CN-121985061-A - Data processing method, system, equipment and medium based on industrial Internet of things

CN121985061ACN 121985061 ACN121985061 ACN 121985061ACN-121985061-A

Abstract

The application relates to a data processing method, system, equipment and medium based on industrial Internet of things. The method comprises the steps of connecting multi-source heterogeneous industrial equipment, collecting original binary data frames, extracting static and dynamic characteristics to form complete characteristic vectors, carrying out protocol matching to generate analysis rules, analyzing the data frames to be structured time sequence data, carrying out multi-dimensional characteristic extraction on the time sequence data, inputting pre-training artificial intelligent model analysis to generate an analysis report containing probability predicted values, adjustment suggestions and classification results and visualizing the analysis report, selecting key data items to carry out hash operation, writing the hash values and associated metadata into alliance blockchain evidence, automatically triggering business logic to generate workflow instructions through intelligent contracts, monitoring protocol matching and model performance, and collecting associated data samples to update a protocol library and a model. The method improves the compatibility, analysis precision and dynamic adaptability of the data processing of the industrial Internet of things by fusing the technologies of dynamic protocol matching, blockchain certification and the like.

Inventors

LIN CHAOFU
XIE SHENG
SHI XIAOHONG
Guo Diqing
WEN DASHENG
CHEN PENG
Xie Baofa

Assignees

江西冠英智能科技股份有限公司
赣州市智能产业创新研究院

Dates

Publication Date: 20260505
Application Date: 20260127

Claims (9)

1. A data processing method based on industrial internet of things, the method comprising: Communication interface connection is carried out on multi-source heterogeneous industrial equipment, and an original binary data frame with a time stamp and a data source identifier is acquired; extracting static features and dynamic features based on the original binary data frame to obtain a complete feature vector, carrying out protocol matching based on the complete feature vector and a preset protocol library to obtain a protocol matching result, and generating a corresponding analysis rule according to the protocol matching result; The method comprises the steps of carrying out multi-dimensional feature extraction processing on structured time sequence data to obtain a multi-dimensional feature matrix, inputting the multi-dimensional feature matrix into a pre-trained artificial intelligent model for analysis processing to obtain an analysis report containing a probability predicted value, an adjustment suggestion and a classification result, wherein the analysis report is used for carrying out visualization processing and is displayed through a visualization platform; Selecting a key data item from the structured time sequence data and the analysis report, carrying out hash operation on the key data item to obtain a hash value, and determining associated metadata corresponding to the key data item, wherein the associated metadata comprises a data time stamp, a data source identifier and an event type; the hash value and the associated metadata are written into a alliance blockchain for certification, and the event type corresponding to the key data item is monitored through an intelligent contract arranged on the alliance blockchain and automatically triggered to execute business logic to generate a digital workflow instruction; And monitoring the protocol matching result and the model analysis performance of the pre-trained artificial intelligent model, and collecting associated data samples corresponding to the unknown protocol or the model analysis performance degradation when the unknown protocol or the model analysis performance degradation is identified, wherein the associated data samples are used for updating a preset protocol library or the pre-trained artificial intelligent model.
2. The method of claim 1, wherein the extracting static features and dynamic features based on the original binary data frame to obtain a complete feature vector, performing protocol matching based on the complete feature vector and a preset protocol library to obtain a protocol matching result, and generating a corresponding parsing rule according to the protocol matching result, and performing parsing processing on the original binary data frame based on the parsing rule to obtain structured time sequence data with semantic tags, comprises: performing frame structure analysis processing on the original binary data frame, extracting frame head and frame tail modes, length field positions, check field types and field offset information, and obtaining static feature vectors; Sequencing each original binary data frame according to the time stamp to obtain a continuous original binary data frame sequence, carrying out statistical analysis processing on the continuous original binary data frame sequence, and calculating a frame length distribution parameter, a byte value information entropy, a time sequence interval distribution rule and a communication interaction mode to obtain a dynamic feature vector; Splicing the static feature vector and the dynamic feature vector to obtain the complete feature vector; Performing multidimensional feature weighted cosine similarity calculation on the complete feature vector and a protocol feature template in the preset protocol library to obtain a similarity score, and comparing the similarity score with a preset matching threshold to obtain the protocol matching result; if the protocol matching result is that the similarity score is lower than the preset matching threshold, the original binary data frames are subjected to grouping processing by adopting a K-Means or DBSCAN density clustering algorithm to obtain a plurality of clusters, the data frames in the clusters are aligned by a sequence comparison algorithm, constant fields and variable fields are identified, field semantics are deduced, and candidate protocol templates and corresponding temporary parsing rules are generated; and taking the standard analysis rule or the temporary analysis rule as the analysis rule, and executing preset analysis operation on the original binary data frame to obtain the structured time sequence data with the semantic tag, wherein the preset analysis operation comprises frame head and frame tail stripping, field segmentation, data type conversion and verification.
3. The method of claim 1, wherein the pre-trained artificial intelligence model comprises an equipment failure prediction model, a process parameter optimization model based on a graph neural network combined with reinforcement learning, and a quality defect recognition model; The method comprises the steps of carrying out multi-dimensional feature extraction processing on the structured time sequence data to obtain a multi-dimensional feature matrix, inputting the multi-dimensional feature matrix into a pre-trained artificial intelligent model for analysis processing to obtain an analysis report containing a probability predicted value, an adjustment suggestion and a classification result, and comprises the following steps: Performing sliding window statistical processing on the structured time sequence data to extract time sequence characteristics comprising mean value, standard deviation and change slope, performing Fourier transform processing on the structured time sequence data to extract frequency domain characteristics comprising main frequency component amplitude and phase, performing statistical calculation on the basis of the structured time sequence data to obtain statistical characteristics comprising skewness, kurtosis and quantile; the time sequence features, the frequency domain features, the statistical features and the association features are subjected to fusion processing, and the multidimensional feature matrix is constructed; inputting the multidimensional feature matrix into a device fault prediction model for analysis and processing to obtain the probability prediction value and a first confidence score of the fault of the target device in a future preset period; Acquiring a factory production flow topological relation, inputting the multidimensional feature matrix and the production flow topological relation into the process parameter optimization model based on the combination of the graph neural network and reinforcement learning for analysis and processing, and obtaining the adjustment recommendation and the second confidence score based on key process parameters; inputting the image data in the structured time sequence data into a quality defect recognition model for analysis and processing to obtain the classification result and a third confidence score; Adopting a D-S evidence theory to perform fusion processing on the first confidence score, the second confidence score and the third confidence score to obtain a comprehensive confidence rating; and carrying out association and integration processing on the probability predicted value, the adjustment suggestion, the classification result and the comprehensive confidence rating to generate the analysis report with the confidence rating.
4. The method of claim 1, wherein the selecting a key data item from the structured time series data and the analysis report, performing a hash operation on the key data item to obtain a hash value, determining associated metadata corresponding to the key data item, wherein the associated metadata includes a data timestamp, a data source identifier, and an event type, writing the hash value and the associated metadata into a federation blockchain for certification, monitoring the event type corresponding to the key data item through an intelligent contract deployed on the federation blockchain, and automatically triggering execution business logic, and generating a digital workflow instruction, comprising: According to a predefined business rule, selecting an equipment state alarm record from the structured time sequence data, and selecting a fault prediction record, a quality defect record and an energy consumption exceeding record marked as a key conclusion from the analysis report as the key data item; carrying out serialization processing on the key data items to obtain serialized key data items, and carrying out operation on the serialized key data items by adopting an SHA-256 encryption hash algorithm to obtain the hash value; Extracting a data time stamp and a data source identifier of the key data item, determining an event type according to the type of the key data item, and taking the data time stamp, the data source identifier and the event type as the associated metadata; Packaging the hash value and the associated metadata into a transaction proposal, and submitting the transaction proposal to a alliance blockchain network consisting of a factory node, an equipment provider node, an endorsement node and a supervision mechanism node; signing the transaction proposal through an endorsement node in the alliance blockchain network to obtain a transaction proposal passing verification, sequencing the transaction proposal passing verification through Raft or PBFT consensus mechanism, packaging the transaction proposal passing verification into a new block and adding the new block into the alliance blockchain; monitoring the event type through a maintenance intelligent contract deployed on the alliance blockchain, and automatically generating a preventive maintenance work order comprising responsible personnel, a processing flow and an expiration date when the event type is monitored to be the fault prediction; When the event type is monitored to be a quality defect, automatically associating the full-flow evidence-storing record of the corresponding production batch to generate a quality traceability report through traceability intelligent contracts deployed on the alliance blockchain; and pushing the preventive maintenance work order and the quality traceability report to a corresponding service management system as the digital workflow instruction.
5. The method of claim 2, wherein the monitoring the protocol matching results and the model analysis performance of the pre-trained artificial intelligence model, when an unknown protocol or the model analysis performance degradation is identified, collecting associated data samples corresponding to the unknown protocol or the model analysis performance degradation, comprises: Monitoring similarity score distribution in the protocol matching result in real time, and counting the original binary data frame duty ratio and cluster generation frequency corresponding to the similarity score lower than the preset matching threshold, wherein when the original binary data frame duty ratio is higher than the preset duty ratio threshold or the cluster generation frequency is higher than the preset frequency threshold, a judgment result is obtained, and the judgment result is that the unknown protocol is identified; Calculating a prediction error, a false alarm rate and a missing alarm rate of the pre-trained artificial intelligent model according to a preset period, and obtaining a judgment result when the prediction error is higher than a preset error threshold, the false alarm rate is higher than a preset false alarm threshold or the missing alarm rate is higher than a preset missing alarm threshold, wherein the judgment result is that the analysis performance of the model is reduced; When the judging result is that the unknown protocol is identified, collecting the original binary data frame and the corresponding feature vector in the corresponding cluster as the associated data sample corresponding to the unknown protocol; And when the judging result is that the analysis performance of the model is reduced, collecting the structured time sequence data, the multidimensional feature matrix and the corresponding analysis report in the period of reduced analysis performance of the model as the associated data sample corresponding to the reduced analysis performance of the model.
6. The method of claim 2, wherein the similarity score is calculated by the formula: Wherein, the Representing the similarity score, wherein the value range is [0,1], Representing the dimensions of the complete feature vector, Represent the first The weight of the individual features is determined, Representing the first of the complete feature vectors to be matched The values of the individual characteristics are chosen, Representing the first of the preset protocol feature templates The numerical value of each feature.
7. A data processing system based on the industrial internet of things, the system comprising: The system comprises a data acquisition and protocol matching module, a structural timing data processing module, a data processing module and a data processing module, wherein the data acquisition and protocol matching module is used for carrying out communication interface connection on multi-source heterogeneous industrial equipment, acquiring an original binary data frame with a time stamp and a data source identifier, extracting static characteristics and dynamic characteristics based on the original binary data frame to obtain a complete characteristic vector; The data analysis and report generation module is used for carrying out multidimensional feature extraction processing on the structured time sequence data to obtain a multidimensional feature matrix, inputting the multidimensional feature matrix into a pre-trained artificial intelligent model for analysis processing to obtain an analysis report containing a probability predicted value, an adjustment suggestion and a classification result, wherein the analysis report is used for carrying out visualization processing and is displayed through a visualization platform; the system comprises a data storage and intelligent contract execution module, a digital workflow instruction, a data management module and a data management module, wherein the data storage and intelligent contract execution module is used for selecting a key data item from the structured time sequence data and the analysis report, carrying out hash operation on the key data item to obtain a hash value, determining associated metadata corresponding to the key data item, wherein the associated metadata comprises a data time stamp, a data source identifier and an event type; The model monitoring and updating module is used for monitoring the protocol matching result and the model analysis performance of the pre-trained artificial intelligent model, and collecting a correlation data sample corresponding to the unknown protocol or the model analysis performance degradation when the unknown protocol or the model analysis performance degradation is identified, wherein the correlation data sample is used for updating a preset protocol library or the pre-trained artificial intelligent model.
8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 6 when the computer program is executed.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.

Description

Data processing method, system, equipment and medium based on industrial Internet of things Technical Field The invention belongs to the technical field of computers, and particularly relates to a data processing method, system, equipment and medium based on industrial Internet of things. Background Along with the deep digital transformation of manufacturing industry, the intelligent industry has become the core direction of industry upgrading, the wide penetration of the internet of things technology promotes the industrial production to develop to intelligence and high efficiency, and the networking cooperation of multi-source heterogeneous industrial equipment becomes a typical characteristic of an intelligent factory. However, a plurality of technical bottlenecks are still faced in the current data processing process based on the industrial internet of things. For example, industrial equipment of different manufacturers adopts various communication protocols such as Modbus, OPC-UA and the like, so that protocol barriers exist in data acquisition of multi-source heterogeneous industrial equipment, the existing data processing method is dependent on a fixed protocol template, unknown protocols are difficult to flexibly adapt, the acquired original binary data frames cannot be efficiently converted into structured time sequence data with semantic tags due to lack of uniform feature extraction and matching mechanisms, and data splitting and semantic deletion are caused. Specifically, in terms of data analysis, the traditional method is single in feature extraction of structural time sequence data, fails to fully integrate time sequence, statistics and other multidimensional features to construct a multidimensional feature matrix, a pre-trained artificial intelligent model is easy to generate performance attenuation when the working condition of a production line is changed, an effective dynamic optimization mechanism is lacked, the generated analysis report is insufficient in accuracy and reliability of prediction and adjustment advice, the safety and traceability of the data are difficult to be considered in the existing scheme, key data items in the structural time sequence data and the analysis report lack of a decentralization evidence storage means, corresponding business logic cannot be automatically triggered through intelligent contracts, digital workflow instructions are generated with hysteresis, and production operation and maintenance efficiency is affected. In addition, the preset protocol library and the artificial intelligent model in the traditional method cannot be updated in time according to the occurrence of an unknown protocol or the performance degradation of the model, and the dynamic change of an industrial scene is difficult to adapt for a long time. Disclosure of Invention Based on the above, it is necessary to provide a data processing method, system, device and medium based on the industrial internet of things, aiming at improving the flexibility, accuracy, safety and dynamic adaptability of the data processing of the industrial internet of things and promoting the efficient operation and industrial upgrading of the intelligent factory. In a first aspect, the present application provides a data processing method based on industrial internet of things, including: Extracting static features and dynamic features based on the original binary data frames to obtain complete feature vectors, carrying out protocol matching based on the complete feature vectors and a preset protocol library to obtain protocol matching results, and generating corresponding analysis rules according to the protocol matching results; The method comprises the steps of carrying out multi-dimensional feature extraction processing on structural time sequence data to obtain a multi-dimensional feature matrix, inputting the multi-dimensional feature matrix into a pre-trained artificial intelligent model for analysis processing to obtain an analysis report containing a probability predicted value, an adjustment suggestion and a classification result, wherein the analysis report is used for carrying out visualization processing and is displayed through a visualization platform; Selecting a key data item from the structured time sequence data and the analysis report, carrying out hash operation on the key data item to obtain a hash value, determining associated metadata corresponding to the key data item, wherein the associated metadata comprises a data timestamp, a data source identifier and an event type; and monitoring a protocol matching result and the model analysis performance of the pre-trained artificial intelligent model, and collecting a correlation data sample corresponding to the unknown protocol or the model analysis performance degradation when the unknown protocol or the model analysis performance degradation is identified, wherein the correlation data sample is used for updating a preset protocol libra