Industrial Internet of Things equipment collaborative management and control system based on data and knowledge driving

Abstract

The invention discloses an industrial Internet of Things equipment collaborative management and control system based on data and knowledge driving. The system comprises six units including a heterogeneous multi-source data fusion interaction unit, an improved knowledge graph construction reasoning unit and an optimized hypergraph neural network feature extraction unit. Heterogeneous data acquisition and fusion are realized through a customized industrial-grade data interface protocol and a self-adaptive coding conversion mechanism; constructing a multi-level knowledge graph by using a semantic relationship mining algorithm and performing logical reasoning; deeply extracting data features by means of an optimized hypergraph neural network; combining equipment collaborative parameter generation, executing and dynamically adjusting a strategy; and monitoring the running state of the system in real time and performing feedback optimization. The method corresponds to six steps of the system, and equipment collaborative full-process management and control are achieved. The problems that a traditional system is difficult in data fusion, low in strategy making and execution efficiency and the like are effectively solved, the industrial Internet of Things equipment collaboration efficiency and production benefits are remarkably improved, and the system is suitable for various industrial production scenes.

Description

Technical Field

[0001] The present invention relates to the field of industrial Internet of Things equipment collaboration, and in particular to a data and knowledge-driven industrial Internet of Things equipment collaborative management and control system. Background Art

[0002] With the widespread application of the Industrial Internet of Things (IIoT) in smart manufacturing, energy management, and other fields, the number of devices in industrial production systems is increasing, and their types are becoming increasingly diverse. Different devices come from a variety of manufacturers and adhere to different data protocols and standards, resulting in significant multi-source and heterogeneous data in industrial scenarios. At the same time, industrial production is placing higher demands on the accuracy, efficiency, and intelligence of equipment collaborative control. Traditional control methods are no longer able to adapt to the trend of digital transformation, and the development of more advanced equipment collaborative control systems is urgent.

[0003] Existing IIoT device management and control technologies suffer from two core flaws. First, at the data processing level, heterogeneous data integration capabilities are weak. Due to the lack of unified data exchange standards, the efficient collection and standardized conversion of structured production data, semi-structured log information, and unstructured image and video data generated by different devices is difficult. Format barriers exist during data transmission and processing, hindering smooth data flow, significantly obscuring the value of the data, and preventing timely and accurate acquisition and analysis of critical information about equipment operations.

[0004] Second, there are deficiencies in the formulation and execution mechanisms for equipment collaboration strategies. Traditional systems often overlook the complex semantic relationships between devices and the coupling relationships between collaboration parameters when formulating equipment collaboration strategies, making it difficult to generate precise strategies that meet actual production needs. Furthermore, during strategy execution, the system is unable to dynamically optimize based on feedback such as the real-time operating status of devices and task progress. This results in low equipment collaboration efficiency, making it difficult to meet the requirements of industrial production for efficient and stable operation, and thus affecting overall production efficiency and quality. Summary of the Invention

[0005] In order to overcome the shortcomings and deficiencies of the existing technology, the present invention provides an industrial Internet of Things equipment collaborative management and control system based on data and knowledge driven.

[0006] The technical solution adopted by the present invention is a data- and knowledge-driven collaborative management and control system for industrial Internet of Things equipment, including: a heterogeneous multi-source data fusion and interaction unit, an improved knowledge graph construction and reasoning unit, an optimized hypergraph neural network feature extraction unit, an equipment collaborative strategy generation unit, a collaborative strategy execution and scheduling unit, and a system operation status monitoring and evaluation unit;

[0007] The heterogeneous multi-source data fusion and interaction unit is used to collect and transmit structured, semi-structured, and unstructured data from different production links and different types of equipment in the Industrial Internet of Things in parallel through a customized industrial-grade data interface protocol, and convert the collected heterogeneous data into a unified data expression format through an adaptive data encoding conversion mechanism for subsequent processing;

[0008] The improved knowledge graph construction reasoning unit is used to identify the semantic associations between data based on the converted unified format data using a semantic relationship mining algorithm, construct an improved knowledge graph with multi-level and multi-dimensional relationships, and use the topological structure of the knowledge graph to perform reasoning analysis on the equipment operation status and potential fault information using logical reasoning rules;

[0009] The optimized hypergraph neural network feature extraction unit is used to take the node and edge information of the improved knowledge graph as input, convert the knowledge graph into a hypergraph structure through a customized hypergraph topology structure mapping function, and then use the optimized hypergraph neural network to deeply extract and abstractly represent the data features in the hypergraph structure.

[0010] The device collaboration strategy generation unit is used to receive the feature representation output by the optimized hypergraph neural network feature extraction unit, and generate collaborative work strategies for different device combinations based on preset collaboration strategy generation rules and various parameters of industrial Internet of Things device collaboration, including device workload and response time threshold;

[0011] The collaborative strategy execution scheduling unit is used to accurately send the strategy instructions to the corresponding industrial Internet of Things devices through the industrial Internet of Things device communication network based on the collaborative work strategy generated by the device collaborative strategy generation unit, and dynamically schedule and adjust the strategy execution process based on the real-time feedback information of the devices;

[0012] The system operation status monitoring and evaluation unit is used to collect real-time operation data of industrial Internet of Things devices during collaborative work, including equipment energy consumption, operation time, and number of failures. Through a preset status evaluation indicator system, the overall operation status of the system is monitored and evaluated, and the evaluation results are fed back to the improved knowledge graph construction reasoning unit and the equipment collaboration strategy generation unit;

[0013] The improved knowledge graph constructs an inference unit, and the following formula is used to construct the improved knowledge graph:

[0014] KG={(N,E,R),Φ(N),Ψ(E)}

[0015] Where KG stands for improved knowledge graph; N is the set of nodes in the knowledge graph, each node represents a device, component, or related concept in the industrial Internet of Things; E is the set of edges, which represent the relationship between nodes; R is the set of relationship types; Φ(N) is the node attribute function, which is used to describe the attribute information of the node, including the device model and manufacturer; Ψ(E) is the edge attribute function, which is used to describe the attribute information of the edge, including the device connection method and data transmission rate; in the reasoning process, the following formula is used:

[0016] I=Ω(KG,Q)

[0017] Among them, I represents the reasoning result; Ω is the reasoning function; Q is the reasoning query statement, which is used to specify the reasoning goal and conditions.

[0018] Furthermore, the mapping function of the optimized hypergraph neural network feature extraction unit that converts the knowledge graph into a hypergraph structure is:

[0019] H = Γ(N, E, θ)

[0020] Where H represents the converted hypergraph; N and E are the node set and edge set of the knowledge graph respectively; θ is the mapping parameter set, which is used to control various parameters in the mapping process; Γ is the hypergraph topology structure mapping function. The feature extraction formula for optimizing the hypergraph neural network is:

[0021] F=Λ(H,W)

[0022] Where F represents the extracted feature vector; W is the weight matrix of the optimized hypergraph neural network; Λ is the feature extraction function.

[0023] Furthermore, the device collaboration strategy generation unit uses the following formula when generating the collaborative work strategy:

[0024] S = Π(P, F, λ)

[0025] Among them, S represents the generated collaborative work strategy; P is the set of parameters for the collaboration of industrial IoT devices, including the device workload P wl , response time threshold P rt ; F is the feature representation output by the optimized hypergraph neural network feature extraction unit; λ is the set of strategy generation weight parameters; Π is the collaborative strategy generation function.

[0026] Furthermore, the collaborative strategy execution scheduling unit adopts the following formula when dynamically scheduling the strategy execution process:

[0027] D=Σ(S,R f , μ)

[0028] Where D represents the dynamic scheduling result; R fA set of real-time feedback information of the device, including the current running state R of the device fs , the progress of the executed task R fp ; μ is a set of scheduling parameters; Σ is a dynamic scheduling function; S represents the generated collaborative work strategy.

[0029] Further, the system running state monitoring and evaluation unit uses the following formula when evaluating the system running state:

[0030] A = Υ(D, M)

[0031] Where A represents the evaluation result of the system running state; D is a set of execution data of the collaborative strategy execution scheduling unit, including the success rate of strategy execution D ps , execution time deviation D td ; M is a preset state evaluation index system; Υ is a state evaluation function.

[0032] Further, the improved knowledge graph construction and reasoning unit uses the following formula when updating the knowledge graph:

[0033] KG new = ρ(KG old , ΔN, ΔE)

[0034] Where KG new represents the updated knowledge graph; KG old represents the original knowledge graph; ΔN is a set of new nodes; ΔE is a set of new edges; ρ is a knowledge graph update function.

[0035] Further, the optimization hypergraph neural network feature extraction unit uses the following formula when optimizing the hypergraph neural network:

[0036] W opt = σ(W init , L, η)

[0037] Where W opt represents the optimized weight matrix; W init represents the initial weight matrix; L is a loss function used to measure the difference between the feature extraction result and the actual data; η is the optimization learning rate; σ is the weight optimization function.

[0038] Further, the device collaborative strategy generation unit uses the following formula when adjusting the collaborative strategy:

[0039] S adj = τ(S, A, ω)

[0040] Where S adjrepresents the adjusted collaborative work strategy; S is the original collaborative work strategy; A represents the system operation status evaluation result; ω is the strategy adjustment parameter set; τ is the collaborative strategy adjustment function.

[0041] Beneficial Effects: This invention proposes a data- and knowledge-driven collaborative management and control system for industrial IoT devices. This system utilizes a heterogeneous multi-source data fusion and interaction unit, a customized industrial-grade data interface protocol, and an adaptive data encoding and conversion mechanism to enable the parallel collection and unified format conversion of structured, semi-structured, and unstructured data. This system breaks down data format barriers, ensures the smooth flow and interaction of data from different devices, fully unlocks data value, and accurately presents device operating status. Regarding device collaborative strategy formulation and execution, the system constructs an improved knowledge graph, utilizes a semantic relationship mining algorithm to identify data semantic associations, performs logical reasoning based on the knowledge graph's topological structure, and deeply analyzes complex relationships between devices. It then deeply extracts data features through an optimized hypergraph neural network and generates precise strategies based on device collaborative parameters. The collaborative strategy execution scheduling unit dynamically schedules strategy execution based on real-time device feedback, ensuring that the strategy adapts to actual production changes. Furthermore, the system's operational status monitoring and evaluation unit collects device operational data in real time and evaluates system status, feeding the results back to the knowledge graph construction and strategy generation stages. This enables continuous system optimization, significantly improving device collaborative efficiency and industrial production benefits, and ensuring efficient and stable industrial production. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a diagram of the system unit composition of the present invention;

[0043] Figure 2 It is a flow chart of the system operation steps of the present invention. DETAILED DESCRIPTION

[0044] It should be noted that, unless there is a conflict, the embodiments in this application and the features described in the embodiments can be combined with each other. The application is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] like Figure 1 As shown in the figure, the data- and knowledge-driven collaborative management and control system for industrial IoT equipment includes: a heterogeneous multi-source data fusion and interaction unit, an improved knowledge graph construction and reasoning unit, an optimized hypergraph neural network feature extraction unit, an equipment collaborative strategy generation unit, a collaborative strategy execution and scheduling unit, and a system operation status monitoring and evaluation unit;

[0046] The heterogeneous multi-source data fusion and interaction unit is used to realize the parallel collection and transmission of structured, semi-structured and unstructured data from different production links and different types of equipment in the Industrial Internet of Things through a customized industrial-grade data interface protocol, and convert the collected heterogeneous data into a unified data expression format through an adaptive data encoding conversion mechanism for subsequent processing;

[0047] Specifically, this unit serves as the data entry point for the collaborative management and control system of industrial IoT devices, primarily responsible for addressing the wide range of data sources and diverse formats in industrial scenarios. Its customized industrial-grade data interface protocol supports at least 10 mainstream industrial communication standards, including but not limited to ModbusTCP, OPCUA, Profinet, and EtherCAT. It can simultaneously collect structured, semi-structured, and unstructured data from different types of devices, including sensors, controllers, and smart meters. The data acquisition rate can be dynamically adjusted within the range of 100KB / s to 10MB / s based on device performance.

[0048] The adaptive data encoding conversion mechanism plays a key role in data format conversion. This mechanism includes over 2,000 built-in data format conversion rules. By automatically identifying metadata in the data header and data structure characteristics, it can convert heterogeneous data into a unified JSON-LD format within 500 milliseconds. This format not only facilitates data storage but also effectively preserves the data's semantic information, providing a standardized data foundation for subsequent improvements to the knowledge graph and enabling seamless data flow across all components of the system.

[0049] The improved knowledge graph construction reasoning unit is used to identify the semantic associations between data based on the converted unified format data using a semantic relationship mining algorithm, construct an improved knowledge graph with multi-level and multi-dimensional relationships, and use the topological structure of the knowledge graph to perform reasoning analysis on information such as equipment operating status and potential faults using logical reasoning rules;

[0050] Specifically, the improved knowledge graph constructs reasoning units based on uniformly formatted data, building a knowledge network with deep semantic associations. The semantic relationship mining algorithm uses a three-level parsing structure. First, vocabulary-level parsing is performed to identify basic information such as device names and parameter terms. Then, sentence-level parsing is performed to determine the grammatical relationships between data. Finally, semantic-level parsing is used to mine deep semantic associations such as device functions, production processes, and parameter dependencies. The constructed knowledge graph has a multi-level architecture and can be divided into a device entity layer, a functional relationship layer, a system process layer, and an enterprise operation layer. It supports the storage of no less than 1 million entity relationship data.

[0051] During the inference process, the topological structure of the knowledge graph and over 1,000 predefined logical reasoning rules are used to analyze equipment operating status and potential risks. The inference engine utilizes a hybrid inference strategy combining forward and backward reasoning. The inference response time for equipment anomalies is less than 2 seconds, and the inference accuracy has been verified to exceed 85%, providing a critical basis for equipment failure prediction and performance optimization.

[0052] The optimized hypergraph neural network feature extraction unit is used to take the node and edge information of the improved knowledge graph as input, convert the knowledge graph into a hypergraph structure through a customized hypergraph topology structure mapping function, and then use the optimized hypergraph neural network to deeply extract and abstractly represent the data features in the hypergraph structure;

[0053] Specifically, an optimized hypergraph neural network feature extraction unit converts the knowledge graph into a hypergraph structure, enabling deep feature extraction of complex device relationships. A customized hypergraph topology mapping function reorganizes knowledge graph nodes and edges to convert traditional graph structures into a hypergraph structure. Each hyperedge can represent multiple relationships between 3-10 nodes. This converted hypergraph structure improves node relationship representation by over 300% compared to traditional graph structures.

[0054] The optimized hypergraph neural network adopts a multi-layered architecture, consisting of no fewer than five convolutional layers and three pooling layers. A multi-head attention mechanism is integrated into the network, extracting node and hyperedge features from eight different dimensions. Using residual connections and batch normalization techniques, it effectively addresses the vanishing gradient problem in deep network training, accelerating network convergence by 40%. The final output feature vector is kept between 200 and 500 dimensions, preserving key information while facilitating computation in subsequent strategy generation units.

[0055] The device collaboration strategy generation unit is used to receive the feature representation output by the optimized hypergraph neural network feature extraction unit, and generate collaborative work strategies for different device combinations based on preset collaboration strategy generation rules and various parameters of industrial Internet of Things device collaboration, such as device workload and response time threshold;

[0056] Specifically, the device collaboration strategy generation unit optimizes features extracted from a hypergraph neural network and combines them with multi-dimensional parameters for IIoT device collaboration to generate a strategy. The parameter system it considers covers 18 core parameters, including device workload (CPU usage, memory usage, etc.), response time threshold (data processing latency requirement), energy consumption index (energy consumption per unit of output), and maintenance cycle. Each parameter is trained using historical data to form a dynamic weight model, with a weight adjustment accuracy of up to 0.01.

[0057] The policy generation rule adopts a hierarchical decision framework, the first layer filters out a set of feasible strategies that meet the basic conditions based on the real-time state of the equipment, the filtering rule contains no less than 50 constraint conditions; the second layer selects the optimal strategy combination from the feasible strategy set through a multi-objective optimization algorithm, the optimization objectives include 6 dimensions such as equipment utilization, energy consumption, production efficiency, etc. The final generated collaborative strategy covers equipment start-stop sequence, task allocation, resource scheduling, etc., and the strategy generation time is controlled within 3 seconds to meet the real-time demand of industrial production.

[0058] The collaborative strategy execution scheduling unit is used for the collaborative working strategy generated by the equipment collaborative strategy generation unit, through the industrial Internet of Things equipment communication network, the strategy instruction is accurately issued to the corresponding industrial Internet of Things equipment, and through the real-time feedback information of the equipment, the dynamic scheduling and adjustment of the strategy execution process are carried out.

[0059] Specifically, the collaborative strategy execution scheduling unit is responsible for accurately issuing the generated strategy to the equipment and dynamically managing the execution process. It interacts with the equipment through industrial Ethernet, 5G and other communication networks, supports TCP / IP, UDP and other protocols, the instruction transmission delay is not more than 80 milliseconds, and the transmission success rate is guaranteed to be more than 99.9%. The unit is built-in priority queue scheduling algorithm, which can be scheduled according to the urgency and importance of the strategy task, to ensure that critical tasks are executed first.

[0060] During the strategy execution process, the unit collects real-time feedback information such as the running state and task progress of the equipment at a frequency of 10 times per second. When detecting equipment abnormalities (such as load overrun, response timeout), the dynamic scheduling mechanism is immediately started. This mechanism adopts a local reconfiguration strategy, compares the current state of the equipment with the expected state of the strategy, calculates the minimum adjustment scheme within 1 second, reallocates tasks or adjusts equipment parameters, and ensures the stability and reliability of the strategy execution.

[0061] The system running state monitoring and evaluation unit is used for real-time collection of running data of industrial Internet of Things equipment during collaborative working, including equipment energy consumption, running time, fault frequency and other parameters, through a preset state evaluation index system, the overall running state of the system is monitored and evaluated, and the evaluation results are fed back to the improvement knowledge graph construction and reasoning unit and the equipment collaborative strategy generation unit.

[0062] Specifically, the system operation state monitoring and evaluation unit collects more than 50 operation data in real time in the collaborative work of industrial Internet of Things devices, including device performance parameters (speed, pressure, temperature, etc.), operation time, fault frequency, task completion rate and other indicators. The data collection frequency can be set within 1-100 times / second according to the device requirements. The preset state evaluation index system determines the weight of each index by using the analytic hierarchy process, and the weight calculation accuracy reaches 4 decimal places. Combined with the fuzzy comprehensive evaluation method, the overall operation state of the system is quantitatively evaluated.

[0063] The evaluation results are presented in the form of quantitative scores (0-100 points) and are divided into three state levels: normal, warning and fault. The evaluation results will be fed back to the improved knowledge graph construction reasoning unit and the device collaborative strategy generation unit in real time, with a feedback delay of no more than 1.5 seconds. When the system state score is lower than the warning threshold, the warning mechanism is automatically triggered to provide data support for system optimization and strategy adjustment, forming a closed-loop management system.

[0064] The output end of the heterogeneous multi-source data fusion interaction unit is connected with the input end of the improved knowledge graph construction reasoning unit, the output end of the improved knowledge graph construction reasoning unit is connected with the input end of the optimized hypergraph neural network feature extraction unit, the output end of the optimized hypergraph neural network feature extraction unit is connected with the input end of the device collaborative strategy generation unit, the output end of the device collaborative strategy generation unit is connected with the input end of the collaborative strategy execution scheduling unit, and the output ends of the collaborative strategy execution scheduling unit and the system operation state monitoring and evaluation unit are respectively connected with the input ends of the improved knowledge graph construction reasoning unit and the device collaborative strategy generation unit.

[0065] Preferably, in the improved knowledge graph construction reasoning unit, the following formula is used when constructing the improved knowledge graph:

[0066] KG={(N,E,R),Φ(N),Ψ(E)}

[0067] Wherein, KG represents the improved knowledge graph; N is a node set in the knowledge graph, each node represents a device, component or related concept in the industrial Internet of Things; E is an edge set, which represents the relationship between nodes; R is a relationship type set; Φ(N) is a node attribute function, which is used to describe the attribute information of the node, such as device model, manufacturer, etc.; Ψ(E) is an edge attribute function, which is used to describe the attribute information of the edge, such as device connection method, data transmission rate, etc. In the reasoning process, the following formula is used:

[0068] I=Ω(KG,Q)

[0069] Wherein, I represents the reasoning result; Ω is a reasoning function; Q is a reasoning query statement, which is used to specify the target and conditions of reasoning.

[0070] Specifically, the improved knowledge graph construction reasoning unit constructs a knowledge graph for the complex relationships between industrial Internet of Things devices using a semantic relationship mining algorithm. This algorithm identifies logical and semantic associations between device entities, parameters, and processes through three levels of lexical, syntactic, and semantic analysis, and constructs a multi-level knowledge graph containing device, system, process, and enterprise layers, which can store more than one million entity relationship data. The reasoning engine uses a forward and backward mixed reasoning strategy based on more than one thousand predefined logical rules to analyze the device operating status and potential risks in real time, with a response time of less than 2 seconds and an accuracy of more than 85%. The incremental update mechanism only reconstructs the relevant subgraph according to the data changes, which improves the update efficiency by 60%, ensuring that the knowledge graph reflects the actual state of the device in real time and provides accurate basis for decision-making.

[0071] Preferably, in the optimization hypergraph neural network feature extraction unit, the mapping function for converting the knowledge graph into a hypergraph structure is:

[0072] H = Γ(N, E, θ)

[0073] Where H represents the converted hypergraph; N and E are the node set and edge set of the knowledge graph, respectively; θ is a set of mapping parameters used to control various parameters in the mapping process; and Γ is a hypergraph topology mapping function. The feature extraction formula of the optimization hypergraph neural network is:

[0074] F = Λ(H, W)

[0075] Where F represents the extracted feature vector; W is the weight matrix of the optimization hypergraph neural network; and Λ is the feature extraction function.

[0076] Specifically, the optimization hypergraph neural network feature extraction unit uses a custom topology mapping function to convert the knowledge graph into a hypergraph structure, and each hyperedge can represent the multi-element relationship of 3-10 nodes, with a 300% improvement in relationship expression capability compared to traditional graph structures. The network architecture includes 5 convolutional layers and 3 pooling layers, integrates 8 attention mechanisms to extract node and hyperedge features from multiple dimensions, and uses residual connection and batch normalization technology to solve the training difficulties of deep networks, accelerating the convergence speed of the model by 40%. The output 200-500 dimensional feature vector greatly reduces the computational complexity while retaining key information, providing efficient and accurate feature representation for device collaboration strategy generation.

[0077] Preferably, in the device collaboration strategy generation unit, the following formula is used when generating a collaborative work strategy:

[0078] S = Π(P, F, λ)

[0079] Where S represents the generated collaborative work strategy; P is a set of parameters for industrial Internet of Things device collaboration, including device work load Pwl , response time threshold P rt ; λ is a set of policy generation weight parameters; Π is a collaborative policy generation function.

[0080] Specifically, the device collaborative policy generation unit constructs an evaluation system covering 18 core parameters, including device workload, response time threshold, energy consumption indicators, etc. The weight of each parameter is dynamically adjusted through historical data training, with a precision of 0.01. A hierarchical decision-making framework is adopted. The first layer filters the feasible strategy set through more than 50 constraint rules, and the second layer uses a multi-objective optimization algorithm to balance the device utilization, energy consumption, production efficiency, etc. in six dimensions. After receiving the feature vector, the strategy combination calculation can be completed within 3 seconds, and the complete collaborative strategy including device start-stop timing, task allocation, and resource scheduling parameters is output, meeting the dual requirements of real-time and accuracy of industrial production.

[0081] Preferably, in the collaborative strategy execution scheduling unit, the following formula is used when dynamically scheduling the strategy execution process:

[0082] D = Σ (S, R f , μ)

[0083] Where D represents the dynamic scheduling result; S is the collaborative working strategy generated by the device collaborative policy generation unit; R f is a set of real-time feedback information of the device, including the current running state R fs , the progress of the executed task R fp , etc.; μ is a set of scheduling parameters; Σ is a dynamic scheduling function.

[0084] Specifically, the collaborative strategy execution scheduling unit transmits strategy instructions through industrial Ethernet, 5G, etc. communication network, supports TCP / IP, UDP protocol, transmission delay control within 80 milliseconds, success rate reaches 99.9%. The priority queue scheduling algorithm is used for task hierarchical management, and the device state is collected at a frequency of 10 times per second. When detecting abnormal conditions such as device load overrun and response timeout, the local reconfiguration mechanism is started within 1 second, the minimum adjustment scheme is generated through state deviation analysis, and the task allocation or device parameters are dynamically adjusted to ensure the stability and reliability of the strategy execution, and reduce the impact of abnormal conditions on production.

[0085] Preferably, in the system running state monitoring and evaluation unit, the following formula is used when evaluating the system running state:

[0086] A = Υ (D, M)

[0087] Where A represents the system running state evaluation result; D is the execution data set of the collaborative strategy execution scheduling unit, including the strategy execution success rate D ps , execution time deviation D tdand so on; M is a preset state evaluation index system; Y is a state evaluation function.

[0088] Specifically, the system operation state monitoring and evaluation unit collects more than 50 equipment operation indicators in real time, and the collection frequency supports dynamic configuration of 1-100 times / second. The analytic hierarchy process is used to determine the index weight, and the calculation accuracy is up to 4 decimal places. Combined with the fuzzy comprehensive evaluation method, a quantitative score of 0-100 points is generated. When the score is lower than the early warning threshold, a multi-level early warning mechanism is triggered, and the evaluation result is fed back to the knowledge graph and strategy generation unit with a 1.5 second delay. This closed-loop management system effectively monitors key indicators such as equipment energy consumption and operation efficiency, provides data support for continuous optimization of the system, and ensures efficient operation of industrial production.

[0089] Preferably, in the improved knowledge graph construction and reasoning unit, the following formula is used when updating the knowledge graph:

[0090] KG new =ρ(KG old ,ΔN,ΔE)

[0091] Where KG new represents the updated knowledge graph; KG old represents the original knowledge graph; ΔN is the set of new nodes; ΔW is the set of new edges; and ρ is the knowledge graph update function.

[0092] Specifically, the dynamic update module of the improved knowledge graph construction and reasoning unit uses an incremental knowledge fusion algorithm to compare the entity identifiers, relationship types and attribute differences between new and old data, and identifies the nodes and edges that need to be updated. When updating, the influence range of data changes is determined first, and the affected subgraph is reconstructed locally to avoid resource consumption caused by full graph update. This algorithm supports processing more than 1000 data change requests per second, and the impact of updating on system performance is controlled within 5%, ensuring that the knowledge graph synchronizes with the state changes of industrial field devices in real time and maintains the accuracy and timeliness of the knowledge network.

[0093] Preferably, in the optimized hypergraph neural network feature extraction unit, the following formula is used when optimizing the hypergraph neural network:

[0094] W opt =σ(W init ,L,η)

[0095] Where W opt represents the optimized weight matrix; W init represents the initial weight matrix; L is the loss function, which is used to measure the difference between the feature extraction result and the actual data; η is the optimization learning rate; and σ is the weight optimization function.

[0096] Specifically, the parameter optimization module of the hypergraph neural network feature extraction unit adopts an adaptive learning rate adjustment strategy, iteratively optimizes network parameters in combination with gradient accumulation and weight decay techniques. By monitoring the change of the loss function, the learning rate step is dynamically adjusted. A large step is used in the early stage of training to speed up convergence, and the step is reduced in the later stage to improve optimization accuracy. The network weight is globally adjusted every 100 training iterations. Through practical verification, the model can reach a stable state within 1000 training rounds. Compared with the fixed parameter training method, the convergence efficiency is improved by 35%, ensuring that the network is always in an efficient operating state.

[0097] Preferably, when adjusting the collaborative strategy in the device collaborative strategy generation unit, the following formula is used:

[0098] S adj = τ (S, A, ω)

[0099] Wherein, S adj represents the adjusted collaborative work strategy; S is the original collaborative work strategy; ω is a set of strategy adjustment parameters; τ is a collaborative strategy adjustment function.

[0100] Specifically, the dynamic adjustment module of the device collaborative strategy generation unit constructs a strategy adaptability evaluation index based on the system running state evaluation result. By comparing the deviation of the strategy execution effect and the expected target, a multi-dimensional decision tree model is used to analyze the adjustment priority. A local fine-tuning strategy is adopted. For different scenarios such as device load fluctuation and task progress delay, an optimization scheme containing parameter correction and task reallocation is generated within 3 seconds, ensuring that the collaborative strategy maintains optimal execution effect in a dynamic production environment. The strategy adjustment coverage rate is more than 98%, improving the flexibility and stability of industrial production.

[0101] As Figure 2 shown, based on the data and knowledge driven industrial internet of things device collaborative management and control system, the system includes the following steps:

[0102] First, through the heterogeneous multi-source data fusion interaction unit, according to the customized industrial level data interface protocol, the structured, semi-structured and unstructured data from different production links and different types of devices in the industrial internet of things are collected in parallel. The adaptive data coding and conversion mechanism is used to convert the collected heterogeneous data into a unified data expression format;

[0103] Second, the improved knowledge graph construction and reasoning unit identifies the semantic association between data based on the converted unified format data, constructs an improved knowledge graph with multi-level and multi-dimensional relationships by means of semantic relationship mining algorithm, and uses logical reasoning rules to analyze and reason the device running state and potential fault information;

[0104] In the third step, the optimized hypergraph neural network feature extraction unit takes the improved node and edge information of the knowledge graph as input, converts the knowledge graph into a hypergraph structure through a self-defined hypergraph topology mapping function, and uses the optimized hypergraph neural network to deeply extract and abstractly represent the data features in the hypergraph structure.

[0105] In the fourth step, the device coordination strategy generation unit receives the feature representation output by the optimized hypergraph neural network feature extraction unit, generates a coordination work strategy for different device combinations according to the preset coordination strategy generation rules and combining various parameters of industrial Internet of Things device coordination, including device work load and response time threshold.

[0106] In the fifth step, the coordination strategy execution and scheduling unit accurately issues the strategy instructions to the corresponding industrial Internet of Things devices through the industrial Internet of Things device communication network according to the coordination work strategy generated by the device coordination strategy generation unit, and dynamically schedules and adjusts the strategy execution process according to the real-time feedback information of the devices.

[0107] In the sixth step, the system running state monitoring and evaluation unit collects the running data of the industrial Internet of Things devices in the coordination work process in real time, including device energy consumption, running time, and fault frequency parameters, monitors and evaluates the overall running state of the system through the preset state evaluation index system, and feeds back the evaluation results to the improved knowledge graph construction and reasoning unit and the device coordination strategy generation unit for subsequent system optimization and strategy adjustment.

[0108] The data and knowledge driven industrial Internet of Things device coordination management system significantly improves the intelligent level and management efficiency of device coordination in the industrial Internet of Things environment with its innovative architecture and core technology.

[0109] The system realizes parallel collection and standardized conversion of structured, semi-structured and unstructured data generated by different production links and different types of devices through the heterogeneous multi-source data fusion interaction unit, using customized industrial-grade data interface protocol and adaptive data coding and conversion mechanism. This design breaks down the barriers of non-uniform data formats in traditional systems, enabling smooth flow and interaction of various device data, providing a high-quality data foundation for subsequent in-depth analysis, thereby overcoming the problem of heterogeneous data difficult to effectively fuse in the background technology.

[0110] In terms of device collaboration strategy formulation, the system constructs an improved knowledge graph and combines it with an optimized hypergraph neural network. The improved knowledge graph construction reasoning unit uses a semantic relationship mining algorithm to identify semantic associations between data and construct a multi-level, multi-dimensional knowledge network, providing a powerful tool for device relationship analysis. The optimized hypergraph neural network feature extraction unit converts the knowledge graph into a hypergraph structure, performs deep feature extraction, and accurately captures the complex collaborative relationships between devices. Based on these technologies, the device collaboration strategy generation unit can combine parameters such as device workload and response time threshold to generate highly adaptive collaboration strategies, overcoming the lack of in-depth analysis in traditional system strategy formulation.

[0111] The system also has the ability to dynamically optimize and continuously improve. The collaborative strategy execution scheduling unit dynamically adjusts the strategy execution process based on real-time feedback from the equipment to ensure that the system is always in the best operating state. The system operation status monitoring and evaluation unit collects parameters such as equipment energy consumption and operating time in real time, evaluates the system status based on a preset indicator system, and feeds the results back to the knowledge graph construction and strategy generation links to form a closed-loop optimization mechanism. This design enables the system to continuously adapt to changes in the industrial production environment and continuously improve the collaborative efficiency of equipment. It effectively solves the problems of rigid policy execution and the inability to dynamically adjust in traditional systems, and comprehensively improves the intelligence and refinement of collaborative management and control of industrial Internet of Things equipment.

[0112] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0113] While embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. Data- and knowledge-driven collaborative management and control system for industrial IoT equipment, characterized by: include: Heterogeneous multi-source data fusion interaction unit, improved knowledge graph construction reasoning unit, optimized hypergraph neural network feature extraction unit, equipment collaboration strategy generation unit, collaboration strategy execution scheduling unit, system operation status monitoring and evaluation unit; The heterogeneous multi-source data fusion and interaction unit is used to collect and transmit structured, semi-structured, and unstructured data from different production links and different types of equipment in the Industrial Internet of Things in parallel through a customized industrial-grade data interface protocol, and convert the collected heterogeneous data into a unified data expression format through an adaptive data encoding conversion mechanism for subsequent processing; The improved knowledge graph construction reasoning unit is used to identify the semantic associations between data based on the converted unified format data using a semantic relationship mining algorithm, construct an improved knowledge graph with multi-level and multi-dimensional relationships, and use the topological structure of the knowledge graph to perform reasoning analysis on the equipment operation status and potential fault information using logical reasoning rules; The optimized hypergraph neural network feature extraction unit is used to take the node and edge information of the improved knowledge graph as input, convert the knowledge graph into a hypergraph structure through a customized hypergraph topology structure mapping function, and then use the optimized hypergraph neural network to deeply extract and abstractly represent the data features in the hypergraph structure.

2. The data and knowledge-driven collaborative management and control system for industrial IoT devices according to claim 1 is characterized in that: The device collaboration strategy generation unit is used to receive the feature representation output by the optimized hypergraph neural network feature extraction unit, and generate collaborative work strategies for different device combinations based on preset collaboration strategy generation rules and various parameters of industrial Internet of Things device collaboration, including device workload and response time threshold; The collaborative strategy execution scheduling unit is used to accurately send the strategy instructions to the corresponding industrial Internet of Things devices through the industrial Internet of Things device communication network based on the collaborative work strategy generated by the device collaborative strategy generation unit, and dynamically schedule and adjust the strategy execution process based on the real-time feedback information of the devices; The system operation status monitoring and evaluation unit is used to collect real-time operation data of industrial Internet of Things devices during collaborative work, including equipment energy consumption, operation time, and number of failures. Through a preset status evaluation indicator system, the overall operation status of the system is monitored and evaluated, and the evaluation results are fed back to the improved knowledge graph construction reasoning unit and the equipment collaboration strategy generation unit; The improved knowledge graph constructs an inference unit, and the following formula is used to construct the improved knowledge graph: KG={(N,E,R),Φ(N),Ψ(E)} Where KG stands for improved knowledge graph; N is the set of nodes in the knowledge graph, each node represents a device, component, or related concept in the industrial Internet of Things; E is the set of edges, which represent the relationship between nodes; R is the set of relationship types; Φ(N) is the node attribute function, which is used to describe the attribute information of the node, including the device model and manufacturer; Ψ(E) is the edge attribute function, which is used to describe the attribute information of the edge, including the device connection method and data transmission rate; in the reasoning process, the following formula is used: I=Ω(KG,Q) Among them, I represents the reasoning result; Ω is the reasoning function; Q is the reasoning query statement, which is used to specify the reasoning goal and conditions.

3. The data and knowledge-driven collaborative management and control system for industrial IoT devices according to claim 1 is characterized in that: The mapping function of the optimized hypergraph neural network feature extraction unit to convert the knowledge graph into a hypergraph structure is: H = Γ(N, E, θ) Where H represents the converted hypergraph; N and E are the node set and edge set of the knowledge graph respectively; θ is the mapping parameter set, which is used to control various parameters in the mapping process; Γ is the hypergraph topology structure mapping function. The feature extraction formula for optimizing the hypergraph neural network is: F=Λ(H,W) Where F represents the extracted feature vector; W is the weight matrix of the optimized hypergraph neural network; Λ is the feature extraction function.

4. The data and knowledge-driven collaborative management and control system for industrial IoT devices according to claim 1 is characterized in that: The device collaboration strategy generation unit generates a collaborative work strategy using the following formula: S = Π(P, F, λ) Among them, S represents the generated collaborative work strategy; P is the set of parameters for the collaboration of industrial IoT devices, including the device workload P wl , response time threshold P rt ; F is the feature representation output by the optimized hypergraph neural network feature extraction unit; λ is the set of strategy generation weight parameters; Π is the collaborative strategy generation function.

5. The data and knowledge-driven collaborative management and control system for industrial IoT devices according to claim 1 is characterized in that: The collaborative strategy execution scheduling unit uses the following formula when dynamically scheduling the strategy execution process: D=Σ(S,R f ,m) Where D represents the dynamic scheduling result; R f It is a real-time feedback information set of the equipment, including the current operating status of the equipment R fs 、The progress of the executed tasks R fp ; μ is the scheduling parameter set; Σ is the dynamic scheduling function; S represents the generated collaborative work strategy.

6. The data and knowledge-driven industrial IoT equipment collaborative management and control system according to claim 1 is characterized in that: The system operation status monitoring and evaluation unit uses the following formula to evaluate the system operation status: A=Y(D,M) Among them, A represents the system operation status evaluation result; D is the execution data set of the collaborative strategy execution scheduling unit, including the strategy execution success rate D ps , execution time deviation D td ; M is the preset state evaluation index system; Υ is the state evaluation function.

7. The data and knowledge-driven collaborative management and control system for industrial IoT devices according to claim 1 is characterized in that: The improved knowledge graph constructs an inference unit and uses the following formula to update the knowledge graph: KG new =ρ(KG old ,ΔN,ΔE) Among them, KG new Represents the updated knowledge graph; KG old represents the original knowledge graph; ΔN is the set of newly added nodes; ΔE is the set of newly added edges; ρ is the knowledge graph update function.

8. The data and knowledge-driven industrial IoT equipment collaborative management and control system according to claim 1 is characterized in that: The optimized hypergraph neural network feature extraction unit uses the following formula when optimizing the hypergraph neural network: W opt =σ(W init ,L,h) Among them, W opt Represents the optimized weight matrix; W init represents the initial weight matrix; L is the loss function, which is used to measure the difference between the feature extraction result and the actual data; η is the optimized learning rate; σ is the weight optimization function.

9. The data and knowledge-driven collaborative management and control system for industrial Internet of Things equipment according to claim 1 is characterized in that: The device coordination strategy generation unit uses the following formula when adjusting the coordination strategy: S adj =τ(S, A, ω) Among them, S adj represents the adjusted collaborative work strategy; S is the original collaborative work strategy; A represents the system operation status evaluation result; ω is the strategy adjustment parameter set; τ is the collaborative strategy adjustment function.

Images