Exception analysis method and system based on blasting data

Abstract

The invention relates to the technical field of anomaly detection, and discloses an anomaly analysis method and system based on blasting data, and the method comprises the steps: collecting the blasting data of each monitoring node, dividing the monitoring node into a plurality of regions, carrying out the statistics of the anomaly density of each region in a preset historical period, and calculating the anomaly density of each region; according to the abnormal density, dynamically setting a region weight of an auto-encoder training sample; reconstructing the current round of data by using the weighted trained auto-encoder, calculating the residual error of each node, and extracting the activation path information of the node of which the residual error is higher than a reconstruction error preliminary screening threshold value; constructing a graph structure model between the monitoring nodes, and calculating an abnormal propagation path according to the correlation between the nodes; and matching the activation path with the propagation path, and determining the abnormal state of the target node. And the accuracy, the stability and the intelligent feedback capability of blasting data anomaly analysis are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of anomaly detection, and in particular to an anomaly analysis method and system based on blasting data. Background Art

[0002] At present, with the continuous improvement of industrial automation and digitalization, data processing systems have been widely used in various fields, especially in complex engineering environments, where real-time data collection, processing and anomaly detection have become key technologies to ensure the safety and stable operation of the system. In recent years, information processing methods based on digital computers have been widely used in blasting engineering, geotechnical engineering and other fields to perform real-time monitoring and anomaly detection of multi-dimensional data such as vibration, acoustic emission, stress, and temperature collected during the blasting process.

[0003] In the existing technology, most methods rely on traditional signal processing algorithms and statistical analysis techniques to preprocess and judge the threshold of data. Although they can detect anomalies to a certain extent, they have problems such as low detection accuracy, sensitivity to noise and data fluctuations, and lack of analysis of the intrinsic structure and spatiotemporal correlation of data when dealing with complex and changeable blasting data environments. On the other hand, with the development of artificial intelligence and machine learning algorithms, deep learning models, especially autoencoders, have shown great advantages in data compression and reconstruction, but their application in blasting data anomaly detection still faces challenges such as insufficient model training, unclear regional focus, and inadequate capture of anomaly propagation laws. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the prior art method for analyzing anomaly in blasting data is not accurate in detecting structural anomalies, is not sufficient in identifying regional features, and is not sufficiently explanatory of anomaly propagation paths.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: an abnormal analysis method based on blasting data, comprising collecting blasting data of each monitoring node, and dividing the monitoring node into multiple areas; Count the abnormal density of each region within the preset historical period, and dynamically set the regional weight of the autoencoder training samples based on the abnormal density; The current round of data is reconstructed using the weighted trained autoencoder, the residual of each monitoring node is calculated, and the activation path information of the nodes whose residual is higher than the initial screening threshold of the reconstruction error is extracted; Build a graph structure model between monitoring nodes and calculate the abnormal propagation path based on the correlation between nodes; The activation path is matched with the propagation path to determine the abnormal state of the target node.

[0007] As a preferred solution of the abnormality analysis method based on blasting data described in the present invention, wherein: the blasting data includes vibration signals, acoustic emission signals, stress data and temperature;

[0008] Dividing the monitoring nodes into multiple areas includes extracting time series features from the blasting data collected by each monitoring node to form a feature sequence corresponding to each node; using the Pearson correlation coefficient to calculate the similarity between any two monitoring nodes and constructing a similarity matrix between nodes;

[0009] The similarity matrix is ​​used as an undirected graph, in which each monitoring node is a node in the graph, and the edges between nodes represent the similarity degree of data features; the undirected graph is divided into several sub-regions by using a spectral clustering algorithm, so that the monitoring nodes in each sub-region have correlation in data behavior.

[0010] As a preferred solution of the anomaly analysis method based on burst data described in the present invention, wherein: the setting of the regional weight of the autoencoder training sample includes counting the number of abnormal nodes in each sub-region in the historical period and calculating the abnormal density of each sub-region; based on the difference between the abnormal density of each sub-region and the average abnormal density of all sub-regions, determining the sample weight coefficient corresponding to each sub-region;

[0011] During the training process of the autoencoder, the sample weight coefficients are applied to the training samples in different sub-regions respectively.

[0012] As a preferred solution of the abnormal analysis method based on burst data described in the present invention, wherein: the reconstruction of the current round of data includes reconstructing the burst data collected in the current round using the trained autoencoder, and calculating the reconstruction error of each monitoring node , the reconstruction error is the difference between the original data and the reconstructed data of the monitoring node.

[0013] As a preferred solution of the abnormal analysis method based on blasting data described in the present invention, wherein: extracting activation path information includes determining whether the reconstruction error of each node exceeds the reconstruction error initial screening threshold , when the reconstruction error exceeds the initial screening threshold , it is determined to be a high residual node; extracting the activation path information of the high residual node from the input layer to the hidden layer in the autoencoder structure , the activation path information is the activation value sequence of each hidden layer of the high residual node in the encoding stage.

[0014] As a preferred solution of the anomaly analysis method based on blasting data described in the present invention, wherein: constructing a graph structure model between monitoring nodes includes: the edge between nodes represents the feature similarity value between any two monitoring nodes;

[0015] The calculation of the abnormal propagation path according to the correlation between nodes includes: for each node determined to be a high residual node , extract the set of adjacent nodes in the graph structure ;exist In the screening, the reconstruction error also exceeds the reconstruction error initial screening threshold The nodes constitute the high residual nodes The initial abnormal propagation path node set ;

[0016] When the high residual node , high residual nodes All adjacent nodes of , node collection When a closed connected subgraph is formed together, it is determined that the propagation path has stable consistency in the graph structure, and the propagation path is marked as a high-confidence propagation path;

[0017] The node set For: with There is a direct connection between any node in the A collection of nodes that are not directly connected to each other;

[0018] In the initial exception propagation path node set The nodes that are not included in the final high-confidence propagation path node set are set as the low-confidence propagation path candidate set ;

[0019] Final abnormal propagation path node set Including, from the high confidence propagation path node set, based on each propagation node With the high residual node The feature similarity value between Sort in descending order and select the top feature similarity values The node of the position; and the low confidence propagation path candidate set Satisfaction-inducing factors identified in The latent inducing abnormal node; wherein the inducing factor It is composed of the neighborhood anomaly density, reconstruction error volatility, and the temporal drift of graph embedding features. express The mean of express The standard deviation of .

[0020] As a preferred solution of the abnormal analysis method based on blasting data described in the present invention, wherein: determining the abnormal state of the target node includes using Jaccard similarity calculation and Similarity ;

[0021] when and When , it is determined to be a structural abnormal node, indicating that the structural abnormal node presents a consistent abnormal trend in both the internal characteristic response of the model and the external structural propagation;

[0022] when and When , it is determined to be a burst abnormal node, indicating that although the burst abnormal node shows significant abnormality in the model, it does not form coordinated diffusion in the structure diagram;

[0023] when When multiple adjacent nodes of the node are the initial abnormal propagation path nodes, it is determined to be a latent abnormal node, indicating that the latent abnormal node is in an abnormal high-incidence environment and requires special attention.

[0024] Anomaly analysis system based on blasting data, including:

[0025] The data module collects the blasting data of each monitoring node and divides the monitoring nodes into multiple areas;

[0026] The training module counts the anomaly density of each region within a preset historical period and dynamically sets the regional weights of the autoencoder training samples based on the anomaly density;

[0027] The reconstruction module reconstructs the current round of data using the weighted trained autoencoder, calculates the residual of each monitoring node, and extracts the activation path information of the nodes whose residual is higher than the initial screening threshold of the reconstruction error;

[0028] The propagation module builds a graph structure model between monitoring nodes and calculates the abnormal propagation path based on the correlation between nodes;

[0029] The confirmation module matches the activation path with the propagation path and calculates the similarity to determine the abnormal state of the target node.

[0030] A computer device comprises a memory and a processor; the memory stores a computer program, wherein the processor implements the steps of any one of the methods of the present invention when executing the computer program.

[0031] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any one of the methods of the present invention.

[0032] The beneficial effects of the present invention are as follows: by constructing a graph structure model based on burst data, introducing regional division and abnormal density guidance mechanisms, the adaptive training optimization of the autoencoder for key areas is realized, and the model's ability to identify high-risk areas is improved. Combined with the reconstruction residual judgment and activation path extraction, the internal response characteristics of the model can be accurately mined. Further, through the propagation path identification and Jaccard similarity matching, the linkage judgment of the structural consistency and propagation trend of the abnormal nodes is realized. Finally, a classification mechanism for structural, burst and latent anomalies is established, which makes the anomaly detection results more interpretable and precise, and significantly improves the accuracy, stability and intelligent feedback capability of the anomaly analysis of burst data. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0034] Figure 1 This is an overall flow chart of the abnormality analysis method based on blasting data provided by the first embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0036] Example 1, reference Figure 1 , is an embodiment of the present invention, providing an abnormality analysis method based on blasting data, comprising:

[0037] S1: Collect blasting data from each monitoring node and divide the monitoring nodes into multiple areas.

[0038] The blasting data include vibration signals, acoustic emission signals, stress data and temperature.

[0039] First, the blasting data collected by each monitoring node is analyzed and processed to extract the time series features to form the feature sequence of each monitoring node, which is recorded as:

[0040]

[0041] in, Indicates Monitoring nodes in A sequence of data at a time step, which may include vibration amplitude, acoustic emission amplitude, stress change or temperature.

[0042] Then, the similarity between any two monitoring nodes is calculated based on the feature sequence of each node, using the Pearson correlation coefficient defined as follows:

[0043]

[0044] in, and Node and The feature mean of all node pairs The calculation results form a symmetric similarity matrix .

[0045] Next, the similarity matrix As an adjacency matrix of an undirected graph, construct the graph structure , where each monitoring node is a node in the graph ,like , then at the node and Create an edge between , the edge weight is The absolute or raw value of . Represents a collection of nodes. Represents a set of edges.

[0046] Finally, based on the graph structure, the spectral clustering algorithm is used to cluster the monitoring nodes to form several sub-regions. The monitoring nodes in each sub-region have strong consistency in historical behavior patterns, providing a regional basis for subsequent training sample weighting.

[0047] Furthermore, by quantifying and similarity modeling the temporal characteristics of blasting data, using graph structures to represent the behavioral correlations between monitoring nodes, and introducing spectral clustering algorithms to complete intelligent spatial regional division, it no longer relies on artificial geographical divisions.

[0048] Furthermore, by constructing a behavioral similarity graph and performing spectral clustering, the model can perform adaptive optimization at a granularity of structural units (sub-regions) that are more consistent with data characteristics during the training phase, significantly improving the sensitivity and learning efficiency of high-risk areas. This lays the foundation for improving the overall anomaly detection accuracy, especially in complex working conditions with stronger robustness and interpretability.

[0049] S2: Count the abnormal density of each region within the preset historical period, and dynamically set the regional weights of the autoencoder training samples based on the abnormal density.

[0050] In each round of detection cycle, the residual of each monitoring node is first calculated according to the reconstruction result of the autoencoder. The original feature vector collected by the monitoring nodes in this round is , the original features include the time series indicators such as vibration, acoustic emission stress, temperature, etc. recorded by the node at multiple time points. Input into the autoencoder to obtain the corresponding reconstructed output , which represents the model's reconstruction of the current state of the node based on historical learning results.

[0051] Set up The residual of each node is:

[0052]

[0053] in, is the selected distance metric. Represents the reconstruction error of each monitoring node.

[0054] Set the initial screening threshold for reconstruction residuals , the initial screening threshold It can be set as the 85% quantile of the reconstructed residual according to the model training error distribution; if , then the node is judged to be an abnormal node. All monitoring nodes are classified according to the area they belong to, and the area number is recorded as , the total number of nodes is In a length of In the sliding time window of The number of abnormal nodes in each round in the region , and calculate the anomaly density of the area in the current time window:

[0055]

[0056] The above abnormal density Indicates that the area is near The frequency of abnormalities in a cycle reflects the risk level. The average abnormal density of all regions is further calculated:

[0057]

[0058] in, Indicates the total number of regions.

[0059] In order to achieve adaptive focusing during regional sample training, a nonlinear function is introduced to set the weight coefficient for regional samples. , which is expressed as follows:

[0060]

[0061] in, is a regulating factor used to control the sensitivity of weight changes. This function has a central compression characteristic. When the abnormal density in a certain area is higher than the average level, The value of increases significantly, thereby increasing the influence of samples in this area during training. On the contrary, when the anomaly density is low, The value decreases, reducing its interference and ensuring the model focuses on learning the features of key areas.

[0062] Furthermore, by introducing a regional weighting mechanism based on anomaly density, the model training process can dynamically perceive the risk distribution of the monitoring area, thereby achieving spatial adaptive adjustment of sample learning. In the blasting monitoring scenario, the probability of anomalies in different regions is significantly different due to differences in geological structure, operation frequency, etc. This design can guide the model to focus on modeling the characteristics of high-risk areas and improve the overall model's ability to characterize abnormal behaviors in key areas and its response accuracy.

[0063] Furthermore, the regional anomaly density is statistically analyzed based on data from multiple cycles to avoid misjudgment caused by a single disturbance, and the training intensity of regional samples is flexibly adjusted through nonlinear functions to make the model training distribution more in line with the actual risk distribution. This strategy not only improves the model's adaptability to diverse regional structures, but also enhances the recognition accuracy of weak anomalies and latent anomalies, providing a more accurate and stable basis for subsequent anomaly type determination.

[0064] S3: After reconstructing the current round of data using the weighted trained autoencoder and calculating the residual of each node, the activation path information of the nodes whose residuals are higher than the initial screening threshold of the reconstruction error is extracted.

[0065] For all high residual nodes, their activation path information is extracted from the trained autoencoder structure. Specifically, the encoding part of the autoencoder is assumed to be Layer composition, The activation value of the layer is recorded as , then The activation path of a node is:

[0066]

[0067] The activation path information It represents the set of characteristic neurons activated during the transmission of the node from the input layer to the hidden layer, which can reflect the response mode of the node in the encoding stage. The activation path information will be compared with the abnormal propagation path in the subsequent steps to determine the consistency matching degree of the structural abnormality.

[0068] Activation Value Use the ReLU activation function to output.

[0069] Furthermore, by extracting activation path information from high residual nodes, an accurate description of the response mechanism of input samples can be formed within the neural network structure, so that it no longer relies on a single numerical judgment, but introduces the hierarchical response process of the model as the basis for abnormal identification. This path reflects the trajectory of node feature propagation and activation in each hidden layer of the autoencoder, which helps to track the transmission link of abnormal information in the model and provide an interpretable basis for subsequent path-level similarity comparison.

[0070] Furthermore, by matching the activation path with the propagation path, the accuracy of anomaly recognition is enhanced, and the model's ability to distinguish complex anomaly structures is also improved. At the same time, the activation value is output using the ReLU function, which further ensures the sparsity and stability of the path expression and avoids the influence of interference features on the final judgment result.

[0071] S4: Build a graph structure model between monitoring nodes and calculate the anomaly propagation path based on the correlation between nodes.

[0072] The behavior similarity of any two nodes i and j is calculated based on the historical behavior data of each monitoring node (such as vibration, acoustic emission, stress, temperature and other time series characteristics). Suppose the data sequences of nodes i and j are , , then its similarity can be expressed using the Pearson correlation coefficient:

[0073]

[0074] Building the graph structure ,in: Indicates all monitoring nodes. If the node and Relevance Greater than the similarity threshold , then an undirected edge is established between the two , the edge represents the feature similarity value .

[0075] This graph is used to characterize the potential anomaly propagation relationship at the data level.

[0076] For the node i that is judged to have high residual in the current round, extract the set of adjacent nodes in its graph structure:

[0077]

[0078] Then in The residual value is also greater than or equal to the initial threshold of the reconstructed residual. , as the abnormal propagation path node set of node i:

[0079]

[0080] in, Indicates high residual nodes The surrounding area is highly correlated with it and the reconstruction error also exceeds A collection of nodes.

[0081] In order to improve the structural rationality of abnormal propagation path identification, After that, when the high residual node , high residual nodes All adjacent nodes of , node collection When a closed connected subgraph is formed together, it is determined that the propagation path has stable consistency in the graph structure, and the propagation path is marked as a high-confidence propagation path;

[0082] The node set For: with There is a direct connection between any node in the A collection of nodes that are not directly connected to each other.

[0083] In the initial exception propagation path node set The nodes in the node set that are not included in the final high-confidence propagation path are called the low-confidence propagation path candidate set. .

[0084] Introducing compound triggers , used to characterize nodes The potential risk of causing abnormal spread. The inducing factors are constructed based on the following three dimensions: Neighborhood abnormal density: refers to the node The proportion of nodes in the adjacent nodes of which the residual state is high; Reconstruction error volatility: a measure of the node The magnitude of the reconstruction error change in different rounds; the degree of temporal drift of graph embedding features: Evaluation nodes The degree to which the embedding vector changes dynamically between different time periods.

[0085] In actual blasting monitoring scenarios, although some nodes do not form significant collaborative propagation paths, they may still trigger latent diffusion behaviors. Relying only on edge weights or residual judgments can easily lead to omissions in the identification of highly sensitive areas. Therefore, this embodiment achieves earlier discovery of abnormal sources by constructing a composite inducing factor.

[0086] Compound triggering factor The construction of adopts a hierarchical structure constraint method, which is defined as follows:

[0087]

[0088] in, Representation Node The abnormal density of the neighborhood, that is, the residual in its adjacent nodes exceeds the threshold proportion; Representation Node The volatility of the reconstruction error is Representation Node The degree of temporal drift of graph embedding features.

[0089] For the low confidence propagation path candidate set in the current round Calculate the inducing factor value of all nodes in , and then based on the overall distribution characteristics of the inducing factors in the set, obtain its mean With standard deviation .

[0090] If a node meets the following conditions:

[0091]

[0092] It is determined to be a latent abnormal node with local abnormal driving characteristics.

[0093] In the node selection process in the high confidence propagation path, based on the relationship between each node and the central node The feature similarity between Sort in descending order and keep the feature similarity value at the top The adjacent nodes of the bit;

[0094] The final exception propagation path node set Including, from the high confidence propagation path node set, based on each propagation node With the high residual node The feature similarity between Sort in descending order and select the top feature similarity values The node of the position; and the low confidence propagation path candidate set Satisfaction-inducing factors identified in The latent abnormal nodes.

[0095] Compared with the strategy of using a fixed threshold, the above judgment conditions have stronger scene adaptability and statistical interpretability, and can dynamically reflect the relative degree of outliers of inducing factors in different rounds and different regions, thereby more effectively identifying latent abnormal sources and avoiding missed detections and misjudgments.

[0096] In the process of constructing the monitoring node graph structure and identifying the abnormal propagation path, the structural consistency judgment is further introduced to improve the stability and recognition accuracy of the propagation path. The graph structure effectively expresses the potential abnormal correlation of each monitoring node in the blasting site at the time series data level, laying the foundation for the construction of the propagation path.

[0097] For each node judged as having high residual in the current round , extract the set of adjacent nodes in the graph structure ;exist In the screening, the reconstruction error also exceeds the reconstruction error initial screening threshold The node that constitutes the initial abnormal propagation path node set of the node ,This step can quickly identify the neighborhood propagation relationships that are strongly associated with high residual nodes and have consistent abnormal trends, and achieve preliminary locking of potential risk diffusion areas.

[0098] Furthermore, in order to improve the structural rationality and credibility of the propagation path, this embodiment introduces a structural consistency determination mechanism: if the node , its adjacent nodes and some second-order adjacent nodes form a closed subgraph structure in the graph structure (such as a triangular closed loop or a four-membered ring structure), the propagation path is determined to have topological consistency and marked as a high-confidence propagation path. This mechanism helps to eliminate misconnections caused by occasional noise or isolated disturbances and improve path stability. After marking as a high-confidence path, the propagation path set The nodes in the The feature similarity value Sort in descending order, keeping only the first The adjacent nodes with the largest propagation intensity are used as the final propagation path set This mechanism improves recognition accuracy while reducing computational complexity, and provides high-quality data support for subsequent activation path comparison and abnormal type determination.

[0099] In the process of constructing the abnormal propagation path, the path relationship is not constructed only based on the correlation between the high residual node and its adjacent nodes, but a structural consistency judgment mechanism and an edge weight optimization screening mechanism for the propagation path nodes are further introduced. On the basis of the initially constructed propagation path, if the target node, its adjacent nodes and the second-order adjacent nodes together form a closed subgraph structure, the path is judged to have topological consistency and is marked as a high-confidence propagation path. Through the introduction of the above dual mechanisms, the propagation path is superior to the existing method of constructing paths based only on adjacent residuals in terms of structural rationality, node representativeness and expression of propagation trends, which significantly improves the accuracy and stability of abnormal propagation identification and provides a more reliable data basis for subsequent abnormal state judgment.

[0100] S5: Match the activation path with the propagation path to determine the abnormal state of the target node.

[0101] The matching judgment of the activation path and the propagation path includes: extracting the target node The activation path set and the abnormal propagation path set formed in the graph structure are used to calculate the similarity between the two It indicates the degree of match between the node's internal response process and the external structural diffusion process in the model.

[0102] Similarity The formula is: The activation path of a node is , the final set of exception propagation paths is , the degree of overlap between the two is calculated by Jaccard similarity, which is defined as follows:

[0103]

[0104] in, represents the cardinality of a set, represents the intersection, represents a union, The value range of is [0, 1]. The closer it is to 1, the more consistent the two paths are.

[0105] The matching judgment between the activation path and the propagation path also includes: setting the similarity judgment threshold , the similarity determination threshold Set to upper limit of experience similarity.

[0106] Combined with the residual initial screening threshold , for nodes The abnormal type is classified and determined.

[0107] when and , it is determined to be a structural abnormal node, indicating that the node presents a consistent abnormal trend in both the internal characteristic response and external structural propagation of the model.

[0108] when and When , it is determined to be a burst abnormal node, indicating that although the node shows significant abnormality in the model, it does not form coordinated diffusion in the structure diagram.

[0109] when And when multiple adjacent nodes of the node are abnormal propagation path nodes ( ),

[0110] If it is determined to be a potential abnormal node, it means that the node is in an abnormal high-incidence environment and needs special attention. Indicates the minimum number of abnormal neighbors required to be judged as a latent anomaly, which is set to 2. Similarity judgment threshold Set to upper limit of experience similarity.

[0111] Furthermore, by introducing a similarity matching mechanism between the activation path and the propagation path, the model makes a comprehensive judgment based not only on the reconstructed residual but also on the abnormal diffusion relationship in the structural association graph when judging the abnormal state of the node. This design realizes the coordinated verification of the internal response of the model (i.e., the encoder activation behavior) and the external structural propagation trend, thereby distinguishing different types of abnormal manifestations, improving the classification accuracy of abnormal identification and the logical interpretability of model judgment.

[0112] Furthermore, by matching the activation path with the propagation path, three typical anomaly types can be effectively identified: structural anomalies, sudden anomalies, and latent anomalies, so that the system can implement differentiated processing for different risk forms during the anomaly identification process. In practical applications, this method not only improves the accuracy of anomaly detection, but also enhances the model's adaptive perception of high residual misjudgments, anomaly diffusion trends, and risk edge areas in complex scenarios, thereby providing more forward-looking and robust decision support for the monitoring system.

[0113] Embodiment 2 is an embodiment of the present invention, which provides an abnormality analysis system based on blasting data, including:

[0114] The data module collects the blasting data of each monitoring node and divides the monitoring nodes into multiple areas.

[0115] The training module counts the anomaly density of each region within a preset historical period and dynamically sets the regional weights of the autoencoder training samples based on the anomaly density.

[0116] The reconstruction module reconstructs the current round of data using the weighted trained autoencoder, calculates the residual of each node, and extracts the activation path information of the nodes whose residuals are higher than the initial screening threshold of the reconstruction error.

[0117] The propagation module builds a graph structure model between monitoring nodes and calculates the abnormal propagation path based on the correlation between nodes.

[0118] The confirmation module matches the activation path with the propagation path and calculates the similarity to determine the abnormal state of the target node.

[0119] Embodiment 3, an embodiment of the present invention, is different from the first two embodiments in that:

[0120] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk.

[0121] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0122] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0123] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0124] Example 4 is an embodiment of the present invention, which provides an abnormality analysis method and system based on blasting data. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0125] Based on the blasting operation scenario in an open-pit mine, a distributed blasting data acquisition system consisting of 36 sensor nodes was constructed, covering an area of ​​about 900 square meters. The data collected by each node include vibration signals (unit: mm / s), acoustic emission signals (unit: dB), stress data (unit: MPa) and temperature information (unit: °C), which are recorded at a frequency of 10 Hz per second to form a multi-dimensional time series data sequence. In order to ensure the breadth and objectivity of the experiment, 30 consecutive days of operation data were selected as the historical reference window, and a total of about 259,200 valid data records were monitored during the test period.

[0126] First, the Pearson correlation coefficient is used to calculate the data feature similarity between all node pairs, and a symmetric similarity matrix is ​​constructed based on this. Then, the 36 nodes are divided into 4 regions through the spectral clustering method. Subsequently, the number of abnormal nodes in each region in the historical period is counted, and the abnormality judgment standard is that the reconstruction error is higher than the historical average residual mean + 1.5 times the standard deviation. According to the abnormal density of each region, the sample weight coefficient is designed using a nonlinear function to adjust the contribution of samples in each region during the autoencoder training process.

[0127] After completing the weighted training, the autoencoder is used to reconstruct the current round of data. For each node, the residual between its original feature and the model output is calculated, and the high residual nodes whose residual is higher than the initial screening threshold T1=0.85T_1 = 0.85T1=0.85 are extracted to further analyze their activation paths. The adjacent nodes of each high residual node are extracted through the graph structure model, and the initial propagation path set is constructed. If the propagation path forms a closed subgraph structure with its adjacent nodes and second-order adjacent nodes, it is identified as a high-confidence propagation path.

[0128] In the process of propagation path screening, the inducing factor GmG_mGm is constructed at the same time, including the neighborhood anomaly density, error volatility and embedding drift degree. The standard deviation outlier detection is performed on the inducing factor to identify the latent induced abnormal nodes and merge them with the high-confidence path nodes to form the final propagation path set. Finally, the matching degree between the activation path and the propagation path is calculated based on the Jaccard similarity, and the abnormal state of the target node is classified and determined.

[0129] In the continuous sampling cycle, the simulation generated 36 nodes and a total of 3,600 sets of blasting time series data. In the current detection round, the system identified 9 high residual nodes and screened 21 candidate nodes when constructing the initial propagation path. After structural connectivity analysis, a high-confidence propagation path set of 12 nodes was formed, of which nodes with similarity values ​​greater than 0.7 accounted for about 66.7%. In the low-confidence path, based on the distribution of inducing factors, 4 potential induced abnormal nodes were identified, bringing the number of final propagation path nodes to 16.

[0130] By comparing with the average propagation pattern of the previous 30 days, it is found that the final propagation path set is more concentrated in high-density areas, and shows stronger connectivity and activation response overlap on the local subgraph. At the same time, the identification results of three types of abnormal nodes are: 4 structural anomalies, 2 sudden anomalies, and 3 latent anomalies, and the coverage rate is about 18% higher than that of traditional residual detection.

[0131] From the experimental data, the proposed autoencoder structure based on dynamic weighted training of abnormal density significantly enhances the model's ability to learn high-frequency abnormal areas. Through the regional weight mechanism, nodes in areas with higher abnormal density receive more attention during the training process, which significantly improves the recognition sensitivity of the model in the current round of detection. Compared with the traditional autoencoder with fixed sample weights, this method improves the residual mean by about 12%, but reduces the false alarm rate by about 15%, reflecting good convergence and discrimination ability.

[0132] At the same time, the closed subgraph structure judgment is introduced in the graph structure construction and propagation path evaluation, which not only ensures the structural consistency of the propagation path, but also avoids the contamination of the propagation path by "single point isolated error" during the node screening stage. By further introducing the composite inducing factor construction mechanism and using dynamic statistical judgment criteria, the limitation of the traditional method of relying on hard thresholds to identify anomalies is broken through, so that the system can identify edge nodes with potential risks, especially in multi-point high-density distribution areas, showing good foresight.

[0133] Jaccard matching is used for collaborative analysis between activation paths and propagation paths, which not only improves the interpretability of anomaly classification, but also provides a semantic basis for subsequent evolution prediction based on distribution trends. Overall, this method takes into account accuracy, stability and interpretability, with a compact technical chain. Its innovations cover multiple levels, including regional dynamic modeling, graph structure propagation construction, and intelligent judgment driven by inducing factors. It shows obvious technical advantages and application potential in real-time safety assessment in blasting operation scenarios.

[0134] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. The abnormal analysis method based on blasting data is characterized by: include, Collect blasting data from each monitoring node and divide the monitoring nodes into multiple areas; Count the abnormal density of each region within the preset historical period, and dynamically set the regional weight of the autoencoder training samples based on the abnormal density; The current round of data is reconstructed using the weighted trained autoencoder, the residual of each monitoring node is calculated, and the activation path information of the nodes whose residual is higher than the initial screening threshold of the reconstruction error is extracted; Build a graph structure model between monitoring nodes and calculate the abnormal propagation path based on the correlation between nodes; The activation path is matched with the propagation path to determine the abnormal state of the target node.

2. The abnormality analysis method based on blasting data according to claim 1, characterized in that: The blasting data includes vibration signals, acoustic emission signals, stress data and temperature; Dividing the monitoring nodes into multiple areas includes extracting time series features from the blasting data collected from each monitoring node to form a feature sequence corresponding to each node; The similarity between any two monitoring nodes is calculated using the Pearson correlation coefficient, and a similarity matrix between nodes is constructed; The similarity matrix is ​​used as an undirected graph, in which each monitoring node is a node in the graph, and the edges between nodes represent the similarity degree of data features; the undirected graph is divided into several sub-regions by using a spectral clustering algorithm, so that the monitoring nodes in each sub-region have correlation in data behavior.

3. The abnormality analysis method based on blasting data according to claim 2, characterized in that: The setting of the regional weights of the autoencoder training samples includes counting the number of abnormal nodes in each sub-region in the historical period and calculating the abnormal density of each sub-region; determining the sample weight coefficient corresponding to each sub-region based on the difference between the abnormal density of each sub-region and the average abnormal density of all sub-regions; During the training process of the autoencoder, the sample weight coefficients are applied to the training samples in different sub-regions respectively.

4. The abnormality analysis method based on blasting data according to claim 3, characterized in that: The reconstructing of the current round of data includes reconstructing the blasting data collected in the current round using the trained autoencoder and calculating the reconstruction error of each monitoring node. , the reconstruction error is the difference between the original data and the reconstructed data of the monitoring node.

5. The abnormality analysis method based on blasting data according to claim 4, characterized in that: Extracting activation path information includes determining whether the reconstruction error of each node exceeds the reconstruction error initial screening threshold , when the reconstruction error exceeds the initial screening threshold , it is determined to be a high residual node; extracting the activation path information of the high residual node from the input layer to the hidden layer in the autoencoder structure , the activation path information is the activation value sequence of each hidden layer of the high residual node in the encoding stage.

6. The abnormality analysis method based on blasting data according to claim 5, characterized in that: Constructing a graph structure model between monitoring nodes includes: the edges between nodes represent the feature similarity values ​​between any two monitoring nodes; The calculation of the abnormal propagation path according to the correlation between nodes includes: for each node determined to be a high residual node , extract the set of adjacent nodes in the graph structure ;exist In the screening, the reconstruction error also exceeds the reconstruction error initial screening threshold The nodes constitute the high residual nodes The initial abnormal propagation path node set ; When the high residual node , high residual nodes All adjacent nodes of , node collection When a closed connected subgraph is formed together, it is determined that the propagation path has stable consistency in the graph structure, and the propagation path is marked as a high-confidence propagation path; The node set For: with There is a direct connection between any node in the A collection of nodes that are not directly connected to each other; In the initial exception propagation path node set The nodes that are not included in the final high-confidence propagation path node set are set as the low-confidence propagation path candidate set ; Final abnormal propagation path node set Including, from the high confidence propagation path node set, based on each propagation node With the high residual node The feature similarity value between Sort in descending order and select the top feature similarity values The node of the position; and the low confidence propagation path candidate set Satisfaction-inducing factors identified in The latent inducing abnormal node; wherein the inducing factor It is composed of the neighborhood anomaly density, reconstruction error volatility, and the temporal drift of graph embedding features. express The mean of express The standard deviation of .

7. The abnormality analysis method based on blasting data according to claim 6, characterized in that: Determining the abnormal state of the target node includes using Jaccard similarity calculation and Similarity ; when and When , it is determined to be a structural abnormal node, indicating that the structural abnormal node presents a consistent abnormal trend in both the internal characteristic response of the model and the external structural propagation; when and When , it is determined to be a burst abnormal node, indicating that although the burst abnormal node shows significant abnormality in the model, it does not form coordinated diffusion in the structure diagram; when When multiple adjacent nodes of the node are the initial abnormal propagation path nodes, it is determined to be a latent abnormal node, indicating that the latent abnormal node is in an abnormal high-incidence environment and requires special attention.

8. An abnormality analysis system based on blasting data using the method according to any one of claims 1 to 7, characterized in that: The data module collects the blasting data of each monitoring node and divides the monitoring nodes into multiple areas; The training module counts the anomaly density of each region within a preset historical period and dynamically sets the regional weights of the autoencoder training samples based on the anomaly density; The reconstruction module reconstructs the current round of data using the weighted trained autoencoder, calculates the residual of each monitoring node, and extracts the activation path information of the nodes whose residual is higher than the initial screening threshold of the reconstruction error; The propagation module builds a graph structure model between monitoring nodes and calculates the abnormal propagation path based on the correlation between nodes; The confirmation module matches the activation path with the propagation path and calculates the similarity to determine the abnormal state of the target node.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the abnormality analysis method based on blasting data described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the abnormality analysis method based on blasting data described in any one of claims 1 to 7 are implemented.

Images