Adaptive Data Fusion and Decision Optimization System for Heterogeneous Industrial Equipment

Abstract

An adaptive data fusion and decision optimization system for heterogeneous industrial equipment, which relates to the field of industrial automation and information technology, includes a cloud. The cloud is communicatively connected to a data acquisition module, a protocol annotation module, a protocol self-consistency module, a detection and decision module, a rapid response compensation module, and a critical trend detection module. The data acquisition module is used to collect multi-source type data of each heterogeneous industrial equipment. The protocol annotation module is used to construct an industrial connection topology diagram and perform initial communication protocol annotation on the industrial connection topology diagram. The protocol self-consistency module is used to analyze the interaction characteristics of each node in the industrial connection topology diagram and automatically match the protocol for protocol-abnormal nodes. The detection and decision module performs real-time detection on the multi-source type data of each node, and according to the monitoring results, selects the rapid response compensation module to perform rapid response compensation operations or the critical trend detection module to perform critical trend detection operations, significantly improving production efficiency and ensuring product quality.

Description

Technical Field

[0001] The present invention relates to the field of industrial automation and information technology, and specifically to an adaptive data fusion and decision optimization system for heterogeneous industrial equipment. Background Art

[0002] A Chinese patent with the publication number CN118967109B discloses an industrial equipment maintenance method and platform based on multi-source heterogeneous data, including the following steps: obtaining multi-source heterogeneous data, including equipment operation and maintenance data, monitoring data, factory operation data, and monitoring image data; generating equipment degradation health indicators based on the operation and maintenance data to quantitatively indicate the equipment health status; extracting features from the factory operation data to generate feature changes; constructing a convolutional neural network model, inputting the monitoring images into the model, and outputting illegal behavior images; and performing maintenance operations on the factory equipment according to the equipment degradation health indicators, the feature changes of the factory operation data, and the illegal behavior images.

[0003] A Chinese patent with the publication number CN112529184B discloses an industrial process optimization decision-making method that integrates domain knowledge and multi-source data, including: using probabilistic soft logic to obtain industrial process domain knowledge and establishing an industrial process domain rule knowledge base; integrating the semantics of multi-source data and the features of multi-source data to form a new industrial process semantic knowledge representation and constructing an industrial process semantic knowledge base; in the posterior regularization framework, using the industrial process domain rule knowledge base and the semantic knowledge base to obtain an optimized decision-making model and a posterior distribution model embedded with domain rule knowledge; and using knowledge distillation technology to transfer the knowledge in the optimized decision-making model embedded with domain rule knowledge to the posterior distribution model.

[0004] Under the current trend of Industry 4.0, industrial production systems have become increasingly complex, and heterogeneous industrial equipment is widely used in various production links. Equipment produced by different manufacturers has different data interfaces, communication protocols, and data formats, which makes it difficult to effectively integrate multi-source data and results in the phenomenon of data islands. For example, traditional programmable logic controllers (PLCs) use specific communication protocols for data transmission, while new intelligent sensors may use wireless communication technologies and different data formats, making it extremely challenging to obtain unified and effective data in an industrial production workshop.

[0005] Meanwhile, the industrial production environment is complex and ever-changing. Factors such as the rapid changes in market demand, the fluctuations in raw material quality, and the aging of equipment require enterprises to make accurate and timely decisions based on real-time data in order to optimize the production process, improve production efficiency, reduce costs, and ensure product quality. However, existing data processing and decision-making systems often cannot adapt to such dynamic changes. For example, the optimization strategies based on historical data (such as genetic algorithms) have a response delay of > 30s and cannot cope with millisecond-level equipment disturbances (such as sudden pressure changes in injection molding machines). It is difficult to achieve efficient fusion of multi-source heterogeneous data and precise decision optimization. Summary of the Invention

[0006] To solve the above technical problems, the purpose of the present invention is to provide an adaptive data fusion and decision optimization system for heterogeneous industrial equipment, including a cloud, which is communicatively connected to a data acquisition module, a protocol annotation module, a protocol self-consistency module, a detection and decision module, a rapid response compensation module, and a critical trend detection module;

[0007] The data acquisition module is used to collect multi-source type data of each heterogeneous industrial equipment, mark the acquisition time, and set the acquisition period;

[0008] The protocol annotation module is used to construct an industrial connection topology diagram and perform initial communication protocol annotation on the industrial connection topology diagram;

[0009] The protocol self-consistency module is used to perform interactive feature analysis on each node in the industrial connection topology diagram and automatically match protocols for protocol-abnormal nodes;

[0010] The detection and decision module is used to perform real-time detection on the multi-source type data of each node and select to execute rapid response compensation operations or critical trend detection operations according to the monitoring results;

[0011] The rapid response compensation module is used to generate a temporary compensation model and a joint compensation model, perform rapid response compensation operations on abnormal nodes according to the temporary compensation model, update the joint compensation model according to the results of the rapid response compensation operations, and then perform joint compensation operations according to the joint compensation model;

[0012] The critical trend detection module is used to extract several critical abnormal trend features of each node, perform critical trend detection operations on each normal node, and perform parameter distribution operations on the rapid response compensation module according to the results of the critical trend detection operations.

[0013] Furthermore, the multi-source type data includes: equipment operation status data (including vibration signals, temperature data, pressure data, etc.), equipment performance data (including production efficiency data, product quality data, etc.), equipment communication data (including baud rate, data frame interval, CRC check bits), and environmental data (including humidity data, dust concentration data, etc.).

[0014] Further, the protocol annotation module constructs an industrial connection topology map. The process of initial communication protocol annotation for the industrial connection topology map includes:

[0015] Obtain the assembly connection relationships of several heterogeneous industrial devices. The several heterogeneous industrial devices include terminal devices such as sensors, actuators, motors, and valves in the production line, edge control centers, human-machine interaction terminals, edge computing nodes, etc. The assembly connection relationships of the several heterogeneous industrial devices include direct connections between heterogeneous terminal devices, connection relationships between terminal devices and edge control centers (such as protocol conversion gateways like PLC integrated with MQTT clients), connection relationships between terminal devices and human-machine interaction terminals (touch screens, industrial computers, large screens in the central control room), connection relationships between terminal devices and edge computing nodes (edge servers, gateways), etc. Use the several heterogeneous industrial devices as nodes of the industrial connection topology map. The nodes are used to store multi-source type data of each heterogeneous industrial device. Use the assembly connection relationships of the several heterogeneous industrial devices as the connection relationships between nodes to construct the industrial connection topology map. Obtain the initial communication protocols between the several heterogeneous industrial devices, and perform initial communication protocol annotation on the connection relationships between the nodes.

[0016] Further, the protocol self-consistency module analyzes the interaction characteristics of each node in the industrial connection topology map. The process of automatic protocol matching for protocol abnormal nodes includes:

[0017] According to the initial communication protocols between the nodes, preset the threshold intervals of various types of interaction indicators between the nodes. Extract the interaction characteristics of the nodes with connection relationships. Obtain various types of interaction indicators between the nodes with connection relationships according to the multi-source type data between the nodes with connection relationships. The type of interaction indicators includes frequent data loss, excessive data delay, parsing success rate, connection establishment time, disconnection and reconnection times, function execution results, etc. Compare the various types of interaction indicators with the corresponding interaction indicator threshold intervals. If there are type interaction indicators not within the corresponding interaction indicator threshold intervals, mark the nodes with connection relationships as protocol abnormal nodes;

[0018] Extract the protocol fingerprint characteristics of the protocol abnormal nodes to obtain the protocol fingerprint characteristics. Construct a protocol fingerprint map library, and use reinforcement learning to construct an automatic protocol matching model according to the protocol fingerprint map library;

[0019] The process of constructing a protocol fingerprint map library includes: constructing a control rule table for an automatic protocol matching model based on the protocol fingerprint map library, and defining the state of the automatic protocol matching model as the protocol fingerprint features of the currently collected node. The features include baud rate, data frame interval, and cyclic redundancy check (CRC check bits). For example, the state s = [9600, 10ms, CRC16] indicates that the currently collected data has a baud rate of 9600, a data frame interval of 10ms, and uses the CRC16 check algorithm. The actions are defined as possible protocol types, such as Modbus, OPCUA, Profibus, etc. A reward function is set, and the reward function is used to evaluate the quality of an agent taking an action in a certain state. If the agent correctly matches the protocol type, a positive reward is given, and if the match is incorrect, a negative reward is given. The reinforcement learning algorithm continuously interacts with the environment, updates the control rule table, obtains the experience scores for each state-action combination, and thus finds the optimal action strategy. When new data of a node is collected, the protocol fingerprint features of the data are used as the state input into the control rule value table, and the action with the highest experience score is selected as the matched protocol type;

[0020] Input the protocol fingerprint features into the automatic protocol matching model, output the communication protocol of the protocol abnormal node according to the automatic protocol matching model, and update the initial communication protocol of the protocol abnormal node according to the communication protocol.

[0021] Furthermore, the process of constructing the fingerprint protocol map includes:

[0022] Based on the idea of simulation learning, use a simulator to generate interactive training data for different communication protocols, perform baud rate analysis, data frame interval analysis, and cyclic redundancy check analysis on the interactive training data for different communication protocols, construct a baud rate probability model, a data frame interval probability model, and a cyclic redundancy check probability model, and combine the baud rate probability model, the data frame interval probability model, and the cyclic redundancy check probability model to construct the fingerprint protocol map.

[0023] Furthermore, the baud rate refers to the data transmission rate, and different industrial protocols may use different baud rates. The probability distribution of the baud rate can be constructed through statistical analysis of the baud rate usage of a large number of known protocols. For example, collect 1000 devices using the Modbus protocol and find that 800 of them use a baud rate of 9600. Then the probability of the 9600 baud rate in the Modbus protocol is 0.8.

[0024] The data frame interval refers to the time interval between two adjacent data frames. Different protocols have different time rules for sending and receiving data frames, so the data frame interval can also be used as a feature of the protocol. The data frame intervals of different protocols can be sampled and statistically analyzed to obtain the probability distribution of the data frame intervals for each protocol.

[0025] Cyclic Redundancy Check (CRC check bits) are used to detect errors during data transmission. Different protocols may use different CRC algorithms and parameters, so the characteristics of CRC check bits can also be used to distinguish protocols. By analyzing parameters such as the CRC generation polynomial and initial value of different protocols, a probability model of CRC check bits is constructed.

[0026] Combining the baud rate, data frame interval, and the probability model of CRC check bits, a protocol fingerprint map can be constructed. Each protocol has a corresponding fingerprint vector, and each element in the vector represents the probability of the protocol in a certain characteristic. For example, for a protocol fingerprint vector [0.8, 0.6, 0.7], it represents that the probability of the protocol using a baud rate of 9600 is 0.8, the probability of the data frame interval within a certain range is 0.6, and the probability of using a certain CRC algorithm is 0.7.

[0027] Furthermore, the process of the detection decision module performing real-time detection on the multi-source type data of each node and selecting to perform a rapid response compensation operation or a critical trend detection operation according to the monitoring results includes:

[0028] Extract the numerical time series of each type of index in the multi-source type data of each node in the current acquisition cycle, preset the standard threshold interval corresponding to each type of index of each node, compare the numerical time series of each type of index with the corresponding standard threshold interval, and obtain the cumulative time when each type of index is not within the corresponding standard threshold interval;

[0029] Preset a cumulative time threshold. If there is a cumulative time of a type of index greater than the cumulative time threshold, generate a sudden abnormal alarm signal and perform a rapid response compensation operation. If the cumulative times of all types of indexes are less than or equal to the cumulative time threshold, mark the node as a normal node and perform a critical trend detection operation.

[0030] Furthermore, the process of the rapid response compensation module generating a temporary compensation model includes:

[0031] Mark the type of index whose corresponding cumulative time is greater than the cumulative time threshold as a key index, mark the node to which the key index belongs as an abnormal node, extract features from the multi-source type data of the abnormal node in the current acquisition cycle based on the key index to obtain a feature vector set. Taking the sudden change of bearing temperature as an example, extract the feature vector set related to the sudden change of bearing temperature from the multi-source type data, including the rate of change of temperature (such as the temperature change amount every 10 seconds), the main frequency of the vibration signal, the harmonic distortion rate of the current, etc. These features can more accurately reflect the abnormal state of the equipment;

[0032] Build a sudden abnormal fault library, which includes a number of historical fault cases. These cases are classified and clustered according to equipment type, fault type, relevant features, etc. For example, for bearing fault cases, they may be clustered according to "sudden temperature change + main vibration frequency" to form different fault categories. Compare the feature data set with the cases in the sudden abnormal fault library by cosine similarity to obtain the cosine similarity between the current feature vector set and the feature vector sets of historical fault cases, and screen out the top k historical fault cases with the highest cosine similarity;

[0033] Build a temporary compensation model based on the decision tree model. The temporary compensation model built based on the decision tree model has a short inference time (usually less than 10 milliseconds) and is suitable for running quickly on the edge side. Extract the compensation strategies of the top k historical fault cases to customize the parameters of the temporary compensation model, and verify the model according to the multi-source type data of abnormal nodes in the current acquisition cycle. For example, if the compensation strategy when the bearing temperature is too high in similar cases adopts a speed reduction of 15% and an increase in lubrication flow rate of 20%, and for every 1°C increase, the speed reduction amplitude increases by 2%, then inject the corresponding compensation parameters into the temporary model. Subsequently, in the model verification process, first select the compensation strategy of the case with the highest cosine similarity from the top k historical fault cases to customize the parameters of the temporary compensation model, and then use the multi-source type data of abnormal nodes in the current acquisition cycle to verify the customized model. Predict the operating state of the equipment (such as temperature change) through the model and compare it with the actual situation. If the deviation between the prediction result and the actual situation exceeds the preset deviation threshold, select the compensation strategy and relevant parameters of the case with the second highest cosine similarity from the top k historical fault cases to customize the parameters of the temporary compensation model, and repeat the above process until the deviation between the prediction result and the actual situation does not exceed the preset deviation threshold to obtain the temporarily compensated model that has completed verification;

[0034] Furthermore, the rapid response compensation module generates a joint compensation model. The process of performing rapid response compensation operations on abnormal nodes according to the temporary compensation model and updating the joint compensation model according to the results of the rapid response compensation operations includes:

[0035] Construct an association rule knowledge graph based on the industrial connection topology diagram. The association rule knowledge graph includes multi-source type data, spatio-temporal feature sequences, and joint compensation strategies between interconnected nodes. Build a joint compensation model based on deep learning. Use the multi-source type data, spatio-temporal feature sequences, and joint compensation strategies as the training set and the test set. Input the training set into the joint compensation model for training until the loss function is trained stably, and save the model parameters. Test the joint compensation model through the test set until it meets the preset requirements, and output the joint compensation model;

[0036] Building a joint compensation model based on deep learning is a complex process that involves multiple steps such as model selection, training, validation, and testing. The following is a detailed supplementary description of this process:

[0037] In this invention, an RBF neural network is selected as the deep learning architecture. After determining the model architecture, Binary Cross-Entropy Loss is selected as the optimization objective. Subsequently, the prepared training set is input into the selected deep learning model for training. During the training process, the weights are continuously updated through the backpropagation algorithm, causing the loss function to gradually decrease until it reaches a stable state. During this period, techniques such as Early Stopping are used to avoid overfitting. In addition to the basic training process, various parameters of the model are tuned through Grid Search. The parameters include the learning rate, batch size, regularization coefficient, etc.

[0038] When the model training is completed and the parameters are adjusted, the final evaluation is carried out through the test set to obtain the evaluation results of the model. The evaluation results include classification metrics such as accuracy, recall rate, F1 score, etc. According to the evaluation results on the test set, it is judged whether the model meets the expected standards. If the requirements are met, the model parameters are saved and ready for deployment; if not ideal, it is necessary to return to a previous stage to re-examine issues such as data quality, model structure, or training strategy.

[0039] Furthermore, the spatio-temporal feature sequences between interconnected nodes, for example: a sudden change in the bearing temperature of device A (ΔT>5℃ / 10s) will cause a decrease in the outlet pressure of the lubricating pump of device B (ΔP<-0.2MPa). According to the device dependency relationship in the industrial connection topology diagram (such as the oil circuit connection between the bearing and the lubricating pump), it is judged that insufficient lubrication of device B is the main cause of the temperature rise of device A, and the wear of the lubricating pump plunger is locked;

[0040] The joint compensation strategy includes:

[0041] Edge side: Immediate stop loss (<10ms). For example, for device A: Based on the temporary compensation model, immediately execute local load reduction (such as reducing the rolling mill speed from 1200rpm to 900rpm to reduce frictional heat); for device B: Trigger the emergency mode of the lubricating pump based on the temporary compensation model (increase the plunger frequency from 50Hz to 60Hz, and short-term overload is allowed for 10 minutes);

[0042] Workshop level: Collaborative adjustment (<5s). For example, the cloud issues a linkage strategy to the workshop controller through the OPC UA protocol: Start the standby lubricating pump C (ready within 30s), and at the same time reduce the load of device A to 50% (to avoid single-pump overload); Adjust the cooling water flow rate of the adjacent device D (compensate for the imbalance of the production line beat caused by the load reduction of device A);

[0043] Encapsulate and deploy the temporary compensation model. The specific process of model encapsulation and deployment includes: encapsulating the customized model into an executable file, such as a TensorFlow Lite micro model. This model file has a small size (e.g., less than 100 KB) and is suitable for running on edge devices. Define an interface for the encapsulated temporary compensation model. For example, subscribe to the Topic of bearing temperature through the MQTT protocol to obtain temperature data in real time, and write the compensation instructions output by the model into the corresponding register through the Modbus protocol to achieve the control of industrial equipment. Finally, set a validity period for the temporary model, such as automatically marking "validity period: 2 minutes". Within the validity period, the model continuously runs and compensates the equipment; after the validity period, the model automatically fails to avoid the adverse impact of outdated compensation strategies on the equipment. Compensate the abnormal nodes according to the temporary compensation model, and preset a compensation time threshold. When the abnormal nodes are marked as normal nodes within the compensation time threshold, collect the compensation parameters and effective data of the temporary compensation model to update the association rule knowledge graph, and perform incremental training on the joint compensation model according to the updated association rule knowledge graph. For example, add an association rule of "bearing temperature rise - speed reduction compensation" to provide richer knowledge support for subsequent fault diagnosis and processing. If the abnormal nodes are not marked as normal nodes within the compensation time threshold, suspend the operation of the abnormal nodes and detect and repair the heterogeneous industrial equipment of the abnormal nodes.

[0044] Furthermore, the process of the rapid response compensation module performing joint compensation operations according to the joint compensation model includes:

[0045] After the abnormal nodes are marked as normal nodes, extract the time features and spatial features of the multi-source type data of the normal nodes and other normal nodes with connection relationships with the normal nodes in the current acquisition cycle to obtain a spatio-temporal feature sequence. Input the multi-source type data and the spatio-temporal feature sequence into the incrementally trained joint compensation model, and output a joint compensation strategy according to the joint compensation model.

[0046] Furthermore, the process of extracting the time features and spatial features of the multi-source type data to obtain a spatio-temporal feature sequence includes:

[0047] Extract the numerical time series of each type of index corresponding to the multi-source type data, input the numerical time series of each type of index into a temporal convolutional neural network, and obtain the numerical change trend features of each type of index according to the trained temporal convolutional network.

[0048] Learn the industrial connection topology diagram through the graph attention network, import the numerical change trend characteristics of various types of indicators of normal nodes and other nodes with connection relationships with normal nodes into the graph attention network, obtain the attention weights of other normal nodes on the target normal node through the attention mechanism, distribute the attention weights to each adjacent normal node, and use neighbor aggregation to generate an aggregated representation of features, generating a spatio-temporal feature sequence of multi-source type data;

[0049] Further, obtain the attention weight of adjacent normal node j to target normal node i The calculation formula is:

[0050] ;

[0051] Among them, () represents an activation function, used to introduce non-linear characteristics, represents a feature vector and The splicing operation of, represents a learnable vector, represents a learnable weight matrix, represents the feature vector of target normal node i, represents the feature vector of adjacent normal node j;

[0052] Through the attention value between target normal node i and adjacent normal node j and the sum of the attention values of target normal node i and all adjacent normal nodes, determine the attention weight , after normalizing the attention weight, use the neighbor aggregation mechanism to update and represent the features of the target normal node, ;

[0053] Among them, represents the final updated representation of target normal node i, represents the activation function, G represents the parameter matrix of feature transformation, j represents the number of adjacent normal node terms, represents the set of adjacent normal nodes of target normal node i.

[0054] Further, the process of the critical trend detection module extracting several critical abnormal trend characteristics of each node includes:

[0055] Extract the moments when nodes are marked as abnormal nodes within the historical collection period, extract the numerical time series and key indicators of various types of indicators between the start timestamp of the historical collection period and the moments, mark other key indicators except the key indicators as non-key indicators, mark the numerical time series as a critical abnormal numerical time series, and mark the historical collection period as a critical abnormal collection period;

[0056] Obtain the standard time series of key indicators within the critical abnormal collection period, perform autocorrelation comparison between the critical abnormal numerical time series of key indicators and the standard time series, obtain the autocorrelation coefficient of key indicators, perform cross-correlation comparison between the critical abnormal numerical time series of key indicators, obtain the cross-correlation coefficient between key indicators, perform cross-correlation comparison between the critical abnormal numerical time series of key indicators and non-key indicators, obtain the cross-correlation coefficient between key indicators and non-key indicators, construct a critical abnormal trend feature based on the autocorrelation coefficient of key indicators, the cross-correlation coefficient between key indicators, and the cross-correlation coefficient between key indicators and non-key indicators, and extract a temporary compensation model for the critical abnormal collection period, and associate the temporary compensation model with the critical abnormal trend feature;

[0057] Construct a trend feature database and store the critical abnormal trend features within several abnormal collection periods of each node into the trend feature database.

[0058] Further, the process of performing autocorrelation comparison includes:

[0059] ;

[0060] Among them, represents the autocorrelation coefficient of type indicator i, represents the value of type i at the t-th moment, represents the value of the standard time series of type indicator i at the t-th moment, and n represents the total number of moments;

[0061] The process of performing cross-correlation comparison includes:

[0062] ;

[0063] Among them, represents the cross-correlation coefficient between type indicator i and type indicator j; represents the value of type indicator i at the t-th moment; represents the average value of type indicator i; represents the value of type indicator j at the t-th moment; represents the average value of type indicator j.

[0064] Further, the process of obtaining the standard time series of each type of indicator in the historical collection period includes:

[0065] Construct a data prediction model based on deep learning. Use the multi-source type data of each node in several historical collection periods as training data to train the data prediction model, and obtain the trained data prediction model. Input the multi-source type data in the critical abnormal collection period into the data prediction model, and obtain the standard time series of each type of indicator according to the output of the data prediction model.

[0066] Further, the process in which the critical trend detection module performs critical trend detection operations on each normal node and issues parameter operations to the rapid response compensation module according to the results of the critical trend detection operations includes:

[0067] Extract the trend characteristics of the numerical time series of each type of indicator in the current collection period of the normal node to obtain the trend characteristics in the current collection period of the normal node (including the autocorrelation coefficient of each type of indicator and the cross-correlation coefficient between each type of indicator). Retrieve and compare in the trend feature database according to the trend characteristics to obtain the similarity between the trend characteristics and several critical abnormal trend characteristics. If the similarity between the trend characteristics and the critical abnormal trend characteristics is greater than the preset similarity threshold, mark the normal node as a critical warning node, customize the parameters of the temporary compensation model for the critical warning node according to the temporary compensation model associated with the critical abnormal trend characteristics, directly skip the process of screening out the top k historical fault cases with the highest cosine similarity, and then perform rapid response compensation operations on the critical warning node according to the temporarily compensated model with customized parameters.

[0068] Compared with the prior art, the beneficial effects of the present invention are:

[0069] 1. The protocol annotation module constructs an industrial connection topology diagram and performs initial communication protocol annotation, enabling the system to clearly master the connection relationship and initial protocol between devices. The protocol self-consistency module performs interactive feature analysis and automatic protocol matching on each node, can timely detect and solve protocol anomaly problems, and ensure smooth and stable communication between devices. It avoids data transmission interruption or errors caused by protocol incompatibility, and improves the compatibility and reliability of the system.

[0070] 2. The detection and decision-making module real-time detects the multi-source type data of each node, and selects to perform rapid response compensation operations or critical trend detection operations according to the monitoring results, realizing intelligent judgment and targeted processing of the device status. It can timely detect sudden anomalies and potential critical trends of devices, take corresponding measures to avoid the occurrence of device failures, and reduce production losses.

[0071] 3. The fast response compensation module generates a temporary compensation model and a joint compensation model, performs fast response compensation operations on abnormal nodes, and updates the joint compensation model according to the operation results. This fast response mechanism can quickly take measures when the device has an abnormality, reduce the impact of the abnormality on production, and at the same time improve the system's ability to handle similar abnormalities and enhance the system's adaptability and self-learning ability by continuously updating the joint compensation model.

[0072] 4. The critical trend detection module extracts the critical abnormal trend features of each node, performs critical trend detection operations on normal nodes, and performs parameter distribution operations on the fast response compensation module according to the detection results. Trend features are generated based on the multi-source type data of the current acquisition cycle, the trend features are retrieved and compared in the trend feature database, and preventive measures are generated according to the comparison results, realizing the prediction of early faults before the industrial device changes from the normal state to the fault state. Thus, preventive measures are taken before the industrial device fault matures, improving the reliability of the device and the continuity of production. Brief Description of the Drawings

[0073] Figure 1 It is a schematic diagram of the adaptive data fusion and decision optimization system for heterogeneous industrial devices according to an embodiment of the present application. Detailed Embodiments

[0074] Next, in conjunction with the drawings in the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of protection of the present application.

[0075] As Figure 1 shown, the adaptive data fusion and decision optimization system for heterogeneous industrial devices includes a cloud, and the cloud is communicatively connected to a data acquisition module, a protocol annotation module, a protocol self-consistency module, a detection and decision module, a fast response compensation module, and a critical trend detection module;

[0076] The data acquisition module is used to collect multi-source type data of each heterogeneous industrial device, mark the acquisition time, and set the acquisition cycle;

[0077] The protocol annotation module is used to construct an industrial connection topology diagram and perform initial communication protocol annotation on the industrial connection topology diagram;

[0078] The protocol self-consistency module is used to perform interactive feature analysis on each node in the industrial connection topology diagram and perform automatic protocol matching on protocol abnormal nodes;

[0079] The detection and decision-making module is used to perform real-time detection on multi-source type data of each node, and select to execute a quick response compensation operation or a critical trend detection operation according to the monitoring results;

[0080] The quick response compensation module is used to generate a temporary compensation model and a joint compensation model, perform a quick response compensation operation on abnormal nodes according to the temporary compensation model, update the joint compensation model according to the results of the quick response compensation operation, and then perform a joint compensation operation according to the joint compensation model;

[0081] The critical trend detection module is used to extract several critical abnormal trend features of each node, perform a critical trend detection operation on each normal node, and perform a parameter distribution operation on the quick response compensation module according to the results of the critical trend detection operation.

[0082] It should be further noted that in the specific implementation process, the multi-source type data includes: device operation status data (including vibration signals, temperature data, pressure data, etc.), device performance data (including production efficiency data, product quality data, etc.), device communication data (including baud rate, data frame interval, CRC check bits), and environmental data (including humidity data, dust concentration data, etc.).

[0083] It should be further noted that in the specific implementation process, the protocol annotation module constructs an industrial connection topology map, and the process of initial communication protocol annotation for the industrial connection topology map includes:

[0084] Obtain the assembly connection relationships of several heterogeneous industrial devices. The several heterogeneous industrial devices include terminal devices such as sensors, actuators, motors, and valves in the production line, edge control centers, human-machine interaction terminals, edge computing nodes, etc. The assembly connection relationships of the several heterogeneous industrial devices include direct connections between heterogeneous terminal devices, connection relationships between terminal devices and edge control centers (such as protocol conversion gateways like PLC integrated with MQTT clients), connection relationships between terminal devices and human-machine interaction terminals (touch screens, industrial computers, large screens in the central control room), connection relationships between terminal devices and edge computing nodes (edge servers, gateways), etc. Use the several heterogeneous industrial devices as nodes of the industrial connection topology map. The nodes are used to store multi-source type data of each heterogeneous industrial device. Use the assembly connection relationships of the several heterogeneous industrial devices as the connection relationships between nodes to construct an industrial connection topology map, obtain the initial communication protocols between the several heterogeneous industrial devices, and perform initial communication protocol annotation on the connection relationships between the nodes.

[0085] It should be further noted that in the specific implementation process, the protocol self-consistency module performs an interactive feature analysis on each node in the industrial connection topology map, and the process of automatic protocol matching for protocol abnormal nodes includes:

[0086] According to the initial communication protocol between each node, preset the threshold intervals of various types of interaction indicators between each node, extract the interaction characteristics of the nodes with connection relationships, obtain various types of interaction indicators between the nodes with connection relationships according to the multi-source type data between the nodes with connection relationships. The type of interaction indicators includes frequent data loss, excessive data delay, parsing success rate, connection establishment time, disconnection and reconnection times, function execution results, etc. Compare each type of interaction indicator with the corresponding interaction indicator threshold interval. If there is a type of interaction indicator that is not within the corresponding interaction indicator threshold interval, mark the nodes with connection relationships as protocol abnormal nodes;

[0087] Extract the protocol fingerprint characteristics of the protocol abnormal nodes, obtain the protocol fingerprint characteristics, construct a protocol fingerprint map library, and construct an automatic protocol matching model using reinforcement learning according to the protocol fingerprint map library;

[0088] The process of constructing the protocol fingerprint map library includes: constructing a control rule table for the automatic protocol matching model according to the protocol fingerprint map library, and defining the state of the automatic protocol matching model as the protocol fingerprint characteristics of the currently collected nodes. It includes characteristics such as baud rate, data frame interval, and cyclic redundancy check (CRC check bit). For example, the state s = [9600, 10ms, CRC16] means that the currently collected data baud rate is 9600, the data frame interval is 10ms, and the CRC16 check algorithm is used. The action is defined as possible protocol types, such as Modbus, OPCUA, Profibus, etc. Set the reward function. The reward function is used to evaluate the quality of the agent taking a certain action in a certain state. If the agent correctly matches the protocol type, a positive reward is given. If the match is incorrect, a negative reward is given. The reinforcement learning algorithm continuously interacts with the environment, updates the control rule table, obtains the experience score of each state-action combination, so as to find the optimal action strategy, so as to find the optimal action strategy. When new data of the node is collected, use the protocol fingerprint characteristics of the data as the state input into the control rule value table, and select the action with the largest experience score as the matching protocol type;

[0089] Input the protocol fingerprint characteristics into the automatic protocol matching model, output the communication protocol of the protocol abnormal node according to the automatic protocol matching model, and update the initial communication protocol of the protocol abnormal node according to the communication protocol.

[0090] It should be further noted that in the specific implementation process, the process of constructing the fingerprint protocol map includes:

[0091] Based on the idea of simulation learning, a simulator is used to generate interactive training data for different communication protocols. The interactive training data for different communication protocols is subjected to baud rate analysis, data frame interval analysis, and cyclic redundancy check analysis, and a baud rate probability model, a data frame interval probability model, and a cyclic redundancy check probability model are constructed. The baud rate probability model, the data frame interval probability model, and the cyclic redundancy check probability model are combined to construct a fingerprint protocol map.

[0092] It should be further noted that the baud rate refers to the data transmission rate, and different industrial protocols may use different baud rates. The probability distribution of the baud rate can be constructed by statistically analyzing the baud rate usage of a large number of known protocols. For example, if 1000 devices using the Modbus protocol are collected and it is found that 800 of them use a baud rate of 9600, then the probability of the 9600 baud rate in the Modbus protocol is 0.8.

[0093] The data frame interval refers to the time interval between two adjacent data frames. Different protocols have different time rules for sending and receiving data frames, so the data frame interval can also be used as a feature of the protocol. The data frame intervals of different protocols can be sampled and statistically analyzed to obtain the probability distribution of the data frame interval for each protocol.

[0094] Cyclic redundancy check (CRC check bits) is used to detect errors during data transmission. Different protocols may use different CRC algorithms and parameters, so the characteristics of the CRC check bits can also be used to distinguish protocols. By analyzing parameters such as the CRC generation polynomial and initial value of different protocols, a probability model of the CRC check bits is constructed.

[0095] Combining the probability models of the baud rate, data frame interval, and CRC check bits, a protocol fingerprint map can be constructed. Each protocol has a corresponding fingerprint vector, and each element in the vector represents the probability of the protocol in a certain feature. For example, for a protocol fingerprint vector [0.8, 0.6, 0.7], it represents that the probability of the protocol using a baud rate of 9600 is 0.8, the probability of the data frame interval within a certain range is 0.6, and the probability of using a certain CRC algorithm is 0.7.

[0096] It should be further noted that in the specific implementation process, the process of the detection decision module performing real-time detection on the multi-source type data of each node and selecting to execute the rapid response compensation operation or the critical trend detection operation according to the monitoring results includes:

[0097] Extract the numerical time series of each type of indicator in the multi-source type data of each node during the current acquisition cycle, preset the standard threshold intervals corresponding to each type of indicator of each node, compare the numerical time series of each type of indicator with the corresponding standard threshold intervals, and obtain the cumulative time when each type of indicator is not within the corresponding standard threshold interval;

[0098] Preset a cumulative time threshold. If there is a type of indicator whose cumulative time is greater than the cumulative time threshold, generate a sudden abnormal alarm signal and perform a quick response compensation operation. If the cumulative times of all types of indicators are less than or equal to the cumulative time threshold, mark the node as a normal node and perform a critical trend detection operation.

[0099] It should be further noted that in the specific implementation process, the process of the quick response compensation module generating a temporary compensation model includes:

[0100] Mark the type of indicator whose corresponding cumulative time is greater than the cumulative time threshold as a key indicator, mark the node to which the key indicator belongs as an abnormal node, extract features from the multi-source type data of the abnormal node during the current acquisition cycle based on the key indicator, and obtain a feature vector set. Taking the sudden change of bearing temperature as an example, extract a feature vector set related to the sudden change of bearing temperature from the multi-source type data, including the change rate of temperature (such as the temperature change amount every 10 seconds), the main frequency of the vibration signal, the harmonic distortion rate of the current, etc. These features can more accurately reflect the abnormal state of the equipment;

[0101] Construct a sudden abnormal fault library, which includes a number of historical fault cases, and these cases are classified and clustered according to equipment type, fault type, related features, etc. For example, for bearing fault cases, they may be clustered according to "temperature mutation + vibration main frequency" to form different fault categories. Compare the feature data set with the cases in the sudden abnormal fault library by cosine similarity, obtain the cosine similarity between the current feature vector set and the feature vector sets of historical fault cases, and screen out the top k historical fault cases with the highest cosine similarity;

[0102] Build a temporary compensation model based on the decision tree model. The inference time of the temporary compensation model built based on the decision tree model is relatively short (usually less than 10 milliseconds), which is suitable for running quickly on the edge side. Extract the compensation strategies of the top k historical fault cases to customize the parameters of the temporary compensation model, and verify the model according to the multi-source type data of abnormal nodes in the current acquisition cycle. For example, if the compensation strategy when the bearing temperature is too high in similar cases adopts a speed reduction of 15% and an increase in lubrication flow of 20%, and for every 1°C increase, the speed reduction amplitude increases by 2%, then inject the corresponding compensation parameters into the temporary model. Subsequently, during the model verification process, first customize the parameters of the temporary compensation model using the compensation strategy of the case with the highest cosine similarity selected from the top k historical fault cases, and then use the multi-source type data of abnormal nodes in the current acquisition cycle to verify the customized model. Predict the operating state of the device (such as temperature change) through the model and compare it with the actual situation. If the deviation between the prediction result and the actual situation exceeds the preset deviation threshold, then select the compensation strategy and related parameters of the case with the second highest cosine similarity from the top k historical fault cases to customize the parameters of the temporary compensation model, and repeat the above process until the deviation between the prediction result and the actual situation does not exceed the preset deviation threshold, and obtain the temporarily compensated model that has completed verification;

[0103] The temporary compensation model on the edge side is generated and executed locally. From anomaly detection to compensation instruction output, it can be controlled within 10 ms (such as through pre-stored lightweight model templates + real-time data matching), and directly drive the device to execute (such as immediately reducing speed, increasing lubrication). In the factory environment, network interruptions caused by wireless signal occlusion, switch failures, etc. occur frequently. If relying on the joint compensation model, the device will lose its protection ability when the network is disconnected;

[0104] The temporary compensation model runs offline, relying on historical cases and model templates pre-stored on the edge side to ensure that compensation operations can be completed even when the network is disconnected;

[0105] At the same time, based on the parameter fine-tuning of similar cases, it is more in line with the current device personality than the general model (such as the compensation difference of bearings under different lubrication conditions), and has a built-in "self-destruction mechanism" to avoid becoming a new fault source. At the same time, through knowledge feedback, the system becomes smarter and smarter during the rescue process. This design breaks through the traditional passive mode of "fault means shutdown" and realizes the key transformation of industrial AI from "post-event analysis" to "real-time life-saving".

[0106] It should be further noted that in the specific implementation process, the fast response compensation module generates a joint compensation model. The process of performing fast response compensation operations on abnormal nodes according to the temporary compensation model and updating the joint compensation model according to the results of the fast response compensation operations includes:

[0107] Construct an association rule knowledge graph based on the industrial connection topology diagram. The association rule knowledge graph includes multi-source type data, spatio-temporal feature sequences, and joint compensation strategies between interconnected nodes. Construct a joint compensation model based on deep learning, use the multi-source type data, spatio-temporal feature sequences, and joint compensation strategies as the training set and the test set, input the training set into the joint compensation model for training until the loss function is trained stably, save the model parameters, test the joint compensation model through the test set until it meets the preset requirements, and output the joint compensation model;

[0108] Constructing a joint compensation model based on deep learning is a complex process that involves multiple steps such as model selection, training, validation, and testing. The following is a detailed supplementary description of this process:

[0109] The present invention selects the RBF neural network as the deep learning architecture. After determining the model architecture, it selects the Binary Cross-Entropy Loss as the optimization objective, and then inputs the prepared training set into the selected deep learning model to start training. During the training process, the weights are continuously updated through the backpropagation algorithm, making the loss function gradually decrease until it reaches a stable state. During this period, techniques such as Early Stopping are used to avoid overfitting. In addition to the basic training process, the various parameters of the model are tuned through Grid Search, and the parameters include the learning rate, batch size, regularization coefficient, etc.

[0110] When the model training is completed and the parameters are adjusted, the final evaluation is carried out through the test set to obtain the evaluation results of the model. The evaluation results include classification metrics such as accuracy, recall rate, F1 score, etc. According to the evaluation results on the test set, it is judged whether the model meets the expected standards. If the requirements are met, the model parameters are saved and ready for deployment; if not, it is necessary to return to a previous stage to re-examine issues such as data quality, model structure, or training strategy.

[0111] It should be further noted that the spatio-temporal feature sequence between interconnected nodes is, for example, a sudden change in the bearing temperature of device A (ΔT>5℃ / 10s) will cause a decrease in the outlet pressure of the lubricating pump of device B (ΔP<-0.2MPa). According to the device dependency relationship in the industrial connection topology diagram (such as the oil circuit connection between the bearing and the lubricating pump), it is judged that the insufficient lubrication of device B is the main cause of the temperature rise of device A, and the wear of the lubricating pump plunger is locked;

[0112] The joint compensation strategies include:

[0113] Edge side: Immediate stop loss (<10 ms). For example, Device A: Based on the temporary compensation model, immediately perform local load reduction (such as reducing the mill speed from 1200 rpm to 900 rpm to reduce frictional heat); Device B: Trigger the emergency mode of the lubrication pump based on the temporary compensation model (increase the plunger frequency from 50 Hz to 60 Hz, short-term overload is allowed for 10 minutes).

[0114] Workshop level: Collaborative adjustment (<5 s). For example, the cloud sends a linkage strategy to the workshop controller through the OPC UA protocol: Start the standby lubrication pump C (ready within 30 s), and at the same time reduce the load of Device A to 50% (to avoid single-pump overload); Adjust the cooling water flow rate of the adjacent Device D (to compensate for the imbalance of the production line rhythm caused by the load reduction of Device A).

[0115] Perform model encapsulation and deployment on the temporary compensation model. The specific process of model encapsulation and deployment includes: Encapsulate the customized model into an executable file, such as a TensorFlow Lite micro model. This model file has a small volume (such as less than 100 KB) and is suitable for running on edge devices. Define an interface for the encapsulated temporary compensation model. For example, subscribe to the Topic of bearing temperature through the MQTT protocol to obtain temperature data in real time, and write the compensation instructions output by the model into the corresponding register through the Modbus protocol to achieve the control of industrial equipment. Finally, set a validity period for the temporary model, such as automatically marking "validity period: 2 minutes". Within the validity period, the model continuously runs and compensates the equipment; after the validity period, the model automatically fails to avoid the adverse impact of outdated compensation strategies on the equipment. Compensate the abnormal nodes according to the temporary compensation model, preset a compensation time threshold. When the abnormal nodes are marked as normal nodes within the compensation time threshold, collect the compensation parameters and effective data of the temporary compensation model to update the association rule knowledge graph, and perform incremental training on the joint compensation model according to the updated association rule knowledge graph. For example, add an association rule of "bearing temperature rise - speed reduction compensation" to provide richer knowledge support for subsequent fault diagnosis and processing. If the abnormal nodes are not marked as normal nodes within the compensation time threshold, suspend the operation of the abnormal nodes and detect and repair the heterogeneous industrial equipment of the abnormal nodes.

[0116] It should be further noted that in the specific implementation process, the process of the rapid response compensation module performing joint compensation operations according to the joint compensation model includes:

[0117] After the abnormal nodes are marked as normal nodes, extract the time features and spatial features of the multi-source type data of the normal nodes and other normal nodes having a connection relationship with the normal nodes in the current acquisition cycle to obtain a spatio-temporal feature sequence. Input the multi-source type data and the spatio-temporal feature sequence into the incrementally trained joint compensation model, and output a joint compensation strategy according to the joint compensation model.

[0118] It should be further noted that in the specific implementation process, the process of extracting time features and space features from multi-source type data to obtain a spatio-temporal feature sequence includes:

[0119] Extract the numerical time series of each type of index corresponding to the multi-source type data, input the numerical time series of each type of index into a temporal convolutional neural network, and obtain the numerical change trend features of each type of index according to the trained temporal convolutional network;

[0120] Learn through a graph attention network on the industrial connection topology graph, import the numerical change trend features of each type of index of normal nodes and other nodes having a connection relationship with the normal nodes into the graph attention network, obtain the attention weights of other normal nodes to the target normal node through the attention mechanism, allocate the attention weights to each adjacent normal node, and use neighbor aggregation to generate an aggregated representation of the features to generate a spatio-temporal feature sequence of the multi-source type data;

[0121] It should be further noted that in the specific implementation process, obtain the attention weight of adjacent normal node j to target normal node i The calculation formula is:

[0122] ;

[0123] Where () represents an activation function, used to introduce non-linear characteristics, represents a feature vector and The concatenation operation of, represents a learnable vector, represents a learnable weight matrix, represents the feature vector of target normal node i, represents the feature vector of adjacent normal node j;

[0124] Determine the attention weight through the ratio of the attention value between target normal node i and adjacent normal node j to the sum of the attention values of target normal node i and all adjacent normal nodes , after normalizing the attention weight, use the neighbor aggregation mechanism to update and represent the features of the target normal node, ;

[0125] Where represents the final updated representation of target normal node i, represents an activation function, G represents the parameter matrix of feature transformation, j represents the number of adjacent normal node terms, Denote the set of adjacent normal nodes of the target normal node i.

[0126] It should be further noted that, in the specific implementation process, the process of the critical trend detection module extracting several critical abnormal trend features of each node includes:

[0127] Extract the moments when the node is marked as an abnormal node within the historical collection period, extract the numerical time series and key indicators of various types of indicators between the start timestamp of the historical collection period and the moment, mark other key indicators except the key indicators as non-key indicators, mark the numerical time series as the critical abnormal numerical time series, and mark the historical collection period as the critical abnormal collection period;

[0128] Obtain the standard time series of the key indicators within the critical abnormal collection period, perform autocorrelation comparison between the critical abnormal numerical time series of the key indicators and the standard time series, obtain the autocorrelation coefficient of the key indicators, perform cross-correlation comparison between the critical abnormal numerical time series of the key indicators, obtain the cross-correlation coefficient between the key indicators, perform cross-correlation comparison between the critical abnormal numerical time series of the key indicators and the non-key indicators, obtain the cross-correlation coefficient between the key indicators and the non-key indicators, construct critical abnormal trend features according to the autocorrelation coefficient of the key indicators, the cross-correlation coefficient between the key indicators, and the cross-correlation coefficient between the key indicators and the non-key indicators, extract the temporary compensation model of the critical abnormal collection period, and associate the temporary compensation model with the critical abnormal trend features;

[0129] Construct a trend feature database, and store the critical abnormal trend features within several abnormal collection periods of each node into the trend feature database.

[0130] It should be further noted that, in the specific implementation process, the process of performing autocorrelation comparison includes:

[0131] ;

[0132] Among them, Denote the autocorrelation coefficient of the type indicator i, Denote the value of the type i at the t-th moment, Denote the value of the standard time series of the type indicator i at the t-th moment, and n denotes the total number of moments;

[0133] The process of performing cross-correlation comparison includes:

[0134] ;

[0135] Among them, Denote the cross-correlation coefficient between the type indicator i and the type indicator j; Denote the value of type index i at the t-th moment; Denote the average value of type index i; Denote the value of type index j at the t-th moment; Denote the average value of type index j.

[0136] It should be further noted that in the specific implementation process, the process of obtaining the standard time series of each type of index in the historical collection period includes:

[0137] Construct a data prediction model based on deep learning, use the multi-source type data of each node in several historical collection periods as training data, train the data prediction model with the training data to obtain the trained data prediction model, input the multi-source type data in the critical abnormal collection period into the data prediction model, and obtain the standard time series of each type of index according to the output of the data prediction model.

[0138] It should be further noted that in the specific implementation process, the process of the critical trend detection module performing critical trend detection operations on each normal node and performing parameter distribution operations on the rapid response compensation module according to the results of the critical trend detection operations includes:

[0139] Extract the trend features of the numerical time series of each type of index in the current collection period of the normal node to obtain the trend features in the current collection period of the normal node (including the autocorrelation coefficients of each type of index and the cross-correlation coefficients between each type of index), perform retrieval and comparison in the trend feature database according to the trend features, obtain the similarity between the trend features and several critical abnormal trend features. If the similarity between the trend features and the critical abnormal trend features is greater than the preset similarity threshold, mark the normal node as a critical warning node, customize the parameters of the temporary compensation model for the critical warning node according to the temporary compensation model associated with the critical abnormal trend feature, directly skip the process of screening out the top k historical fault cases with the highest cosine similarity, and then perform rapid response compensation operations on the critical warning node according to the temporarily compensated model with customized parameters.

[0140] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. Adaptive data fusion and decision optimization system for heterogeneous industrial equipment, characterized by: It includes a cloud, wherein the cloud is connected to a data acquisition module, a protocol annotation module, a protocol self-consistent module, a detection decision module, a rapid response compensation module and a critical trend detection module; The data acquisition module is used to collect multi-source data of various heterogeneous industrial equipment, mark the collection time, and set the collection cycle; The protocol annotation module is used to construct an industrial connection topology map and perform initial communication protocol annotation on the industrial connection topology map; The protocol self-consistent module is used to analyze the interactive features of each node in the industrial connection topology diagram, extract the protocol fingerprint features of the protocol abnormal nodes, obtain the protocol fingerprint features, build a protocol fingerprint map library, build an automatic protocol matching model based on the protocol fingerprint map library using reinforcement learning, input the protocol fingerprint features into the automatic protocol matching model, output the communication protocol of the protocol abnormal node, and update the initial communication protocol of the protocol abnormal node according to the communication protocol; The detection decision module is used to perform real-time detection of multi-source data of each node, and select to perform rapid response compensation operation or critical trend detection operation according to the monitoring results; The rapid response compensation module is used to generate a temporary compensation model, build an association rule knowledge graph according to the industrial connection topology graph, the association rule knowledge graph includes multi-source type data, spatiotemporal feature sequences and joint compensation strategies between interconnected nodes, build a joint compensation model based on deep learning, perform rapid response compensation operations on abnormal nodes according to the temporary compensation model, update the joint compensation model according to the results of the rapid response compensation operation, and then perform joint compensation operations according to the joint compensation model; The critical trend detection module is used to extract several critical abnormal trend features of each node, perform critical trend detection operations on each normal node, and issue parameters to the rapid response compensation module according to the critical trend detection operation results.

2. The adaptive data fusion and decision optimization system for heterogeneous industrial equipment according to claim 1 is characterized in that: The protocol annotation module constructs an industrial connection topology map. The process of initial communication protocol annotation on the industrial connection topology map includes: Obtain the assembly connection relationship of several heterogeneous industrial equipment, use several heterogeneous industrial equipment as nodes of an industrial connection topology map, use the assembly connection relationship of several heterogeneous industrial equipment as the connection relationship between nodes, construct an industrial connection topology map, obtain the initial communication protocol between several heterogeneous industrial equipment, and mark the connection relationship between each node with the initial communication protocol.

3. The adaptive data fusion and decision optimization system for heterogeneous industrial equipment according to claim 2 is characterized in that: The process of the protocol self-consistent module analyzing the interactive characteristics of each node in the industrial connection topology diagram includes: According to the initial communication protocol, the threshold intervals of various types of interaction indicators between nodes are preset, and the interaction features of the nodes with connected relationships are extracted to obtain various types of interaction indicators. If the type of interaction indicator is not within the corresponding interaction indicator threshold interval, the node with the connected relationship is marked as a protocol abnormal node.

4. The adaptive data fusion and decision optimization system for heterogeneous industrial equipment according to claim 3 is characterized in that: The process of building a fingerprint protocol graph includes: Based on the idea of ​​simulation learning, a simulator is used to generate interactive training data of different communication protocols. Baud rate analysis, data frame interval analysis and cyclic redundancy check analysis are performed on the interactive training data of different communication protocols. The baud rate probability model, data frame interval probability model and cyclic redundancy check probability model are constructed. The baud rate probability model, data frame interval probability model and cyclic redundancy check probability model are combined to construct a fingerprint protocol map.

5. The adaptive data fusion and decision optimization system for heterogeneous industrial equipment according to claim 4 is characterized in that: The detection decision module performs real-time detection on multi-source data of each node, and the process of selecting to perform rapid response compensation operation or critical trend detection operation according to the monitoring results includes: Extract the numerical time series sequence of each type of indicator in the multi-source type data of each node in the current collection cycle, preset the standard threshold interval corresponding to each type of indicator of each node, compare the numerical time series sequence of each type of indicator with the corresponding standard threshold interval, and obtain the cumulative time that each type of indicator is not in the corresponding standard threshold interval; A cumulative time threshold is preset. If the cumulative time of a node's type indicator is greater than the cumulative time threshold, a sudden abnormal alarm signal is generated and a rapid response compensation operation is performed. If the cumulative time of each type of indicator of the node is less than or equal to the cumulative time threshold, the node is marked as a normal node and a critical trend detection operation is performed.

6. The adaptive data fusion and decision optimization system for heterogeneous industrial equipment according to claim 5 is characterized in that: The process of generating a temporary compensation model by the rapid response compensation module includes: Mark the type indicator whose corresponding cumulative time is greater than the cumulative time threshold as a key indicator, mark the node to which the key indicator belongs as an abnormal node, and extract features from multi-source type data of the abnormal node in the current collection cycle based on the key indicator to obtain a feature vector set; Constructing a sudden abnormal fault library, the sudden abnormal fault library includes a number of historical fault cases, performing cosine similarity comparison between the feature data set and the number of historical fault cases, and screening out the top k historical fault cases with the highest cosine similarity; A temporary compensation model is constructed, and the compensation strategies of the top k historical fault cases are extracted to customize the parameters of the temporary compensation model. At the same time, the temporary compensation model is verified according to the multi-source type data of the abnormal nodes to obtain a verified temporary compensation model.

7. The adaptive data fusion and decision optimization system for heterogeneous industrial equipment according to claim 6 is characterized in that: The rapid response compensation module generates a joint compensation model, performs a rapid response compensation operation on abnormal nodes according to the temporary compensation model, and updates the joint compensation model according to the result of the rapid response compensation operation, including: The multi-source data, spatiotemporal feature sequences and joint compensation strategies are used as training sets and test sets, the training sets are input into the joint compensation model for training until the loss function training is stable, and the model parameters are saved, the joint compensation model is tested by the test set until it meets the preset requirements, and the joint compensation model is output; The temporary compensation model is encapsulated and deployed, and abnormal nodes are compensated according to the temporary compensation model. The compensation time threshold is preset. When the abnormal nodes within the compensation time threshold are marked as normal nodes, the compensation parameters and effective data of the temporary compensation model are collected to update the association rule knowledge graph, and the joint compensation model is incrementally trained according to the updated association rule knowledge graph.

8. The adaptive data fusion and decision optimization system for heterogeneous industrial equipment according to claim 7 is characterized in that: The process of the rapid response compensation module performing joint compensation operations according to the joint compensation model includes: When an abnormal node is marked as a normal node, time features and spatial features are extracted from multi-source type data of the normal node and other normal nodes connected to the normal node in the current acquisition cycle to obtain a spatiotemporal feature sequence. The multi-source type data and the spatiotemporal feature sequence are input into the joint compensation model after incremental training, and a joint compensation strategy is output according to the joint compensation model.

9. The adaptive data fusion and decision optimization system for heterogeneous industrial equipment according to claim 8 is characterized in that: The process of extracting several critical abnormal trend features of each node by the critical trend detection module includes: Extract the moment when the node is marked as an abnormal node in the historical collection period, extract the numerical time series sequence and key indicators of various types of indicators between the start timestamp of the historical collection period and the moment, mark other key indicators other than the key indicators as non-key indicators, mark the numerical time series sequence as a critical abnormal numerical time series sequence, and mark the historical collection period as a critical abnormal collection period; Obtaining a standard time series sequence of key indicators within a critical anomaly acquisition period, performing an autocorrelation comparison between the critical anomaly value time series sequence of the key indicators and the standard time series sequence, obtaining an autocorrelation coefficient of the key indicators, performing a cross-correlation comparison between the critical anomaly value time series sequences between the key indicators, obtaining a mutual correlation coefficient between the key indicators, performing a cross-correlation comparison between the critical anomaly value time series sequences of the key indicators and non-key indicators, obtaining a mutual correlation coefficient between the key indicators and the non-key indicators, constructing a critical anomaly trend feature according to the autocorrelation coefficient of the key indicators, the mutual correlation coefficient between the key indicators, and the mutual correlation coefficient between the key indicators and the non-key indicators, and extracting a temporary compensation model for the critical anomaly acquisition period, and associating the temporary compensation model with the critical anomaly trend feature; A trend feature database is constructed, and critical abnormal trend features within several abnormal collection cycles of each node are stored in the trend feature database.

10. The adaptive data fusion and decision optimization system for heterogeneous industrial equipment according to claim 9 is characterized in that: The critical trend detection module performs a critical trend detection operation on each normal node, and sends parameters to the rapid response compensation module according to the critical trend detection operation result, including: Obtain the trend characteristics of the normal node in the current acquisition cycle, search and compare in the trend characteristic database according to the trend characteristics, obtain the similarity between the trend characteristics and several critical abnormal trend characteristics, if the similarity between the trend characteristics and the critical abnormal trend characteristics is greater than the preset similarity threshold, mark the normal node as a critical warning node, customize the parameters of the temporary compensation model of the critical warning node according to the temporary compensation model associated with the critical abnormal trend characteristics, and then perform a rapid response compensation operation on the critical warning node.

Images