Remote monitoring method and system for fastening state of anchor bolt

Abstract

The invention discloses an anchor bolt fastening state remote monitoring method and system, and relates to the technical field of anchor bolt state monitoring based on an Internet of Things operating system, a fastening feature vector is determined by fusing sensor collection, digital conversion and preprocessing operation and removing noise, and if the feature value of the fastening feature vector exceeds a preset feature threshold value, the anchor bolt fastening state is monitored. If yes, grouping adjacent nodes to form a dynamic linkage structure, exchanging state abstracts under an edge computing framework, generating a comprehensive sensing matrix, performing identification by applying a machine learning model to obtain a fastening behavior mode of anchor bolt nodes, monitoring a data channel load according to the fastening behavior mode, and obtaining key abnormal data; according to the technical scheme, the target transmission path needing to be monitored is identified, the abnormal early warning signal is extracted, the stability and potential risk of the data flow in the target transmission path are evaluated, the systematic risk evaluation result is determined, efficient information sharing and load optimization are achieved, and the real-time performance and accuracy of structure monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of anchor bolt status monitoring technology based on Internet of Things operating system, and particularly to a method and system for remote monitoring of anchor bolt fastening status. Background Technology

[0002] Anchor bolts, as indispensable connecting components in engineering structures, play a crucial role in ensuring safety in critical infrastructure such as bridges, buildings, and railways. Their fastening condition directly affects the stability and service life of the overall structure. With the continuous expansion of infrastructure scale, the need for anchor bolt monitoring is becoming increasingly prominent, making it an important research area for ensuring public safety and engineering quality. How to achieve efficient and real-time monitoring of the status of a large number of anchor bolts is a major issue that urgently needs to be addressed.

[0003] However, existing monitoring methods often struggle to balance coverage and response speed when dealing with large-scale anchor bolt groups, especially in complex environments where insufficient coordination between data acquisition and transmission leads to low monitoring efficiency. A deeper problem lies in the lack of comprehensive analytical capabilities for the overall behavior of anchor bolt groups, making it difficult to capture the mutual influences between groups and thus failing to promptly identify potential systemic risks. This limitation renders monitoring inadequate for large-scale, multi-regional engineering scenarios.

[0004] Against this backdrop, remote monitoring of anchor bolt fastening status faces significant technical challenges. The primary difficulty lies in constructing a network architecture capable of covering multiple anchor bolt nodes and achieving information sharing. Due to the widespread distribution and complex environment of anchor bolts, monitoring information from a single node needs to be linked with other nodes to form a dynamic perception of the overall status. Insufficient linkage directly leads to information silos. Furthermore, these information silos exacerbate another critical issue: the pressure on data processing and transmission. In large-scale anchor bolt clusters, if all data relies on remote transmission without local processing capabilities, network congestion will occur, potentially affecting the real-time nature of monitoring. For example, on a large bridge, thousands of anchor bolts are distributed in different areas. If the status data of each anchor bolt is directly uploaded to a central server, not only will the transmission efficiency be low, but network latency may also cause missed early warning opportunities for critical anomalies.

[0005] Therefore, how to achieve efficient information sharing among nodes in a widely distributed group of anchor bolts, while balancing the pressure of data processing and transmission, has become a key problem that this research urgently needs to solve. Solving this problem will directly affect the reliability and response speed of engineering structure safety monitoring, and is of great significance for ensuring the long-term stability of infrastructure. Summary of the Invention

[0006] This invention provides a method and system for remote monitoring of anchor bolt fastening status, so as to achieve efficient information sharing between nodes in a widely distributed anchor bolt group, while balancing the pressure of data processing and transmission.

[0007] This invention provides a method for remote monitoring of anchor bolt fastening status, comprising: Data on the anchor bolt fastening status is collected by sensors deployed on the anchor bolt structure nodes, and the anchor bolt fastening status data is converted into digital format to obtain the initial state dataset of the anchor bolt structure nodes. Based on the initial state dataset, preprocessing operations are performed in the local processor, noise removal methods are used to remove interference, and tight feature vectors are determined. If the eigenvalue of the fastening feature vector exceeds the preset feature threshold, then the adjacent anchor bolt structure nodes are grouped by a grouping algorithm to obtain a dynamic linkage structure containing anchor bolt structure groups. Based on the shared information of the structural nodes within the anchor bolt structure group in the dynamic linkage structure, a state summary required for the anchor bolt structure nodes is generated, and the state summary is exchanged under the edge computing framework to generate a comprehensive perception matrix after information sharing. Based on the comprehensive perception matrix, a machine learning model is applied to perform feature extraction and pattern recognition to obtain the fastening behavior pattern describing the anchor bolt structure group. Based on the fastening behavior pattern, the data channel load is monitored. If the data channel load is lower than a preset load threshold, key abnormal data is extracted from the data channel load and uploaded to the server to obtain the target transmission path that needs to be monitored in real time. Based on the target transmission path, an anomaly warning signal is obtained, and based on the anomaly warning signal, the stability and potential risks of the data flow in the target transmission path are assessed to determine the systemic risk assessment result.

[0008] According to the remote monitoring method for anchor bolt fastening status of the present invention, the step of performing a preprocessing operation on a local processor based on the initial state dataset, removing interference using a noise removal method, and determining the fastening feature vector includes: Based on the initial state dataset, potential interference signals present in the initial state dataset are preliminarily screened to determine the basic data set after preliminary screening. Based on the initial filtered basic data set, a preprocessing operation is performed in the local processor, and a noise removal method is used for smoothing to obtain a smoothed data set after removing interference. Based on the smoothed data set, the data points in the smoothed data set are normalized to obtain a standardized data matrix; Based on the standardized data matrix, core data elements related to fastening features are extracted. If the distribution of the core data elements is within a preset feature range, then the core data elements within the preset feature range are taken as a valid feature data group. Based on the effective feature data set, the corresponding feature vector representation is constructed using a vector mapping method to obtain the fastening feature vector.

[0009] According to the remote monitoring method for anchor bolt fastening status of the present invention, if the feature value of the fastening feature vector exceeds a preset feature threshold, then adjacent anchor bolt structure nodes are grouped by a grouping algorithm to obtain a dynamic linkage structure containing anchor bolt structure groups, including: If the noise level of the fastening feature vector exceeds a preset range, a noise reduction method is used to clean the fastening feature vector to obtain a cleaned feature vector. If the purified feature vector exceeds the preset feature threshold, then the adjacent anchor bolt structure nodes are clustered using a grouping algorithm to determine the correlation between the anchor bolt structure nodes and form an anchor bolt structure group. Based on the anchor bolt structure group, the dynamic linkage relationship between the anchor bolt structure nodes is identified, wherein the dynamic linkage relationship includes the interaction mode of each anchor bolt structure node within the group; Based on the dynamic linkage relationship, the node association strength within the anchor bolt structure group is analyzed to obtain the association strength distribution within the anchor bolt structure group; Based on the correlation strength distribution, the dynamic linkage mode of the anchor bolt structure group is determined, and the dynamic linkage structure of the anchor bolt structure group is obtained.

[0010] According to the remote monitoring method for anchor bolt fastening status of the present invention, the step of generating a state summary required for the anchor bolt structure nodes based on the shared information of the structural nodes within the anchor bolt structure group in the dynamic linkage structure, exchanging the state summaries under an edge computing framework, and generating a comprehensive sensing matrix after information sharing includes: Based on the dynamic linkage structure, the shared information of the structural nodes within the anchor bolt structure group is obtained, and the shared information is initially classified using a distributed processing method to obtain a preliminarily sorted data set. Based on the pre-organized data set, to meet the interaction requirements between local nodes of the edge computing framework, edge computing technology is used to localize the pre-organized data set and generate a state summary. If the state digest meets the preset integrity conditions, the state digest is distributed to the corresponding group nodes of the edge computing framework through the framework transmission mechanism to obtain the distributed state information. Based on the distributed state information, a matrix-based processing method is used to construct an intermediate layer data structure for information sharing and determine the preliminary matrix for comprehensive perception. Based on the preliminary matrix of the comprehensive perception, the data dimension is reduced by applying the principal component analysis algorithm to obtain the comprehensive perception matrix.

[0011] According to the remote monitoring method for anchor bolt fastening status of the present invention, the step of using a machine learning model to extract features and recognize patterns based on the comprehensive sensing matrix to obtain a fastening behavior pattern describing the anchor bolt structure group includes: Based on the comprehensive perception matrix, a standardized matrix dataset is obtained by cleaning and standardizing the data using data preprocessing tools. Based on the standardized matrix dataset, a machine learning model is applied to perform feature extraction and pattern recognition to determine the distribution set of fastening behavior features; Based on the distribution set of the fastening behavior characteristics, the relationship between the characteristics is identified through the feature association analysis module to determine the connection strength within the anchor bolt structure group; If the connection strength exceeds a preset strength threshold, the fastening behavior is classified and mapped based on the distribution set of the fastening behavior characteristics to obtain a preliminary classification result of the fastening behavior pattern. Based on the preliminary classification results of the fastening behavior patterns, the fastening behavior patterns are integrated and optimized through the pattern construction module to determine the fastening behavior patterns.

[0012] According to the remote monitoring method for anchor bolt fastening status of the present invention, the step of monitoring the data channel load based on the fastening behavior pattern, and if the data channel load is lower than a preset load threshold, extracting key abnormal data from the data channel load and uploading the key abnormal data to the server to obtain the target transmission path that needs to be monitored in real time, includes: Based on the aforementioned fastening behavior pattern, the load of the data channel is continuously collected by the monitoring system to obtain the data channel load characterizing the load change. If the load of the data channel is lower than a preset load threshold, an abnormal data filtering mechanism is triggered to extract key abnormal data from the data channel load and to prioritize and mark the key abnormal data to determine the abnormal content of the priority marking. Based on the abnormal content marked with the priority, an optimization protocol is used to compress the data packet to obtain the compressed data packet. The compressed data packet is uploaded to the server through a pre-established transmission path, and a confirmation signal indicating that the upload is complete is obtained. Upon receiving the confirmation signal, the compressed data packet is decompressed and parsed to obtain the restored abnormal data content; Based on the restored abnormal data content, the real-time monitoring module performs data comparison and status updates to determine the target transmission path that needs to be monitored in real time.

[0013] According to the remote monitoring method for anchor bolt fastening status of the present invention, the step of acquiring an abnormal early warning signal based on the target transmission path, and assessing the stability and potential risks of the data stream in the target transmission path based on the abnormal early warning signal to determine the systemic risk assessment result includes: Data stream information is obtained from the target transmission path, and based on the data stream information, possible abnormal fluctuations are filtered to obtain an abnormal warning signal; Based on the abnormal early warning signal and combined with the interaction records between the anchor bolt structure nodes, the abnormal propagation path in the node linkage is analyzed to determine the chain reaction mode. Based on the chain reaction mode, the linkage parameters between the anchor bolt structure nodes are adaptively adjusted to obtain the optimized parameter configuration. Based on the optimized parameter configuration, the stability and potential risks of the data flow in the target transmission path are assessed, and the systemic risk assessment results are determined.

[0014] The present invention also provides a remote monitoring system for anchor bolt fastening status, comprising: The status data acquisition module is used to collect anchor bolt fastening status data through sensors deployed on the anchor bolt structure node, convert the anchor bolt fastening status data into digital format, and obtain the initial status dataset of the anchor bolt structure node. The feature vector determination module is used to perform preprocessing operations on the local processor based on the initial state dataset, remove interference using a noise removal method, and determine the compact feature vector; The grouping module is used to group adjacent anchor bolt structure nodes by a grouping algorithm if the feature value of the fastening feature vector exceeds a preset feature threshold, so as to obtain a dynamic linkage structure containing anchor bolt structure groups. The matrix generation module is used to generate the state summary required by the anchor structure node based on the shared information of the structural nodes in the anchor structure group in the dynamic linkage structure, and to exchange the state summary under the edge computing framework to generate a comprehensive perception matrix after information sharing. The behavior pattern recognition module is used to perform feature extraction and pattern recognition based on the comprehensive perception matrix and apply a machine learning model to obtain the fastening behavior pattern describing the anchor bolt structure group. The transmission path monitoring module is used to monitor the data channel load based on the tightening behavior pattern. If the data channel load is lower than a preset load threshold, key abnormal data is extracted from the data channel load and uploaded to the server to obtain the target transmission path that needs to be monitored in real time. The risk assessment module is used to obtain abnormal early warning signals based on the target transmission path, and to assess the stability and potential risks of the data flow in the target transmission path based on the abnormal early warning signals, so as to determine the systemic risk assessment results.

[0015] This invention discloses a remote monitoring method and system for anchor bolt fastening status. Addressing the unique challenges of distributed structures where fastening status data acquisition is susceptible to noise interference, insufficient inter-node linkage, poor information sharing, and high data transmission load leading to delays in real-time risk assessment, the method integrates sensor acquisition, digital conversion, and preprocessing to remove noise and determine fastening feature vectors. If the feature value of the fastening feature vector exceeds a preset feature threshold, adjacent nodes are grouped to form a dynamic linkage structure. Subsequently, status summaries are exchanged within an edge computing framework to generate a comprehensive perception matrix. Machine learning models are applied for feature extraction and behavior pattern recognition to identify the fastening behavior patterns of anchor bolt nodes. Based on these patterns, the data channel load is monitored, key anomaly data is obtained from the data channel load, target transmission paths requiring monitoring are identified, and anomaly warning signals are extracted. The stability and potential risks of the data flow in the target transmission path are assessed, ultimately determining the systemic risk assessment result. This effectively solves the problems of noise interference and insufficient linkage, achieves efficient information sharing and load optimization, improves the real-time performance and accuracy of structural monitoring, significantly reduces risk assessment delays, and enhances the overall system robustness. This invention enables efficient information sharing among nodes in a widely distributed group of anchor bolts, while balancing the pressure of data processing and transmission. Attached Figure Description

[0016] Figure 1 This is one of the flowcharts of the remote monitoring method for anchor bolt fastening status provided in the embodiments of the present invention; Figure 2 This is the second flowchart of the remote monitoring method for anchor bolt fastening status provided in this embodiment of the invention; Figure 3 This is the third flowchart of the remote monitoring method for anchor bolt fastening status provided in this embodiment of the invention; Figure 4 This is the fourth flowchart of the remote monitoring method for anchor bolt fastening status provided in this embodiment of the invention; Figure 5 This is the fifth flowchart of the remote monitoring method for anchor bolt fastening status provided in this embodiment of the invention; Figure 6 This is the sixth flowchart of the remote monitoring method for anchor bolt fastening status provided in this embodiment of the invention; Figure 7 This is the seventh flowchart of the remote monitoring method for anchor bolt fastening status provided in this embodiment of the invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Reference Figure 1 This invention provides a method for remote monitoring of anchor bolt fastening status, comprising the following steps: Step 100: Collect anchor bolt fastening status data by sensors deployed on the anchor bolt structure node, convert the anchor bolt fastening status data into digital format, and obtain the initial status dataset of the anchor bolt structure node. Sensors deployed on anchor bolt structural nodes collect characteristic data reflecting the actual fastening state of the anchor bolts. These sensors, serving as a crucial interface between physical and digital information, are used at key nodes most sensitive to changes in structural mechanical behavior to continuously monitor multidimensional physical parameters such as preload changes, vibration spectrum characteristics, structural micro-strain, and acoustic emission signals. The sensors capture these dynamic mechanical state signals through their built-in sensitive elements and convert them into corresponding continuous analog electrical signals, thus completing the mapping from the physical world to the electrical signal domain and constructing a quantitative assessment of the anchor bolt structural health status.

[0019] Subsequently, an integrated high-precision analog-to-digital converter module converts the analog electrical signal output by the sensor into a discrete data format suitable for digital system processing. This achieves the transformation of the signal from the continuous time domain to a discrete digital sequence, effectively eliminating the drawback of analog signals being susceptible to interference during transmission. After processing steps such as sampling, quantization, and encoding, the original mechanical state information is transformed into a timestamped digital sequence, resulting in an initial state dataset reflecting the fastening state of the anchor bolt structure nodes at a specific time point. This initial state dataset contains key feature information about the anchor bolt fastening state.

[0020] Step 200: Based on the initial state dataset, perform preprocessing operations in the local processor, use noise removal methods to remove interference, and determine the compact feature vector; After successfully constructing the initial state dataset, the next stage is data refinement and feature extraction. This step is completed in a locally deployed embedded processor to cleanse the initial dataset, which contains various environmental noises and measurement interferences. Since the raw signals acquired in the field inevitably contain electromagnetic interference, mechanical vibration background noise, and transient pulses—non-target information—these interfering components severely obscure the effective features reflecting the true fastening state of the anchor bolts. Therefore, advanced digital signal processing algorithms are employed as noise removal methods. Through techniques such as frequency domain filtering, wavelet threshold denoising, or adaptive filtering, these random fluctuations and errors unrelated to the essence of anchor bolt fastening are identified and removed, thereby improving the signal-to-noise ratio. After data purification, core information characterizing the anchor bolt fastening health status is identified from the purified smoothed signal, thereby determining the fastening feature vector. Through various algorithms such as time-domain analysis, frequency-domain analysis, or joint time-frequency analysis, statistical features such as mean, variance, peak value, and kurtosis, or dynamic features such as resonant frequency, spectral energy distribution, and damping ratio, are extracted from the waveform signal. These feature parameters, which characterize the anchor bolt's mechanical behavior from different dimensions, are organically combined and encapsulated to obtain a multi-dimensional fastening feature vector. This fastening feature vector transforms complex time-series data into a standardized mathematical vector that can be directly recognized and processed by machine learning models.

[0021] Step 300: If the feature value of the fastening feature vector exceeds the preset feature threshold, then the adjacent anchor bolt structure nodes are grouped by a grouping algorithm to obtain a dynamic linkage structure containing the anchor bolt structure group. After obtaining the fastening feature vector, the key feature values ​​in the fastening feature vector are compared in real time with preset feature thresholds that have been calibrated through extensive experiments and mechanical models. These preset feature thresholds represent the critical point at which the anchor bolt transitions from a securely fastened state to a potentially loose or failed state. When the feature value of an anchor bolt node exceeds the preset feature threshold, the anchor bolt node is marked as an abnormal state, triggering the structural correlation analysis process. This signifies that monitoring has shifted from judging the isolated state of a single node to evaluating the overall collaborative performance of the structure.

[0022] At this point, to assess the potential impact of the anomalous node on the overall structural stability, a grouping algorithm is used. Based on the mechanical transmission path of the structure and the dynamic response correlation between nodes, and comprehensively considering the structural topology, load distribution, and the consistency of vibration modes exhibited in historical data, adjacent anchor bolt structural nodes that are mechanically closely coupled with the initial anomalous node and exhibit significant linkage effects are identified. One or more anchor bolt structural groups are generated and defined as dynamic linkage structures. These dynamic linkage structures reflect the force flow transmission path in the structure under mechanical loads and the dynamic dependence of mutual constraints and influences between nodes.

[0023] Step 400: Based on the shared information of the structural nodes in the anchor structure group in the dynamic linkage structure, generate the state summary required by the anchor structure node, and exchange the state summary under the edge computing framework to generate a comprehensive perception matrix after information sharing. After constructing a dynamic linkage structure that reflects mechanical relationships, a state summary for each member node in the dynamic linkage structure is first generated. The core information selected from multiple dimensions such as the historical state, real-time feature vector, health score, and behavioral consistency with other nodes in the group carries the metadata for evaluating the state of the anchor bolt structure node.

[0024] Subsequently, this process autonomously and in parallel at the edge, within the edge computing framework, involves the exchange of state summaries between adjacent nodes. This avoids the latency and bandwidth pressure associated with uploading massive amounts of raw data to the cloud, achieving real-time performance. Through this point-to-point, distributed information interaction, each anchor structure node not only knows its own state but also instantly acquires the health status of its mechanically related anchors, thus forming a localized, shared sensing network.

[0025] Finally, all the summary information exchanged within the anchor bolt group is gathered and integrated into a comprehensive sensing matrix. This comprehensive sensing matrix is ​​a structured data model in which rows and columns correspond to different nodes in the group and their multi-dimensional status indicators. It includes real-time readings of each independent anchor bolt structural node. Through the correlation and comparison between data, the overall health status of the entire group as a dynamic and interconnected structure, the balance of load distribution, and the transmission path of potential risks can be revealed.

[0026] Step 500: Based on the comprehensive perception matrix, a machine learning model is applied to perform feature extraction and pattern recognition to obtain a fastening behavior pattern describing the anchor bolt structure group. After constructing a comprehensive perception matrix that can reflect the internal state of the anchor bolt structure group in a panoramic way, the multidimensional and correlated information contained in the comprehensive perception matrix is ​​given to a well-trained machine learning model for analysis. The entire matrix is ​​processed as a complete digital image representing the instantaneous state of the structural system. Through a complex and nonlinear internal computing network, deep feature extraction is automatically performed from the matrix. This process can uncover high-order statistical laws and complex correlation features hidden within the data that surpass human expert experience.

[0027] Furthermore, the machine learning model leverages its powerful pattern recognition capabilities to differentiate and categorize these extracted deep features. It compares and matches the data patterns presented by the comprehensive perception matrix with the massive historical data patterns learned during its training phase, including normal tightness, varying degrees of looseness, and typical response patterns under various loading conditions. This comparison is a probabilistic, similarity-based reasoning process aimed at identifying the inherent patterns that best represent the overall behavior of the current group from the complex fluctuations in data. Ultimately, the output of this analysis process is a tightness behavior pattern that reflects how the entire anchor group works collaboratively as a dynamically linked structure at the current moment. For example, it might identify a collaborative pattern where loosening of the central node leads to a uniform increase in load on surrounding nodes, or an evolutionary pattern where some nodes within the group experience synchronous stiffness degradation due to fatigue.

[0028] Step 600: Monitor the data channel load based on the fastening behavior pattern. If the data channel load is lower than a preset load threshold, extract key abnormal data from the data channel load and upload the key abnormal data to the server to obtain the target transmission path that needs to be monitored in real time. After identifying the fastening behavior patterns of anchor bolt structure groups using a machine learning model, the load on the current data channel used for data transmission is dynamically evaluated based on the identified patterns. Key performance indicators such as instantaneous bandwidth, data throughput, and transmission latency are comprehensively considered to ensure the efficiency and stability of the communication link. When the current data channel load is determined to be below a preset load threshold, indicating sufficient redundancy in the communication link, instead of indiscriminately uploading all raw data or all feature vectors, a data filtering and priority transmission mechanism is initiated to extract key anomaly data from the rich data stored locally. This key anomaly data includes feature vector fragments that lead to the currently identified abnormal fastening behavior patterns, key time-series data that trigger alarms, and a quantitative assessment summary of the severity and development trend of the anomaly pattern. This extraction process significantly reduces the data volume while ensuring information fidelity.

[0029] Subsequently, uploading the abnormal data to a remote server alleviates the long-term average pressure on network bandwidth, allowing limited communication resources to serve a wider range of monitoring nodes; and ensures that the cloud server can receive fault warning information in real time. Through this series of judgments and screenings, an efficient, reliable target transmission path that requires real-time monitoring was obtained.

[0030] Step 700: Obtain an anomaly warning signal based on the target transmission path, and assess the stability and potential risks of the data flow in the target transmission path based on the anomaly warning signal to determine the systemic risk assessment result.

[0031] After establishing the target transmission path and starting to receive abnormal early warning signals, a comprehensive assessment of the operational health and risks of the entire monitoring system is conducted. Continuous monitoring of the target transmission path is initiated to evaluate the stability of the data flow, including key communication indicators such as transmission delay, packet loss rate, and signal integrity.

[0032] Subsequently, a fusion analysis process is conducted, combining abnormal early warning signals of the physical state with data flow stability assessments at the digital level. This aims to estimate the potential risks faced by the entire monitoring system, determine the severity of structural consequences that the abnormal behavior patterns of the current anchor bolt group may cause, and consider the instability of the data transmission link. For example, a moderately severe abnormal signal, if superimposed on a highly unstable transmission path, may be judged to have a higher overall risk level because the system cannot guarantee the effective delivery of subsequent early warning information. Finally, by combining the abnormal early warning signals and data flow stability assessments, a quantitative systemic risk assessment result is determined. This systemic risk assessment result is a global diagnostic report of structural state risk and communication risk, which includes the risk outcome of whether the anchor bolt structural nodes are safe, as well as the monitoring and perception of the structure's safety.

[0033] This invention discloses a remote monitoring method and system for anchor bolt fastening status. Addressing the unique challenges of distributed structures where fastening status data acquisition is susceptible to noise interference, insufficient inter-node linkage, poor information sharing, and high data transmission load leading to delays in real-time risk assessment, the method integrates sensor acquisition, digital conversion, and preprocessing to remove noise and determine fastening feature vectors. If the feature value of the fastening feature vector exceeds a preset feature threshold, adjacent nodes are grouped to form a dynamic linkage structure. Subsequently, status summaries are exchanged within an edge computing framework to generate a comprehensive sensing matrix. Machine learning models are applied for feature extraction and behavioral pattern recognition to identify the fastening behavior patterns of anchor bolt nodes. Based on these patterns, the data channel load is monitored, key anomaly data is obtained from the data channel load, the target transmission path requiring monitoring is identified, and anomaly warning signals are extracted. The stability and potential risks of the data flow in the target transmission path are assessed, ultimately determining the systemic risk assessment result. This effectively solves the problems of noise interference and insufficient linkage, achieves efficient information sharing and load optimization, improves the real-time performance and accuracy of structural monitoring, significantly reduces risk assessment delays, and enhances the overall system robustness. This invention enables efficient information sharing among nodes in a widely distributed group of anchor bolts, while balancing the pressure of data processing and transmission.

[0034] In one embodiment, please refer to Figure 2The step of performing preprocessing operations on the local processor based on the initial state dataset, removing interference using noise removal methods, and determining the compact feature vector includes: Step 201: Based on the initial state dataset, perform preliminary screening of potential interference signals in the initial state dataset to determine the basic data set after preliminary screening; Step 202: Based on the basic data set after preliminary screening, perform preprocessing operations in the local processor, and use noise removal methods to smooth the data set to obtain a smoothed data set after removing interference. Step 203: Based on the smoothed data set, perform a normalization operation on the data points of the smoothed data set to obtain a standardized data matrix; Step 204: Based on the standardized data matrix, extract the core data elements related to the fastening features. If the distribution of the core data elements is within a preset feature range, then the core data elements within the preset feature range are taken as a valid feature data group. Step 205: Based on the effective feature data set, construct the corresponding feature vector representation using a vector mapping method to obtain the fastening feature vector.

[0035] In the data preprocessing and feature extraction stage, based on the initial state dataset, various atypical fluctuations or potential interference signals that significantly deviate from the normal response range are initially screened to identify and eliminate significant abnormal data points caused by instantaneous sensor misreading, sudden environmental interference, or other accidental factors, thereby determining a reliable preliminary screening base dataset. Subsequently, based on this base dataset, in-depth preprocessing operations are performed on the local processor. Advanced noise removal methods are used to smooth the data sequence, effectively filtering out high-frequency random fluctuations and background noise while retaining low-frequency trends and key signal characteristics that reflect the true mechanical state of the anchor bolts. This results in a smoothed dataset after interference removal, improving the signal-to-noise ratio and quality of the data.

[0036] Next, each data point in the smoothed dataset is normalized to eliminate inconsistencies in data dimensions and numerical ranges caused by individual sensor differences, initial installation conditions, or changes in environmental benchmarks. The data in the smoothed dataset is transformed to a unified scale, generating a standardized data matrix. This ensures that feature parameters from different sources and of different magnitudes can be compared and fused during subsequent feature extraction.

[0037] Based on this standardized data matrix, core data elements most closely related to the anchor bolt fastening status are further extracted. These core data elements, selected through theoretical analysis and experimental verification, are key parameters that most sensitively reflect changes in preload and connection status. Their distribution is then assessed. If the values ​​of these core data elements fall within a preset characteristic range based on safe operating conditions, they are considered valid feature data sets representing a normal or controllable state. Finally, based on these valid feature data sets, a specific vector mapping method is used to construct a mathematically unified feature vector representation, resulting in a fastening feature vector. This fastening feature vector comprehensively describes the current fastening status of the anchor bolt, thus providing a fastening feature vector for advanced analysis and pattern recognition.

[0038] In this embodiment, through a multi-level data processing flow, data refinement and feature construction are achieved, a standardized generation process from raw data to feature vectors is established, and multi-dimensional features are constructed into a single vector. This compresses the data volume to suit edge computing and transmission, and provides standardized and optimized input for subsequent machine learning model learning.

[0039] In one embodiment, please refer to Figure 3 If the eigenvalue of the fastening feature vector exceeds a preset feature threshold, then adjacent anchor bolt structure nodes are grouped using a grouping algorithm to obtain a dynamic linkage structure containing anchor bolt structure groups, including: Step 301: If the noise level of the fastening feature vector exceeds a preset range, a noise reduction method is used to clean the fastening feature vector to obtain a cleaned feature vector. Step 302: If the purified feature vector exceeds the preset feature threshold, then the adjacent anchor bolt structure nodes are clustered using a grouping algorithm to determine the correlation between the anchor bolt structure nodes and form an anchor bolt structure group. Step 303: Based on the anchor bolt structure group, identify the dynamic linkage relationship between the anchor bolt structure nodes, wherein the dynamic linkage relationship includes the interaction mode of each anchor bolt structure node within the group; Step 304: Based on the dynamic linkage relationship, analyze the node association strength within the anchor bolt structure group to obtain the association strength distribution within the anchor bolt structure group; Step 305: Based on the correlation strength distribution, determine the dynamic linkage mode of the anchor bolt structure group to obtain the dynamic linkage structure of the anchor bolt structure group.

[0040] First, the quality of the constructed compact feature vector is evaluated. If the noise level of the feature vector exceeds the preset allowable range, it indicates that the data may be affected by interference that has not been completely eliminated. At this time, a denoising process is performed to clean the feature vector itself. The denoising process identifies and suppresses random fluctuations and abnormal frequency points, thereby obtaining a cleaned feature vector with a significantly improved signal-to-noise ratio and more reliable data quality.

[0041] Next, the purified feature vector is compared with a preset feature threshold. If the feature value of the purified feature vector exceeds the safety threshold, a grouping algorithm is triggered. This algorithm clusters adjacent anchor bolt structural nodes to analyze the similarity and correlation of each node in terms of mechanical response, vibration mode, and other characteristics. This determines the inherent correlation between anchor bolt structural nodes based on common mechanical behavior, forming an anchor bolt structural group with collaborative working characteristics. Based on this, the complex interactions within the anchor bolt structural group are further explored. By analyzing time-series data and response modes, the dynamic linkage between anchor bolt structural nodes is identified. This dynamic linkage reveals the complex interaction patterns of how nodes within the group influence each other, coordinate deformation, or transfer stress under external loads, reflecting the overall mechanical behavior of the structural system.

[0042] Subsequently, based on the identified dynamic linkage relationships, the correlation strength between nodes within the anchor bolt structure group was quantitatively analyzed. By calculating indicators such as the correlation coefficient of characteristic responses between nodes, the tightness of the connection between each pair of nodes was evaluated, thereby obtaining the relational topology and correlation strength distribution within the anchor bolt structure group. Finally, by synthesizing the correlation strength distribution, the dynamic linkage pattern exhibited by the anchor bolt structure group was determined through pattern recognition and structural dynamics analysis. The dynamic linkage pattern describes the collaborative working mechanism of the anchor bolt structure group as a whole under stress, and defining the dynamic linkage pattern as the dynamic linkage structure of the anchor bolt structure group provides a key model for understanding the overall behavior of the system.

[0043] In this embodiment, functional grouping based on actual mechanical behavior is achieved through correlation analysis using a clustering algorithm, making group division more accurate. Furthermore, through quantitative analysis of dynamic linkage relationships and correlation strength distribution, the abstract concept of structural linkage is transformed into an analyzable dynamic linkage structure.

[0044] In one embodiment, please refer to Figure 4 The process involves generating a state summary of the anchor structure nodes based on shared information within the anchor structure group in the dynamic linkage structure, exchanging the state summaries under an edge computing framework, and generating a comprehensive perception matrix after information sharing, including: Step 401: Based on the dynamic linkage structure, obtain the shared information of the structural nodes in the anchor bolt structure group, and use a distributed processing method to perform preliminary classification of the shared information to obtain a preliminary sorted data set; Step 402: Based on the pre-organized data set, for the interaction requirements between local nodes of the edge computing framework, the pre-organized data set is localized using edge computing technology to generate a state summary. Step 403: If the state digest meets the preset integrity conditions, the state digest is distributed to the corresponding group nodes of the edge computing framework through the framework transmission mechanism to obtain the distributed state information. Step 404: Based on the distributed state information, construct the intermediate layer data structure for information sharing using a matrix processing method, and determine the preliminary matrix for comprehensive perception; Step 405: Based on the preliminary matrix of the comprehensive perception, the principal component analysis algorithm is applied to reduce the dimensionality of the data to obtain the comprehensive perception matrix.

[0045] First, based on the established dynamic linkage structure, the shared state information of all structural nodes within the anchor bolt structure group is obtained, resulting in shared information. This shared information includes real-time feature vectors, historical behavior records, and local diagnostic results of the anchor bolt structure nodes. Then, a distributed processing approach is used to initially classify and integrate this heterogeneous shared information, classifying it according to its data type, physical meaning, and urgency, forming a pre-organized data set. Subsequently, to meet the requirements of efficient and low-latency interaction between local nodes under the edge computing framework, edge computing technology is used to localize the pre-organized data set at the network edge. This process is not a simple data forwarding, but rather, at each node close to the data source, the content most representative of its core state is extracted from the complex local data to generate a state summary, greatly reducing the amount of data to be transmitted.

[0046] Next, the generated state digest is verified. If it meets the preset conditions for data integrity and consistency, the framework transmission mechanism is triggered, reliably distributing the state digests of each node to all other group nodes corresponding to this dynamic linkage structure within the edge computing framework. Through this step, each node can obtain its own state, as well as the state of its associated nodes, resulting in a distributed state information that is fully shared within the group.

[0047] Based on the distributed state information, a matrix-based intermediate data structure for information sharing is constructed. This intermediate data structure aligns and organizes the multidimensional state information of different nodes in a standardized format, forming a preliminary comprehensive perception matrix reflecting the overall state of the group at the current moment. Finally, this preliminary comprehensive perception matrix may have high dimensionality and information redundancy. Principal component analysis and other data dimensionality reduction algorithms are applied to extract features and condense information, retaining the most critical variance information of the group state and eliminating redundant and noisy dimensions, ultimately obtaining a comprehensive perception matrix with significantly improved information density.

[0048] In this embodiment, the distributed processing capabilities of the edge computing framework enable local data interaction, significantly improving real-time performance and robustness. Furthermore, by using a comprehensive perception matrix, the complex structural group states are transformed into a mathematical representation that can be directly and efficiently processed by machine algorithms, thereby improving data processing efficiency.

[0049] In one embodiment, please refer to Figure 5 Based on the comprehensive perception matrix, a machine learning model is applied to perform feature extraction and pattern recognition to obtain a fastening behavior pattern describing the anchor bolt structure group, including: Step 501: Based on the comprehensive perception matrix, clean and standardize the data using a data preprocessing tool to obtain a standardized matrix dataset; Step 502: Based on the normalized matrix dataset, apply a machine learning model to perform feature extraction and pattern recognition to determine the distribution set of fastening behavior features; Step 503: Based on the distribution set of the fastening behavior characteristics, the relationship between the characteristics is identified through the feature association analysis module to determine the connection strength within the anchor bolt structure group; Step 504: If the connection strength exceeds a preset strength threshold, then based on the distribution set of the fastening behavior characteristics, the fastening behavior is classified and mapped to obtain a preliminary classification result of the fastening behavior pattern. Step 505: Based on the preliminary division results of the fastening behavior rules, the fastening behavior rules are integrated and optimized through the pattern construction module to determine the fastening behavior pattern.

[0050] The constructed integrated sensing matrix is ​​processed by using data preprocessing tools to clean up any missing values ​​and outliers that may exist in the matrix, and the values ​​in the matrix are standardized to eliminate analytical biases caused by different units, ultimately resulting in a standardized matrix dataset.

[0051] Next, the normalized matrix dataset is input into a pre-trained machine learning model. This model, utilizing its deep neural network structure, automatically extracts key features characterizing the working state of the anchor bolt group from the matrix data. Through its powerful pattern recognition capabilities, it identifies the underlying regularities in these features, thus determining a comprehensive feature distribution set describing the group's fastening behavior. The feature correlation analysis module analyzes the interdependencies between elements within this fastening behavior feature distribution set. By calculating statistical indicators such as correlation and covariance between features, the degree of common influence of different features on the group's state can be quantified, thereby determining the strength of the intrinsic mechanical connections between nodes within the anchor bolt structure group.

[0052] The identified connection strength is compared with a preset strength threshold. When the connection strength exceeds the preset threshold, it indicates the existence of significant collaborative behavior patterns within the group. Based on the distribution set of sticky behavior features, intelligent algorithms such as cluster analysis are used to automatically classify and map complex sticky behaviors, thus obtaining preliminary classification results of different behavioral patterns within the group and achieving a preliminary structured understanding of group behavior. Finally, the pattern construction module further integrates and optimizes the preliminary classification results. This module eliminates contradictions and redundancies in the classification results, merges similar behavioral categories, and extracts core behavioral features to ultimately determine the sticky behavior patterns.

[0053] This embodiment establishes an end-to-end automated analysis process from data cleaning to pattern construction, transforming complex structural behavior analysis into a standardized and repeatable computational process. Secondly, by introducing a dual analysis mechanism combining machine learning and feature association analysis, it is possible to identify the feature distribution on the surface and uncover hidden causal relationships between features and dynamic connections within groups, thereby improving the accuracy of state diagnosis.

[0054] In one embodiment, please refer to Figure 6 The monitoring of data channel load based on the tightening behavior pattern, if the data channel load is lower than a preset load threshold, extracts key abnormal data from the data channel load and uploads the key abnormal data to the server to obtain the target transmission path that needs to be monitored in real time, including: Step 601: Based on the fastening behavior pattern, continuously collect data channel load information through the monitoring system to obtain the data channel load characterizing the load change. Step 602: If the data channel load is lower than a preset load threshold, an abnormal data filtering mechanism is triggered to extract key abnormal data from the data channel load and to prioritize the key abnormal data to determine the abnormal content of the priority-marked data. Step 603: Based on the abnormal content marked with the priority, compress it using an optimization protocol to obtain a compressed data packet; Step 604: Upload the compressed data packet to the server through the pre-established transmission path and obtain a confirmation signal indicating that the upload is complete; Step 605: Upon receiving the confirmation signal, the compressed data packet is decompressed and parsed to obtain the restored abnormal data content; Step 606: Based on the restored abnormal data content, and in conjunction with the real-time monitoring module, perform data comparison and status updates to determine the target transmission path that needs to be monitored in real time.

[0055] Based on identified tight behavior patterns, a load sensing module integrated into the monitoring network continuously and dynamically collects data channel load data. It tracks key indicators such as network bandwidth utilization, data transmission rate, and queue latency in real time to obtain accurate data channel load information characterizing the real-time status of the communication link. When the current data channel load is determined to be below a preset load threshold, an anomaly data filtering mechanism is immediately triggered. This mechanism extracts anomalies from the current load data stream that match the identified tight behavior patterns—these are identified as critical anomalies. Subsequently, based on the type, severity, and development trend of the anomalies, the critical anomalies are given multi-level priority markings to determine the anomalies with priority markings, ensuring that risk signals receive the most timely response.

[0056] Based on the prioritized abnormal data, an optimized protocol designed for industrial monitoring data can be used for compression. This process eliminates redundant information to the greatest extent possible while ensuring the core information remains intact, ultimately generating a compressed data packet that is easy to transmit quickly. This compressed data packet is then uploaded to a remote server via a pre-established, optimized, and reliable transmission path. The upload process is continuously monitored until a confirmation signal from the server confirms the upload is complete, thus ensuring that the critical data has been successfully delivered.

[0057] After receiving the confirmation signal, the server decompresses and parses the delivered compressed data packet to restore the original structure and content of the data, obtaining the restored abnormal data content. Finally, based on the restored abnormal data content, the real-time monitoring module quickly compares and comprehensively analyzes it with the current status and historical behavior patterns. This allows for the dynamic assessment and confirmation of which transmission paths(s) are most effective in ensuring the real-time and reliable uploading of such critical abnormal data, thereby ultimately determining the target transmission paths that need to be prioritized and monitored in real time.

[0058] In this embodiment, by dynamically confirming and optimizing the target transmission path, the monitoring system has the ability to self-recognize and optimize the status of the communication link, and can automatically select the best data upload path, thereby building a remote monitoring data transmission system.

[0059] In one embodiment, please refer to Figure 7 The step of acquiring an anomaly warning signal based on the target transmission path, and assessing the stability and potential risks of the data stream in the target transmission path based on the anomaly warning signal to determine the systemic risk assessment result includes: Step 701: Obtain data stream information from the target transmission path; based on the data stream information, filter out any possible abnormal fluctuations to obtain an abnormal warning signal. Step 702: Based on the abnormal warning signal and combined with the interaction records between the anchor bolt structure nodes, analyze the abnormal propagation path in the node linkage to determine the chain reaction mode. Step 703: Based on the chain reaction mode, adaptively adjust the linkage parameters between the anchor bolt structure nodes to obtain the optimized parameter configuration; Step 704: Based on the optimized parameter configuration, assess the stability and potential risks of the data flow in the target transmission path, and determine the systemic risk assessment result.

[0060] Real-time data stream information is continuously acquired from the established target transmission path. This data stream information is scanned and analyzed to identify and filter out abnormal fluctuations that deviate from the normal pattern. These fluctuations may indicate potential hardware failures, communication interference, or structural damage. Anomaly warning signals are generated based on these abnormal fluctuations; for example, if the peak value of an abnormal fluctuation exceeds a preset fluctuation threshold, an anomaly warning signal is generated. Subsequently, the anomaly warning signals are correlated with the interaction records between anchor bolt structure nodes stored in the historical database. This analysis examines the transmission trajectory of anomalies between nodes and the propagation path of anomalies in node linkages. This reveals the chain reaction mechanism of how a single node failure affects other nodes through the data link, ultimately determining the chain reaction pattern.

[0061] Based on the identified chain reaction patterns, the linkage parameters between anchor bolt structure nodes are dynamically adjusted, such as node correlation weights, early warning thresholds, or data sampling frequencies. Dynamic self-correction is performed based on real-time identified risk patterns, resulting in an optimized parameter configuration that enhances the monitoring system's environmental adaptability and risk response capabilities. Finally, based on the optimized parameter configuration, a comprehensive assessment of the target transmission path's operational status is conducted. From a reliability perspective, potential risks such as single points of failure, bandwidth bottlenecks, and security vulnerabilities in the path are comprehensively analyzed to determine a fully quantified systemic risk assessment result that reflects the monitoring system's health.

[0062] In this embodiment, a parameter adaptive adjustment mechanism based on a chain reaction model is used to achieve self-learning and optimization capabilities. It can dynamically adjust the monitoring strategy according to the real-time risk model, which significantly improves the level of intelligence and responsiveness.

[0063] The remote monitoring system for anchor bolt fastening status provided by the present invention is described below. The remote monitoring system for anchor bolt fastening status described below can be referred to in correspondence with the remote monitoring method for anchor bolt fastening status described above.

[0064] The present invention also provides a remote monitoring system for anchor bolt fastening status, comprising: The status data acquisition module is used to collect anchor bolt fastening status data through sensors deployed on the anchor bolt structure node, convert the anchor bolt fastening status data into digital format, and obtain the initial status dataset of the anchor bolt structure node. The feature vector determination module is used to perform preprocessing operations on the local processor based on the initial state dataset, remove interference using a noise removal method, and determine the compact feature vector; The grouping module is used to group adjacent anchor bolt structure nodes by a grouping algorithm if the feature value of the fastening feature vector exceeds a preset feature threshold, so as to obtain a dynamic linkage structure containing anchor bolt structure groups. The matrix generation module is used to generate the state summary required by the anchor structure node based on the shared information of the structural nodes in the anchor structure group in the dynamic linkage structure, and to exchange the state summary under the edge computing framework to generate a comprehensive perception matrix after information sharing. The behavior pattern recognition module is used to perform feature extraction and pattern recognition based on the comprehensive perception matrix and apply a machine learning model to obtain the fastening behavior pattern describing the anchor bolt structure group. The transmission path monitoring module is used to monitor the data channel load based on the tightening behavior pattern. If the data channel load is lower than a preset load threshold, key abnormal data is extracted from the data channel load and uploaded to the server to obtain the target transmission path that needs to be monitored in real time. The risk assessment module is used to obtain abnormal early warning signals based on the target transmission path, and to assess the stability and potential risks of the data flow in the target transmission path based on the abnormal early warning signals, so as to determine the systemic risk assessment results.

[0065] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method of remotely monitoring the state of an anchor fastening, characterized by, The method comprises the following steps: Collecting anchor fastening state data through sensors deployed on anchor structure nodes, converting the anchor fastening state data into digital format to obtain an initial state data set of the anchor structure nodes; Performing preprocessing operations in a local processor based on the initial state data set, removing interference using a noise removal method, and determining a fastening feature vector; If the eigenvalue of the fastening feature vector exceeds a preset feature threshold, grouping adjacent anchor structure nodes using a grouping algorithm to obtain a dynamic linkage structure containing an anchor structure group; Based on the shared information of the structure nodes within the anchor structure group in the dynamic linkage structure, generating a state summary required by the anchor structure nodes, exchanging the state summary under an edge computing framework, and generating a comprehensive perception matrix after information sharing; Based on the comprehensive perception matrix, applying a machine learning model for feature extraction and pattern recognition to obtain a fastening behavior pattern describing the anchor structure group; Based on the fastening behavior pattern, monitoring data channel load, if the data channel load is lower than a preset load threshold, extracting key abnormal data from the data channel load, uploading the key abnormal data to a server, and obtaining a target transmission path that needs real-time monitoring; Based on the target transmission path, obtaining an abnormal early warning signal, and based on the abnormal early warning signal, evaluating the data flow stability and potential risks in the target transmission path to determine a systematic risk assessment result.

2. The method of claim 1, wherein The method comprises the following steps: Based on the initial state data set, performing preliminary screening on the potential interference signals in the initial state data set to determine a basic data set after preliminary screening; Based on the basic data set after preliminary screening, performing preprocessing operations in a local processor, and using a noise removal method for smoothing processing to obtain a smooth data set after removing interference; Based on the smooth data set, performing normalization operations on the data points of the smooth data set to obtain a standardized data matrix; Based on the standardized data matrix, extracting core data elements related to fastening features, if the distribution of the core data elements is within a preset feature range, regarding the core data elements within the preset feature range as effective feature data groups; Based on the effective feature data groups, constructing corresponding feature vector representations through vector mapping methods to obtain the fastening feature vector.

3. The method of claim 1, wherein If the noise level of the fastening feature vector exceeds a preset range, using a denoising processing method to purify the fastening feature vector to obtain a purified feature vector. ​ If the purified feature vector exceeds the preset feature threshold, a clustering operation is performed on adjacent anchor structure nodes by a grouping algorithm to determine the relevance between the anchor structure nodes and form an anchor structure group; Based on the anchor structure group, a dynamic linkage relationship between the anchor structure nodes is identified, wherein the dynamic linkage relationship includes an interaction mode of each anchor structure node within the group; Based on the dynamic linkage relationship, the node correlation strength within the anchor structure group is analyzed to obtain the correlation strength distribution within the anchor structure group; Based on the correlation strength distribution, a dynamic linkage mode of the anchor structure group is determined to obtain a dynamic linkage structure of the anchor structure group.

4. The method of claim 1, wherein Based on the shared information of the structure nodes within the anchor structure group in the dynamic linkage structure, a state summary required by the anchor structure node is generated, and the state summary is exchanged under an edge computing framework to generate a comprehensive perception matrix after information sharing, including: Based on the dynamic linkage structure, the shared information of the structure nodes within the anchor structure group is obtained, and the shared information is preliminarily classified in a distributed processing manner to obtain a preliminarily sorted data set; Based on the preliminarily sorted data set, the interaction demand between local nodes of the edge computing framework is adopted to perform local processing on the preliminarily sorted data set by using edge computing technology to generate a state summary; If the state summary meets the preset integrity condition, the state summary is distributed to the corresponding group nodes of the edge computing framework by a framework delivery mechanism to obtain distributed state information; Based on the distributed state information, an intermediate layer data structure for information sharing is constructed by using a matrix processing method to determine a preliminary matrix of comprehensive perception; Based on the preliminary matrix of comprehensive perception, a principal component analysis algorithm is applied to reduce the dimension of data to obtain the comprehensive perception matrix.

5. The method of claim 1, wherein Based on the comprehensive perception matrix, a machine learning model is applied for feature extraction and pattern recognition to obtain a fastening behavior mode describing the anchor structure group, including: Based on the comprehensive perception matrix, cleaning and standardization processing are performed by a data preprocessing tool to obtain a normalized matrix data set; Based on the normalized matrix data set, a machine learning model is applied for feature extraction and pattern recognition to determine a distribution set of fastening behavior features; Based on the distribution set of fastening behavior features, the interrelationship between features is identified by a feature correlation analysis module to determine the contact strength within the anchor structure group; If the contact strength exceeds a preset strength threshold, the fastening behavior is classified and mapped based on the distribution set of fastening behavior features to obtain a preliminary division result of the fastening behavior rule; Based on the preliminary division result of the fastening behavior rule, the fastening behavior rule is integrated and optimized by a mode construction module to determine the fastening behavior mode.

6. The method of claim 1, wherein The monitoring data channel load is monitored based on the fastening behavior mode, if the data channel load is lower than a preset load threshold, key abnormal data in the data channel load is extracted and uploaded to a server, and a target transmission path that needs real-time monitoring is obtained, including: Based on the fastening behavior mode, the load of the data channel is continuously collected by the monitoring system to obtain the data channel load representing the load change; If the data channel load is lower than a preset load threshold, an abnormal data screening mechanism is triggered to extract key abnormal data from the data channel load, and the key abnormal data is marked with priority to determine the abnormal content marked with priority; Based on the abnormal content marked with priority, an optimized protocol is used for compression processing to obtain compressed data packets; Through the pre-established transmission path, the compressed data packets are uploaded to the server to obtain an upload completion confirmation signal; When the confirmation signal is received, the compressed data packets are decompressed and parsed to obtain the restored abnormal data content; Based on the restored abnormal data content, data comparison and state update are performed in combination with the real-time monitoring module to determine the target transmission path that needs real-time monitoring.

7. The method of claim 1, wherein The abnormal early warning signal is obtained based on the target transmission path, and based on the abnormal early warning signal, the data flow stability and potential risk in the target transmission path are evaluated to determine the systematic risk evaluation result, including: Data flow information is obtained from the target transmission path, and based on the data flow information, abnormal fluctuations that may exist are screened to obtain an abnormal early warning signal; Based on the abnormal early warning signal, the interaction record between the anchor structure nodes is analyzed to determine the abnormal propagation path in the node linkage, so as to determine the chain reaction mode; Based on the chain reaction mode, the linkage parameters between the anchor structure nodes are adaptively adjusted to obtain optimized parameter configurations; Based on the optimized parameter configurations, the data flow stability and potential risk in the target transmission path are evaluated to determine the systematic risk evaluation result.

8. A system for remote monitoring of the state of anchoring, characterized in that it comprises: Including: A state data acquisition module is configured to collect anchor fastening state data through sensors deployed on anchor structure nodes, convert the anchor fastening state data into digital format, and obtain an initial state data set of the anchor structure nodes; A feature vector determination module is configured to perform preprocessing operations in a local processor based on the initial state data set, remove interference using a noise removal method, and determine a fastening feature vector; A grouping module is configured to group adjacent anchor structure nodes using a grouping algorithm if the feature value of the fastening feature vector exceeds a preset feature threshold, and obtain a dynamic linkage structure containing an anchor structure group; A matrix generation module is configured to generate a state summary required by the anchor structure nodes based on shared information of the structure nodes in the anchor structure group in the dynamic linkage structure, exchange the state summary under an edge computing framework, and generate a comprehensive perception matrix after information sharing; a behavior pattern recognition module configured to apply a machine learning model to perform feature extraction and pattern recognition based on the comprehensive perception matrix, to obtain a fastening behavior pattern describing the anchor structure group; a transmission path monitoring module configured to monitor a data channel load based on the fastening behavior pattern, and if the data channel load is lower than a preset load threshold, extract key abnormal data from the data channel load, and upload the key abnormal data to a server to obtain a target transmission path requiring real-time monitoring; a risk assessment module configured to obtain an abnormal early warning signal based on the target transmission path, and based on the abnormal early warning signal, assess data flow stability and potential risks in the target transmission path to determine a systematic risk assessment result.

Images