A method and device for acquiring heterogeneous information based on feature mapping

Abstract

The present invention discloses a method and device for acquiring heterogeneous information based on feature mapping, wherein the method comprises: collecting raw data in various industrial control systems to obtain raw data information; performing data dimension reduction processing on the raw data information to obtain raw data information after dimension reduction; performing data fusion processing on the raw data information after dimension reduction by feature extraction to obtain fused data information; inputting the fused data information into a fusion storage module for storage based on time sequence. In the embodiment of the present invention, the storage and management of raw data are greatly simplified; and unified fusion between different physical dimension data structures based on dimensionless feature data is achieved; and direct exposure of raw sensitive data is avoided when it is disclosed to the outside.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial data acquisition, and in particular to a method and device for acquiring heterogeneous information based on feature mapping. Background Art

[0002] Industrial automation and informatization have enabled more and more industrial departments to achieve global control of the production process through industrial control systems (such as ERP / MES / DCS / SCADA systems); on the other hand, affected by technical factors such as product process flow and upstream and downstream supporting capabilities, industrial control systems often differ significantly during on-site implementation. Taking chemical industry control as an example, there are more than 200 common industrial field control bus types, and the collected data involves more than 40 physical dimensions. From the perspective of the upstream and downstream of the chemical supply chain, there are up to more than 60 supporting equipment manufacturers, and each has its own focus on technical solutions. Therefore, the on-site process control of a typical industrial control system is an application scenario with very obvious heterogeneous characteristics. In addition, from a non-technical perspective, medium and large industrial departments generally split industrial control systems into different supply conditions under the considerations of preventing technology monopoly, technology confidentiality, and control security, and different suppliers carry out equipment installation and commissioning separately, further exacerbating the problem of effective management of data types in industrial control systems.

[0003] Therefore, how to efficiently collect information on massive heterogeneous states of different physical dimensions in industrial control systems, and at the same time achieve privacy protection for the confidentiality and security requirements of core industrial process data through data feature learning methods, will become an important basic support method for the effective interconnection of information in the future direction of Industry 4.0. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art. The present invention provides a method and device for obtaining heterogeneous information based on feature mapping, which greatly simplifies the storage and management of original data; and realizes that different physical dimension data structures can be unified and integrated based on dimensionless feature data; and when it is externalized, it avoids direct exposure of original sensitive data.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a method for acquiring heterogeneous information based on feature mapping, the method comprising:

[0006] Collect the original data in various industrial control systems to obtain original data information;

[0007] Performing data dimensionality reduction processing on the original data information to obtain original data information after dimensionality reduction;

[0008] Performing data fusion processing on the original data information after dimensionality reduction by using a feature extraction method to obtain fused data information;

[0009] The fused data information is input into a fusion storage module for storage based on time sequence.

[0010] Optionally, the method further includes:

[0011] Based on the corresponding mathematical model, the priori data model in the data model library is called to perform calculation processing on the fused data information stored in the fusion storage module according to different types of eigenvalues, and the calculation results are output;

[0012] Determining whether the calculation result meets the expected result;

[0013] If not, the original data information after dimensionality reduction is re-processed for data fusion based on the calculation result by using a feature extraction method to obtain fused data information.

[0014] Optionally, the corresponding mathematical model is constructed according to the application scenario requirements, and has an operation structure, calculation parameters, normal operation and error recovery required by the application scenario.

[0015] Optionally, the prior data model includes templates of corresponding mathematical models, callable sub-functions, function libraries and other model types with prior knowledge.

[0016] Optionally, the collecting of raw data in various industrial control systems to obtain raw data information includes:

[0017] Determine the collection method of the original data to be collected, wherein the collection method includes contact collection and non-contact collection;

[0018] When it is determined that the collection method of the original data to be collected is contact collection, the original data of various industrial control systems are collected based on the contact method to obtain original data information, where the original data information is recordable digital information;

[0019] When it is determined that the collection method of the original data to be collected is non-contact collection, the parameters of the non-contact collection equipment are optimized according to the actual industrial site conditions to obtain the processed non-contact collection equipment, wherein the non-contact collection equipment includes industrial vision equipment, industrial remote sensing equipment, and industrial physical property conversion device equipment;

[0020] The processed non-contact acquisition device collects the original data of various industrial control systems to obtain original data information, which is image information.

[0021] Optionally, performing data dimensionality reduction processing on the original data information to obtain the original data information after dimensionality reduction includes:

[0022] When the original data information is obtained by contact collection, losslessly compressing and reducing the dimension of the recordable digital information to obtain the original data information after the dimension reduction;

[0023] When the original data information is obtained by non-contact collection, the image information is compressed and dimensionally reduced to obtain the original data information after dimension reduction; wherein, compressing and reducing the dimension of the image information includes image denoising, compression and pooling processing.

[0024] Optionally, the step of performing data fusion processing on the original data information after dimensionality reduction by using a feature extraction method to obtain fused data information includes:

[0025] The reduced-dimensional raw data of different physical elements involved are used as input data after removing the physical dimensions and input into the data fusion model;

[0026] Taking the fusion classification objective function as the output value of the data fusion model to obtain a fusion classification result;

[0027] Based on the neural network model, the fusion mapping process between the input data and the fusion classification result is completed in the data fusion model, and all parameters generated in the fusion mapping process are retained as feature values ​​to obtain fused data information.

[0028] Optionally, the fused data information is a physically dimensionless eigenvalue.

[0029] In addition, an embodiment of the present invention further provides a heterogeneous information acquisition device based on feature mapping, the device comprising:

[0030] Data acquisition module: used to collect raw data in various industrial control systems and obtain raw data information;

[0031] Data processing module: used to perform data dimensionality reduction processing on the original data information to obtain the original data information after dimensionality reduction;

[0032] Data fusion module: used to perform data fusion processing on the original data information after dimension reduction by using feature extraction method to obtain fused data information;

[0033] Data storage module: used to input the fused data information into the fusion storage module for storage based on time sequence.

[0034] Optionally, the device further comprises:

[0035] The operation processing module is used to call the prior data model in the data model library based on the corresponding mathematical model to perform operation processing on the fused data information stored in the fusion storage module according to different types of characteristic values, and output the calculation results;

[0036] Judgment module: used to judge whether the calculation result meets the expected result; and if it is judged that it does not meet the expected result, the original data information after dimensionality reduction is re-processed by feature extraction based on the calculation result to obtain the fused data information.

[0037] In the embodiments of the present invention, the storage and management of raw data are greatly simplified; and different physical dimension data structures can be unified and integrated based on dimensionless characteristic data, so that the collected data retains the characteristic meaning of the original data, and can avoid the problem that the data structure is too large and difficult to maintain and store due to the physical dimensions carried by the parameters of the industrial control system; and when it is externally exposed, the direct exposure of the original sensitive data is avoided. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 is a flow chart of a method for acquiring heterogeneous information based on feature mapping in an embodiment of the present invention;

[0040] Figure 2 It is a schematic diagram of the structural composition of a heterogeneous information acquisition device based on feature mapping in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Example

[0043] See also Figure 1 , Figure 1 It is a flowchart of a method for acquiring heterogeneous information based on feature mapping in an embodiment of the present invention.

[0044] like Figure 1 As shown, a method for acquiring heterogeneous information based on feature mapping, the method comprising:

[0045] S11: Collect the original data in various industrial control systems to obtain original data information;

[0046] In the specific implementation process of the present invention, the raw data in various industrial control systems are collected to obtain raw data information, including: determining the collection method of the raw data to be collected, wherein the collection method includes contact collection and non-contact collection; when it is determined that the collection method of the raw data to be collected is contact collection, the raw data of various industrial control systems are collected based on contact to obtain raw data information, and the raw data information is recordable digital information; when it is determined that the collection method of the raw data to be collected is non-contact collection, the parameters of the non-contact collection equipment are optimized according to the actual industrial site conditions to obtain the processed non-contact collection equipment, wherein the non-contact collection equipment includes industrial vision equipment, industrial remote sensing equipment, and industrial physical property conversion device equipment; based on the processed non-contact collection equipment, the raw data of various industrial control systems are collected to obtain raw data information, and the raw data information is image information.

[0047] Specifically, in industrial control systems, there are many types of data, some of which can be directly collected using a contact method, while others require a non-contact method. Therefore, before collecting data, it is first necessary to determine the collection method of the original data to be collected, which includes a contact collection method and a non-contact collection method. When it is determined that the collection method of the original data to be collected is contact collection, the original data of various industrial control systems are then collected according to the contact collection method to obtain the original data information, and the original data information is recordable digital information.

[0048] When it is determined that the collection method of the original data to be collected is non-contact collection, it is necessary to optimize the parameters of the non-contact collection equipment according to the actual industrial site conditions to obtain the processed non-contact collection equipment, where the non-contact collection equipment includes industrial vision equipment, industrial remote sensing equipment, and industrial physical property conversion device equipment; then use the processed non-contact collection equipment to collect the original data of various industrial control systems to obtain the original data information; these original data information are image information.

[0049] S12: Performing data dimensionality reduction processing on the original data information to obtain original data information after dimensionality reduction;

[0050] In the specific implementation process of the present invention, the data dimensionality reduction processing is performed on the original data information to obtain the original data information after dimensionality reduction, including: when the original data information is obtained by contact collection, the recordable digital information is losslessly compressed and processed by dimensionality reduction to obtain the original data information after dimensionality reduction; when the original data information is obtained by non-contact collection, the image information is compressed and processed by dimensionality reduction to obtain the original data information after dimensionality reduction; wherein, the compression and dimensionality reduction processing of the image information includes image denoising, compression and pooling processing.

[0051] Specifically, the original data information is subjected to data dimensionality reduction processing, wherein the form of the original data information is different, and the method of dimensionality reduction of the collected data is also different. When the original data information is obtained by contact collection, that is, when it is recordable digital information, it is necessary to perform lossless compression and dimensionality reduction processing on the original data information, wherein the lossless compression processing is a digital compression algorithm. After the original data information is compressed by the digital compression algorithm, a data space mapping algorithm is used to perform space mapping processing, and finally lossless compression and dimensionality reduction processing are achieved, thereby obtaining the original data information after dimensionality reduction.

[0052] When the original data information is obtained by non-contact collection, the original data information is compressed and dimensionally reduced to obtain the original data information after dimension reduction; because the original data information is image information, the compression and dimension reduction processing includes: denoising the image information, then compressing it, further performing pooling processing, and finally performing space mapping processing through a data space mapping algorithm to finally obtain the original data information after dimension reduction.

[0053] S13: performing data fusion processing on the original data information after dimensionality reduction by using a feature extraction method to obtain fused data information;

[0054] In the specific implementation process of the present invention, the feature extraction method is used to perform data fusion processing on the original data information after dimensionality reduction to obtain the fused data information, including: using the original data after dimensionality reduction of different physical elements involved as input data after removing the physical dimension, and inputting it into the data fusion model; using the fusion classification objective function as the output value of the data fusion model to obtain the fusion classification result; completing the fusion mapping process between the input data and the fusion classification result in the data fusion model based on the neural network model, and retaining all parameters generated in the fusion mapping process as feature values ​​to obtain the fused data information.

[0055] Furthermore, the fused data information is a physically dimensionless eigenvalue.

[0056] Specifically, data fusion processing is performed on the original data information after dimensionality reduction, mainly data fusion processing is performed on the original data information after dimensionality reduction collected by contact and the original data information after dimensionality reduction collected by non-contact, and the data is fused through feature extraction method.

[0057] The reduced-dimensional original data of different physical elements (the reduced-dimensional original data information collected in contact and the reduced-dimensional original data information collected in non-contact) are used as input data after removing the physical dimensions, and are all input into the data fusion model; the fusion classification objective function is then used as the output value of the data fusion model to obtain the fusion classification result; then the fusion mapping process between the input data and the fusion classification result is completed in the data fusion model according to the neural network model or other classification network model, and all parameters generated in the fusion mapping process are retained as feature values ​​to obtain the fused data information.

[0058] Furthermore, the fused data information is a dimensionless eigenvalue.

[0059] S14: inputting the fused data information into a fusion storage module for storage based on time sequence;

[0060] In the specific implementation process of the present invention, after the fused data information is obtained after the data fusion is completed, the fused data information retaining the eigenvalues ​​with or without physical dimensions is input into the fusion storage module in chronological order for storage.

[0061] S15: Based on the corresponding mathematical model, calling the priori data model in the data model library performs calculation processing on the fused data information stored in the fusion storage module according to different types of eigenvalues, and outputs the calculation result;

[0062] In the specific implementation process of the present invention, the corresponding mathematical model is constructed according to the application scenario requirements, and has the operation structure, calculation parameters, normal operation and error recovery required by the application scenario.

[0063] Furthermore, the a priori data model includes templates of corresponding mathematical models, callable sub-functions, function libraries and other model types with a priori knowledge.

[0064] Specifically, according to specific needs, the corresponding mathematical model is used to call the prior data model in the data model library to perform operations on the fused data information stored in the fusion storage module according to different types of eigenvalues, and the calculation results are output; when performing operations on different types of eigenvalues, the operation process includes but is not limited to the execution and result acquisition of various algebraic operations, geometric mapping and other mathematical models; the output process of the mathematical model includes but is not limited to design and interpretation. The design process refers to the construction of a mathematical model with various operation structures, calculation parameters, normal operation and error recovery according to the requirements of the application scenario. The interpretation process refers to the interpretation of all operation results, and when the operation results meet the requirements, the correct conclusion of the mathematical model output is given; the corresponding prior data model is stored in the data model library, and the prior data model includes the template of the corresponding mathematical model, callable sub-functions, function library and other model types with prior knowledge.

[0065] S16: Determine whether the calculation result meets the expected result;

[0066] In the specific implementation process of the present invention, it is determined whether the calculation result meets the expected result. When it is determined that the calculation result does not meet the expected result, the calculation result is fed back to step S13. In step S13, the original data information after dimensionality reduction is re-processed for data fusion according to the calculation result using a feature extraction method to obtain fused data information.

[0067] S17: If it is in compliance, the merged data information is connected to the public network.

[0068] During the specific implementation of the present invention, if the calculation result meets the expectation, the integrated data information will be connected to the public network; and according to the mathematical principle of zero-knowledge proof, the industrial control system does not need to present the original data to the application party, and it can prove that the industrial control system corresponding to the heterogeneous data acquisition device has the capabilities required by the application party.

[0069] Through the above embodiments, the problem of heterogeneous collection when common industrial control systems are interconnected is solved: the information bus and storage processing methods of the industrial sector need to be optimized according to the characteristics of the industry, resulting in dozens of common industrial information buses and hundreds of information systems, which can generate tens of thousands of industrial information data combinations. Therefore, there is a need for a unified information collection and processing device that can collect data without physical dimensions for various heterogeneous industrial environments, so that the collected data retains the original data feature meaning, and can avoid the problem that the data structure is too large and difficult to maintain and store if the parameters of the industrial control system carry physical dimensions; feature learning and privacy protection issues: the industrial sector has extremely high privacy protection requirements for industrial core key data, and the core data is not allowed to be accessed through the public network; on the other hand, the industrial Internet requires the collected heterogeneous data to have interconnection capabilities; therefore, it is necessary to extract the key features of the key original data that is not allowed to access the public network based on the main features of the data, so as to form a conclusion that does not expose the original data and can fully express the main features of the information.

[0070] Therefore, the implementation of the above embodiment can be used to perform unified dimension removal processing on various types of heterogeneous information collected in the industrial site according to the industrial site operation rules. Heterogeneous information refers to a data set formed by various types of information including but not limited to industrial optical data, remote control and remote sensing data, various types of industrial bus data, enterprise financial systems, sales systems, etc. Industrial site operation rules refer to various types of knowledge expected to be obtained from the above data sets, including but not limited to abnormal alarms, production forecasts and other behaviors; in the process of parameter convergence calculation of the artificial intelligence network, there is no direct physical dimension mapping relationship between the original data and the classification results, and the intermediate calculation parameters do not have any physical meaning. Therefore, this project uses two processing steps of dimensionality reduction and fusion of industrial heterogeneous information, extracts the dimensionless characteristics of the original data of industrial control based on the neural network method, and performs data fusion based on the dimensionless characteristics, which greatly simplifies the storage and management of the original data. For example, the financial system, ERP management system, MES / DCS / SCADA production system of an enterprise often generate different structural data, and their data fusion communication is usually the main information contradiction in the operation of the enterprise. The present invention realizes unified information collection and processing of various types of heterogeneous data through two steps: dimensionality reduction and data fusion. Data structures of different physical dimensions can be unified and fused based on dimensionless feature data. In view of the industrial privacy protection demand that typical key data cannot exchange information through the network, the present invention avoids direct exposure of original sensitive data by extracting features from key data. Based on the extracted data features, the basic principle of zero-knowledge proof is used to perform model analysis in the user planning model, thereby obtaining a complete proof of the capabilities of the extracted data features.

[0071] In the embodiments of the present invention, the storage and management of raw data are greatly simplified; and different physical dimension data structures can be unified and integrated based on dimensionless characteristic data, so that the collected data retains the characteristic meaning of the original data, and can avoid the problem that the data structure is too large and difficult to maintain and store due to the physical dimensions carried by the parameters of the industrial control system; and when it is externally exposed, the direct exposure of the original sensitive data is avoided.

[0072] Example

[0073] See also Figure 2 , Figure 2 It is a schematic diagram of the structural composition of a heterogeneous information acquisition device based on feature mapping in an embodiment of the present invention.

[0074] like Figure 2 As shown, a heterogeneous information acquisition device based on feature mapping, the device includes:

[0075] Data acquisition module 21: used to collect raw data in various industrial control systems to obtain raw data information;

[0076] In the specific implementation process of the present invention, the raw data in various industrial control systems are collected to obtain raw data information, including: determining the collection method of the raw data to be collected, wherein the collection method includes contact collection and non-contact collection; when it is determined that the collection method of the raw data to be collected is contact collection, the raw data of various industrial control systems are collected based on contact to obtain raw data information, and the raw data information is recordable digital information; when it is determined that the collection method of the raw data to be collected is non-contact collection, the parameters of the non-contact collection equipment are optimized according to the actual industrial site conditions to obtain the processed non-contact collection equipment, wherein the non-contact collection equipment includes industrial vision equipment, industrial remote sensing equipment, and industrial physical property conversion device equipment; based on the processed non-contact collection equipment, the raw data of various industrial control systems are collected to obtain raw data information, and the raw data information is image information.

[0077] Specifically, in industrial control systems, there are many types of data, some of which can be directly collected using a contact method, while others require a non-contact method. Therefore, before collecting data, it is first necessary to determine the collection method of the original data to be collected, which includes a contact collection method and a non-contact collection method. When it is determined that the collection method of the original data to be collected is contact collection, the original data of various industrial control systems are then collected according to the contact collection method to obtain the original data information, and the original data information is recordable digital information.

[0078] When it is determined that the collection method of the original data to be collected is non-contact collection, it is necessary to optimize the parameters of the non-contact collection equipment according to the actual industrial site conditions to obtain the processed non-contact collection equipment, where the non-contact collection equipment includes industrial vision equipment, industrial remote sensing equipment, and industrial physical property conversion device equipment; then use the processed non-contact collection equipment to collect the original data of various industrial control systems to obtain the original data information; these original data information are image information.

[0079] Data processing module 22: used to perform data dimension reduction processing on the original data information to obtain the original data information after dimension reduction;

[0080] In the specific implementation process of the present invention, the data dimensionality reduction processing is performed on the original data information to obtain the original data information after dimensionality reduction, including: when the original data information is obtained by contact collection, the recordable digital information is losslessly compressed and processed by dimensionality reduction to obtain the original data information after dimensionality reduction; when the original data information is obtained by non-contact collection, the image information is compressed and processed by dimensionality reduction to obtain the original data information after dimensionality reduction; wherein, the compression and dimensionality reduction processing of the image information includes image denoising, compression and pooling processing.

[0081] Specifically, the original data information is subjected to data dimensionality reduction processing, wherein the form of the original data information is different, and the method of dimensionality reduction of the collected data is also different. When the original data information is obtained by contact collection, that is, when it is recordable digital information, it is necessary to perform lossless compression and dimensionality reduction processing on the original data information, wherein the lossless compression processing is a digital compression algorithm. After the original data information is compressed by the digital compression algorithm, a data space mapping algorithm is used to perform space mapping processing, and finally lossless compression and dimensionality reduction processing are achieved, thereby obtaining the original data information after dimensionality reduction.

[0082] When the original data information is obtained by non-contact collection, the original data information is compressed and dimensionally reduced to obtain the original data information after dimension reduction; because the original data information is image information, the compression and dimension reduction processing includes: denoising the image information, then compressing it, further performing pooling processing, and finally performing space mapping processing through a data space mapping algorithm to finally obtain the original data information after dimension reduction.

[0083] Data fusion module 23: used for performing data fusion processing on the original data information after dimension reduction by using feature extraction method to obtain fused data information;

[0084] In the specific implementation process of the present invention, the feature extraction method is used to perform data fusion processing on the original data information after dimensionality reduction to obtain the fused data information, including: using the original data after dimensionality reduction of different physical elements involved as input data after removing the physical dimension, and inputting it into the data fusion model; using the fusion classification objective function as the output value of the data fusion model to obtain the fusion classification result; completing the fusion mapping process between the input data and the fusion classification result in the data fusion model based on the neural network model, and retaining all parameters generated in the fusion mapping process as feature values ​​to obtain the fused data information.

[0085] Furthermore, the fused data information is a physically dimensionless eigenvalue.

[0086] Specifically, data fusion processing is performed on the original data information after dimensionality reduction, mainly data fusion processing is performed on the original data information after dimensionality reduction collected by contact and the original data information after dimensionality reduction collected by non-contact, and the data is fused through feature extraction method.

[0087] The reduced-dimensional original data of different physical elements (the reduced-dimensional original data information collected in contact and the reduced-dimensional original data information collected in non-contact) are used as input data after removing the physical dimensions, and are all input into the data fusion model; the fusion classification objective function is then used as the output value of the data fusion model to obtain the fusion classification result; then the fusion mapping process between the input data and the fusion classification result is completed in the data fusion model according to the neural network model or other classification network model, and all parameters generated in the fusion mapping process are retained as feature values ​​to obtain the fused data information.

[0088] Furthermore, the fused data information is a dimensionless eigenvalue.

[0089] Data storage module 24: used for inputting the fused data information into the fusion storage module for storage based on time sequence;

[0090] In the specific implementation process of the present invention, after the fused data information is obtained after the data fusion is completed, the fused data information retaining the eigenvalues ​​with or without physical dimensions is input into the fusion storage module in chronological order for storage.

[0091] The calculation processing module 25 is used to call the prior data model in the data model library based on the corresponding mathematical model to perform calculation processing on the fused data information stored in the fusion storage module according to different types of characteristic values, and output the calculation results;

[0092] In the specific implementation process of the present invention, the corresponding mathematical model is constructed according to the application scenario requirements, and has the operation structure, calculation parameters, normal operation and error recovery required by the application scenario.

[0093] Furthermore, the a priori data model includes templates of corresponding mathematical models, callable sub-functions, function libraries and other model types with a priori knowledge.

[0094] Specifically, according to specific needs, the corresponding mathematical model is used to call the prior data model in the data model library to perform operations on the fused data information stored in the fusion storage module according to different types of eigenvalues, and the calculation results are output; when performing operations on different types of eigenvalues, the operation process includes but is not limited to the execution and result acquisition of various algebraic operations, geometric mapping and other mathematical models; the output process of the mathematical model includes but is not limited to design and interpretation. The design process refers to the construction of a mathematical model with various operation structures, calculation parameters, normal operation and error recovery according to the requirements of the application scenario. The interpretation process refers to the interpretation of all operation results, and when the operation results meet the requirements, the correct conclusion of the mathematical model output is given; the corresponding prior data model is stored in the data model library, and the prior data model includes the template of the corresponding mathematical model, callable sub-functions, function library and other model types with prior knowledge.

[0095] Judgment module 26: used to judge whether the calculation result meets the expected result; and if it is judged that it does not meet the expected result, the original data information after dimensionality reduction is re-processed by feature extraction based on the calculation result to obtain fused data information.

[0096] In the specific implementation process of the present invention, it is determined whether the calculation result meets the expected result. When it is determined that the calculation result does not meet the expected result, the calculation result is fed back to the data fusion module 23. In the data fusion module 23, the original data information after dimensionality reduction is re-processed for data fusion according to the calculation result using a feature extraction method to obtain fused data information.

[0097] During the specific implementation of the present invention, if the calculation result meets the expectation, the integrated data information will be connected to the public network; and according to the mathematical principle of zero-knowledge proof, the industrial control system does not need to present the original data to the application party, and it can prove that the industrial control system corresponding to the heterogeneous data acquisition device has the capabilities required by the application party.

[0098] Through the above embodiments, the problem of heterogeneous collection when common industrial control systems are interconnected is solved: the information bus and storage processing methods of the industrial sector need to be optimized according to the characteristics of the industry, resulting in dozens of common industrial information buses and hundreds of information systems, which can generate tens of thousands of industrial information data combinations. Therefore, there is a need for a unified information collection and processing device that can collect data without physical dimensions for various heterogeneous industrial environments, so that the collected data retains the original data feature meaning, and can avoid the problem that the data structure is too large and difficult to maintain and store if the parameters of the industrial control system carry physical dimensions; feature learning and privacy protection issues: the industrial sector has extremely high privacy protection requirements for industrial core key data, and the core data is not allowed to be accessed through the public network; on the other hand, the industrial Internet requires the collected heterogeneous data to have interconnection capabilities; therefore, it is necessary to extract the key features of the key original data that is not allowed to access the public network based on the main features of the data, so as to form a conclusion that does not expose the original data and can fully express the main features of the information.

[0099] Therefore, the implementation of the above embodiment can be used to perform unified dimension removal processing on various types of heterogeneous information collected in the industrial site according to the industrial site operation rules. Heterogeneous information refers to a data set formed by various types of information including but not limited to industrial optical data, remote control and remote sensing data, various types of industrial bus data, enterprise financial systems, sales systems, etc. Industrial site operation rules refer to various types of knowledge expected to be obtained from the above data sets, including but not limited to abnormal alarms, production forecasts and other behaviors; in the process of parameter convergence calculation of the artificial intelligence network, there is no direct physical dimension mapping relationship between the original data and the classification results, and the intermediate calculation parameters do not have any physical meaning. Therefore, this project uses two processing steps of dimensionality reduction and fusion of industrial heterogeneous information, extracts the dimensionless characteristics of the original data of industrial control based on the neural network method, and performs data fusion based on the dimensionless characteristics, which greatly simplifies the storage and management of the original data. For example, the financial system, ERP management system, MES / DCS / SCADA production system of an enterprise often generate different structural data, and their data fusion communication is usually the main information contradiction in the operation of the enterprise. The present invention realizes unified information collection and processing of various types of heterogeneous data through two steps: dimensionality reduction and data fusion. Data structures of different physical dimensions can be unified and fused based on dimensionless feature data. In view of the industrial privacy protection demand that typical key data cannot exchange information through the network, the present invention avoids direct exposure of original sensitive data by extracting features from key data. Based on the extracted data features, the basic principle of zero-knowledge proof is used to perform model analysis in the user planning model, thereby obtaining a complete proof of the capabilities of the extracted data features.

[0100] In the embodiments of the present invention, the storage and management of raw data are greatly simplified; and different physical dimension data structures can be unified and integrated based on dimensionless characteristic data, so that the collected data retains the characteristic meaning of the original data, and can avoid the problem that the data structure is too large and difficult to maintain and store due to the physical dimensions carried by the parameters of the industrial control system; and when it is externally exposed, the direct exposure of the original sensitive data is avoided.

[0101] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0102] In addition, the above is a detailed introduction to a method and device for acquiring heterogeneous information based on feature mapping provided in an embodiment of the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for acquiring heterogeneous information based on feature mapping, characterized in that: The method comprises: Collect the original data in various industrial control systems to obtain original data information; Performing data dimensionality reduction processing on the original data information to obtain original data information after dimensionality reduction; Performing data fusion processing on the original data information after dimensionality reduction by using a feature extraction method to obtain fused data information; Inputting the fused data information into a fusion storage module for storage based on time sequence; The raw data in various industrial control systems are collected to obtain raw data information, including: Determine the collection method of the original data to be collected, wherein the collection method includes contact collection and non-contact collection; When it is determined that the collection method of the original data to be collected is contact collection, the original data of various industrial control systems are collected based on the contact collection method to obtain original data information, where the original data information is recordable digital information; When it is determined that the collection method of the original data to be collected is non-contact collection, the parameters of the non-contact collection equipment are optimized according to the actual industrial site conditions to obtain the processed non-contact collection equipment, wherein the non-contact collection equipment includes industrial vision equipment, industrial remote sensing equipment, and industrial physical property conversion device equipment; The processed non-contact acquisition device collects the raw data of various industrial control systems to obtain raw data information, wherein the raw data information is image information; The performing data dimensionality reduction processing on the original data information to obtain the original data information after dimensionality reduction includes: When the original data information is obtained by contact collection, losslessly compressing and reducing the dimension of the recordable digital information to obtain the original data information after the dimension reduction; When the original data information is obtained by non-contact collection, compressing and reducing the dimension of the image information to obtain the original data information after dimensionality reduction; wherein compressing and reducing the dimension of the image information includes image denoising, compression and pooling processing; The method of performing data fusion processing on the original data information after dimensionality reduction by using a feature extraction method to obtain fused data information includes: The reduced-dimensional raw data of different physical elements involved are used as input data after removing the physical dimensions and input into the data fusion model; Taking the fusion classification objective function as the output value of the data fusion model to obtain a fusion classification result; Based on the neural network model, a fusion mapping process between the input data and the fusion classification result is completed in the data fusion model, and all parameters generated in the fusion mapping process are retained as feature values ​​to obtain fused data information; The fused data information is a physical dimensionless eigenvalue.

2. The heterogeneous information acquisition method according to claim 1, characterized in that: The method further comprises: Based on the corresponding mathematical model, the priori data model in the data model library is called to perform calculation processing on the fused data information stored in the fusion storage module according to different types of eigenvalues, and the calculation results are output; Determining whether the calculation result meets the expected result; If not, then based on the calculation result, the original data information after dimensionality reduction is re-processed with data fusion by using a feature extraction method to obtain fused data information.

3. The heterogeneous information acquisition method according to claim 2, characterized in that: The corresponding mathematical model is constructed according to the application scenario requirements, and has an operation structure, calculation parameters, normal operation and error recovery required by the application scenario.

4. The heterogeneous information acquisition method according to claim 2, characterized in that: The prior data model includes templates of corresponding mathematical models, callable sub-functions, function libraries and other model types with prior knowledge.

5. A heterogeneous information acquisition device based on feature mapping, characterized in that: The device comprises: Data acquisition module: used to collect raw data in various industrial control systems and obtain raw data information; Data processing module: used to perform data dimensionality reduction processing on the original data information to obtain the original data information after dimensionality reduction; Data fusion module: used to perform data fusion processing on the original data information after dimension reduction by using feature extraction method to obtain fused data information; Data storage module: used for inputting the fused data information into the fusion storage module for storage based on time sequence; The raw data in various industrial control systems are collected to obtain raw data information, including: Determine the collection method of the original data to be collected, wherein the collection method includes contact collection and non-contact collection; When it is determined that the collection method of the original data to be collected is contact collection, the original data of various industrial control systems are collected based on the contact collection method to obtain original data information, where the original data information is recordable digital information; When it is determined that the collection method of the original data to be collected is non-contact collection, the parameters of the non-contact collection equipment are optimized according to the actual industrial site conditions to obtain the processed non-contact collection equipment, wherein the non-contact collection equipment includes industrial vision equipment, industrial remote sensing equipment, and industrial physical property conversion device equipment; The processed non-contact acquisition device collects the raw data of various industrial control systems to obtain raw data information, wherein the raw data information is image information; The performing data dimensionality reduction processing on the original data information to obtain the original data information after dimensionality reduction includes: When the original data information is obtained by contact collection, losslessly compressing and reducing the dimension of the recordable digital information to obtain the original data information after the dimension reduction; When the original data information is obtained by non-contact collection, compressing and reducing the dimension of the image information to obtain the original data information after dimensionality reduction; wherein compressing and reducing the dimension of the image information includes image denoising, compression and pooling processing; The method of performing data fusion processing on the original data information after dimensionality reduction by using a feature extraction method to obtain fused data information includes: The reduced-dimensional raw data of different physical elements involved are used as input data after removing the physical dimensions and input into the data fusion model; Taking the fusion classification objective function as the output value of the data fusion model to obtain a fusion classification result; Based on the neural network model, a fusion mapping process between the input data and the fusion classification result is completed in the data fusion model, and all parameters generated in the fusion mapping process are retained as feature values ​​to obtain fused data information; The fused data information is a physical dimensionless eigenvalue.

6. The heterogeneous information acquisition device according to claim 5, characterized in that: The device also includes: The operation processing module is used to call the prior data model in the data model library based on the corresponding mathematical model to perform operation processing on the fused data information stored in the fusion storage module according to different types of characteristic values, and output the calculation results; Judgment module: used to judge whether the calculation result meets the expected result; and if it is judged that it does not meet the expected result, the original data information after dimensionality reduction is re-processed by feature extraction based on the calculation result to obtain the fused data information.

Images