Industrial equipment communication method and system based on industrial internet

Abstract

The invention discloses an industrial equipment communication method and system based on the industrial internet, and the method comprises the following steps: S1, collecting the multi-modal data of industrial equipment, and generating a preliminary multi-modal fusion feature; s2, performing event-driven analysis on the preliminary multi-modal fusion features to generate event sensing data; s3, constructing a semantic-driven communication protocol, and generating a semantic data queue with a priority identifier; s4, compressing the semantic data queue and carrying out time consistency calibration; s5, optimizing a communication frequency and bandwidth resource allocation strategy in combination with a mixed firefly optimization algorithm and a dynamic differential evolution algorithm; s6, constructing an industrial equipment communication topology model, and predicting path delay and bandwidth occupation indexes of the communication network in combination with the space-time multi-task network; and S7, constructing a zero-sum game model to analyze the potential confrontation relationship in the communication process, and outputting a secure communication scheme. According to the invention, semantic-driven communication, a hybrid optimization algorithm and the like are adopted, and intelligent communication of industrial equipment is realized.

Description

Technical Field

[0001] The present invention relates to the field of industrial equipment communication technology, and in particular to an industrial equipment communication method and system based on the Industrial Internet. Background Art

[0002] With the rapid development of industrial Internet technology, communication and collaboration between industrial equipment have gradually become an important foundation in the field of intelligent manufacturing. In the industrial Internet environment, data interaction between devices not only involves traditional communication technologies, but also needs to deal with issues such as the fusion of multimodal data, dynamic optimization of complex networks, and security protection during communication. Existing industrial equipment communication technologies mainly focus on the following aspects, including data collection and fusion, event analysis, communication protocol optimization, resource allocation and path selection, and secure communication. However, these technologies still have many shortcomings in the face of increasingly complex industrial scenarios.

[0003] First, in the actual application of the Industrial Internet, industrial equipment generates a wide variety of multimodal data, including vibration data, temperature and humidity data, electromagnetic signal data, and equipment operation status data. These data are highly heterogeneous and dynamic in time and space dimensions, and traditional single-modal data processing methods are difficult to effectively fuse multimodal data. In addition, the dynamic change characteristics of multimodal data require more efficient fusion technology to generate reliable equipment feature data in real time to support subsequent analysis and decision-making. However, existing multimodal data fusion technologies often have problems such as information loss, low fusion efficiency, and high computing resource consumption when processing complex dynamic data scenarios, and cannot adapt to scenarios with high real-time and high reliability requirements for industrial equipment.

[0004] Secondly, existing event-driven analysis technologies also have obvious deficiencies when dealing with complex event identification in industrial scenarios. Events in industrial equipment communications are sparsely distributed, that is, valuable event information only accounts for a very small proportion of the data stream. Traditional event detection algorithms are easily interfered by irrelevant data when identifying these sparse events, resulting in reduced recognition accuracy. In addition, existing event analysis technologies are usually based on fixed perception frequencies, which can easily lead to resource waste or failure to capture key events in a timely manner in dynamic industrial scenarios. The lack of an adaptive perception frequency optimization mechanism makes industrial equipment communications inefficient when dealing with sudden events.

[0005] At the communication protocol level, traditional communication protocols are usually based on static rules and fixed priorities, and cannot dynamically allocate resources according to the importance of data. The data transmitted in industrial equipment communication has different priorities and semantic values. The existing technology is insufficient in identifying and dynamically scheduling data priorities, which can easily lead to unreasonable allocation of communication resources and the real-time performance of high-priority tasks cannot be guaranteed. In addition, traditional communication protocols do not make sufficient use of the semantic information of data content, making it difficult to optimize for the special semantic requirements in industrial equipment communication.

[0006] In terms of resource allocation and path optimization, existing optimization methods usually adopt a single-objective optimization model, such as minimizing communication delay or maximizing bandwidth utilization. This single-objective optimization strategy is difficult to balance multiple key performance indicators such as communication efficiency, resource consumption, and network load. At the same time, the path selection algorithms in the existing technology are usually based on static network topology structures and cannot adapt to the dynamically changing communication needs and equipment status in industrial scenarios. In addition, traditional optimization algorithms often have problems such as slow convergence speed and difficulty in obtaining global optimal solutions when dealing with large-scale network optimization problems, which limits their application in complex industrial communication networks.

[0007] In terms of secure communications, the risk of adversarial attacks in industrial equipment communications is increasing, such as data theft, replay attacks, and masquerade attacks. Existing security protection technologies usually rely on fixed keys or simple encryption strategies, which are difficult to dynamically adapt to complex and changing attack scenarios. In addition, traditional attack detection and protection mechanisms are often independent of communication paths and resource optimization processes, lacking deep coordination with the communication process, making it difficult to strike a balance between security and efficiency.

[0008] Therefore, how to provide an industrial equipment communication method and system based on the Industrial Internet is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0009] One purpose of the present invention is to propose an industrial equipment communication method and system based on the industrial Internet. The present invention adopts methods such as multimodal data fusion, semantic-driven communication protocol, hybrid optimization algorithm, graph neural network, spatiotemporal multitask network and game theory security strategy to build an intelligent industrial equipment communication system, which realizes dynamic fusion of multimodal data, efficient optimization of communication resources, intelligent planning of paths and security of communication process. Through technical means such as accurate event identification, dynamic resource allocation and key generation optimization, the present invention has the advantages of high efficiency, security and adaptability, and can be widely used in equipment collaborative communication and task execution scenarios in complex industrial environments.

[0010] An industrial device communication method based on the industrial Internet according to an embodiment of the present invention includes the following steps:

[0011] S1. Collect multimodal data of industrial equipment, and dynamically fuse the multimodal data to generate preliminary multimodal fusion features;

[0012] S2. performing event-driven analysis on the preliminary multimodal fusion features, identifying perception events in industrial scenarios and optimizing perception frequencies through a sparse event detection algorithm, and generating event perception data;

[0013] S3, constructing a semantic-driven communication protocol, using a semantic embedding model to semantically process the event perception data, and quantifying the importance of the event perception data through a semantic entropy algorithm to generate a semantic data queue with a priority identifier;

[0014] S4, compressing the semantic data queue with the priority identifier, and using the multimodal synchronization mechanism to calibrate the time consistency of the compressed semantic data, and outputting the synchronized semantic data;

[0015] S5. Based on the synchronized semantic data, the hybrid firefly optimization algorithm and the dynamic differential evolution algorithm are combined to optimize the communication frequency and bandwidth resource allocation strategy to generate the optimal resource allocation plan;

[0016] S6. According to the optimal resource allocation plan, an industrial equipment communication topology model based on graph neural network is constructed, and the path delay and bandwidth occupancy indicators of the communication network are predicted by combining the spatiotemporal multi-task network to generate a dynamically optimized communication path;

[0017] S7. Based on the dynamically optimized communication path and combined with the dynamic security strategy of game theory, a zero-sum game model is constructed to analyze the potential adversarial relationship in the communication process. Deep game learning is used to optimize the dynamic key generation and exchange strategy, and a secure communication solution is output.

[0018] Optionally, the multimodal data includes vibration data, temperature and humidity data, electromagnetic signal data and equipment operation status data.

[0019] Optionally, the S2 specifically includes:

[0020] S21, record the preliminary multimodal fusion features generated by S1 as ,in Indicates Multimodal feature data, Represents the total number of multimodal features, and the set of feature values ​​of adjacent time slices is recorded as ,in Indicates the number of time slices, Indicates Multimodal eigenvalues ​​of time slices;

[0021] S22. Feature value set for adjacent time slices The features in the image are used to extract the baseline pattern, and a set of stable reference feature patterns are selected to form the basic pattern set. , select the eigenvalue of the median position of the basic mode set , and then for each time slice Calculate the maximum value of the eigenvalue set ;

[0022] S23. Based on the sparse event detection algorithm, define the sparse activation function:

[0023] ;

[0024] in, Indicates The sparse activation value of the time slice, Represents the absolute value operator;

[0025] S24. Select threshold As the event judgment standard, Time Slice Marked as event time slices, the event time slice set is defined as ;

[0026] S25: extract event attributes from the corresponding feature value set in the event time slice set, and map the feature value in each event time slice into an event set. ,in Indicates the number of events identified, for each event Assigning importance parameters , is a non-negative real number used to measure the suddenness of an event;

[0027] S26. According to the event collection The distribution of event importance parameters in , defines the perception frequency optimization function:

[0028] ;

[0029] in, represents the optimized perceived frequency, represents the exponential function, Indicates the maximum value operation. represents the median operation, Used to ensure that the logarithmic input is positive;

[0030] S27, through the optimized perception frequency Adaptively adjust the event feature collection, and finally optimize the perception frequency The filtered event features are integrated into event perception data.

[0031] Optionally, the S3 specifically includes:

[0032] S31, constructing a semantically driven communication protocol, the semantically driven communication protocol consisting of a semantic processing layer and a transmission scheduling layer; the semantic processing layer performs semantic parsing on event perception data, including semantic embedding, semantic entropy quantization and priority identification generation; the transmission scheduling layer allocates communication resources according to the priority in the semantic data queue;

[0033] S32, using a semantic embedding model to semantically process the event perception data to generate a set of semantic vectors, wherein the semantic embedding model includes an embedding layer, a context encoding layer, and a feature extraction layer;

[0034] The embedding layer maps the input event perception data vector to a high-dimensional continuous semantic space to generate an initial semantic vector;

[0035] The context encoding layer uses a self-attention mechanism to perform contextual encoding on the initial semantic vector to generate a semantic vector, wherein the semantic vector contains the contextual relationship between events;

[0036] The feature extraction layer performs dimension reduction and feature compression on the semantic vector;

[0037] S33, clustering the semantic vector set generated by the semantic embedding model to form several semantic subsets based on semantic similarity, where each semantic subset Include events with high semantic similarity;

[0038] S34. Introduce a semantic entropy algorithm in the semantic-driven communication protocol to quantify the importance of events in each semantic subset:

[0039] ;

[0040] in, Representing a semantic subset The semantic entropy value of represents the exponential function, Representing semantic vector The norm of and represents the weight coefficient;

[0041] S35. According to the semantic entropy value The size of is used to assign priority to events in the semantic subset. When the semantic entropy value is low, it means that the events in the semantic subset contain events with prominent features, and the events in the semantic subset are marked as high priority; when the semantic entropy value is high, it means that the semantic distribution of events in the semantic subset is relatively balanced, and the semantic subset is marked as low priority.

[0042] S36. Arrange the semantic events with priority identifiers from high to low priority to generate a semantic data queue with priority identifiers.

[0043] Optionally, the S5 specifically includes:

[0044] S51. Define a multi-objective optimization problem and set the communication frequency and bandwidth resource allocation strategy As the optimization variable, minimize the comprehensive optimization objective:

[0045] ;

[0046] in, represents the comprehensive optimization objective function, represents the communication delay, represents the communication energy consumption, Indicates bandwidth utilization. represents the delay dynamic weight parameter, represents the dynamic weight parameter of energy consumption, Indicates the bandwidth dynamic weight parameter, which is dynamically adjusted through the real-time collected network status to meet ;

[0047] S52, initialize the population, each individual in the population Indicates the communication frequency and bandwidth resource allocation strategy A set of strategy combinations, randomly generate the initial population and calculate the initial fitness values ​​of the individuals in the population:

[0048] ;

[0049] in, Represents an individual The initial fitness value of

[0050] S53. The firefly optimization algorithm is used in the hybrid optimization strategy, and the dynamic attraction factor and adaptive step size are used to update the firefly position:

[0051] ;

[0052] in, Indicates The current location of each individual, Indicates The current location of each individual, Indicates the maximum step range, represents the attraction factor that is dynamically adjusted over time, represents the adaptive step size, Table norm;

[0053] S54, using the preferred individuals generated in the firefly optimization stage as the input population of the dynamic differential evolution algorithm to perform local optimization:

[0054] ;

[0055] in, represents a variant individual, , and represents a randomly selected individual from the population, represents the nonlinear scaling factor, Indicates the current iteration number, Indicates the maximum number of iterations;

[0056] S55. Verify the candidate solutions generated by the dynamic differential evolution algorithm, and the optimization results meet the constraints:

[0057] ;

[0058] in, Indicates the lower limit of the communication frequency, Indicates the upper limit of the communication frequency, Indicates the lower limit of bandwidth resources. Indicates the upper limit of bandwidth resources. represents the communication load, Indicates the maximum communication load of the system;

[0059] S56. Combine the optimization results of the hybrid firefly optimization algorithm and the dynamic differential evolution algorithm to output the optimal communication frequency and optimal bandwidth resource allocation strategy , integrate and generate the optimal resource allocation plan.

[0060] Optionally, the S6 specifically includes:

[0061] S61. Construct an industrial equipment communication topology model based on graph neural network, with industrial equipment as nodes of the network and communication links as edges between nodes to form a topology graph model; add dynamic features to each node, including the communication frequency, bandwidth allocation and device status of the equipment; assign weights to each edge to represent the communication delay, bandwidth occupancy and load characteristics of the link;

[0062] S62. Use graph neural networks to model the communication topology model of industrial equipment, and optimize the attribute representation of nodes and edges through the propagation and aggregation mechanism of node features, including: assigning dynamic features to each node as input in the initial stage, recursively propagating the information of adjacent nodes through a multi-layer nested network structure, and aggregating the features of neighboring nodes and itself, gradually updating the characteristics of nodes and edges, and generating feature representations that reflect the local and global attributes of nodes;

[0063] S63. Building a spatiotemporal multi-task network based on the graph neural network model, wherein the structure of the spatiotemporal multi-task network includes an input layer, a temporal modeling layer, a spatial modeling layer, and a task decomposition layer;

[0064] The input layer receives the features of nodes and edges output by the graph neural network as input;

[0065] The time series modeling layer adopts a time series module based on a self-attention mechanism to analyze the time series features of each node and extract the communication change pattern in the time dimension;

[0066] The spatial modeling layer uses a graph convolutional network to analyze the topological relationship between nodes and capture the dynamic changes of node interaction characteristics and edge weights in the spatial dimension;

[0067] The task decomposition layer is composed of a shared feature layer and a task-specific layer, where:

[0068] The shared feature layer is used to extract common features related to all tasks and generate a globally shared feature representation by integrating spatial and temporal characteristics;

[0069] The task-specific layer performs task-specific modeling on the path delay and bandwidth occupancy indicators in the communication network, and generates optimized prediction results for the path delay and bandwidth occupancy indicators respectively;

[0070] S64. Based on the optimization prediction results of the spatiotemporal multi-task network, the communication path is dynamically optimized. By analyzing the performance indicators of each path, the path with the lowest delay and reasonable bandwidth occupancy is selected to generate a dynamically optimized communication path, and the link load is balanced.

[0071] Optionally, the S7 specifically includes:

[0072] S71. Based on the dynamically optimized communication path and combined with the dynamic security strategy of game theory, the communication tasks of legitimate devices and the destructive behavior of potential attackers are regarded as the two parties of the game, and the device optimizes the communication path and dynamically updates the key as the strategy, and the attacker attempts to steal data or destroy communication as the confrontation strategy, forming a zero-sum game model. The goal of the zero-sum game model is to maximize communication security and minimize the probability of potential data leakage or communication failure through the optimization of device strategies;

[0073] S72. Define a state space in a zero-sum game model. The communication path, current key state, and attack behavior characteristics of the device are used as a game state set. The device's policy actions include path adjustment and key update. The attacker's policy actions include path detection and data theft. The state transition is determined by the device's policy decision and the attacker's behavior, forming a dynamic interactive relationship of the game.

[0074] S73. Use deep neural networks to represent the game state space, optimize the device's decision-making strategy through the policy gradient method, and dynamically predict the attacker's behavior pattern based on historical game data;

[0075] S74. Generate a random key using the communication path, the device's environmental status and timestamp as input features, dynamically evaluate the security of the current key through the feedback mechanism of the zero-sum game model, perform real-time replacement of vulnerable keys, and complete key distribution and exchange through a secure channel;

[0076] S75. Introduce attack detection mechanism in the communication process, identify potential attack behaviors through timestamp verification, path behavior analysis and device status monitoring, respond to replay attacks and spoofing attacks in a timely manner, and trigger key update and path reallocation strategies;

[0077] S76. Integrate the optimized communication path and dynamic key to output a secure communication solution.

[0078] An industrial equipment communication system based on the industrial Internet according to an embodiment of the present invention includes the following modules:

[0079] Multimodal data acquisition module: used to collect multimodal data of industrial equipment and generate preliminary multimodal fusion features;

[0080] Event perception module: used to analyze preliminary multimodal fusion features and optimize perception frequency through sparse event detection algorithm to generate event perception data;

[0081] Semantic processing module: used to build a semantic-driven communication protocol, use the semantic embedding model to semantically process event perception data, and generate a semantic data queue with priority;

[0082] Data synchronization module: used to compress semantic data and calibrate time consistency;

[0083] Resource optimization module: used to optimize communication frequency and bandwidth allocation and generate the optimal resource allocation plan;

[0084] Topology modeling and path optimization module: used to build the communication topology model of industrial equipment, and combine the spatiotemporal multi-task network to predict the path delay and bandwidth occupancy indicators of the communication network, and generate a dynamically optimized communication path;

[0085] Security policy module: used to build a zero-sum game model, optimize key generation and exchange strategies, and output secure communication solutions.

[0086] The beneficial effects of the present invention are:

[0087] First, the present invention can dynamically fuse the multimodal data of industrial equipment, effectively integrate multi-source heterogeneous data such as vibration data, temperature and humidity data, electromagnetic signal data, and equipment operation status data, and generate fusion features with high reliability and real-time performance, providing basic data support for subsequent event analysis and communication optimization. The realization of multimodal data fusion improves the adaptability and data accuracy of communication tasks between industrial equipment.

[0088] Secondly, through the sparse event detection algorithm and adaptive perception frequency optimization mechanism, the present invention can accurately identify sparse events in industrial scenarios and dynamically adjust the perception frequency, thereby improving the accuracy of event detection and reducing the redundant use of communication resources. The design of event-driven perception ensures the efficiency and stability of the device in handling sudden events, meeting the high real-time requirements of complex industrial environments.

[0089] In addition, the present invention uses semantically driven communication protocols and semantic entropy algorithms to semantically process event perception data, generate semantic data queues with priority identifiers, and realize dynamic allocation of communication resources and intelligent task scheduling. By mining the priority of semantic data, the present invention significantly improves the processing efficiency of communication tasks and ensures the reliable transmission of high-priority tasks.

[0090] In terms of resource optimization and path planning, the present invention adopts a hybrid firefly optimization algorithm and a dynamic differential evolution algorithm to efficiently optimize the communication frequency and bandwidth resource allocation problem. The generated optimal resource allocation scheme comprehensively considers multiple performance indicators such as communication delay, bandwidth utilization and energy consumption, ensuring the efficiency of the communication process and the rationality of resource utilization. At the same time, based on the graph neural network to build an industrial equipment communication topology model, and combined with the spatiotemporal multi-task network to predict the delay and bandwidth occupancy of the communication path, the present invention realizes dynamic optimization of communication path selection, ensuring the rapid transmission of communication tasks and the balanced distribution of link loads.

[0091] Finally, the present invention has significant advantages in communication security. By combining the game theory dynamic security strategy to build a zero-sum game model and using deep game learning to dynamically optimize the key generation and exchange strategy, the present invention achieves security protection of the communication process. The combination of dynamic key generation and attack detection mechanism not only improves the system's ability to resist attacks, but also ensures the security and reliability of data transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0093] Figure 1 This is an overall flow chart of an industrial equipment communication method based on the industrial Internet proposed by the present invention;

[0094] Figure 2 This is a flow chart of a sparse event detection algorithm for an industrial equipment communication method based on the Industrial Internet proposed in the present invention. DETAILED DESCRIPTION

[0095] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0096] refer to Figure 1 , an industrial equipment communication method based on industrial Internet, comprising the following steps:

[0097] S1. Collect multimodal data of industrial equipment, and dynamically fuse the multimodal data to generate preliminary multimodal fusion features;

[0098] S2. performing event-driven analysis on the preliminary multimodal fusion features, identifying perception events in industrial scenarios and optimizing perception frequencies through a sparse event detection algorithm, and generating event perception data;

[0099] S3, constructing a semantic-driven communication protocol, using a semantic embedding model to semantically process the event perception data, and quantifying the importance of the event perception data through a semantic entropy algorithm to generate a semantic data queue with a priority identifier;

[0100] S4, compressing the semantic data queue with the priority identifier, and using the multimodal synchronization mechanism to calibrate the time consistency of the compressed semantic data, and outputting the synchronized semantic data;

[0101] S5. Based on the synchronized semantic data, the hybrid firefly optimization algorithm and the dynamic differential evolution algorithm are combined to optimize the communication frequency and bandwidth resource allocation strategy to generate the optimal resource allocation plan;

[0102] S6. According to the optimal resource allocation plan, an industrial equipment communication topology model based on graph neural network is constructed, and the path delay and bandwidth occupancy indicators of the communication network are predicted by combining the spatiotemporal multi-task network to generate a dynamically optimized communication path;

[0103] S7. Based on the dynamically optimized communication path and combined with the dynamic security strategy of game theory, a zero-sum game model is constructed to analyze the potential adversarial relationship in the communication process. Deep game learning is used to optimize the dynamic key generation and exchange strategy, and a secure communication solution is output.

[0104] In this embodiment, the multimodal data includes vibration data, temperature and humidity data, electromagnetic signal data and equipment operation status data.

[0105] In this implementation, S2 specifically includes:

[0106] S21, record the preliminary multimodal fusion features generated by S1 as ,in Indicates Multimodal feature data, Represents the total number of multimodal features, and the set of feature values ​​of adjacent time slices is recorded as ,in Indicates the number of time slices, Indicates Multimodal eigenvalues ​​of time slices;

[0107] S22. Feature value set for adjacent time slices The features in the image are used to extract the baseline pattern, and a set of stable reference feature patterns are selected to form the basic pattern set. , select the eigenvalue of the median position of the basic mode set , and then for each time slice Calculate the maximum value of the eigenvalue set ;

[0108] S23. Based on the sparse event detection algorithm, define the sparse activation function:

[0109] ;

[0110] in, Indicates The sparse activation value of the time slice, Represents the absolute value operator;

[0111] S24. Select threshold As the event judgment standard, Time Slice Marked as event time slices, the event time slice set is defined as ;

[0112] S25: extract event attributes from the corresponding feature value set in the event time slice set, and map the feature value in each event time slice into an event set. ,in Indicates the number of events identified, for each event Assigning importance parameters , is a non-negative real number used to measure the suddenness of an event;

[0113] S26. According to the event collection The distribution of event importance parameters in , defines the perception frequency optimization function:

[0114] ;

[0115] in, represents the optimized perceived frequency, represents the exponential function, Indicates the maximum value operation. represents the median operation, Used to ensure that the logarithmic input is positive;

[0116] S27, through the optimized perception frequency Adaptively adjust the event feature collection, and finally optimize the perception frequency The filtered event features are integrated into event perception data.

[0117] In this implementation, S3 specifically includes:

[0118] S31, constructing a semantically driven communication protocol, the semantically driven communication protocol consisting of a semantic processing layer and a transmission scheduling layer; the semantic processing layer performs semantic parsing on event perception data, including semantic embedding, semantic entropy quantization and priority identification generation; the transmission scheduling layer allocates communication resources according to the priority in the semantic data queue;

[0119] S32, using a semantic embedding model to semantically process the event perception data to generate a set of semantic vectors, wherein the semantic embedding model includes an embedding layer, a context encoding layer, and a feature extraction layer;

[0120] The embedding layer maps the input event perception data vector to a high-dimensional continuous semantic space to generate an initial semantic vector;

[0121] The context encoding layer uses a self-attention mechanism to perform contextual encoding on the initial semantic vector to generate a semantic vector, wherein the semantic vector contains the contextual relationship between events;

[0122] The feature extraction layer performs dimension reduction and feature compression on the semantic vector;

[0123] S33, clustering the semantic vector set generated by the semantic embedding model to form several semantic subsets based on semantic similarity, where each semantic subset Include events with high semantic similarity;

[0124] S34. Introduce a semantic entropy algorithm in the semantic-driven communication protocol to quantify the importance of events in each semantic subset:

[0125] ;

[0126] in, Representing a semantic subset The semantic entropy value of represents the exponential function, Representing semantic vector The norm of and represents the weight coefficient;

[0127] S35. According to the semantic entropy value The size of is used to assign priority to events in the semantic subset. When the semantic entropy value is low, it means that the events in the semantic subset contain events with prominent features, and the events in the semantic subset are marked as high priority; when the semantic entropy value is high, it means that the semantic distribution of events in the semantic subset is relatively balanced, and the semantic subset is marked as low priority.

[0128] S36. Arrange the semantic events with priority identifiers from high to low priority to generate a semantic data queue with priority identifiers.

[0129] In this implementation manner, S5 specifically includes:

[0130] S51. Define a multi-objective optimization problem and set the communication frequency and bandwidth resource allocation strategy As the optimization variable, minimize the comprehensive optimization objective:

[0131] ;

[0132] in, represents the comprehensive optimization objective function, represents the communication delay, represents the communication energy consumption, Indicates bandwidth utilization. represents the delay dynamic weight parameter, represents the dynamic weight parameter of energy consumption, Indicates the bandwidth dynamic weight parameter, which is dynamically adjusted through the real-time collected network status to meet ;

[0133] S52, initialize the population, each individual in the population Indicates the communication frequency and bandwidth resource allocation strategy A set of strategy combinations, randomly generate the initial population and calculate the initial fitness values ​​of the individuals in the population:

[0134] ;

[0135] in, Represents an individual The initial fitness value of

[0136] S53. The firefly optimization algorithm is used in the hybrid optimization strategy, and the dynamic attraction factor and adaptive step size are used to update the firefly position:

[0137] ;

[0138] in, Indicates The current location of each individual, Indicates The current location of each individual, Indicates the maximum step range, represents the attraction factor that is dynamically adjusted over time, represents the adaptive step size, Table norm;

[0139] S54, using the preferred individuals generated in the firefly optimization stage as the input population of the dynamic differential evolution algorithm to perform local optimization:

[0140] ;

[0141] in, represents a variant individual, , and represents a randomly selected individual from the population, represents the nonlinear scaling factor, Indicates the current iteration number, Indicates the maximum number of iterations;

[0142] S55. Verify the candidate solutions generated by the dynamic differential evolution algorithm, and the optimization results meet the constraints:

[0143] ;

[0144] in, Indicates the lower limit of the communication frequency, Indicates the upper limit of the communication frequency, Indicates the lower limit of bandwidth resources. Indicates the upper limit of bandwidth resources. represents the communication load, Indicates the maximum communication load of the system;

[0145] S56. Combine the optimization results of the hybrid firefly optimization algorithm and the dynamic differential evolution algorithm to output the optimal communication frequency and optimal bandwidth resource allocation strategy , integrate and generate the optimal resource allocation plan.

[0146] In this implementation manner, S6 specifically includes:

[0147] S61. Construct an industrial equipment communication topology model based on graph neural network, with industrial equipment as nodes of the network and communication links as edges between nodes to form a topology graph model; add dynamic features to each node, including the communication frequency, bandwidth allocation and device status of the equipment; assign weights to each edge to represent the communication delay, bandwidth occupancy and load characteristics of the link;

[0148] S62. Use graph neural networks to model the communication topology model of industrial equipment, and optimize the attribute representation of nodes and edges through the propagation and aggregation mechanism of node features, including: assigning dynamic features to each node as input in the initial stage, recursively propagating the information of adjacent nodes through a multi-layer nested network structure, and aggregating the features of neighboring nodes and itself, gradually updating the characteristics of nodes and edges, and generating feature representations that reflect the local and global attributes of nodes;

[0149] S63. Building a spatiotemporal multi-task network based on the graph neural network model, wherein the structure of the spatiotemporal multi-task network includes an input layer, a temporal modeling layer, a spatial modeling layer, and a task decomposition layer;

[0150] The input layer receives the features of nodes and edges output by the graph neural network as input;

[0151] The time series modeling layer adopts a time series module based on a self-attention mechanism to analyze the time series features of each node and extract the communication change pattern in the time dimension;

[0152] The spatial modeling layer uses a graph convolutional network to analyze the topological relationship between nodes and capture the dynamic changes of node interaction characteristics and edge weights in the spatial dimension;

[0153] The task decomposition layer is composed of a shared feature layer and a task-specific layer, where:

[0154] The shared feature layer is used to extract common features related to all tasks and generate a globally shared feature representation by integrating spatial and temporal characteristics;

[0155] The task-specific layer performs task-specific modeling on the path delay and bandwidth occupancy indicators in the communication network, and generates optimized prediction results for the path delay and bandwidth occupancy indicators respectively;

[0156] S64. Based on the optimization prediction results of the spatiotemporal multi-task network, the communication path is dynamically optimized. By analyzing the performance indicators of each path, the path with the lowest delay and reasonable bandwidth occupancy is selected to generate a dynamically optimized communication path, and the link load is balanced.

[0157] In this implementation manner, the S7 specifically includes:

[0158] S71. Based on the dynamically optimized communication path and combined with the dynamic security strategy of game theory, the communication tasks of legitimate devices and the destructive behavior of potential attackers are regarded as the two parties of the game, and the device optimizes the communication path and dynamically updates the key as the strategy, and the attacker attempts to steal data or destroy communication as the confrontation strategy, forming a zero-sum game model. The goal of the zero-sum game model is to maximize communication security and minimize the probability of potential data leakage or communication failure through the optimization of device strategies;

[0159] S72. Define a state space in a zero-sum game model. The communication path, current key state, and attack behavior characteristics of the device are used as a game state set. The device's policy actions include path adjustment and key update. The attacker's policy actions include path detection and data theft. The state transition is determined by the device's policy decision and the attacker's behavior, forming a dynamic interactive relationship of the game.

[0160] S73. Use deep neural networks to represent the game state space, optimize the device's decision-making strategy through the policy gradient method, and dynamically predict the attacker's behavior pattern based on historical game data;

[0161] S74. Generate a random key using the communication path, the device's environmental status and timestamp as input features, dynamically evaluate the security of the current key through the feedback mechanism of the zero-sum game model, perform real-time replacement of vulnerable keys, and complete key distribution and exchange through a secure channel;

[0162] S75. Introduce attack detection mechanism in the communication process, identify potential attack behaviors through timestamp verification, path behavior analysis and device status monitoring, respond to replay attacks and spoofing attacks in a timely manner, and trigger key update and path reallocation strategies;

[0163] S76. Integrate the optimized communication path and dynamic key to output a secure communication solution.

[0164] refer to Figure 2 , an industrial equipment communication system based on industrial Internet, including the following modules:

[0165] Multimodal data acquisition module: used to collect multimodal data of industrial equipment and generate preliminary multimodal fusion features;

[0166] Event perception module: used to analyze preliminary multimodal fusion features and optimize perception frequency through sparse event detection algorithm to generate event perception data;

[0167] Semantic processing module: used to build a semantic-driven communication protocol, use the semantic embedding model to semantically process event perception data, and generate a semantic data queue with priority;

[0168] Data synchronization module: used to compress semantic data and calibrate time consistency;

[0169] Resource optimization module: used to optimize communication frequency and bandwidth allocation and generate the optimal resource allocation plan;

[0170] Topology modeling and path optimization module: used to build the communication topology model of industrial equipment, and combine the spatiotemporal multi-task network to predict the path delay and bandwidth occupancy indicators of the communication network, and generate a dynamically optimized communication path;

[0171] Security policy module: used to build a zero-sum game model, optimize key generation and exchange strategies, and output secure communication solutions.

[0172] Embodiment 1:

[0173] In order to verify the feasibility of the present invention in implementation, the present invention is applied to a certain intelligent manufacturing factory. The factory is mainly engaged in the production and assembly of precision electronic equipment. Vibration sensors, temperature and humidity sensors, electromagnetic signal acquisition modules and equipment operation status monitoring systems are distributed in the workshop. The equipment needs to work together in real time and efficiently to ensure the stability and efficiency of the production process. In the past, the factory had problems such as high data transmission delay, unreasonable allocation of communication resources and insufficient security in equipment communication, which seriously affected the production efficiency and stability of the equipment. The present invention optimizes equipment communication by transforming the production environment of the factory and applying an industrial equipment communication method based on the industrial Internet.

[0174] In the factory, the method of the present invention is deployed in the factory's communication control center, and a multimodal data acquisition module is used to collect vibration data, temperature and humidity data, electromagnetic signal data, and equipment operation status data. After dynamic fusion, these data generate preliminary multimodal fusion features, and identify key events such as high temperature alarms, vibration anomalies, and electromagnetic interference through sparse event detection algorithms. For these events, the event data is semantically processed through a semantically driven communication protocol, and a semantic data queue with a priority identifier is generated. The communication scheduling system gives priority to high-priority events according to the semantic data queue, thereby effectively avoiding delays in key tasks.

[0175] The resource allocation optimization module combines the hybrid firefly optimization algorithm and the dynamic differential evolution algorithm to optimize the frequency allocation and bandwidth resources for the wireless communication network in the factory. During the peak production period, due to the increase in data transmission demand between devices, the module can optimize the communication path in real time and dynamically adjust the bandwidth occupancy between devices to avoid network congestion. At the same time, the security policy module uses game theory dynamic security strategies to fully protect the communication process, build a zero-sum game model to analyze the confrontational behavior between potential attackers and devices, and optimize the key generation and exchange strategy through deep game learning to ensure the secure transmission of data.

[0176] By deploying the method of the present invention, the communication efficiency and security of the test factory have been significantly improved. During the peak production period, the factory compared and tested the differences between the traditional communication method and the method of the present invention in key performance indicators, including communication delay, bandwidth utilization, energy consumption and security.

[0177] Table 1 Comparison of the communication optimization effects of industrial equipment between the traditional communication method and the method of the present invention

[0178] Test indicators Traditional communication methods Method of the present invention Improvement ratio Average communication delay (ms) 25 12 52% High priority task delay (ms) 8 2 75% Bandwidth utilization (%) 68 92 24% Communication energy consumption (kWh) 145 105 27% Replay attack interception rate (%) 80 98 18% Data integrity verification rate (%) 95 99 4% Device status abnormality detection rate (%) 87 95 8%

[0179] From the data analysis in Table 1 above, it can be seen that the present invention is significantly superior to traditional communication methods in multiple key performance indicators, and comprehensively improves the efficiency and security of industrial equipment communication.

[0180] First, in terms of average communication delay, the method of the present invention reduces the delay from 25 milliseconds in traditional communication to 12 milliseconds, a reduction of 52%. This significant improvement stems from the efficient processing capabilities of the present invention in communication resource allocation and path optimization, ensuring faster and more stable data transmission between devices. In terms of high-priority task processing, the present invention further highlights its superiority, shortening the delay of high-priority tasks to 2 milliseconds, which is 75% faster than traditional methods, ensuring that critical tasks can be completed with extremely high real-time performance.

[0181] Secondly, in terms of bandwidth utilization, the method of the present invention improves bandwidth utilization from 68% of traditional communication to 92% by optimizing the communication resource allocation strategy, an increase of 24%. This efficient bandwidth utilization ensures the reasonable allocation of network resources and efficient collaboration between devices, thereby significantly reducing network congestion.

[0182] In terms of energy consumption optimization, the present invention has shown significant energy-saving effects. The communication energy consumption is reduced from 145 kWh of the traditional method to 105 kWh, and the energy saving ratio is 27%. This improvement not only reduces the operating cost of industrial communications, but also meets the needs of modern industry for green and low-carbon development.

[0183] In terms of security, the present invention has achieved a replay attack interception rate of 98%, which is 18% higher than the traditional method. This improvement is due to the application of game theory dynamic security strategy and deep game learning, which effectively protects against adversarial behaviors during communication. In addition, the data integrity verification rate has been increased from the traditional 95% to 99%, further ensuring the reliability of data transmission. In terms of device status anomaly detection rate, the present invention has reached 95%, which is 8% higher than the traditional method, indicating that it has higher accuracy in abnormal event identification.

[0184] Therefore, the present invention achieves significant performance improvement by optimizing the key technologies of industrial equipment communication. It demonstrates high efficiency in communication delay, bandwidth utilization and energy consumption optimization, and has excellent performance in key task real-time and communication security. At the same time, the reliability of communication is further guaranteed by replay attack interception and data integrity verification, and the ability to detect abnormal status of equipment is significantly improved. These improvements not only meet the needs of complex industrial scenarios for efficient and secure communication, but also have the advantages of energy saving and environmental protection, verifying the excellent performance and wide value of the present invention in practical applications.

[0185] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An industrial equipment communication method based on the industrial Internet, characterized in that: The steps include: S1. Collect multimodal data of industrial equipment, and dynamically fuse the multimodal data to generate preliminary multimodal fusion features; S2. performing event-driven analysis on the preliminary multimodal fusion features, identifying perception events in industrial scenarios and optimizing perception frequencies through a sparse event detection algorithm, and generating event perception data; S3, constructing a semantic-driven communication protocol, using a semantic embedding model to semantically process the event perception data, and quantifying the importance of the event perception data through a semantic entropy algorithm to generate a semantic data queue with a priority identifier; S4, compressing the semantic data queue with the priority identifier, and using the multimodal synchronization mechanism to calibrate the time consistency of the compressed semantic data, and outputting the synchronized semantic data; S5. Based on the synchronized semantic data, the hybrid firefly optimization algorithm and the dynamic differential evolution algorithm are combined to optimize the communication frequency and bandwidth resource allocation strategy to generate the optimal resource allocation plan; S6. According to the optimal resource allocation plan, an industrial equipment communication topology model based on graph neural network is constructed, and the path delay and bandwidth occupancy indicators of the communication network are predicted by combining the spatiotemporal multi-task network to generate a dynamically optimized communication path; S7. Based on the dynamically optimized communication path and combined with the dynamic security strategy of game theory, a zero-sum game model is constructed to analyze the potential adversarial relationship in the communication process. Deep game learning is used to optimize the dynamic key generation and exchange strategy, and a secure communication solution is output.

2. According to claim 1, an industrial equipment communication method based on industrial Internet is characterized in that: The multimodal data includes vibration data, temperature and humidity data, electromagnetic signal data and equipment operation status data.

3. According to the industrial Internet-based industrial equipment communication method of claim 1, it is characterized in that: The S2 specifically includes: S21, record the preliminary multimodal fusion features generated by S1 as ,in Indicates Multimodal feature data, Represents the total number of multimodal features, and the set of feature values ​​of adjacent time slices is recorded as ,in Indicates the number of time slices, Indicates Multimodal eigenvalues ​​of time slices; S22. Feature value set for adjacent time slices The features in the image are used to extract the baseline pattern, and a set of stable reference feature patterns are selected to form the basic pattern set. , select the eigenvalue of the median position of the basic mode set , and then for each time slice Calculate the maximum value of the eigenvalue set ; S23. Based on the sparse event detection algorithm, define the sparse activation function: ； in, Indicates The sparse activation value of the time slice, Represents the absolute value operator; S24. Select threshold As the event judgment standard, Time Slice Marked as event time slices, the event time slice set is defined as ; S25: extract event attributes from the corresponding feature value set in the event time slice set, and map the feature value in each event time slice into an event set. ,in Indicates the number of events identified, for each event Assigning importance parameters , is a non-negative real number used to measure the suddenness of an event; S26. According to the event collection The distribution of event importance parameters in , defines the perception frequency optimization function: ； in, represents the optimized perceived frequency, represents the exponential function, Indicates the maximum value operation. represents the median operation, Used to ensure that the logarithmic input is positive; S27, through the optimized perception frequency Adaptively adjust the event feature collection, and finally optimize the perception frequency The filtered event features are integrated into event perception data.

4. According to claim 1, an industrial equipment communication method based on industrial Internet is characterized in that: The S3 specifically includes: S31, constructing a semantically driven communication protocol, the semantically driven communication protocol consisting of a semantic processing layer and a transmission scheduling layer; the semantic processing layer performs semantic parsing on event perception data, including semantic embedding, semantic entropy quantization and priority identification generation; the transmission scheduling layer allocates communication resources according to the priority in the semantic data queue; S32, using a semantic embedding model to semantically process the event perception data to generate a set of semantic vectors, wherein the semantic embedding model includes an embedding layer, a context encoding layer, and a feature extraction layer; The embedding layer maps the input event perception data vector to a high-dimensional continuous semantic space to generate an initial semantic vector; The context encoding layer uses a self-attention mechanism to perform contextual encoding on the initial semantic vector to generate a semantic vector, wherein the semantic vector contains the contextual relationship between events; The feature extraction layer performs dimension reduction and feature compression on the semantic vector; S33, clustering the semantic vector set generated by the semantic embedding model to form several semantic subsets based on semantic similarity, where each semantic subset Include events with high semantic similarity; S34. Introduce a semantic entropy algorithm in the semantic-driven communication protocol to quantify the importance of events in each semantic subset: ； in, Representing a semantic subset The semantic entropy value of represents the exponential function, Representing semantic vector The norm of and represents the weight coefficient; S35. According to the semantic entropy value The size of is used to assign priority to events in the semantic subset. When the semantic entropy value is low, it means that the events in the semantic subset contain events with prominent features, and the events in the semantic subset are marked as high priority; when the semantic entropy value is high, it means that the semantic distribution of events in the semantic subset is relatively balanced, and the semantic subset is marked as low priority. S36. Arrange the semantic events with priority identifiers from high to low priority to generate a semantic data queue with priority identifiers.

5. The industrial equipment communication method based on the industrial Internet according to claim 1 is characterized in that: The S5 specifically includes: S51. Define a multi-objective optimization problem and set the communication frequency and bandwidth resource allocation strategy As the optimization variable, minimize the comprehensive optimization objective: ； in, represents the comprehensive optimization objective function, Indicates the communication delay, represents the communication energy consumption, Indicates bandwidth utilization. represents the delay dynamic weight parameter, represents the dynamic weight parameter of energy consumption, Indicates the bandwidth dynamic weight parameter, which is dynamically adjusted through the real-time collected network status to meet ; S52, initialize the population, each individual in the population Indicates the communication frequency and bandwidth resource allocation strategy A set of strategy combinations, randomly generate the initial population and calculate the initial fitness values ​​of the individuals in the population: ； in, Represents an individual The initial fitness value of S53. The firefly optimization algorithm is used in the hybrid optimization strategy, and the dynamic attraction factor and adaptive step size are used to update the firefly position: ； in, Indicates The current location of each individual, Indicates The current location of each individual, Indicates the maximum step range, represents the attraction factor that is dynamically adjusted over time, represents the adaptive step size, Table norm; S54, using the preferred individuals generated in the firefly optimization stage as the input population of the dynamic differential evolution algorithm to perform local optimization: ； in, represents a variant individual, , and represents a randomly selected individual from the population, represents the nonlinear scaling factor, Indicates the current iteration number, Indicates the maximum number of iterations; S55. Verify the candidate solutions generated by the dynamic differential evolution algorithm, and the optimization results meet the constraints: ； in, Indicates the lower limit of the communication frequency, Indicates the upper limit of the communication frequency, Indicates the lower limit of bandwidth resources. Indicates the upper limit of bandwidth resources. represents the communication load, Indicates the maximum communication load of the system; S56. Combine the optimization results of the hybrid firefly optimization algorithm and the dynamic differential evolution algorithm to output the optimal communication frequency and optimal bandwidth resource allocation strategy , integrate and generate the optimal resource allocation plan.

6. The industrial equipment communication method based on the industrial Internet according to claim 1 is characterized in that: The S6 specifically includes: S61. Construct an industrial equipment communication topology model based on graph neural network, with industrial equipment as nodes of the network and communication links as edges between nodes to form a topology graph model; add dynamic features to each node, including the communication frequency, bandwidth allocation and device status of the equipment; assign weights to each edge to represent the communication delay, bandwidth occupancy and load characteristics of the link; S62. Use graph neural networks to model the communication topology model of industrial equipment, and optimize the attribute representation of nodes and edges through the propagation and aggregation mechanism of node features, including: assigning dynamic features to each node as input in the initial stage, recursively propagating the information of adjacent nodes through a multi-layer nested network structure, and aggregating the features of neighboring nodes and itself, gradually updating the characteristics of nodes and edges, and generating feature representations that reflect the local and global attributes of nodes; S63. Building a spatiotemporal multi-task network based on the graph neural network model, wherein the structure of the spatiotemporal multi-task network includes an input layer, a temporal modeling layer, a spatial modeling layer, and a task decomposition layer; The input layer receives the features of nodes and edges output by the graph neural network as input; The time series modeling layer adopts a time series module based on a self-attention mechanism to analyze the time series features of each node and extract the communication change pattern in the time dimension; The spatial modeling layer uses a graph convolutional network to analyze the topological relationship between nodes and capture the dynamic changes of node interaction characteristics and edge weights in the spatial dimension; The task decomposition layer is composed of a shared feature layer and a task-specific layer, where: The shared feature layer is used to extract common features related to all tasks and generate a globally shared feature representation by integrating spatial and temporal characteristics; The task-specific layer performs task-specific modeling on the path delay and bandwidth occupancy indicators in the communication network, and generates optimized prediction results for the path delay and bandwidth occupancy indicators respectively; S64. Based on the optimization prediction results of the spatiotemporal multi-task network, the communication path is dynamically optimized. By analyzing the performance indicators of each path, the path with the lowest delay and reasonable bandwidth occupancy is selected to generate a dynamically optimized communication path, and the link load is balanced.

7. The industrial equipment communication method based on the industrial Internet according to claim 1 is characterized in that: The S7 specifically includes: S71. Based on the dynamically optimized communication path and combined with the dynamic security strategy of game theory, the communication tasks of legitimate devices and the destructive behavior of potential attackers are regarded as the two parties of the game, and the device optimizes the communication path and dynamically updates the key as the strategy, and the attacker attempts to steal data or destroy communication as the confrontation strategy, forming a zero-sum game model. The goal of the zero-sum game model is to maximize communication security and minimize the probability of potential data leakage or communication failure through the optimization of device strategies; S72. Define a state space in a zero-sum game model. The communication path, current key state, and attack behavior characteristics of the device are used as a game state set. The device's policy actions include path adjustment and key update. The attacker's policy actions include path detection and data theft. The state transition is determined by the device's policy decision and the attacker's behavior, forming a dynamic interactive relationship of the game. S73. Use deep neural networks to represent the game state space, optimize the device's decision-making strategy through the policy gradient method, and dynamically predict the attacker's behavior pattern based on historical game data; S74. Generate a random key using the communication path, the device's environmental status and timestamp as input features, dynamically evaluate the security of the current key through the feedback mechanism of the zero-sum game model, perform real-time replacement of vulnerable keys, and complete key distribution and exchange through a secure channel; S75. Introduce attack detection mechanism in the communication process, identify potential attack behaviors through timestamp verification, path behavior analysis and device status monitoring, respond to replay attacks and spoofing attacks in a timely manner, and trigger key update and path reallocation strategies; S76. Integrate the optimized communication path and dynamic key to output a secure communication solution.

8. An industrial equipment communication system based on the industrial Internet, executing the industrial equipment communication method based on the industrial Internet according to any one of claims 1 to 7, characterized in that: Includes the following modules: Multimodal data acquisition module: used to collect multimodal data of industrial equipment and generate preliminary multimodal fusion features; Event perception module: used to analyze preliminary multimodal fusion features and optimize perception frequency through sparse event detection algorithm to generate event perception data; Semantic processing module: used to build a semantic-driven communication protocol, use the semantic embedding model to semantically process event perception data, and generate a semantic data queue with priority; Data synchronization module: used to compress semantic data and calibrate time consistency; Resource optimization module: used to optimize communication frequency and bandwidth allocation and generate the optimal resource allocation plan; Topology modeling and path optimization module: used to build the communication topology model of industrial equipment, and combine the spatiotemporal multi-task network to predict the path delay and bandwidth occupancy indicators of the communication network, and generate a dynamically optimized communication path; Security policy module: used to build a zero-sum game model, optimize key generation and exchange strategies, and output secure communication solutions.

Images