment collection frequency to 1.25 times, the orange warning adjusts the collection frequency to 1.5 times, and the red warning adjusts the collection frequency to 2 times.

[0123] In the present invention, the monitoring data outlier discrimination model is deployed on the edge device to solve the delay problem caused by the transmission of large amounts of data, and data judgment is performed locally so that the staff can grasp the health status of the roadbed in a timely manner. A railway infrastructure monitoring knowledge graph is constructed, and the threshold of the outlier discrimination model is not for a single value, but is adjusted according to the category label of the input data to achieve multi-index discrimination, and the model has diversified capabilities. A sequence outlier discrimination method based on time freshness is constructed, and the influence weight of the monitoring sequence data is determined in the time dimension according to the principle of "near large and far small". An outlier discrimination model based on Fourier and Monte Carlo is constructed, and the confidence interval is determined through the Monte Carlo simulation idea to eliminate pseudo-abnormal phenomena caused by external factors and ensure the accuracy of the returned abnormal alarm information. According to the returned abnormal judgment information, the edge device acquisition frequency is controlled in a targeted and real-time manner, and the frequency is adjusted adaptively to provide richer and more effective monitoring information.

[0124] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the contents disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary technical means in the art that are not disclosed in the present application.

[0125] It should be understood that the present application is not limited to the exact construction that has been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof.

Claims

1. A multi-level control method for edge device monitoring frequency based on abnormal value identification, characterized in that: The steps include: Extract entities from railway infrastructure monitoring knowledge and build a railway infrastructure monitoring knowledge graph; Collect railway infrastructure monitoring data, deploy models on edge devices, and perform data preprocessing on sensor data; The outlier discrimination model of the sequence considering time freshness and the outlier discrimination model based on Fourier and Monte Carlo are used for anomaly discrimination respectively; The knowledge graph is used to find the abnormal value discrimination threshold for the current monitoring data that needs to be judged, and the threshold judgment is performed. Based on the judgment result, the edge device monitoring frequency is controlled.

2. The multi-level control method for edge device monitoring frequency based on abnormal value identification according to claim 1 is characterized in that: In the steps of using the sequence outlier discrimination model considering time freshness and the outlier discrimination model based on Fourier and Monte Carlo to perform anomaly discrimination, the process of using the sequence outlier discrimination model to perform anomaly discrimination is: Split the sequence data; Establish a sequence outlier discrimination model that considers the time factor weighting; Identify sequence outliers.

3. The multi-level control method for edge device monitoring frequency based on abnormal value identification according to claim 1 is characterized in that: In the steps of using the sequence outlier discrimination model considering time freshness and the outlier discrimination model based on Fourier and Monte Carlo to perform outlier discrimination, the process of using the outlier discrimination model based on Fourier and Monte Carlo to perform outlier discrimination is: Construct multiple super Latin sampling datasets; The constructed data sets were subjected to multi-level Fourier series expansion, and confidence intervals were formed using Monte Carlo simulation; Use confidence intervals to identify outliers.

4. The multi-level control method for edge device monitoring frequency based on abnormal value identification according to claim 1 is characterized in that: In the steps of using the sequence outlier discrimination model considering time freshness and the outlier discrimination model based on Fourier and Monte Carlo to perform outlier discrimination: If the judgment shows an abnormality, the abnormal value judgment threshold is searched based on the knowledge graph; If the display is judged to be normal, it is returned to the user end.

5. The multi-level control method for edge device monitoring frequency based on abnormal value identification according to claim 1, characterized in that: In the step of using the knowledge graph to find the abnormal value discrimination threshold for the current monitoring data to be judged, making a threshold judgment, and controlling the edge device monitoring frequency according to the judgment result: If the threshold judgment shows an abnormality, abnormality identification is performed; If the threshold value is normal, the monitoring frequency of the current monitoring data will be increased.

6. The multi-level control method for edge device monitoring frequency based on abnormal value identification according to claim 5 is characterized in that: In the step of performing abnormality identification when the threshold judgment shows abnormality: According to the railway infrastructure monitoring knowledge graph, highly correlated monitoring indicators are mobilized to perform multi-model anomaly identification, and based on the identification results, an anomaly identification risk alarm model is constructed to return the anomaly report and risk level to the user end.

7. The multi-level control method for edge device monitoring frequency based on abnormal value identification according to claim 5, characterized in that: In the step of adjusting the monitoring frequency of the current monitoring data when the threshold value is normal: The monitoring frequency of the current monitoring data is adjusted through the adaptive adjustment module, and the data preprocessing and abnormality identification are repeated for the current sensor data collected after the next round of monitoring frequency adjustment.

Images