Industrial algorithm real-time verification platform, method and system based on cloud-edge-end collaboration

Abstract

The present invention relates to the field of cloud-edge-end collaboration technology, and discloses a real-time verification platform, method and system for industrial algorithms based on cloud-edge-end collaboration. The platform closely combines key processes such as data collection, model training, and verification, significantly improving real-time performance and efficiency. Real-time data collection and processing of sensors in industrial scenarios, coupled with online training and verification of models, effectively promote rapid response to changes in the production environment, fully support the data processing and algorithm verification needs of industrial production, and provide a solid technical foundation for realizing intelligent and automated production.

Description

Technical Field

[0001] The present invention relates to the field of cloud-edge-end collaboration technology, and specifically to a real-time verification platform, method and system for industrial algorithms based on cloud-edge-end collaboration. Background Art

[0002] With the continuous development of the industrial field and the improvement of intelligence, the collection, processing and algorithm verification of industrial data have become increasingly important. Traditional industrial control systems often rely on centralized data processing, which may lead to problems such as data transmission delays and security. In order to meet these challenges, edge computing and cloud computing technologies have gradually become an effective means to solve data processing and algorithm verification problems in the industrial field. However, in traditional industrial production, algorithm verification is usually carried out offline, which limits its real-time and adaptability, making it difficult to monitor and analyze data in real time and unable to adapt to changes in industrial needs. Summary of the invention

[0003] The present invention provides an industrial algorithm real-time verification platform, method and system based on cloud-edge-end collaboration to solve the problem that in traditional industrial production, algorithm verification is usually carried out in an offline manner, which limits its real-time and adaptability, makes it difficult to monitor and analyze data in real time, and cannot adapt to changes in industrial needs.

[0004] In order to achieve the above object, the present invention is implemented by the following technical solutions:

[0005] In a first aspect, the present invention provides an industrial algorithm real-time verification platform based on cloud-edge-end collaboration, comprising:

[0006] The real-time data acquisition unit is used to collect the measured values ​​of industrial field sensors in real time and store them in the database of the edge server;

[0007] The privacy protection unit is used to encrypt the data stored on the platform using a two-way privacy protection mechanism and is also used to encrypt the verification model itself;

[0008] The model online training and verification unit is used to use the data in the edge server to perform online training and verification on the model uploaded by the user, and to associate the model with the corresponding data in the database by binding the input and output parameters of the model, and to train the model. After obtaining the model parameters, the model is verified online using real-time data to obtain the final model;

[0009] The algorithm evaluation unit performs real-time statistics and evaluation on the output results of the final model by setting performance indicators.

[0010] In a second aspect, the present application provides a real-time verification method for industrial algorithms based on cloud-edge-end collaboration, including:

[0011] The real-time data acquisition unit is used to collect the measurement values ​​of industrial field sensors in real time and store them in the database of the edge server;

[0012] A two-way privacy protection mechanism is adopted through the privacy protection unit to encrypt the data stored on the platform and the verification model itself;

[0013] The model online training and verification unit is used to train and verify the model uploaded by the user online using the data in the edge server. The model is associated with the corresponding data in the database by binding the input and output parameters of the model to train the model. After obtaining the model parameters, the model is verified online using real-time data to obtain the final model.

[0014] The algorithm evaluation unit is used to set performance indicators to conduct real-time statistics and evaluation of the output results of the final model.

[0015] In a third aspect, the present application provides a real-time verification system for industrial algorithms based on cloud-edge-end collaboration, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the method described in the first aspect.

[0016] The present invention has the following beneficial effects:

[0017] The real-time verification platform for industrial algorithms based on cloud-edge-end collaboration provided by this application closely integrates key processes such as data collection, model training, and verification, significantly improving real-time performance and efficiency. Real-time data collection and processing of sensors in industrial scenarios, coupled with online training and verification of models, effectively promote rapid response to changes in the production environment, fully support the data processing and algorithm verification needs of industrial production, and provide a solid technical foundation for realizing intelligent and automated production.

[0018] In the further technical solution, the cloud visualization unit provides an intuitive and user-friendly interface, covering model management and performance display functions. Users can easily upload and manage models, and monitor model performance and its ranking in real time, thereby assisting decision making and the implementation of optimization measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the platform structure of an industrial algorithm real-time verification platform based on cloud-edge-end collaboration in a preferred embodiment of the present invention;

[0020] Figure 2 It is a schematic diagram of the operating steps of the cloud-edge-end collaborative industrial algorithm real-time verification platform of the preferred embodiment of the present invention. DETAILED DESCRIPTION

[0021] The technical solution of the present invention is described clearly and completely below. 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.

[0022] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the usual meanings understood by persons with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one" or "one" do not indicate quantity restrictions, but indicate the existence of at least one. Words such as "connect" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0023] See also Figure 1 The present invention provides a real-time verification platform for industrial algorithms based on cloud-edge-end collaboration, including:

[0024] The real-time data acquisition unit is used to collect the measured values ​​of industrial field sensors in real time and store them in the database of the edge server;

[0025] The privacy protection unit is used to encrypt the data stored on the platform using a two-way privacy protection mechanism and is also used to encrypt the verification model itself;

[0026] The model online training and verification unit is used to use the data in the edge server to perform online training and verification on the model uploaded by the user, and to associate the model with the corresponding data in the database by binding the input and output parameters of the model, and to train the model. After obtaining the model parameters, the model is verified online using real-time data to obtain the final model;

[0027] The algorithm evaluation unit performs real-time statistics and evaluation on the output results of the final model by setting performance indicators.

[0028] In this embodiment, the corresponding relationship with the corresponding data in the database is achieved by correlating and matching the variable names input by the model with the variable names in the database.

[0029] The above-mentioned industrial algorithm real-time verification platform based on cloud-edge-end collaboration closely combines key processes such as data collection, model training, and verification, significantly improving real-time performance and efficiency. Real-time data collection and processing of sensors in industrial scenarios, coupled with online training and verification of models, effectively promote rapid response to changes in the production environment, fully support the data processing and algorithm verification needs of industrial production, and provide a solid technical foundation for realizing intelligent and automated production.

[0030] Optionally, the above-mentioned platform also includes an edge monitoring unit, which is used to monitor the key parameters of the edge server and industrial field sensors in real time, run the model with the highest CPI score in the model online training and verification unit, and calculate the performance indicators of the equipment based on the key parameters and the model with the highest CPI score to perform health monitoring, soft measurement and optimization control processing.

[0031] In this optional implementation, the edge-end monitoring unit includes a comprehensive monitoring module and an enterprise-end Web display interface;

[0032] The integrated monitoring module is responsible for providing functions such as health monitoring, soft measurement and optimization control. It tracks the operating status of edge devices in real time, evaluates device health by analyzing device data, performs soft measurement analysis to obtain key performance indicators, and implements optimization control strategies based on these indicators to improve system efficiency.

[0033] The enterprise-side Web display interface provides users with an intuitive and easy-to-operate interface for displaying monitoring data, analysis results and control options, enabling users to intuitively manage and optimize the operation process of industrial systems.

[0034] Optionally, the above platform also includes a cloud visualization unit for displaying the evaluation results of the algorithm evaluation unit.

[0035] In this optional implementation, the cloud visualization unit consists of a user management module, a model management module, a model result display module, and a model performance display module. The user management module supports basic operations such as user registration, logout, login, and exit. The model management module provides the functions of uploading, downloading, configuring, and sending training and verification instructions for the model. The model result display module is responsible for obtaining the model operation results from the edge server in real time and displaying them in the form of charts through the Web interface. The model performance display module displays the model performance data in real time and ranks the user model by performance, which enhances the user experience and the interactivity of the platform.

[0036] Optionally, the privacy protection unit includes a first encryption unit for encrypting To encrypt the data stored on the platform, the following relationship is satisfied:

[0037] ;

[0038] In the formula, Indicates the encrypted data. Query function for the data to be protected. Represents data stored on the platform, represents noise, where the encrypted data as follows:

[0039] ;

[0040] In the formula, represents the privacy parameter, Indicates the sensitivity of the data query function to be protected, used to indicate adjacent data sets and The maximum query result difference between represents the Laplace distribution.

[0041] The privacy protection unit includes a second encryption unit for encrypting the data through a homomorphic encryption function The verification model itself is encrypted to satisfy the following relationship:

[0042] ;

[0043] in, Represents the encrypted model, vector For Model Parameters, where , homomorphic encryption function Encrypt each parameter independently:

[0044] ;

[0045] When performing homomorphic encryption calculations in the cloud, the encrypted model is applied to the encrypted data to obtain encrypted prediction results. :

[0046] ;

[0047] in, Is the input data Features, Represents homomorphic multiplication operation, the result is the predicted value in encrypted state, express.

[0048] Finally, the cloud can send the encrypted calculation results back to the edge server, and the edge server uses the corresponding private key to decrypt the results. :

[0049] ;

[0050] in, is with Decryption function for the pair.

[0051] In this way, by encrypting the calculation results, both the data security at the edge and the user's model security are protected.

[0052] Optionally, the model online training and verification unit includes a model configuration parameter loading module, a model input and output parameter binding module, a model training module and a model verification module;

[0053] The model configuration parameter loading module is used to load the model parameters set by the user to the edge server, build the model based on the model parameters, and load the model parameters. :

[0054]

[0055] Model input and output parameter binding module, used to bind the model input parameters and output parameters Bind to the corresponding data in the database:

[0056] ;

[0057] The model training module is used to start the training engine after parameter loading and data binding, perform model training on the cloud server, and obtain model parameters. , and send it down for storage to the edge server;

[0058] Model validation module for leveraging real-time data Real-time validation of the model:

[0059] ;

[0060] in, It’s an algorithm Real-time output results.

[0061] Optionally, the performance indicators include a three-level indicator evaluation system, which are the top-level indicators , Middle-level indicators and underlying indicators The top-level indicator is used to reflect the comprehensive performance of the entire model, which is calculated by summarizing the scores of all middle-level indicators; the middle-level indicators are divided into accuracy, efficiency and resource utilization, and the top-level indicator Satisfies the following relationship:

[0062] ;

[0063] In the formula, , , are the weights of the three middle-level indicators, , , The three middle-level indicators are scores, and the bottom-level indicators are the specific manifestations of each middle-level indicator. The specific manifestation is that the accuracy includes the precision rate. , recall rate and F1 score, efficiency includes response time and throughput , resource usage includes CPU usage and memory usage , the scores of the middle-level indicators are:

[0064] ;

[0065] ;

[0066] ;

[0067] In the formula, , , , , , and They are all weights of the underlying indicators.

[0068] Optionally, the above platform also includes an optimization suggestion generation module, which matches corresponding improvement suggestions according to the analysis results and the scores of the middle-level indicators, and finally generates suggestions for improving model performance to help users optimize their algorithms. In one example, When the score is lower than a certain value, the matching suggestion is "Model accuracy is too low, improve model accuracy". Similarly, suggestions for the other two indicators can be obtained.

[0069] See also Figure 2 , the above platform, the steps when using it are as follows:

[0070] S1: The user logs in to the platform website and registers or logs in;

[0071] S2: Upload the user model to the cloud and configure the model parameters;

[0072] S3: Bind the model input and output parameters with the data variables in the cloud database;

[0073] S4: Upload the training data required by the edge server to the cloud through the privacy protection unit;

[0074] S5: Perform model training, environment detection, and necessary environment configuration in the cloud;

[0075] S6: After passing through the privacy protection unit, the trained model is sent to the edge server for real-time verification and performance evaluation;

[0076] S7: The model verification results are transmitted back to the cloud, and the model operation results and user model performance rankings are displayed in real time through the cloud platform.

[0077] This application also provides a real-time verification method for industrial algorithms based on cloud-edge-end collaboration, including:

[0078] The real-time data acquisition unit is used to collect the measurement values ​​of industrial field sensors in real time and store them in the database of the edge server;

[0079] A two-way privacy protection mechanism is adopted through the privacy protection unit to encrypt the data stored on the platform and the verification model itself;

[0080] The model online training and verification unit is used to train and verify the model uploaded by the user online using the data in the edge server. The model is associated with the corresponding data in the database by binding the input and output parameters of the model to train the model. After obtaining the model parameters, the model is verified online using real-time data to obtain the final model.

[0081] The algorithm evaluation unit is used to set performance indicators to conduct real-time statistics and evaluation of the output results of the final model.

[0082] Optionally, the method further comprises:

[0083] An edge monitoring unit is used to monitor the key parameters of edge servers and industrial field sensors in real time, and health monitoring, soft measurement and optimization control processing are performed based on the key parameters.

[0084] The above-mentioned real-time verification method of industrial algorithms based on cloud-edge-end collaboration can implement the various embodiments of the above-mentioned real-time verification platform of industrial algorithms based on cloud-edge-end collaboration, and can achieve the same beneficial effects, which will not be repeated here.

[0085] The embodiment of the present application also provides an industrial algorithm real-time verification system based on cloud-edge-end collaboration, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above method when executing the computer program. The industrial algorithm real-time verification system based on cloud-edge-end collaboration can implement various embodiments of the above industrial algorithm real-time verification method based on cloud-edge-end collaboration, and can achieve the same beneficial effects, which will not be described here.

[0086] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A real-time verification platform for industrial algorithms based on cloud-edge-end collaboration, characterized in that: include: The real-time data acquisition unit is used to collect the measured values ​​of industrial field sensors in real time and store them in the database of the edge server; The privacy protection unit is used to encrypt the data stored on the platform using a two-way privacy protection mechanism, and is also used to encrypt the industrial algorithm model itself; The model online training and verification unit is used to associate the industrial algorithm model with the corresponding data in the database by binding the input and output parameters of the model, train the industrial algorithm model, obtain the parameters of the industrial algorithm model, and then use real-time data to perform online verification of the industrial algorithm model to obtain the final industrial algorithm model; The algorithm evaluation unit performs real-time statistics and evaluation on the output results of the final industrial algorithm model by setting performance indicators; Wherein, the privacy protection unit is located in the edge server, the model online training is implemented in the cloud server, the model verification is implemented in the edge server, and the algorithm evaluation unit is located in the edge server; It also includes an edge monitoring unit, which is used to monitor key parameters of edge servers and industrial field sensors in real time, run the industrial algorithm model with the highest CPI score in the model online training and verification unit, and calculate the performance indicators of the equipment based on the key parameters and the industrial algorithm model with the highest CPI score; The performance indicators include a three-level indicator evaluation system, namely, the top-level indicators , Middle-level indicators and underlying indicators The top-level indicator is used to reflect the comprehensive performance of the entire industrial algorithm model, which is calculated by summarizing the scores of all middle-level indicators; the middle-level indicators are divided into accuracy, efficiency and resource utilization, and the top-level indicator Satisfies the following relationship: ； In the formula, , , are the weights of the three middle-level indicators, , , The bottom-level indicators are the scores of the three middle-level indicators. The bottom-level indicators are the specific manifestations of each middle-level indicator. Among them, accuracy includes precision rate. , recall rate and score F1, efficiency includes response time and throughput , resource usage includes CPU usage and memory usage , the scores of the middle-level indicators are: ； ； ； In the formula, , , , , , and They are all weights of the underlying indicators.

2. The real-time verification platform for industrial algorithms based on cloud-edge-end collaboration according to claim 1 is characterized in that: A cloud-based visualization unit is also included for displaying evaluation results of the algorithm evaluation unit.

3. The real-time verification platform for industrial algorithms based on cloud-edge-end collaboration according to claim 1 is characterized in that: The privacy protection unit includes a first encryption unit for To encrypt the data stored on the platform, the following relationship is satisfied: ； In the formula, Indicates the encrypted data. Query function for the data to be protected. Represents data stored on the platform, represents noise, where the encrypted data as follows: ； In the formula, represents the privacy parameter, Indicates the sensitivity of the data query function to be protected, used to indicate adjacent data sets and The maximum query result difference between represents the Laplace distribution.

4. The real-time verification platform for industrial algorithms based on cloud-edge-end collaboration according to claim 1 is characterized in that: The privacy protection unit includes a second encryption unit for encrypting The verification industrial algorithm model itself is encrypted to satisfy the following relationship: ； in, Represents the encrypted industrial algorithm model, vector For Model Parameters, where , homomorphic encryption function Encrypt each parameter independently: ； When performing homomorphic encryption calculations in the cloud, the encrypted industrial algorithm model is applied to the encrypted data to obtain encrypted prediction results. : ； in, Is the input data Features, Represents a homomorphic multiplication operation, the result of which is the predicted value in an encrypted state.

5. The real-time verification platform for industrial algorithms based on cloud-edge-end collaboration according to claim 1 is characterized in that: The model online training and verification unit includes an industrial algorithm model configuration parameter loading module, an industrial algorithm model input and output parameter binding module, an industrial algorithm model training module and an industrial algorithm model verification module; The industrial algorithm model configuration parameter loading module is used to load the industrial algorithm model and parameters set by the user. Vector to edge server: In the formula, To load the function; The industrial algorithm model input and output parameter binding module is used to bind the input parameters of the industrial algorithm model and output parameters Bind to the corresponding data in the database: ； In the formula, Represents an industrial algorithm model; The industrial algorithm model training module is used to start the training engine after parameter loading and data binding, perform industrial algorithm model training on the cloud server, and obtain model parameters. , and send it down for storage to the edge server; The industrial algorithm model validation module is used to utilize real-time data Real-time verification of industrial algorithm models: ； in, It’s an algorithm Real-time output results.

6. A real-time verification method for industrial algorithms based on cloud-edge-end collaboration, characterized in that: include: The real-time data acquisition unit is used to collect the measurement values ​​of industrial field sensors in real time and store them in the database of the edge server; The privacy protection unit uses a two-way privacy protection mechanism to encrypt the data stored on the platform and the industrial algorithm model itself; The model online training and verification unit is used to perform online training and verification on the industrial algorithm model uploaded by the user using the data in the edge server. The industrial algorithm model is associated with the corresponding data in the database by binding the input and output parameters of the industrial algorithm model to train the industrial algorithm model. After obtaining the parameters of the industrial algorithm model, the industrial algorithm model is verified online using real-time data to obtain the final model. Adopting the algorithm evaluation unit, by setting performance indicators, the output results of the final industrial algorithm model are counted and evaluated in real time; Wherein, the privacy protection unit is located in the edge server, the model online training is implemented in the cloud server, the model verification is implemented in the edge server, and the algorithm evaluation unit is located in the edge server; Use edge monitoring units to monitor key parameters of edge servers and industrial field sensors in real time, run the industrial algorithm model with the highest CPI score in the model online training and verification unit, and calculate the performance indicators of the equipment based on the key parameters and the industrial algorithm model with the highest CPI score; The performance indicators include a three-level indicator evaluation system, namely, the top-level indicators , Middle-level indicators and underlying indicators The top-level indicator is used to reflect the comprehensive performance of the entire industrial algorithm model, which is calculated by summarizing the scores of all middle-level indicators; the middle-level indicators are divided into accuracy, efficiency and resource utilization, and the top-level indicator Satisfies the following relationship: ； In the formula, , , are the weights of the three middle-level indicators, , , The bottom-level indicators are the scores of the three middle-level indicators. The bottom-level indicators are the specific manifestations of each middle-level indicator. Among them, accuracy includes precision rate. , recall rate and score F1, efficiency includes response time and throughput , resource usage includes CPU usage and memory usage , the scores of the middle-level indicators are: ； ； ； In the formula, , , , , , and They are all weights of the underlying indicators.

7. The real-time verification method of industrial algorithms based on cloud-edge-end collaboration according to claim 6 is characterized in that: The method further comprises: An edge monitoring unit is used to monitor the key parameters of edge servers and industrial field sensors in real time, and health monitoring, soft measurement and optimization control processing are performed based on the key parameters.

8. A real-time verification system for industrial algorithms based on cloud-edge-end collaboration, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the method according to any one of claims 6 to 7.

Images