A device operation and maintenance intelligent management method and system

Abstract

The application discloses a kind of equipment operation and maintenance intelligent management method and system, it is related to data management technical field, it is difficult to analyze equipment data from multiple angles, simultaneously lack of equipment relationship graph, and the analysis of potential fault data correlation, cannot clearly understand the correlation between fault equipment and other equipment, it is difficult to accurately assess the influence range of fault technical problem;Multi-dimensional sensor, operation log collector and edge computing node are deployed in edge layer, so that equipment state data, equipment running video stream data and operation log data can be preliminarily processed and analyzed locally.In the cloud layer, a multi-dimensional data analysis model for risk data and key equipment node data is established, which can analyze equipment data from multiple angles.Through equipment relationship graph analysis, the influence range of potential faults can be clearly understood.

Description

Technical Field

[0001] This invention belongs to the field of data management, specifically a method and system for intelligent management of equipment operation and maintenance. Background Technology

[0002] Traditional equipment maintenance relies heavily on manual inspections, requiring maintenance personnel to periodically visit the equipment site. Taking a large factory as an example, with numerous and widely distributed pieces of equipment, inspection personnel need to spend a significant amount of time traversing the factory grounds to check the operational status of each piece. For instance, in a factory with hundreds of pieces of equipment, a comprehensive inspection could take several days, which is not only inefficient but also prone to quality issues due to fatigue, leading to missed or incorrect inspections. When equipment malfunctions, traditional maintenance models often rely on on-site personnel to discover and report the problem. If the malfunction occurs outside of working hours or in remote areas, it may not be addressed promptly.

[0003] The rise of IoT technology has laid the foundation for intelligent equipment operation and maintenance. Therefore, an intelligent equipment operation and maintenance management solution is needed. By installing various sensors on equipment, real-time operational data can be collected, and massive amounts of equipment operation and maintenance data can be stored, managed, and analyzed using big data technology.

[0004] Existing intelligent equipment operation and maintenance management solutions struggle to analyze equipment data from multiple perspectives. Relying solely on single factors or simple rules for fault identification easily leads to misjudgments or omissions. The lack of equipment relationship maps and analysis of the correlations between potential fault data makes it difficult to clearly understand the relationships between faulty equipment and other equipment, hindering accurate assessment of the fault's impact. For example, in a complex production system, a fault in a critical piece of equipment may affect the normal operation of other connected equipment, but due to the lack of an equipment relationship map, it is impossible to quickly determine the scope of affected equipment. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an intelligent equipment operation and maintenance management method and system to solve the technical problems of difficulty in analyzing equipment data from multiple perspectives, easy misjudgment or omission, lack of equipment relationship map, and analysis of the correlation of potential fault data, making it impossible to clearly understand the relationship between faulty equipment and other equipment, and making it difficult to accurately assess the impact range of the fault.

[0006] To address the above problems, a first aspect of the present invention provides an intelligent equipment operation and maintenance management method and system, comprising:

[0007] Multi-dimensional sensors, operation log collectors, and edge computing nodes are deployed at the edge layer. The edge computing nodes manage the device links within the signal connection range and set up a time-series database to store device status data, device operation video stream data, and operation log data. The edge computing nodes establish a device relationship graph based on the device nodes of the managed device links and filter key device nodes.

[0008] A risk data detection model is established at the edge layer to analyze device data in the time-series database. The risk data and key device node data are uploaded to the cloud layer, and the cloud layer extracts the hash value of the operation log of the key device node and stores it in the blockchain.

[0009] Establish a multi-dimensional data analysis model for risk data and key equipment node data in the cloud layer, analyze equipment data, identify potential fault data of equipment nodes, and generate ordinary maintenance work orders;

[0010] The impact range of potential faults is analyzed by analyzing the device relationship graph. At the same time, when multiple device nodes with potential fault data are detected in the same device link, the correlation of potential fault data is analyzed. When the impact range of the fault is greater than the threshold or when there is a correlation between potential fault data, an alarm is triggered and an expedited maintenance work order is generated.

[0011] Optionally, in one example of the above aspects, the edge computing nodes establish a device relationship graph based on the device nodes of the managed device links, and filter key device nodes, including the following steps:

[0012] Add device node data, device load and alarm records to the device relationship graph in the device link;

[0013] Establish edges between device nodes that have connections or data transmission relationships;

[0014] Calculate the degree centrality and betweenness centrality of each device node in the device relationship graph, and then take a weighted average of the degree centrality and betweenness centrality to obtain the importance coefficient of the device node.

[0015] Where, degree centrality Cd = d(v) / (N−1), d(v) is the number of edges directly connecting node v, and N is the total number of nodes in the graph; betweenness centrality Cv = σst is the total number of paths from node s to t, and σst(v) is the number of paths from node s to t that pass through node v.

[0016] Within the same device link, the importance coefficients of the device nodes are sorted from high to low, and the top 10% of the device nodes are selected as critical device nodes.

[0017] Optionally, in one example of the above aspects, a risk data detection model is established at the edge layer to analyze device data in a time-series database, including the following steps:

[0018] A risk data detection model is established at the edge layer, including three levels of risk detection and analysis;

[0019] A three-tiered risk detection and analysis strategy is set up, which includes: sequentially detecting the device's operation log data, device status data, and device operation video stream data at the three tiers. If any tier detects risk data, all data from the device during the same period are judged as risk data, and the detection is completed; otherwise, the risk detection and analysis continues to the next tier.

[0020] Optionally, in one example of the above aspects, the first-level risk detection includes:

[0021] The edge computing node obtains standard operation process data from the device attribute data of the management device node, extracts operation keywords and timestamps from the operation log data in the time series database, arranges the operation keywords according to the operation time sequence, generates the actual operation process, compares the actual operation process with the standard operation process data, and detects the actual operation process in the standard operation process data.

[0022] The second level of risk detection includes: using a machine learning model pre-trained on historical fault data to detect whether there is abnormal data in the equipment status data and to determine the risk data;

[0023] The third level of risk detection includes: using a Faster R-CNN recognition model pre-trained on historical operational image data during equipment failure, capturing images of the equipment's operational video stream data every 50 frames, and detecting whether there are fault points in the captured image data; judging the risk data based on whether there are fault points.

[0024] Optionally, in one example of the above aspects, the cloud layer extracts the hash value of the operation logs of key device nodes and stores it on the blockchain, including the following steps:

[0025] Extract the operation logs of key device nodes and merge operation logs with an operation time interval less than a threshold.

[0026] Add a hash value label = Hash(operation content + timestamp + device ID) to the operation log and store it on the blockchain.

[0027] Optionally, in one example of the above aspects, a multi-dimensional data analysis model for risk data and key equipment node data is established in the cloud layer to analyze the equipment data and identify potential fault data of equipment nodes, including the following steps:

[0028] For risk data and key equipment node data, a multi-dimensional data analysis model for equipment nodes is established, including: a pre-trained Faster R-CNN recognition model with edge layer replication for verification training, a risk coefficient checking model for operation log data, and a multi-dimensional analysis algorithm for equipment status data;

[0029] Based on the detection results of the model and algorithm, a weighted average is calculated. If the weighted average result is greater than the threshold, the device node is judged to have a potential fault; otherwise, it is judged not to have a potential fault.

[0030] Optionally, in one example of the above aspects, analyzing the impact range of a potential fault through a device relationship graph includes the following steps:

[0031] Extract historical fault data from the same device link where the time interval between device node failures is less than a preset threshold, count the number of failures Sn of the corresponding device node, and the number of failures Sc of the corresponding device node with a time interval between failures less than the preset threshold. Also, count the number of fault nodes in the corresponding number of failures and calculate the average number of fault nodes pc in the corresponding number of failures.

[0032] Obtain the importance coefficient of the corresponding equipment node in the equipment relationship graph, and analyze the impact range coefficient of the corresponding potential faults:

[0033]

[0034] Wherein, Pb is the impact range coefficient of the corresponding potential fault, p0 is the total number of devices in the same device link, Ft is the importance coefficient of the corresponding device node in the device relationship graph, and Fu is the analysis result of the multi-dimensional data analysis model of the device node.

[0035] Optionally, in one example of the above aspects, when potential fault data is detected in multiple device nodes within the same device link, analyzing the correlation of the potential fault data includes the following steps:

[0036] By extracting historical fault data from the same device link, data where the time interval between device node failures is less than a preset threshold are identified, and correlation labels are applied between these data with time intervals less than the preset threshold.

[0037] An LSTM time series model is trained using labeled data, and the LSTM time series model is used to detect whether there is a correlation between potential fault data of device nodes.

[0038] According to another aspect of this disclosure, an intelligent equipment operation and maintenance management system is provided, characterized in that the system adopts the intelligent equipment operation and maintenance management method described above to achieve intelligent equipment operation and maintenance management.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] This invention establishes a multi-dimensional data analysis model at the cloud layer for risk data and key equipment node data, enabling analysis of equipment data from multiple perspectives. By comprehensively analyzing this data, potential fault data of equipment nodes can be identified more accurately, avoiding misjudgments or omissions that may result from single-factor analysis. Analyzing the impact range of potential faults through equipment relationship graphs clearly reveals the correlation between faulty equipment and other equipment. Accurately assessing the impact range of faults helps maintenance personnel rationally allocate maintenance resources and time. Faults with a small impact range can be scheduled for maintenance during off-peak production periods; while faults with a large impact range require immediate action to minimize the impact on production.

[0041] This invention deploys multi-dimensional sensors, operation log collectors, and edge computing nodes at the edge layer, enabling preliminary processing and analysis of device status data, device operation video stream data, and operation log data locally. The edge computing nodes perform initial data filtering and analysis, uploading only risk data and data from critical device nodes to the cloud layer. This reduces the amount of data uploaded to the cloud, lowers the data processing pressure and network bandwidth consumption in the cloud, and improves the overall system efficiency. A device relationship graph is established based on the device nodes in the managed device links, clearly displaying the relationships between devices and facilitating subsequent analysis. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0044] Figure 2 This is a schematic diagram of the method for identifying potential fault data of device nodes according to the present invention. Detailed Implementation

[0045] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0046] Please see Figures 1-2 The first aspect of this invention provides an intelligent equipment operation and maintenance management method and system, including:

[0047] Deploy multi-dimensional sensors, operation log collectors, and edge computing nodes at the edge layer. The edge computing nodes manage the device links within the signal connection range and set up an InfluxDB time-series database to store device status data, device operation video stream data, and operation log data. The edge computing nodes establish a device relationship graph based on the device nodes of the managed device links and filter key device nodes.

[0048] A risk data detection model is established at the edge layer to analyze device data in the time-series database. The risk data and key device node data are uploaded to the cloud layer, and the cloud layer extracts the hash value of the operation log of the key device node and stores it in the blockchain.

[0049] Establish a multi-dimensional data analysis model for risk data and key equipment node data in the cloud layer, analyze equipment data, identify potential fault data of equipment nodes, and generate ordinary maintenance work orders;

[0050] The impact range of potential faults is analyzed by analyzing the device relationship graph. At the same time, when multiple device nodes with potential fault data are detected in the same device link, the correlation of potential fault data is analyzed. When the impact range of the fault is greater than the threshold or when there is a correlation between potential fault data, an alarm is triggered and an expedited maintenance work order is generated.

[0051] Specifically, in this embodiment, multi-dimensional sensors, operation log collectors, and edge computing nodes are deployed at the edge layer, enabling device status data, device operation video stream data, and operation log data to be preliminarily processed and analyzed locally.

[0052] Edge computing nodes perform preliminary screening and analysis of data, uploading only risk data and data from key equipment nodes to the cloud layer. This reduces the amount of data uploaded to the cloud, lowers the data processing pressure and network bandwidth usage in the cloud, and improves the overall system operating efficiency.

[0053] Building a device relationship graph based on the device nodes in the managed device chain clearly shows the connections between devices. This helps in analyzing the overall system architecture and device dependencies, facilitating more accurate analysis of the impact scope later. Identifying key device nodes allows for focused analysis of the operational status of critical devices, enabling the rational allocation of maintenance resources and improving operational efficiency.

[0054] Establishing a risk data detection model at the edge layer enables timely detection of potential risks during equipment operation. Since the detection is performed locally, it is unaffected by factors such as network latency, allowing for faster risk identification and corresponding countermeasures. Deploying an InfluxDB time-series database on the edge computing nodes to store device data—a database specifically designed for storing and processing time-series data—provides highly efficient data writing and querying performance. This facilitates convenient querying and analysis of historical device data, providing strong support for equipment fault diagnosis, performance evaluation, and predictive maintenance.

[0055] The hash values ​​of operation logs from key equipment nodes are extracted at the cloud layer and stored on the blockchain. Blockchain features decentralization, immutability, and traceability. Storing the hash values ​​of operation logs on the blockchain ensures their integrity and authenticity, preventing data tampering. For example, in the event of equipment malfunction or a safety incident, the authenticity of the operation logs can be verified using the hash values ​​on the blockchain, providing a reliable basis for accident investigation and liability determination.

[0056] The cloud layer receives risk data and key device node data uploaded from the edge layer, enabling more comprehensive and in-depth analysis and processing. By comprehensively analyzing data uploaded from multiple edge computing nodes, potential problems in device operation can be identified, providing a more scientific basis for enterprise equipment operation and maintenance decisions.

[0057] By establishing a multi-dimensional data analysis model for risk data and key equipment node data in the cloud layer, equipment data can be analyzed from multiple perspectives. Through comprehensive analysis of this data, potential fault data of equipment nodes can be identified more accurately, avoiding misjudgments or omissions that may result from single-factor analysis. The system can automatically generate routine maintenance work orders based on the identified potential fault data, reducing manual intervention and improving work efficiency. Maintenance personnel can promptly schedule maintenance work based on these work orders, preventing further deterioration of faults and reducing equipment downtime and repair costs.

[0058] Analyzing the impact range of potential faults through equipment relationship diagrams provides a clear understanding of the relationships between faulty equipment and other equipment. Accurately assessing the impact range helps maintenance personnel rationally allocate maintenance resources and time. Faults with a small impact range can be scheduled for maintenance during off-peak production periods; while faults with a large impact range require immediate action to minimize the impact on production.

[0059] In the same device chain, when potential fault data is detected across multiple device nodes, analyzing the correlation between this data can reveal the inherent connections between the faults. For example, multiple device failures may be caused by the same reason, such as power issues, network failures, or external environmental factors. Analyzing these correlations helps to identify potential complex fault modes in advance. When the potential fault data from multiple devices are correlated, it may indicate that a more serious system failure is imminent. By triggering alarms, maintenance personnel can take preventative and maintenance measures in advance to avoid the escalation and spread of the fault.

[0060] A threshold is set for the scope of a fault's impact. When the scope of the fault's impact exceeds the threshold, or when potential fault data is correlated, the system automatically triggers an alarm and generates an expedited maintenance work order. This mechanism ensures that maintenance personnel can promptly address and handle major faults, reducing their impact on production.

[0061] Generating expedited maintenance work orders allows maintenance personnel to prioritize urgent fault handling tasks, improving the timeliness and effectiveness of fault resolution. Compared to regular maintenance work orders, expedited maintenance work orders have a higher priority, ensuring that faults in critical equipment are resolved promptly and guaranteeing normal production operations.

[0062] Multi-dimensional data analysis and correlation analysis at the cloud layer provide rich data support for operation and maintenance decisions. Operation and maintenance personnel can formulate more scientific and reasonable operation and maintenance strategies based on the analysis results; by accurately identifying potential faults and handling them in a timely manner, preventive maintenance is achieved, reducing the probability of sudden equipment failures and lowering equipment maintenance costs and production losses.

[0063] In this embodiment, the threshold corresponding to the impact range of a potential fault is set in the following way:

[0064] Historical detection data showing multiple device nodes failing within a time interval of less than a preset threshold, where critical device nodes fail simultaneously, are collected. The impact range coefficient of the corresponding potential fault is calculated, and the minimum value is taken as the threshold.

[0065] In one embodiment of the present invention, the edge computing nodes establish a device relationship graph based on the device nodes of the managed device links, and screen key device nodes, including the following steps:

[0066] Add device node data, device load and alarm records to the device relationship graph in the device link;

[0067] Device attribute data includes: device ID, type, model, IP address, geographical location, system to which it belongs, and standard operating procedures;

[0068] Establish edges between device nodes that have connections or data transmission relationships;

[0069] Calculate the degree centrality and betweenness centrality of each device node in the device relationship graph, and then take a weighted average of the degree centrality and betweenness centrality to obtain the importance coefficient of the device node.

[0070] Where, degree centrality Cd = d(v) / (N−1), d(v) is the number of edges directly connecting node v, and N is the total number of nodes in the graph; betweenness centrality Cv = σst is the total number of paths from node s to t, and σst(v) is the number of paths from node s to t that pass through node v.

[0071] Within the same device link, the importance coefficients of the device nodes are sorted from high to low, and the top 10% of the device nodes are selected as critical device nodes.

[0072] In this embodiment, the connection relationships between device nodes include:

[0073] Physical connections, such as switch-server connections in a network topology.

[0074] Logical dependencies, such as microservice A depending on database B.

[0075] Data transmission relationships include:

[0076] Data stream transmission, such as sensor → gateway → cloud platform.

[0077] The device with the most connections is identified by calculating the number of direct neighbors of a device node using degree centrality.

[0078] Betweenness centrality measures the frequency with which a node acts as a "bridge," and is applicable to critical path nodes in logical dependencies. A weighted average of degree centrality and betweenness centrality is used, with weights of 0.4 and 0.6, respectively.

[0079] In one embodiment of the present invention, a risk data detection model is established at the edge layer to analyze device data in a time-series database, including the following steps:

[0080] A risk data detection model is established at the edge layer, including three levels of risk detection and analysis;

[0081] A three-tiered risk detection and analysis strategy is set up, which includes: sequentially detecting the device's operation log data, device status data, and device operation video stream data at the three tiers. If any tier detects risk data, all data from the device during the same period are judged as risk data, and the detection is completed; otherwise, the risk detection and analysis continues to the next tier.

[0082] In one embodiment of the present invention, the first-level risk detection includes:

[0083] The edge computing node obtains standard operation process data from the device attribute data of the management device node, extracts operation keywords and timestamps from the operation log data in the time series database, arranges the operation keywords according to the operation time sequence, generates the actual operation process, compares the actual operation process with the standard operation process data, and detects the actual operation process in the standard operation process data.

[0084] The second level of risk detection includes: using a machine learning model pre-trained on historical fault data to detect whether there is abnormal data in the equipment status data and to determine the risk data;

[0085] The third level of risk detection includes: using a Faster R-CNN recognition model pre-trained on historical operational image data during equipment failure, capturing images of the equipment's operational video stream data every 50 frames, and detecting whether there are fault points in the captured image data; judging the risk data based on whether there are fault points.

[0086] In one embodiment of the present invention, the cloud layer extracts the hash value of the operation log of key device nodes and stores it in the blockchain, including the following steps:

[0087] Extract the operation logs of key device nodes and merge operation logs with an operation time interval less than a threshold.

[0088] Add a hash value label = Hash(operation content + timestamp + device ID) to the operation log and store it on the blockchain; the hash uses the SHA-256 algorithm.

[0089] In one embodiment of the present invention, a multi-dimensional data analysis model for risk data and key equipment node data is established in the cloud layer to analyze equipment data and identify potential fault data of equipment nodes, including the following steps:

[0090] For risk data and key equipment node data, a multi-dimensional data analysis model for equipment nodes is established, including: a pre-trained Faster R-CNN recognition model with edge layer replication for verification training, a risk coefficient checking model for operation log data, and a multi-dimensional analysis algorithm for equipment status data;

[0091] Based on the detection results of the model and algorithm, a weighted average is calculated. If the weighted average result is greater than the threshold, the device node is judged to have a potential fault; otherwise, it is judged not to have a potential fault.

[0092] In one embodiment of the present invention, the pre-trained Faster R-CNN recognition model for edge layer replication, the operation log data risk coefficient checking model, and the device status data multidimensional analysis algorithm are constructed in the following manner:

[0093] The pre-trained Faster R-CNN recognition model copied from the edge layer is validated and trained using historical running image data of the corresponding device nodes with annotated abnormal regions. If the time of the detected abnormal region in the video stream exceeds a preset threshold, the risk coefficient of the device running video stream is 1; otherwise, it is 0.

[0094] The operation log data risk coefficient inspection model includes: defining risky operation steps for device nodes, summarizing continuous operation steps of operation log data and corresponding time-series features within a preset time period using a large language model, performing cosine similarity detection on the summarized continuous operation steps and the defined risky operation steps in turn, and taking the maximum value of the cosine similarity as the operation risk coefficient.

[0095] The multidimensional analysis algorithm for equipment status data is as follows:

[0096]

[0097] Where St is the equipment status risk coefficient, Ei is the information entropy of the i-th type of equipment status data detection sensor, M0 is the average value of historical detection data of the i-th type of equipment status data detection sensor when the equipment is running normally, and Mi is the detection value of the i-th type of equipment status data detection sensor.

[0098] In this embodiment, the pre-trained Faster R-CNN recognition model, operation log data risk coefficient check model, and device status data multidimensional analysis algorithm copied from the edge layer are weighted and averaged according to the detection results of the model and algorithm. This includes: setting the weight of the device operation video stream risk coefficient to 0.3, normalizing the operation risk coefficient and the device status risk coefficient respectively, setting the weight of the normalized operation risk coefficient to 0.3, and setting the weight of the device status risk coefficient to 0.4; when the device is running normally, the mean and minimum of the historical detection data are used as a threshold. If the weighted average is greater than the threshold, it is determined that the device node has a potential fault; otherwise, it is determined that there is no potential fault.

[0099] In one embodiment of the present invention, the impact range of a potential fault is analyzed through a device relationship graph, including the following steps:

[0100] Extract historical fault data from the same device link where the time interval between device node failures is less than a preset threshold, count the number of failures Sn of the corresponding device node, and the number of failures Sc of the corresponding device node with a time interval between failures less than the preset threshold. Also, count the number of fault nodes in the corresponding number of failures and calculate the average number of fault nodes pc in the corresponding number of failures.

[0101] Obtain the importance coefficient of the corresponding equipment node in the equipment relationship graph, and analyze the impact range coefficient of the corresponding potential faults:

[0102]

[0103] Wherein, Pb is the impact range coefficient of the corresponding potential fault, p0 is the total number of devices in the same device link, Ft is the importance coefficient of the corresponding device node in the device relationship graph, and Fu is the analysis result of the multi-dimensional data analysis model of the device node.

[0104] In this embodiment, the threshold for the impact range coefficient of the potential fault is set in the following way:

[0105] Historical detection data showing multiple device nodes failing within a time interval of less than a preset threshold, where critical device nodes fail simultaneously, are collected. The impact range coefficient of the corresponding potential fault is calculated, and the minimum value is taken as the threshold.

[0106] In one embodiment of the present invention, when potential fault data is detected in multiple device nodes within the same device link, the correlation of the potential fault data is analyzed, including the following steps:

[0107] By extracting historical fault data from the same device link, data where the time interval between device node failures is less than a preset threshold are identified, and correlation labels are applied between these data with time intervals less than the preset threshold.

[0108] An LSTM time series model is trained using labeled data, and the LSTM time series model is used to detect whether there is a correlation between potential fault data of device nodes.

[0109] In one optional embodiment, an intelligent equipment operation and maintenance management system is provided, characterized in that the system employs an intelligent equipment operation and maintenance management method as described above to achieve intelligent equipment operation and maintenance management, including:

[0110] Edge layer:

[0111] Deploy multi-dimensional sensors, operation log collectors, and edge computing nodes. The edge computing nodes manage the device links within the signal connection range and set up a time-series database to store device status data, device operation video stream data, and operation log data. The edge computing nodes establish a device relationship graph based on the device nodes of the managed device links and filter key device nodes.

[0112] Establish a risk data detection model, analyze equipment data in the time series database, and upload risk data and key equipment node data to the cloud layer;

[0113] Cloud layer:

[0114] Extract the hash values ​​of the operation logs of key equipment nodes and store them on the blockchain; establish a multi-dimensional data analysis model for risk data and key equipment node data, analyze equipment data, identify potential fault data of equipment nodes, and generate ordinary maintenance work orders;

[0115] The impact range of potential faults is analyzed by analyzing the device relationship graph. At the same time, when multiple device nodes with potential fault data are detected in the same device link, the correlation of potential fault data is analyzed. When the impact range of the fault is greater than the threshold or when there is a correlation between potential fault data, an alarm is triggered and an expedited maintenance work order is generated.

[0116] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for intelligent management of equipment operation and maintenance, characterized in that, include: Multi-dimensional sensors, operation log collectors, and edge computing nodes are deployed at the edge layer. The edge computing nodes manage the device links within the signal connection range and set up a time-series database to store device status data, device operation video stream data, and operation log data. The edge computing nodes establish a device relationship graph based on the device nodes of the managed device links and filter key device nodes. A risk data detection model is established at the edge layer to analyze device data in the time-series database. The risk data and key device node data are uploaded to the cloud layer, and the cloud layer extracts the hash value of the operation log of the key device node and stores it in the blockchain. Establish a multi-dimensional data analysis model for risk data and key equipment node data in the cloud layer, analyze equipment data, identify potential fault data of equipment nodes, and generate ordinary maintenance work orders; The impact range of potential faults is analyzed by analyzing the device relationship graph. At the same time, when multiple device nodes with potential fault data are detected in the same device link, the correlation of potential fault data is analyzed. When the impact range of the fault is greater than the threshold or when there is a correlation between potential fault data, an alarm is triggered and an expedited maintenance work order is generated. The edge computing nodes establish a device relationship graph based on the device nodes in the managed device links, and key device nodes are selected, including the following steps: Add device node data, device load and alarm records to the device relationship graph in the device link; Establish edges between device nodes that have connections or data transmission relationships; Calculate the degree centrality and betweenness centrality of each device node in the device relationship graph, and then take a weighted average of the degree centrality and betweenness centrality to obtain the importance coefficient of the device node. Where, degree centrality Cd = d(v) / (N−1), d(v) is the number of edges directly connecting node v, and N is the total number of nodes in the graph; betweenness centrality Cv = σst is the total number of paths from node s to t, and σst(v) is the number of paths from node s to t that pass through node v. In the same device link, the importance coefficients of the device nodes are sorted from high to low, and the top 10% of the device nodes are selected as critical device nodes. Analyzing the impact range of potential faults using equipment relationship diagrams includes the following steps: Extract historical fault data from the same device link where the time interval between device node failures is less than a preset threshold, count the number of failures Sn of the corresponding device node, and the number of failures Sc of the corresponding device node with a time interval between failures less than the preset threshold. Also, count the number of fault nodes in the corresponding number of failures and calculate the average number of fault nodes pc in the corresponding number of failures. Obtain the importance coefficient of the corresponding equipment node in the equipment relationship graph, and analyze the impact range coefficient of the corresponding potential faults: Wherein, Pb is the impact range coefficient of the corresponding potential fault, p0 is the total number of devices in the same device link, Ft is the importance coefficient of the corresponding device node in the device relationship graph, and Fu is the analysis result of the multi-dimensional data analysis model of the device node.

2. The intelligent equipment operation and maintenance management method according to claim 1, characterized in that, A risk data detection model is established at the edge layer, and device data in the time-series database is analyzed, including the following steps: A risk data detection model is established at the edge layer, including three levels of risk detection and analysis; A three-tiered risk detection and analysis strategy is set up, which includes: sequentially detecting the device's operation log data, device status data, and device operation video stream data at the three tiers. If any tier detects risk data, all data from the device during the same period are judged as risk data, and the detection is completed; otherwise, the risk detection and analysis continues to the next tier.

3. The intelligent equipment operation and maintenance management method according to claim 2, characterized in that, The first level of risk detection includes: The edge computing node obtains standard operation process data from the device attribute data of the management device node, extracts operation keywords and timestamps from the operation log data in the time series database, arranges the operation keywords according to the operation time sequence, generates the actual operation process, compares the actual operation process with the standard operation process data, and detects the actual operation process in the standard operation process data. The second level of risk detection includes: using a machine learning model pre-trained on historical fault data to detect whether there is abnormal data in the equipment status data and to determine the risk data; The third level of risk detection includes: using a Faster R-CNN recognition model pre-trained on historical operating image data during equipment failure, capturing images of the equipment's operating video stream data every 50 frames, and detecting whether there are fault points in the captured image data; judging the risk data based on whether there are fault points.

4. The intelligent equipment operation and maintenance management method according to claim 1, characterized in that, The hash values ​​of the operation logs of key device nodes are extracted from the cloud layer and stored in the blockchain, including the following steps: Extract the operation logs of key device nodes and merge operation logs with an operation time interval less than a threshold. Add a hash value label = Hash(operation content + timestamp + device ID) to the operation log and store it on the blockchain.

5. The intelligent equipment operation and maintenance management method according to claim 1, characterized in that, Establish a multi-dimensional data analysis model for risk data and key equipment node data in the cloud layer, analyze equipment data, and identify potential fault data of equipment nodes, including the following steps: For risk data and key equipment node data, a multi-dimensional data analysis model for equipment nodes is established, including: a pre-trained Faster R-CNN recognition model with edge layer replication for verification training, a risk coefficient checking model for operation log data, and a multi-dimensional analysis algorithm for equipment status data; Based on the detection results of the model and algorithm, a weighted average is calculated. If the weighted average result is greater than the threshold, the device node is judged to have a potential fault; otherwise, it is judged not to have a potential fault.

6. The intelligent equipment operation and maintenance management method according to claim 5, characterized in that, The pre-trained Faster R-CNN recognition model, operation log data risk coefficient checking model, and equipment status data multidimensional analysis algorithm, which validate the edge layer replication of the training, are constructed in the following way: The pre-trained Faster R-CNN recognition model copied from the edge layer is validated and trained using historical running image data of the corresponding device nodes with annotated abnormal regions. If the time of the detected abnormal region in the video stream exceeds a preset threshold, the risk coefficient of the device running video stream is 1; otherwise, it is 0. The operation log data risk coefficient inspection model includes: defining risky operation steps for device nodes, summarizing continuous operation steps of operation log data and corresponding time-series features within a preset time period using a large language model, performing cosine similarity detection on the summarized continuous operation steps and the defined risky operation steps in turn, and taking the maximum value of the cosine similarity as the operation risk coefficient. The multidimensional analysis algorithm for equipment status data is as follows: Where St is the equipment status risk coefficient, Ei is the information entropy of the i-th type of equipment status data detection sensor, M0 is the average value of historical detection data of the i-th type of equipment status data detection sensor when the equipment is running normally, and Mi is the detection value of the i-th type of equipment status data detection sensor.

7. The intelligent equipment operation and maintenance management method according to claim 1, characterized in that, When potential fault data is detected in multiple device nodes within the same device link, the correlation of the potential fault data is analyzed, including the following steps: By extracting historical fault data from the same device link, data where the time interval between device node failures is less than a preset threshold are identified, and correlation labels are applied between these data with time intervals less than the preset threshold. An LSTM time series model is trained using labeled data, and then used to detect whether there is a correlation between potential fault data of device nodes.

8. An intelligent management system for equipment operation and maintenance, characterized in that, The system employs an intelligent equipment operation and maintenance management method as described in any one of claims 1-7 to achieve intelligent equipment operation and maintenance management, including: Edge layer: Deploy multi-dimensional sensors, operation log collectors, and edge computing nodes. The edge computing nodes manage the device links within the signal connection range and set up a time-series database to store device status data, device operation video stream data, and operation log data. The edge computing nodes establish a device relationship graph based on the device nodes of the managed device links and filter key device nodes. Establish a risk data detection model, analyze equipment data in the time series database, and upload risk data and key equipment node data to the cloud layer; Cloud layer: Extract the hash values ​​of the operation logs of key equipment nodes and store them on the blockchain; establish a multi-dimensional data analysis model for risk data and key equipment node data, analyze equipment data, identify potential fault data of equipment nodes, and generate ordinary maintenance work orders; The impact range of potential faults is analyzed by analyzing the device relationship graph. At the same time, when multiple device nodes with potential fault data are detected in the same device link, the correlation of potential fault data is analyzed. When the impact range of the fault is greater than the threshold or when there is a correlation between potential fault data, an alarm is triggered and an expedited maintenance work order is generated.

Images