An intelligent management method and system for equipment based on big data processing

Abstract

The application relates to the technical field of industrial internet and intelligent manufacturing, and discloses a device intelligent management method and system based on big data processing, which comprises the following steps: acquiring original data sets and user operation data of a device; constructing a three-dimensional network topology graph according to the original data sets to obtain a topological structure description reflecting dynamic interaction between devices; performing dynamic correlation analysis and behavior mode classification by using a graph neural network and a support vector machine according to the topological structure description and the user operation data to obtain a behavior prediction score and an optimized permission scheme; and generating a final dynamic permission configuration through resource scheduling and access control rules according to the optimized permission scheme. The method can realize real-time, accurate and adaptive control of device access permissions, and significantly improves the system security and resource utilization efficiency in an industrial internet environment.

Description

Technical Field

[0001] This invention relates to the field of industrial internet and intelligent manufacturing, and in particular to a method and system for intelligent equipment management based on big data processing. Background Technology

[0002] In the fields of industrial internet and smart manufacturing, equipment management is a core element in driving production efficiency and safe operation. With the widespread adoption of IoT technology, the interaction between physical devices and digital systems is becoming increasingly complex, highlighting the growing importance of equipment management. Reasonable equipment access control not only ensures production safety but also optimizes resource utilization.

[0003] In existing technologies, device management typically relies on static permission allocation mechanisms or simple authentication systems. Common device access control methods mainly include Role-Based Access Control (RBAC). This method predefines the mapping relationship between roles and operation permissions through static configuration of digital permission data, and assigns users to corresponding roles, thereby achieving centralized management of device access permissions. This mechanism has a clear structure, is relatively simple to configure, and is suitable for system environments with relatively stable permission structures.

[0004] Existing technologies lack in-depth fusion analysis of inter-device topology relationships and user behavior patterns, resulting in an inability to adaptively adjust permission allocation according to dynamic changes in device and user states. This leads to significant deficiencies in security, efficiency, and resource utilization. In summary, existing technologies have limitations in permission dynamism, real-time response capabilities, and the fusion processing of complex data relationships. In particular, the lack of deep integration with next-generation information technologies limits their applicability in dynamic industrial environments. Summary of the Invention

[0005] This invention provides a device intelligent management method and system based on big data processing to solve the limitations in terms of dynamic permissions, real-time response capabilities, and the fusion processing of complex data relationships.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides a device intelligent management method based on big data processing, comprising:

[0007] Obtain the raw dataset of the devices and construct a three-dimensional network topology map to obtain a topology description that reflects the dynamic interaction between devices;

[0008] Based on the topology description, the spatial distribution and operating status changes among the processing devices are analyzed to determine the correlation strength value of device collaboration.

[0009] Based on the correlation strength value, user behavior patterns are classified and it is determined whether they conform to the preset normal operation range to obtain a behavior prediction score.

[0010] If the behavior prediction score is higher than the preset device load threshold, then the real-time device status index is extracted to obtain the current device load and collaborative task details;

[0011] Based on the current device load and the details of the collaborative task, update the dynamic weights in the topology description to determine the potential interference of access requests on network interactions;

[0012] Based on the potential interference level and the behavior prediction score, an access control decision matrix is ​​generated and it is determined whether the interference value is lower than the security risk threshold to obtain a preliminary permission allocation scheme.

[0013] Based on the preliminary permission allocation scheme and the association strength value, the load distribution and resource utilization parameters of the fusion device collaborative services are adjusted to determine the service load association strength for permission optimization.

[0014] Based on the business load correlation strength, the resource utilization parameters are adjusted by integrating efficiency optimization rules to obtain the final dynamic permission generation result.

[0015] Secondly, the present invention provides an intelligent device management system based on big data processing, comprising:

[0016] The topology awareness and construction module acquires the raw dataset of devices and constructs a 3D network topology map to obtain a topology description that reflects the dynamic interaction between devices.

[0017] The collaborative relationship analysis module, based on the topology description, processes the spatial distribution and operating status changes among devices to determine the association strength value of device collaborative work;

[0018] The user behavior assessment module, in conjunction with the correlation strength value, classifies user behavior patterns and determines whether they conform to the preset normal operation range, thereby obtaining a behavior prediction score.

[0019] If the behavior prediction score is higher than the preset equipment load threshold, the equipment status monitoring module extracts real-time equipment status indicators to obtain the current equipment load and collaborative task details.

[0020] The interference assessment module updates the dynamic weights in the topology description based on the current device load and the details of the collaborative task, and determines the potential interference level of the access request on network interaction.

[0021] The access control decision module generates an access control decision matrix based on the potential interference level and the behavior prediction score, and determines whether the interference value is lower than the security risk threshold to obtain a preliminary permission allocation scheme.

[0022] The permission optimization module determines the service load association strength for permission optimization by integrating the load distribution and resource utilization parameters of the device collaborative services based on the preliminary permission allocation scheme and the association strength value.

[0023] The resource scheduling and permission generation module adjusts the resource utilization parameters based on the business load correlation strength and integrates efficiency optimization rules to obtain the final dynamic permission generation result.

[0024] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the device intelligent management method based on big data processing described in any one of the above.

[0025] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform any of the above-described intelligent device management methods based on big data processing.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] (1) This invention collects multi-dimensional device data from the Industrial Internet in real time, constructs a dynamic three-dimensional network topology map, and calculates the dynamic interaction weights and association strengths between devices based on graph neural networks. This enables accurate perception and visualization of dynamic interactions between devices, significantly improving the real-time performance and accuracy of device management in the Industrial Internet, and providing a reliable data foundation for subsequent permission decisions.

[0028] (2) This invention combines the correlation strength value and user operation data, uses support vector machine (SVM) to classify user behavior patterns and detect anomalies, generates behavior prediction scores, and dynamically adjusts the permission policy based on the scores and device load status, thereby realizing intelligent evaluation and risk prediction of user behavior, effectively preventing abnormal operations and unauthorized access, and improving system security and response agility.

[0029] (3) Based on the business load correlation strength and resource utilization parameters, this invention dynamically generates permission configuration through efficiency optimization rules and integrates access control rules to achieve fine-grained permission management, realizing a high degree of coordination between permission configuration and resource status, significantly improving the execution efficiency of device collaborative tasks and the overall resource utilization of the system, and supporting the high concurrency and high dynamic business needs in the industrial Internet environment. Attached Figure Description

[0030] Figure 1This is a schematic diagram of a device intelligent management method based on big data processing provided in the first embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram of a device intelligent management system based on big data processing provided in the second embodiment of the present invention. Detailed Implementation

[0032] 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.

[0033] Reference Figure 1 The first embodiment of the present invention provides a device intelligent management method based on big data processing, comprising the following steps:

[0034] S11, Obtain the original dataset of the devices and construct a three-dimensional network topology diagram to obtain a topology description that reflects the dynamic interaction between devices;

[0035] S12, Based on the topology description, process the spatial distribution and operating status changes among devices, and determine the association strength value of device collaborative work;

[0036] S13, Combine the correlation strength value to classify user behavior patterns and determine whether they conform to the preset normal operation range to obtain a behavior prediction score;

[0037] S14, if the behavior prediction score is higher than the preset device load threshold, then extract the real-time device status index to obtain the current device load and collaborative task details;

[0038] S15, based on the current device load and the details of the collaborative task, update the dynamic weights in the topology description to determine the potential interference level of the access request on network interaction;

[0039] S16, Based on the potential interference level and the behavior prediction score, generate an access control decision matrix and determine whether the interference value is lower than the security risk threshold to obtain a preliminary permission allocation scheme;

[0040] S17. Based on the preliminary permission allocation scheme and the association strength value, adjust the load distribution and resource utilization parameters of the integrated device collaborative services to determine the service load association strength for permission optimization.

[0041] S18. Based on the business load association strength, integrate efficiency optimization rules to adjust the resource utilization parameters and obtain the final dynamic permission generation result.

[0042] In step S11, the original dataset of the devices is obtained, and a three-dimensional network topology map is constructed to obtain a topology description reflecting the dynamic interaction between devices, including:

[0043] S1101 obtains the device identifier, operating status, connection strength, data traffic, communication protocol, device location, interaction frequency, access request frequency, data flow efficiency, data type, and timestamp from the Industrial Internet through a preset acquisition interface to obtain the raw dataset;

[0044] S1102, Based on the original dataset, construct an initial topological structure description containing nodes and edges;

[0045] S1103, if the connection strength between nodes in the initial topology description is greater than the preset connection strength threshold, then calculate the dynamic interaction weight and update the topology to obtain a dynamic interaction topology description.

[0046] S1104. Based on the topology description of the dynamic interaction, a three-dimensional network topology diagram is drawn according to the device location and topology level to obtain a topology description that reflects the dynamic interaction between devices.

[0047] In step S1101, the device identifier, operating status, connection strength, data traffic, communication protocol, device location, interaction frequency, access request frequency, data flow efficiency, data type, and timestamp are obtained from the Industrial Internet through a preset acquisition interface to obtain the raw dataset.

[0048] In one embodiment, 100 devices in a factory upload data through an industrial IoT gateway. The devices are identified by a unique ID such as "DEV001", the operating status includes "normal" or "fault", the connection strength is represented by a signal strength value such as -60dBm, the data flow is MB per second, the communication protocol is such as MQTT, the device location is latitude and longitude coordinates, the interaction frequency is the number of interactions per minute, the data type includes sensor readings or control commands, and the timestamp is accurate to milliseconds.

[0049] In step S1102, an initial topological structure description containing nodes and edges is constructed based on the original dataset.

[0050] It should be noted that, based on the original dataset, a graph database such as Neo4j is used to store the device identifier, connection strength, interaction frequency, and device location, constructing a topology structure containing nodes and edges. Here, the device identifier is used as a node, connection strength as an edge, and interaction frequency and location as attributes.

[0051] In one embodiment, the connection strength between DEV001 and DEV002 is 0.8, the interaction frequency is 10 times per minute, and their locations are coordinates (100, 200) in area A and (150, 250) in area B of the factory, respectively. The initial topology reflects the static relationships between devices with nodes and edges, resembling a directed graph, and the edges between nodes are weighted according to the connection strength.

[0052] In step S1103, if the connection strength between nodes in the initial topology description is greater than a preset connection strength threshold, then the dynamic interaction weight is calculated and the topology is updated to obtain a dynamic interaction topology description.

[0053] It should be noted that the preset connection strength threshold is a predefined critical value used to filter effective or important connections in the network topology. It is calculated by collecting connection strength data between all nodes in the target application scenario over a predetermined time period, calculating their mean and standard deviation, and setting the connection strength threshold as the mean plus twice the standard deviation. The dynamic interaction weight is calculated based on the interaction frequency and timestamps in the original dataset, obtained by multiplying the interaction frequency by a freshness factor based on the timestamp. The freshness factor calculation formula is as follows:

[0054]

[0055] Where p represents the freshness factor, The difference between the current time and the interaction timestamp, in units of 10 ... . This is the attenuation coefficient, with units of... Based on half-life settings, ,in The desired half-life.

[0056] In one embodiment, analysis of historical monitoring data from the factory network over the past 30 days revealed a mean connection strength of 0.5 and a standard deviation of 0.1, resulting in a calculated connection strength threshold of 0.7. The connection strength between DEV001 and DEV002 was 0.8, with an interaction frequency of 10 times per minute. Timestamps indicated that the most recent interaction occurred 10 minutes prior to the current time. Based on business requirements analysis, a half-life of 5 minutes was set, and the calculated... The calculated freshness factor is 0.25. The dynamic interaction weight is calculated to be 2.5. By updating the topology in real time and adjusting the edge weights, the real-time changes in interactions between devices are reflected.

[0057] In step S1104, based on the topology description of the dynamic interaction, a three-dimensional network topology diagram is drawn according to the device location and topology level to obtain a topology description reflecting the dynamic interaction between devices.

[0058] It should be noted that, for the topology description of the dynamic interaction, a visualization tool such as Three.js is used to draw a 3D network topology diagram based on the device locations and topology levels, resulting in a topology description reflecting the dynamic interaction between devices. Here, device locations determine the spatial coordinates of nodes, and topology levels reflect the logical levels between devices, such as sensor layers, gateway layers, and cloud layers.

[0059] In one embodiment, DEV001 is a sensor node located at (100, 200, 0) in 3D space, and DEV002 is a gateway node located at (150, 250, 50). Edges are represented by varying shades of color based on dynamic interaction weights; an edge with a weight of 0.72 is displayed in dark blue. When the edge weight between DEV001 and DEV002 drops to 0.3, maintenance personnel can quickly locate signal interference issues through visualization, optimize equipment layout, or adjust communication protocols. This method, combining dynamic weights and visualization, supports efficient management and optimized scheduling of devices in the Industrial Internet, reduces maintenance costs, and improves system stability.

[0060] In step S12, based on the topology description, the spatial distribution and operating status changes among the processing devices are analyzed to determine the association strength value for device collaboration, including:

[0061] S1201, Based on the topology description, calculate the spatial distance and state change characteristics between nodes to obtain node feature vectors;

[0062] S1202, if the spatial distance of the node feature vectors is less than a preset spatial distance threshold and the frequency of the running state change is higher than a preset state change frequency threshold, then calculate the dynamic interaction weight to obtain a weighted topology description.

[0063] S1203, for the weighted topology description, aggregate node degrees and collaboration modes to determine the association strength value of device collaboration.

[0064] In step S1201, based on the topology description, the spatial distance and state change characteristics between nodes are calculated to obtain node feature vectors.

[0065] It should be noted that, based on the spatial location and operational status described in the topology, a graph neural network is used to calculate the spatial distance between nodes. A two-layer GraphSAGE model is employed, using node features such as spatial distance and state change frequency, and edge features such as interaction frequency as inputs. The node representation is updated by aggregating neighbor node information, and finally, a fully connected layer outputs the association strength value. The model uses a mean squared error loss function and an Adam optimizer, and is trained in a supervised manner on historical device collaboration relationship data. The node feature vector is a multi-dimensional vector, whose dimensions mainly include: the normalized value of spatial distance, the normalized value of state change frequency, and the device's own operational status encoding, etc.

[0066] In one embodiment, a graph neural network is used to compute node feature vectors. Based on spatial location, the Euclidean distance between DEV101 (50,100,10) and DEV201 (80,120,20) is calculated to be 36.6 meters and normalized to the [0,1] interval. Simultaneously, the frequency of operational state changes is calculated by analyzing the device state timestamp sequence, determining the number of state changes per unit time, and then normalizing this value. The graph neural network combines the aforementioned spatial distance and state change features to generate node feature vectors reflecting the spatial associations and state dependencies between devices. This method effectively captures the physical layout and dynamic characteristics of the devices, supporting dynamic analysis of system behavior.

[0067] In step S1202, if the spatial distance of the node feature vectors is less than a preset spatial distance threshold and the frequency of the running state change is higher than a preset state change frequency threshold, then the dynamic interaction weight is calculated to obtain a weighted topology description.

[0068] It should be noted that the preset spatial distance threshold is used to determine whether the spatial distance between two device nodes in the topology network is close enough to generate a meaningful collaborative relationship. This threshold is set as the 95th percentile of the historical collaborative device pair distances, based on the statistical distribution of effective communication distances between devices, obtained by collecting spatial location data and operational status change data of all device nodes over a historical period. The preset state change frequency threshold measures the frequency of changes in device operational status. This threshold is set as the mean plus one standard deviation, based on the statistical distribution of state change frequency under normal operating conditions, obtained by recording the actual collaborative working relationships and communication quality indicators between devices. The dynamic interaction weight is calculated by multiplying the interaction frequency in the original dataset by the timestamp freshness factor.

[0069] In one embodiment, the Euclidean distance between DEV101 (50,100,10) and DEV201 (80,120,20) is calculated to be 36.6 meters. Based on historical data analysis over the past 90 days, the 95th percentile of the Euclidean distance between all cooperating device pairs is 45 meters, thus satisfying the condition. The historical state change frequency is calculated to have a mean of 8 times / minute and a standard deviation of 2 times / minute, resulting in a state change frequency threshold of 10 times / minute. The interaction frequency between DEV101 and DEV201 is 15 times / minute, which is higher than the preset state change threshold. The timestamp shows that the most recent interaction occurred 10 minutes ago, resulting in a freshness factor of 0.25. Therefore, the dynamic interaction weight is calculated to be 3.75. This quantitatively calculated weight is used to update the weighted topology description, accurately reflecting the real-time nature and intensity of interactions between devices, ensuring that the topology can adapt to frequently changing interaction patterns between devices, thereby improving the flexibility and accuracy of network management.

[0070] In step S1203, for the weighted topology description, the node degree and cooperation mode are aggregated to determine the association strength value of device cooperation.

[0071] It should be noted that the graph neural network aggregates node degree and collaboration patterns to determine the strength of association. The graph neural network adopts a graph convolutional network with 2-3 layers and 64-128 hidden units per layer. ELU is used as the activation function, and the output layer uses the Sigmoid function to limit the association strength value to the range [0,1]. Historical device operation data is collected, including device topology, interaction frequency, and collaborative work records. The ground truth of association strength is labeled based on the actual collaborative task completion effect and the communication quality between devices. The data is divided into a training set (70%), a validation set (15%), and a test set (15%), divided in chronological order to avoid data leakage. Node degree reflects the number of device connections, and collaboration patterns analyze the data flow between devices. Global length is used as the global topology indicator. The model is trained as follows: the mean squared error loss function is used; the Adam optimizer is used, and the learning rate is determined in the range [0.001, 0.01] through grid search; 5-fold cross-validation is used to optimize parameters such as hidden layer dimension, number of attention heads, and dropout rate. Training stops when the validation set loss no longer decreases for 5 consecutive epochs. Graph neural networks are periodically retrained with new data to maintain the accuracy of model predictions.

[0072] In one embodiment, based on historical data from the past six months, DEV101 connects to three devices with a temperature reading of 3, DEV102 with a temperature reading of 2, an interaction frequency of 15 times / minute, and a communication latency of 50ms. DEV101 sends temperature data to DEV201, and DEV201 forwards instructions to the cloud, forming a collaborative workflow. A graph neural network aggregates this information and calculates a high correlation strength between DEV101 and DEV201, indicating close collaboration between them. This analysis reveals the functional dependencies between devices, facilitating optimized resource scheduling.

[0073] In step S13, the user behavior pattern is classified based on the association strength value, and it is determined whether it conforms to a preset normal operation range to obtain a behavior prediction score, including:

[0074] S1301, combine the association strength value with the original dataset to extract state change features and generate the user behavior pattern feature vector;

[0075] S1302, if the deviation of the user behavior pattern feature vector from the preset normal range is greater than the preset behavior deviation threshold, then calculate the dynamic weighting factor to obtain the weighted behavior pattern description.

[0076] S1303, based on the weighted behavior pattern description and the normal range boundary conditions, if the user behavior pattern conforms to the normal operation range, a behavior prediction score is obtained.

[0077] In step S1301, the association strength value is combined with the original dataset to extract state change features and generate the user behavior pattern feature vector.

[0078] It should be noted that the initial behavior dataset is generated by retrieving the operation logs and current access requests corresponding to the user identifier from the database, extracting timestamps, interaction frequency, data traffic, and spatial location. The association strength value is then combined with this initial behavior dataset. The association strength value, based on historical data, reflects the normal pattern of user operation on the device; for example, U001 typically operates DEV301 2-4 times per minute. Feature extraction employs temporal feature extraction. The frequency change rate is obtained by calculating the relative change rate between the current interaction frequency and the historical mean; the data traffic fluctuation rate is obtained by calculating the ratio of the standard deviation to the mean of the data traffic using a sliding window; and spatial movement features are obtained by calculating the Euclidean distance of the user's location change and their movement speed. This constitutes a user behavior pattern feature vector containing the frequency change rate, data traffic fluctuation rate, spatial movement speed, and association strength value.

[0079] In one embodiment, within a single factory, the operation log records user U001's operation of production line equipment via terminal device DEV301. The timestamp is 2025-09-15 22:30:00, the average interaction frequency is 3 times / minute, the standard deviation is 0.5 times / minute, the average data flow is 1.5MB / s, and the standard deviation is 0.25MB / s. The spatial location is the factory coordinates (20, 50, 5). This data is collected in real time through an industrial IoT gateway and stored in a relational database such as MySQL. If the timestamp shows that U001's operation frequency suddenly increases to 6 times / minute within 10 minutes, the state extraction calculation yields a frequency change rate of 6 times / minute and a data flow fluctuation rate of 0, resulting in a user behavior pattern feature vector of [6.0, 0, 0, 0.85].

[0080] In step S1302, if the deviation of the user behavior pattern feature vector from the preset normal range is greater than the preset behavior deviation threshold, then a dynamic weighting factor is calculated to obtain a weighted behavior pattern description.

[0081] It should be noted that the preset normal range refers to the baseline values ​​of behavioral characteristics and their reasonable fluctuation range determined by long-term monitoring and learning of the historical behavioral data of legitimate users under normal and stable system operation. This is achieved by collecting behavioral data of all normal operations of the user group within the past predetermined time period and calculating the mean of each behavioral characteristic. and standard deviation Set the normal range to The preset behavioral deviation threshold is a critical value used to determine whether a user's behavioral deviation is significant enough to trigger a security alert or further analysis mechanism. It is typically set to 20% of the normal range. The dynamic weighting factor is calculated using a quantifiable decay function; the more recently an anomaly occurred, the higher its importance and the greater its weight.

[0082] In one embodiment, the normal range is defined as an interaction frequency of 2-4 times / minute and a data flow of 1-2 MB / s. A preset behavioral deviation threshold, for example, is 0.4 times / minute. If the detected current behavioral pattern feature vector of user U001 is 6 times / minute, exceeding the preset behavioral deviation threshold of 0.4 times / minute, anomaly detection is triggered. Assuming the timestamp of this abnormal behavior is 10 seconds from the current time, an attenuation coefficient is calculated. ,in If the half-life is set to 10 minutes, then The calculated value is 0.001155, and the dynamic weighting factor, based on the time decay function, is 0.9885, resulting in a weighted behavioral pattern description [5.93, 0, 0, 0.84]. This weighted vector highlights recent anomalous behavior, and its value decays over time, thus its contribution to subsequent comprehensive security assessments gradually decreases.

[0083] In step S1303, the weighted behavior pattern description is combined with the normal range boundary conditions. If the user behavior pattern conforms to the normal operation range, a behavior prediction score is obtained.

[0084] It should be noted that a Support Vector Machine (SVM) classifier is used, combined with normal range boundary conditions, to determine whether user behavior is normal and generate a behavior prediction score. The SVM classifier is a pre-trained machine learning model. Its training process is as follows: First, 10,000 historical behavior records are collected, including 8,000 normal records labeled as 1 and 2,000 abnormal records labeled as 0. These are then divided into a training set (70%), a validation set (15%), and a test set (15%) in chronological order. The user behavior pattern feature vectors are used as input features, and a grid search combined with 5-fold cross-validation is used to optimize the hyperparameters, including the RBF kernel function and a penalty function C=10. In the validation... The parameter "scale" will automatically adjust based on the number of features, and will select the parameter combination that yields the highest accuracy and largest AUC area on the validation set.

[0085] In one embodiment, the weighted feature vector [5.93,0,0,0.84] is input into a trained SVM classifier, which outputs a behavior prediction score of 0.3. This score indicates that the current behavior has a 30% confidence level of belonging to the normal mode, which is lower than the set normal threshold of 0.7, and is therefore judged as abnormal behavior. The classifier accurately distinguishes between normal and abnormal operations by analyzing the bias and weighting factors.

[0086] In step S14, if the behavior prediction score is higher than a preset normal behavior threshold and the device load is higher than a preset device load threshold, then real-time device status indicators are extracted to obtain the current device load and collaborative task details, including:

[0087] S1401, if the behavior prediction score is higher than the preset normal behavior threshold and the device load is higher than the preset device load threshold, then extract the real-time device status index and determine the task scheduling adjustment scheme according to the task allocation status and task priority.

[0088] S1402, Based on the task scheduling adjustment scheme, extract the data flow efficiency and device collaboration frequency to obtain task optimization execution parameters;

[0089] S1403, based on the task optimization execution parameters, combined with resource utilization and data processing latency, determine the execution order of device collaborative tasks, and obtain the current device load and collaborative task details.

[0090] In step S1401, if the behavior prediction score is higher than the preset normal behavior threshold and the device load is higher than the preset device load threshold, then the real-time device status index is extracted, and a task scheduling adjustment scheme is determined based on the task allocation status and task priority.

[0091] It should be noted that the normal behavior threshold is a critical score used to trigger the task scheduling adjustment mechanism. The mean is calculated by statistically analyzing the predicted scores of historical normal user behavior. and standard deviation Set the normal behavior threshold as The device load threshold is a critical value that a user judges as indicating that a device is under abnormal load. It is calculated by taking the mean of the predicted scores based on the statistical distribution of historical normal user behavior. and standard deviation Set the device load threshold to When the behavior prediction score is greater than the normal behavior threshold and the device load is greater than the device load threshold, it indicates that the user behavior is normal but the system load is too high, and task scheduling needs to be optimized. This dual judgment mechanism ensures that resource adjustments are only triggered when they are truly needed.

[0092] In one embodiment, a preset normal behavior threshold is 0.75, and a load threshold is 0.80. When a user behavior prediction score of 0.8 is detected and the load of DEV301 reaches 0.85, a high load alarm is triggered, initiating task scheduling adjustments. DEV301 is currently executing a high-priority task T001, producing critical components, and the task allocation status shows that it occupies 80% of the production line resources. The adjustment plan prioritizes reducing the resource allocation of the low-priority task T002, reducing its load from 20% to 10%, thereby freeing up resources for DEV301 and optimizing equipment operating efficiency.

[0093] In step S1402, based on the task scheduling adjustment scheme, the data flow efficiency and device collaboration frequency are extracted to obtain task optimization execution parameters.

[0094] It should be noted that data transfer efficiency refers to the average data transmission rate between processing units or storage nodes within the system during task execution, typically measured in MB / s (the amount of data successfully transmitted per unit time). This parameter reflects the task's consumption of system I / O and network bandwidth resources. Device coordination frequency refers to the average frequency, such as times per minute, of communication or coordination interactions between multiple devices to achieve a common goal during the task execution cycle. This parameter reveals the strength of the task's distributed coordination and its dependence on system communication resources. These optimization parameters collectively constitute the key dimensions for evaluating task execution performance and resource requirements—the task optimization execution parameters.

[0095] In one embodiment, the data transfer efficiency of task T001 is calculated by statistically analyzing the total amount of data successfully transmitted at the gateway within the most recent time window, such as 5 minutes, and dividing it by the window duration; for example, the result is 2.0 MB / s. The device collaboration frequency is calculated by analyzing communication records between devices, statistically analyzing the number of effective interactions that occur within the same time window when devices related to task T001, such as DEV301 and DEV302, collaborate to complete the task, and converting this number to a frequency per minute; for example, the result is 5 times / minute, reflecting the intensity of data interaction between DEV301 and other devices. These parameters provide a basis for subsequent resource allocation.

[0096] In step S1403, based on the task optimization execution parameters, the execution order of the device collaborative tasks is determined by combining resource utilization and data processing latency, and the current device load and collaborative task details are obtained.

[0097] It should be noted that, to ensure the optimized parameters reflect the latest system status, the extraction of data flow efficiency and device collaboration frequency must be highly real-time. The system gateway updates and pushes log data every second to support subsequent dynamic scheduling analysis. Resource allocation aims to maximize resource utilization and minimize overall processing latency, making comprehensive decisions based on the aforementioned parameters.

[0098] In one embodiment, a greedy algorithm based on priority and resource utilization is employed. This algorithm always prioritizes scheduling tasks with the highest current priority and whose required resources are within the currently available resource range. The task execution order is determined by considering data processing latency. Input task execution optimization parameters, such as for T001, include a data flow rate of 2.0 MB / s, a collaboration frequency of 5 times / minute, and real-time device status such as DEV301 CPU resource utilization of 85% and a data processing latency of 50 milliseconds, which is lower than the preset latency threshold of 100 milliseconds, indicating sufficient processing capacity. Task T001 is prioritized for execution, while task T002 is postponed to the next cycle, reducing the load on DEV301 to 75% and improving overall production efficiency. The preset latency threshold refers to the maximum allowable latency set by the system to ensure timely processing of collaborative tasks. It is primarily based on the average latency level obtained from long-term operation under normal load conditions, such as the historical 95th percentile latency.

[0099] In step S15, based on the current device load and the details of the collaborative task, the dynamic weights in the topology description are updated to determine the potential interference level of the access request on network interaction, including:

[0100] S1501, Based on the current device load and the details of the collaborative task, a topology vector is constructed;

[0101] S1502, For the topology vector, update the dynamic weights of node connections, and obtain a weighted topology by combining the access request frequency in the original dataset;

[0102] In S1503, if the dynamic weight in the weighted topology is higher than the preset dynamic weight threshold, the resource allocation scheme is adjusted and the network interaction priority is determined.

[0103] In S1504, the delayed data is processed based on the network interaction priority, and the potential interference level of the access request on the network interaction is calculated.

[0104] In step S1501, a topology vector is constructed based on the current device load and the details of the collaborative task.

[0105] It should be noted that the topology description vector includes device load, task allocation status, and communication strength between nodes, in the form of [78%, T101, 70%, 4 times / minute, 1.5MB / s], describing the role and status of DEV401 in the production line network. The initial topology is built based on node connection data, reflecting the physical and logical connections between devices.

[0106] In one embodiment, within an Industrial Internet of Things (IIoT) scenario, the equipment status monitoring module collects real-time data from production line equipment via an IIoT gateway, generating equipment load and collaborative task details. In a factory production line, equipment DEV401 has a real-time load of 78%, and the collaborative task details show it is currently executing task T101, involving the production of key components and occupying 70% of the production line resources. Node connection data records the communication relationships between devices through the gateway; for example, the interaction frequency between DEV401 and DEV402 is 4 times per minute, with a data transmission rate of 1.5 MB / s. This data provides the foundation for constructing a topology description vector. DEV401 is connected to DEV402 and DEV403 through the gateway, forming a star topology, with DEV401 as the central node.

[0107] In step S1502, for the topology vector, the dynamic weights of node connections are updated, and a weighted topology is obtained by combining the access request frequency in the original dataset.

[0108] It should be noted that the dynamic weights are calculated by a graph neural network model. This model uses a temporal graph neural network with 2-3 layers and 64-128 hidden units per layer. The Sigmoid function is used to limit the dynamic weight values ​​to the range [0,1]. Historical device communication data is collected as training data, including temporal topology changes, communication intensity fluctuations, and load level changes. The true values ​​of the dynamic weights are labeled based on actual network performance indicators. The training set (first 70% of the time period), validation set (middle 15%), and test set (last 15%) are divided according to the time series. Historical and real-time communication data between nodes, such as communication intensity and load level, are used as input features. Historical access request frequency is used as edge features, and the feature change trend of the most recent N time slices is used as temporal features. Model training is performed using the Huber loss function to balance mean squared error and absolute error. The AdamW optimizer is used, and the weight decay coefficient is determined by grid search. Bayesian optimization is used to search for the learning rate, hidden layer dimension, and number of attention heads. Layer normalization and dropout rate (dropout=0.2-0.5) are used to prevent overfitting. The updated connection weights are output and, together with the frequency of node access requests, form a weighted topology to more accurately reflect the current actual state and criticality of node connections in the network.

[0109] In one embodiment, the access request frequency of device node DEV401 is 10 times per second. The communication characteristics of it and its neighboring node DEV402 are input into a pre-trained graph neural network model, and the dynamic weight of the connection between DEV401 and DEV402 is updated to 0.85. This connection plays a crucial role in the current network, and combined with the node's access request frequency, a weighted topology is generated.

[0110] In step S1503, if the dynamic weight in the weighted topology is higher than the preset dynamic weight threshold, the resource allocation scheme is adjusted and the network interaction priority is determined.

[0111] It should be noted that the preset dynamic weight threshold is a critical value used to determine whether a connection in the network is a critical connection. It is set based on statistical analysis of historical weighted topology data. For example, the dynamic weight values ​​are sorted by percentage, and values ​​above the 90th percentile are selected as the threshold to filter out the most important top 10% of connections.

[0112] In one embodiment, the dynamic weight of connection (DEV401-DEV402) is 0.85, higher than the preset critical connection weight threshold of 0.80. T101 is a high-priority task with a data transfer efficiency of 2.2 MB / s, better than the average of 1.8 MB / s, indicating that DEV401 has strong network transmission capabilities. The adjustment scheme prioritizes resource allocation for T101, reducing the resource consumption of the low-priority task T102 on DEV402 from 15% to 8%, thereby freeing up bandwidth for DEV401. Network interaction priorities are determined accordingly, increasing the communication priority of DEV401 to ensure stable data transmission for critical tasks.

[0113] In step S1504, the delayed data is processed based on the network interaction priority, and the potential interference level of the access request on the network interaction is calculated.

[0114] It should be noted that the potential interference level is a quantitative parameter used to assess the negative impact of external access requests consuming network resources such as bandwidth and the number of connections on high-priority task interactions. It is calculated by combining the business importance defined by network interaction priorities with the real-time performance status reflected by processing latency data. The calculated interference level will directly guide subsequent optimization decisions, such as dynamically limiting the request rate of non-critical devices to ensure the service quality of core interactions and the overall stability of the production line network.

[0115] In one embodiment, the current processing latency of high-priority device DEV401 is 40 milliseconds, indicating that its current processing capacity is sufficient. When calculating the interference level, the bandwidth utilization rate (60%) of DEV401's requests and the priority (high) of their respective interactions are first analyzed, and the following rules are applied: if the bandwidth utilization rate is >70% and the priority is high, the interference level is "high"; if the bandwidth utilization rate is between 40% and 70% and the priority is high, the interference level is "medium"; if the bandwidth utilization rate is <40% or the priority is not high, the interference level is "low". Therefore, the interference level of DEV401 is rated as "medium". To further reduce potential interference, the optimizer generates an adjustment strategy: reducing the access request frequency of device DEV402, which may cause resource contention with DEV401, from 8 times per second to 5 times per second, thereby effectively reducing its interference on the critical path and improving overall network efficiency.

[0116] In step S16, an access control decision matrix is ​​generated based on the potential interference level and the behavior prediction score, and it is determined whether the interference value is lower than the security risk threshold to obtain a preliminary permission allocation scheme, including:

[0117] S1601, Based on the potential interference level and the behavior prediction score, construct a user behavior feature vector to obtain an initial access control decision matrix;

[0118] S1602, for the initial access control decision matrix, update the potential interference level, and combine the access request frequency and data flow efficiency in the original dataset to obtain a weighted decision matrix;

[0119] S1603, if the interference value of the weighted decision matrix is ​​lower than the preset security risk threshold, the permission allocation scheme is adjusted according to the resource allocation priority and network interaction priority to obtain a preliminary permission allocation scheme.

[0120] In step S1601, based on the potential interference level and the behavior prediction score, a user behavior feature vector is constructed to obtain an initial access control decision matrix.

[0121] It should be noted that the user behavior feature vector is a multi-dimensional data structure used to quantify the characteristics of a user's access behavior within a specific time window. Its dimensions include at least: user identity, access request frequency, behavior prediction score, and potential interference level. The initial access control decision matrix is ​​a dynamic permission mapping table generated driven by the user behavior feature vector. Its rows typically represent users or user groups, columns represent protected devices or resources, and matrix element values ​​represent the initial access permission level granted to the user in the current context. Generating this matrix is ​​a core step in implementing dynamic fine-grained access control.

[0122] In one embodiment, in an Industrial Internet of Things (IIoT) scenario, a user behavior feature vector is constructed using user identification and access time series to achieve access control decisions. In a factory production line, operator A's identification is UID-501. Their access time series records their operational behavior over the past hour, such as initiating two access requests per minute, involving production data queries for device DEV401. A behavior prediction score is generated based on historical operation patterns; UID-501's prediction score is 0.9, indicating highly reliable behavior. The user behavior feature vector is thus constructed in the form [UID-501, 2 times / minute, 0.9], reflecting the operator's access frequency and reliability. An initial access control decision matrix is ​​generated based on this vector. The matrix records the access permissions of different users to the devices; for example, UID-501 has read and write permissions for DEV401, but only read-only permissions for DEV402.

[0123] In step S1602, the potential interference level is updated for the initial access control decision matrix, and a weighted decision matrix is ​​obtained by combining the access request frequency and the data flow efficiency in the original dataset.

[0124] It should be noted that logistic regression is used to update the potential interference level. Logistic regression analysis uses access request frequency, behavior prediction score, and data flow efficiency as input features, and performs Z-score standardization on all values. For example, based on historical data, the mean access request frequency is approximately 3 times / minute, and the standard deviation is approximately 1.5 times / minute; therefore, 2 times / minute, after standardization, is approximately -0.67. Using historical access control decision data, including normal and abnormal access patterns, the output interference level probability value represents the potential impact of the current access request on system stability. The formula for calculating the interference level probability value is as follows:

[0125] ,

[0126] Where G represents the probability value of the interference level. The feature weights are obtained through training on historical data using a standard logistic regression process. ...

[0127] In one embodiment, DEV401's network bandwidth is 10MB / s, and UID-501's request bandwidth is 1MB / s, accounting for 10%. Combined with a behavior prediction score of 0.9 (normalized to 0.67), an access request frequency of 2 times / minute (normalized to -0.67), and a data flow efficiency of 2.5MB / s (normalized to 0.625), the logistic regression output interference level is 0.553, lower than the security risk threshold of 0.7, indicating a low request risk. The interference level adjustment factor is calculated to be 1.2235. In the initial decision matrix, UID-501's permission value for DEV401 is 0.65. Multiplying this by the interference level adjustment factor of 1.2235 and weighting the result, a new permission value of 0.80 is obtained, resulting in a weighted decision matrix.

[0128] In step S1603, if the interference value of the weighted decision matrix is ​​lower than the preset security risk threshold, the permission allocation scheme is adjusted according to the resource allocation priority and network interaction priority to obtain a preliminary permission allocation scheme.

[0129] It should be noted that the preset security risk threshold is a critical value used to determine the security risk of user access requests. Its establishment is primarily based on the statistical distribution of interference levels in historical attack or abnormal operation data, such as selecting the median of historical data, and choosing a high percentile value, such as the 95th percentile, that can cover the vast majority of normal operations while effectively identifying anomalies as the threshold. The preliminary permission allocation scheme is a dynamic and temporary permission configuration, which can be further optimized in subsequent steps based on real-time device load.

[0130] In one embodiment, the interference value of user UID-501's request update to device DEV401 is 0.3, which is lower than the security risk threshold of 0.5, indicating that its request is safe and reliable. The permission allocation scheme is adjusted based on resource allocation priority and network interaction priority. DEV401 runs a high-priority task T101, which has a high network interaction priority. Since UID-501's request is directly related to T101, its read / write permissions are prioritized. After the initial permission allocation scheme is optimized, UID-501's access permissions to DEV401 remain unchanged, but the read-only permission for DEV402 is further restricted, allowing only critical data queries, freeing up 0.2MB / s bandwidth for T101. This adjustment ensures sufficient resources for critical tasks while reducing network interference from unnecessary requests. DEV401's load is 75%, close to a high-load state. The scheme prioritizes reducing the access frequency of low-priority user UID-503 from 3 times per minute to 1 time per minute, ensuring the stable execution of UID-501's high-priority requests. This multi-faceted collaborative access control approach ensures the efficient operation of the production line network by dynamically adjusting access permissions and resource allocation.

[0131] In step S17, based on the preliminary permission allocation scheme and the association strength value, the load distribution and resource utilization parameters of the fusion device collaborative services are adjusted to determine the service load association strength for permission optimization, including:

[0132] S1701, Obtain the association strength value, extract the load distribution data of device collaborative services from the preliminary permission allocation scheme and group them to obtain a classification set of service loads;

[0133] S1702, Based on the classification set of the service load, obtain the resource utilization rate and calculate the matching degree with the load distribution data to obtain the matching parameters of the load resources;

[0134] S1703, If the matching parameters of the load resources are lower than the preset matching threshold, the task allocation balance and permission adjustment range are adjusted to obtain an optimized permission allocation configuration.

[0135] S1704, Based on the optimized permission allocation configuration, dynamically adjust the resource allocation dynamics and business load balancing, and determine the business load correlation strength of the permission optimization.

[0136] In step S1701, the association strength value is obtained, and the load distribution data of device collaborative services is extracted from the preliminary permission allocation scheme and grouped to obtain a classification set of service loads.

[0137] It should be noted that in industrial IoT scenarios, optimizing permission allocation requires combining device collaboration services and load distribution to ensure efficient resource utilization. Association strength values ​​reflect the closeness of user-device interaction, while load distribution data describes the task allocation of devices. K-means clustering is used to group the association strength values ​​and load distribution to obtain a set of service load categories. The number of clusters K is set to 3 (high, medium, low load) based on prior knowledge of the service load type, or automatically determined from historical data using the silhouette coefficient method.

[0138] In one embodiment, in a factory production line, device DEV401 runs high-priority task T101. The association strength value is generated based on the user UID-501's access frequency of 2 times / minute and historical behavior confidence of 0.9, calculated to be 0.85, indicating a strong association. Load distribution data is obtained through DEV401's CPU utilization of 70% and memory utilization of 60%, reflecting the device's high load state. Using the K-means algorithm, the load data of DEV401 and DEV402 are divided into high-load and low-load groups. DEV401 is classified into the high-load group because it runs task T101, consuming 8MB / s of bandwidth; DEV402 is classified into the low-load group, consuming only 3MB / s of bandwidth. The classification set provides a basis for subsequent matching degree calculation.

[0139] In step S1702, based on the classification set of the service load, the resource utilization rate is obtained and the matching degree with the load distribution data is calculated to obtain the matching parameters of the load resources.

[0140] It should be noted that the formula for calculating the matching degree between the load distribution data and resource utilization is as follows:

[0141] ,

[0142] in, and The weights are calculated based on historical task execution data. The direct impact of bandwidth usage on system performance is weighted at 0.6, and the impact of response time on user experience is weighted at 0.4. This indicates the percentage of bandwidth used. This represents the response time smoothing function, where t represents the response time. This represents the baseline response time; a value of 1 indicates a response time better than the baseline. When the response time deteriorates, the value is [value missing]. It exhibits exponential decay, distinguishing different degrees of deterioration: slight deterioration ,get The contribution rate is moderately low; the situation has seriously deteriorated. ,get The contribution rate has decreased significantly.

[0143] In one embodiment, resource utilization and device coordination data are obtained for a categorized set of service loads. Resource utilization includes the bandwidth usage ratio of device DEV401 at 80%, a response time of 200 milliseconds, and a baseline response time of 150 milliseconds. Device coordination is based on the degree of coordination between device UID-501 and task T101, with a coordination coefficient of 0.9. This coefficient can be calculated using historical collaborative task completion efficiency or inter-device communication quality indicators. A weighted average method is used to comprehensively calculate the resource utilization indicators, where the bandwidth component contributes 0.8, the time component contributes 0.72, the weighted sum is 0.768, and the final weighted sum of the matching parameters divided by the coordination coefficient is 0.853.

[0144] In step S1703, if the matching parameters of the load resources are lower than the preset matching threshold, the task allocation balance and permission adjustment range are adjusted to obtain an optimized permission allocation configuration.

[0145] It should be noted that the preset matching threshold is a benchmark value used to determine whether the current resource allocation of the system meets the business load requirements. It is calculated by collecting all load resource matching parameter data during the system's historical stable operation period and determining the statistical distribution characteristics of the matching parameters, including the mean. Standard deviation And percentiles, set the matching degree threshold to the 80th percentile of the historical matching parameter data.

[0146] In one embodiment, 10,000 samples of matching parameter data were collected, and the mean was calculated. Standard deviation 80th percentile P 80 =0.82, meaning 80% of historical matching parameters are below this value, is set as the matching threshold. Because the matching parameter 0.8 is below the matching threshold, the system determines that the current resource allocation cannot meet the load requirements and needs to initiate a permission allocation optimization process. The optimization process dynamically adjusts the balance of task allocation based on the priority of device or user permissions for tasks and their business response latency. For example, UID-501 has a priority of 0.9 for task T101, and its response latency is 200ms, which is better than UID-502's 300ms. Based on this, the system adjusts the permission allocation strategy: maintaining UID-501's read and write permissions to the DEV401 device, while reducing the access frequency of UID-502 from 3 times / minute to 1 time / minute, thereby ensuring the stable operation of the high-priority task T101. Based on the above adjustments, the system finally generates an optimized permission allocation configuration.

[0147] In step S1704, based on the optimized permission allocation configuration, the dynamics of resource allocation and the balancing of business load are dynamically adjusted to determine the business load correlation strength of permission optimization.

[0148] In one embodiment, resource allocation dynamics and service load balancing are dynamically adjusted by combining a data throughput efficiency of 2.5 MB / s and a network interaction frequency of 2 times / minute. The DEV401 prioritizes bandwidth allocation for task T101, freeing up 0.3 MB / s for high-priority requests. The service load correlation strength optimized for permissions is recalculated to 0.88, reflecting more efficient resource matching. This approach, through multi-faceted collaboration, ensures production line task priority and improves resource utilization efficiency.

[0149] In step S18, based on the service load correlation strength, the resource utilization parameters are adjusted by integrating efficiency optimization rules to obtain the final dynamic permission generation result, including:

[0150] S1801, The service load correlation strength data is grouped to obtain a classified service load set;

[0151] S1802, For the classified service load set, calculate the load balancing parameters;

[0152] S1803, If the load balancing parameter is lower than the preset load balancing threshold, the resource allocation dynamics are adjusted to obtain the dynamically adjusted resource allocation configuration.

[0153] S1804, combine the resource allocation configuration with the preset access control precision requirements to generate dynamic permission configuration and determine the final dynamic permission generation result.

[0154] In step S1801, the service load correlation strength data is grouped to obtain a classified service load set.

[0155] It should be noted that the K-means clustering algorithm is used to group the business load classification data. The number of clusters K is set to 3 (high, medium, and low load) based on prior knowledge of the business load type, or it can be automatically determined from historical data using the silhouette coefficient method. This distinguishes different priority load types, providing a basis for subsequent differentiated resource scheduling. The algorithm was chosen based on its effectiveness in processing numerical data, its fast convergence speed, and its ease of understanding, making it suitable for load classification scenarios in the Industrial Internet of Things (IIoT) based on multi-dimensional indicators such as permission priority and response latency.

[0156] In one embodiment, within an Industrial Internet of Things (IIoT) scenario, workload classification data is analyzed based on permission priority and business response latency to optimize task allocation for production line equipment. Permission priority reflects the importance of a user's access to a specific task, while business response latency measures the real-time performance of task execution. Assume that in a factory production line, equipment DEV301 is running the critical task T201. User UID-601 has a permission priority of 0.95 and a response latency of 150ms, which is better than UID-602's priority of 0.7 and response latency of 250ms. Based on this data, a K-means clustering algorithm is used to group the workload classification data. Equipment is divided into a high-priority task group and a normal task group. DEV301, running task T201, is assigned to the high-priority group, while DEV302, running the normal task T202, is assigned to the normal group. This grouping provides a clear basis for subsequent resource allocation.

[0157] In step S1802, load balancing parameters are calculated for the classified service load set.

[0158] It should be noted that the load balancing parameter is a comprehensive indicator used to reflect the rationality of system resource allocation among different load groups and the efficiency of device coordination. The load balancing parameter is calculated by combining bandwidth utilization and memory utilization using a weighted average method. Bandwidth utilization and memory utilization are each assigned a 50% weight, reflecting their equal importance to resource balance. Device coordination is included as an adjustment factor in the calculation; the lower the value of this parameter, the more uneven the resource utilization.

[0159] In one embodiment, resource utilization and device synergy data are obtained for the categorized set of service loads. Resource utilization can be calculated using the bandwidth utilization of DEV301 (75%) and memory utilization (65%), while device synergy is based on the correlation of 0.92 between UID-601 and task T201. A weighted average method is used, assigning a 50% weight to bandwidth utilization and a 50% weight to memory utilization, to calculate a load balancing parameter of 0.7.

[0160] In step S1803, if the load balancing parameter is lower than the preset load balancing threshold, the resource allocation dynamics are adjusted to obtain the dynamically adjusted resource allocation configuration.

[0161] It should be noted that the preset load balancing threshold is a critical value used to determine whether the system resource allocation is in a balanced state. Load balancing parameter data of the system during historical stable operation periods, such as the past 30 days, are collected, and the statistical distribution characteristics of the load balancing parameters, such as mean μ, standard deviation σ, percentiles, etc., are calculated. The load balancing threshold is set as the 10th percentile of the historical load balancing parameters, and the minimum load balancing parameters required to ensure that the system performance is not less than 95% are determined.

[0162] In one embodiment, based on historical data analysis over the past 30 days, 9000 sample points of load balancing parameters were collected, and the mean μ was calculated to be 0.78, the standard deviation σ to be 0.05, and the 10th percentile P... 10 =0.70, meaning 90% of historical load balancing parameters are higher than this value, is set as the load balancing threshold. If the preset load balancing threshold is 0.7, and the load balancing parameter is lower than the threshold, resource allocation needs to be adjusted. Based on a data throughput efficiency of 3MB / s and a network interaction frequency of 1.5 times / minute, the bandwidth allocation of DEV301 is dynamically adjusted, prioritizing the T201 task and freeing up 0.2MB / s for other high-priority requests. This adjustment improves the stability of task execution.

[0163] In step S1804, the resource allocation configuration is combined with the preset access control precision requirements to generate a dynamic permission configuration and determine the final dynamic permission generation result.

[0164] It should be noted that, based on dynamically adjusted resource allocation configurations and combined with access control precision requirements, a decision tree algorithm is used to generate dynamic permission configurations. The access control precision requirements are a set of pre-defined policy rules to ensure that permission allocation accurately matches task requirements, such as defining the maximum permission range corresponding to tasks of different priorities. The decision tree algorithm is generated using the CART algorithm, splitting nodes based on information gain characteristics such as user priority and device load rate. Each node ultimately corresponds to permission policies such as 'read / write', 'read-only', and 'deny'. This decision tree model can be trained by collecting historical best permission allocation records. Through hierarchical judgment, for example, first determining if the user priority is high (if yes, branch one; otherwise, branch two); then, in branch one, determining if the device load exceeds the device load threshold... and so on, combining device coordination and user behavior to generate fine-grained permission configurations. This approach ensures efficient execution of production line tasks while optimizing the balance of resource utilization, providing a flexible permission management solution for industrial IoT scenarios.

[0165] In one embodiment, UID-601 (user identifier) ​​retains read and write permissions to DEV301 (device identifier) ​​due to its high priority, while the access frequency of UID-602 is reduced from 2 times / minute to 0.5 times / minute to reduce the load on DEV301. The decision tree algorithm analyzes features such as priority, response latency, and device load, and generates permission rules based on the aforementioned decision rules to ensure that task T201 (task identifier) ​​is executed first. This process improves the targeting of permission allocation and task execution efficiency. The dynamic adjustment of data flow efficiency and network interaction frequency is reflected in real-time monitoring and rapid response. When task T201 of DEV301 is under high load, the system can dynamically allocate more bandwidth to ensure that the data flow efficiency remains stable above 2.8MB / s.

[0166] In summary, this invention discloses an intelligent equipment management method and system based on big data processing. By collecting real-time equipment operation data and user operation behavior, a dynamic three-dimensional network topology is constructed. Graph neural networks and support vector machines are integrated for behavior pattern recognition and permission prediction, and access control policies are dynamically generated based on business load and resource status. This invention deeply integrates next-generation information technologies such as the Internet of Things, big data analytics, and artificial intelligence, achieving real-time, precise, and adaptive control of equipment access permissions. It significantly improves system security, resource utilization efficiency, and collaborative operation capabilities in the industrial internet environment, demonstrating strong engineering applicability and promotional value.

[0167] Reference Figure 2 The second embodiment of the present invention provides a device intelligent management system based on big data processing, comprising:

[0168] The topology awareness and construction module acquires the raw dataset of devices and constructs a 3D network topology map to obtain a topology description that reflects the dynamic interaction between devices.

[0169] The collaborative relationship analysis module, based on the topology description, processes the spatial distribution and operating status changes among devices to determine the association strength value of device collaborative work;

[0170] The user behavior assessment module, in conjunction with the correlation strength value, classifies user behavior patterns and determines whether they conform to the preset normal operation range, thereby obtaining a behavior prediction score.

[0171] If the behavior prediction score is higher than the preset normal behavior threshold and the device load is higher than the preset device load threshold, the device status monitoring module extracts real-time device status indicators to obtain the current device load and collaborative task details.

[0172] The interference assessment module updates the dynamic weights in the topology description based on the current device load and the details of the collaborative task, and determines the potential interference level of the access request on network interaction.

[0173] The access control decision module generates an access control decision matrix based on the potential interference level and the behavior prediction score, and determines whether the interference value is lower than the security risk threshold to obtain a preliminary permission allocation scheme.

[0174] The permission optimization module determines the service load association strength for permission optimization by integrating the load distribution and resource utilization parameters of the device collaborative services based on the preliminary permission allocation scheme and the association strength value.

[0175] The resource scheduling and permission generation module adjusts the resource utilization parameters based on the business load correlation strength and integrates efficiency optimization rules to obtain the final dynamic permission generation result.

[0176] It should be noted that the device intelligent management system based on big data processing provided in this embodiment of the invention is used to execute all the process steps of the device intelligent management method based on big data processing in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0177] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a device intelligent management program based on big data processing. When the processor executes the computer program, it implements the steps in the various embodiments of the device intelligent management method based on big data processing described above, for example... Figure 1The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above system embodiments, such as the permission optimization module.

[0178] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0179] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0180] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0181] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function, such as sound playback, image playback, etc. The data storage area may store data created based on the use of the mobile phone, such as audio data, phonebook, etc. In addition, the memory may include high-speed random access memory and non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0182] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0183] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0184] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for intelligent equipment management based on big data processing, characterized in that, include: Obtain the raw dataset of the devices and construct a three-dimensional network topology map to obtain a topology description that reflects the dynamic interaction between devices; Based on the topology description, the spatial distribution and operating status changes among the processing devices are analyzed to determine the correlation strength value of device collaboration. Based on the correlation strength value, user behavior patterns are classified and it is determined whether they conform to the preset normal operation range to obtain a behavior prediction score. If the behavior prediction score is higher than the preset normal behavior threshold and the device load is higher than the preset device load threshold, then extract the real-time device status index to obtain the current device load and collaborative task details; Based on the current device load and the details of the collaborative task, update the dynamic weights in the topology description to determine the potential interference of access requests on network interactions; Based on the potential interference level and the behavior prediction score, an access control decision matrix is ​​generated and it is determined whether the interference value is lower than the security risk threshold to obtain a preliminary permission allocation scheme. Based on the preliminary permission allocation scheme and the association strength value, the load distribution and resource utilization parameters of the collaborative services of the converged devices are adjusted to determine the service load association strength for permission optimization. Based on the business load correlation strength, the resource utilization parameters are adjusted by integrating efficiency optimization rules to obtain the final dynamic permission generation result; Specifically, based on the topology description, the spatial distribution and operational status changes among processing devices are considered to determine the association strength value for device collaboration, including: Based on the topological description, the spatial distance and state change characteristics between nodes are calculated to obtain node feature vectors; If the spatial distance of the node feature vectors is less than a preset spatial distance threshold and the frequency of changes in the running state is higher than a preset state change frequency threshold, then the dynamic interaction weight is calculated to obtain a weighted topology description. Based on the weighted topology description, the node degree and collaboration mode are aggregated to determine the association strength value of device collaboration. Specifically, based on the correlation strength value, user behavior patterns are classified, and it is determined whether they conform to a preset normal operating range to obtain a behavior prediction score, including: The association strength value is combined with the original dataset to extract state change features and generate the user behavior pattern feature vector. If the deviation of the user behavior pattern feature vector from the preset normal range is greater than the preset behavior deviation threshold, then a dynamic weighting factor is calculated to obtain a weighted behavior pattern description. Based on the weighted behavior pattern description and the normal range boundary conditions, if the user's behavior pattern conforms to the normal operating range, a behavior prediction score is obtained.

2. The intelligent equipment management method based on big data processing according to claim 1, characterized in that, Obtain the raw data set of the devices and construct a 3D network topology map to obtain a topology description reflecting the dynamic interactions between devices, including: The raw dataset is obtained by acquiring device identifier, operating status, connection strength, data traffic, communication protocol, device location, interaction frequency, access request frequency, data flow efficiency, data type, and timestamp from the Industrial Internet through a preset acquisition interface. Based on the original dataset, construct an initial topological structure description containing nodes and edges; If the connection strength between nodes in the initial topology description is greater than the preset connection strength threshold, then the dynamic interaction weight is calculated and the topology is updated to obtain a dynamic interaction topology description. Based on the dynamic interaction topology description, a three-dimensional network topology diagram is drawn according to the device location and topology hierarchy, resulting in a topology structure description that reflects the dynamic interaction between devices.

3. The intelligent equipment management method based on big data processing according to claim 2, characterized in that, If the behavior prediction score is higher than a preset normal behavior threshold and the device load is higher than a preset device load threshold, then real-time device status indicators are extracted to obtain the current device load and collaborative task details, including: If the behavior prediction score is higher than the preset normal behavior threshold and the device load is higher than the preset device load threshold, then extract the real-time device status index and determine the task scheduling adjustment scheme according to the task allocation status and task priority. Based on the task scheduling adjustment scheme, the data flow efficiency and device collaboration frequency are extracted to obtain task optimization execution parameters; Based on the optimized execution parameters of the task, the execution order of the device collaborative tasks is determined by combining resource utilization and data processing latency, and the current device load and collaborative task details are obtained.

4. The intelligent equipment management method based on big data processing according to claim 3, characterized in that, Based on the current device load and the details of the collaborative task, update the dynamic weights in the topology description to determine the potential interference level of access requests on network interactions, including: Based on the current device load and the details of the collaborative task, a topology vector is constructed. For the aforementioned topology vector, the dynamic weights of node connections are updated, and a weighted topology is obtained by combining the access request frequency in the original dataset. If the dynamic weight in the weighted topology is higher than the preset dynamic weight threshold, the resource allocation scheme is adjusted and the network interaction priority is determined. By prioritizing network interactions, delayed data is processed, and the potential interference of access requests on network interactions is calculated.

5. The intelligent equipment management method based on big data processing according to claim 4, characterized in that, Based on the potential interference level and the behavior prediction score, an access control decision matrix is ​​generated, and it is determined whether the interference value is lower than the security risk threshold to obtain a preliminary permission allocation scheme, including: Based on the potential interference level and the behavior prediction score, a user behavior feature vector is constructed to obtain the initial access control decision matrix; For the initial access control decision matrix, the potential interference level is updated, and a weighted decision matrix is ​​obtained by combining the access request frequency and the data flow efficiency in the original dataset; If the interference value of the weighted decision matrix is ​​lower than the preset security risk threshold, the permission allocation scheme is adjusted according to the resource allocation priority and network interaction priority to obtain a preliminary permission allocation scheme.

6. The intelligent equipment management method based on big data processing according to claim 1, characterized in that, Based on the preliminary permission allocation scheme and the association strength value, the load distribution and resource utilization parameters of the collaborative services of the converged devices are adjusted to determine the service load association strength for permission optimization, including: Obtain the association strength value, extract the load distribution data of device collaboration services from the preliminary permission allocation scheme and group them to obtain a set of service load categories; Based on the classification set of the business load, the resource utilization rate is obtained and the matching degree with the load distribution data is calculated to obtain the matching parameters of the load resources. If the matching parameters of the load resources are lower than the preset matching threshold, the task allocation balance and permission adjustment range are adjusted to obtain an optimized permission allocation configuration. Based on the optimized permission allocation configuration, the dynamics of resource allocation and the business load balancing are dynamically adjusted to determine the business load correlation strength of permission optimization.

7. The intelligent equipment management method based on big data processing according to claim 1, characterized in that, Based on the business load correlation strength, the resource utilization parameters are adjusted by integrating efficiency optimization rules to obtain the final dynamic permission generation result, including: The service load correlation strength data is grouped to obtain a classified service load set; For the classified set of service loads, the load balancing parameters are calculated. If the load balancing parameter is lower than the preset load balancing threshold, the resource allocation dynamics are adjusted to obtain a dynamically adjusted resource allocation configuration. The resource allocation configuration is combined with the preset access control precision requirements to generate dynamic permission configuration and determine the final dynamic permission generation result.

8. A device intelligent management system based on big data processing, characterized in that, The method for implementing the intelligent device management method based on big data processing as described in any one of claims 1 to 7 includes: The topology awareness and construction module acquires the raw dataset of devices and constructs a 3D network topology map to obtain a topology description that reflects the dynamic interaction between devices. The collaborative relationship analysis module, based on the topology description, processes the spatial distribution and operating status changes among devices to determine the association strength value of device collaborative work; The user behavior assessment module, in conjunction with the correlation strength value, classifies user behavior patterns and determines whether they conform to the preset normal operation range, thereby obtaining a behavior prediction score. If the behavior prediction score is higher than the preset normal behavior threshold and the device load is higher than the preset device load threshold, the device status monitoring module extracts real-time device status indicators to obtain the current device load and collaborative task details. The interference assessment module updates the dynamic weights in the topology description based on the current device load and the details of the collaborative task, and determines the potential interference level of the access request on network interaction. The access control decision module generates an access control decision matrix based on the potential interference level and the behavior prediction score, and determines whether the interference value is lower than the security risk threshold to obtain a preliminary permission allocation scheme. The permission optimization module adjusts the load distribution and resource utilization parameters of the collaborative services of the converged devices based on the preliminary permission allocation scheme and the association strength value, and determines the service load association strength for permission optimization. The resource scheduling and permission generation module adjusts the resource utilization parameters based on the business load correlation strength and integrates efficiency optimization rules to obtain the final dynamic permission generation result.

Images