Edge computing device processing method and system for industrial internet
By introducing edge computing nodes into industrial IoT systems, data can be collected and analyzed in real time, edge prediction models can be built, and real-time decision-making and equipment status optimization can be achieved. This solves the problems of data transmission latency and bandwidth consumption, improves the real-time performance and security of the system, and promotes the intelligence and efficiency of the industrial internet.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-21
- Publication Date
- 2026-04-28
AI Technical Summary
Traditional industrial IoT systems suffer from data transmission latency and high network bandwidth consumption, leading to difficulties in operation and maintenance and limited system scalability.
By adding edge computing nodes between IoT devices and the cloud, industrial data is collected in real time, preprocessed and analyzed, and edge prediction models are built to achieve real-time decision-making and optimized scheduling, dynamically adjust equipment status, and optimize model parameters through multi-layer sensor algorithms to achieve closed-loop intelligent edge operation.
It reduces data transmission latency and network bandwidth consumption, improves system real-time performance and responsiveness, lowers operation and maintenance costs and network security risks, and promotes the intelligent and efficient development of the Industrial Internet.
Smart Images
Figure CN121934987A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial internet technology, and in particular to a processing method and system for edge computing devices used in the industrial internet. Background Technology
[0002] Traditional Industrial Internet of Things (IIoT) systems face challenges such as difficulties in connecting heterogeneous devices, network security risks, network isolation limitations, and high costs associated with cloud-edge collaboration, leading to operational difficulties and limited system scalability. To address these challenges, an edge computing device processing approach based on an IoT edge computing architecture offers a novel solution. This method adds edge computing nodes between IoT devices and the cloud, enabling local data processing and analysis, thereby reducing data transmission latency and network bandwidth consumption. Edge computing nodes can integrate various sensors and actuators to achieve real-time data acquisition and analysis from devices, while simultaneously supporting local decision-making and control.
[0003] Furthermore, edge computing nodes can reduce the amount of data to only contain key information through preprocessing and filtering, thereby alleviating the pressure on cloud servers. This edge computing-based processing method not only improves the real-time performance and responsiveness of the system, but also reduces operation and maintenance costs and cybersecurity risks, bringing new opportunities for the development of the Industrial Internet. Summary of the Invention
[0004] In view of the problems existing in the Industrial Internet of Things, this invention is proposed.
[0005] Therefore, the problem to be solved by this invention is how to reduce data transmission latency and network bandwidth consumption.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:
[0007] In a first aspect, embodiments of the present invention provide an edge computing device processing method for the Industrial Internet, comprising: collecting industrial data in real time through edge-side sensors and preprocessing it; constructing and training an edge prediction model based on the preprocessed data and a multilayer perceptron algorithm to achieve real-time analysis and prediction of industrial data; outputting control decisions and optimization scheduling in real time at the edge side according to the real-time analysis results, adjusting the equipment operating status, and outputting dynamic adjustment strategies; inputting the dynamic adjustment strategies into the edge prediction model for prediction, updating the edge prediction model parameters according to the prediction results, and sending them back to each edge node to achieve closed-loop intelligent edge operation processing.
[0008] As a preferred embodiment of the edge computing device processing method for the Industrial Internet described in this invention, the industrial data includes equipment operating status data, product processing data, and environmental monitoring data; the control decision includes the following steps: classifying the latency sensitivity of the control decision based on the complexity of the operating status; the specific formula for the complexity of the operating status is as follows:
[0009]
[0010] Where C represents the complexity of the running state, n represents the number of factors, and w i For the weights of each factor, l i The values for each factor.
[0011] When the latency sensitivity of control decisions is Level 1, execution is performed directly at the edge, reducing data transmission and processing time and improving response speed. When the latency sensitivity of control decisions is Level 2, in-depth analysis of industrial data is conducted, and the analysis data is transmitted to the cloud for processing. When the latency sensitivity of control decisions is Level 3, data is sent to the central server for analysis, and the analysis results are returned to local edge nodes for processing. Tasks that cannot be processed by local edge nodes are broken down into several sub-tasks, which are executed in a distributed manner through collaboration between edge nodes. The calculation results of each node are summarized to obtain the final analysis results. The results are then sent back to the corresponding edge nodes through cloud analysis to guide edge-side operations, output control decisions, and optimize scheduling.
[0012] As a preferred embodiment of the edge computing device processing method for the Industrial Internet described in this invention, the relevant formula for the control decision is as follows:
[0013] u * =argmin[σ(Y,Y * )+λΩ(u)],u∈U
[0014] subject_to g(x,u)≤0,h(x,u)=0
[0015] Among them, u * Let U be the optimal control decision vector, and σ(Y,Y) be the set of control strategies. * ) represents the difference between the predicted target and the actual output, λ is the regularization coefficient, Ω(u) is the penalty term for the constrained optimization problem, g(x,u) is the state variable inequality constraint function, and h(x,u) is the state variable equality constraint function.
[0016] The specific formula for the dynamic adjustment strategy is as follows:
[0017]
[0018] Where, x t+1 Let x be the state variable at time t+1. t Let be the state variable at time t. Let ξ be the optimal control decision at time t. t Let f be the external disturbance at time t, and f be the state transition function.
[0019] As a preferred embodiment of the edge computing device processing method for the Industrial Internet described in this invention, the edge prediction model includes the following steps: constructing an edge prediction model using a multilayer perceptron algorithm to meet real-time analysis and prediction needs; training the edge prediction model using a training set, and adjusting the weights and biases of the edge prediction model using a backpropagation algorithm; evaluating the edge prediction model using a test set, and optimizing the edge prediction model based on the evaluation results, adjusting the model's hyperparameters to improve model performance; the hyperparameters include the learning rate, the number of hidden layer nodes, and the activation function; deploying the trained edge prediction model to an edge computing device, acquiring new industrial data, and obtaining corresponding analysis results and predicted values through pre-inference and prediction.
[0020] As a preferred embodiment of the edge computing device processing method for the Industrial Internet described in this invention, the relevant formulas of the edge prediction model are as follows:
[0021]
[0022] Where Y is the edge prediction model, g is the activation function, and N is the edge prediction model. L Let W be the layer number of layer L. (i) Let a be the weight matrix of the i-th layer. (i-1) For the output of the (i-1)th layer, b (i-1) is the bias vector of the (i-1)th layer.
[0023] The specific formula for the prediction result is as follows:
[0024]
[0025] Among them, Y t+k The prediction result at time t+k is... κ(x) represents the transitive relationship between state variables and control decisions. t+k () is used to correct the prediction results.
[0026] As a preferred embodiment of the edge computing device processing method for the Industrial Internet described in this invention, the prediction results include: when the prediction error is greater than or equal to a threshold ε, the edge prediction model parameters are updated; if the prediction result is good, the edge prediction model is sent back to each edge node, and small batches of parameters are updated at regular intervals; if the prediction result is average, the edge prediction model is adjusted, and parameters are updated in batches; if the prediction result is poor, the edge prediction model is immediately reconstructed; if the prediction confidence is low, the edge prediction model is assigned a high priority; if it is a core edge node, the updated edge prediction model is prioritized for distribution to this node; if the network quality is good, the model is updated as a whole; if the network quality is poor, only key parameters are updated; when the prediction error is less than the threshold ε, the edge prediction model is directly sent back to each edge node, realizing closed-loop intelligent edge operation processing; the specific formula for the threshold is as follows:
[0027]
[0028] Where ε is the threshold, α is the weighting factor, and y r For the actual value, y p The predicted value is n, and the number of factors is n.
[0029] As a preferred embodiment of the edge computing device processing method for the Industrial Internet described in this invention, the first level of sensitivity is that the control system is highly sensitive to latency and requires decision-making and execution to be completed within milliseconds; the second level of sensitivity is that the control system is generally sensitive to latency and requires decision-making and execution time to be within a few milliseconds to tens of milliseconds; the third level of sensitivity is that the control system is low sensitive to latency and requires decision-making and execution time to be within tens of milliseconds to hundreds of milliseconds.
[0030] Secondly, embodiments of the present invention provide an edge computing device processing system for the Industrial Internet, comprising: a collection module for collecting industrial data in real time through edge-side sensors and performing preprocessing; a construction module for constructing and training an edge prediction model based on the preprocessed data and a multilayer perceptron algorithm to achieve real-time analysis and prediction of industrial data; an adjustment module for outputting control decisions and optimized scheduling in real time at the edge side based on the real-time analysis results, adjusting the equipment operating status, and outputting dynamic adjustment strategies; and a processing module for inputting the dynamic adjustment strategies into the edge prediction model for prediction, updating the edge prediction model parameters based on the prediction results, and sending them back to each edge node to achieve closed-loop intelligent edge operation processing.
[0031] Thirdly, embodiments of the present invention provide a computer device including a memory and a processor, wherein the memory stores a computer program, and the computer program instructions, when executed by the processor, implement the steps of the edge computing device processing method for the Industrial Internet as described in the first aspect of the present invention.
[0032] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the edge computing device processing method for the Industrial Internet as described in the first aspect of the present invention.
[0033] The beneficial effects of this invention are as follows: This invention achieves intelligent production processes by real-time acquisition, analysis, and prediction of industrial data, and real-time output of control decisions and optimized scheduling at the edge; real-time acquisition and preprocessing of data by edge sensors ensures data reliability and accuracy; the edge prediction model built based on the multilayer sensor algorithm enables real-time analysis and prediction of industrial data, identifying potential problems in advance; real-time output of control decisions and optimized scheduling adjusts equipment operating status, improving production efficiency and quality; simultaneously, closed-loop intelligent edge operation processing achieves production optimization and automated control, promoting intelligent, efficient, and sustainable development of the production process. Attached Figure Description
[0034] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:
[0035] Figure 1 This is a flowchart of the edge computing device processing method for the Industrial Internet in Example 1.
[0036] Figure 2 This is an internal structural diagram of the computer device used in the edge computing device processing method for the Industrial Internet in Example 1. Detailed Implementation
[0037] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0038] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0039] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.
[0040] Example 1
[0041] Reference Figures 1-2 This is the first embodiment of the present invention, which provides a processing method for an edge computing device in the Industrial Internet, including,
[0042] S1: Real-time acquisition and preprocessing of industrial data via edge-side sensors.
[0043] Specifically, industrial data includes equipment operating status data, product processing data, and environmental monitoring data; heterogeneous data acquisition involves deploying edge data acquisition gateways to support access from various industrial field devices, and real-time acquisition of production data from heterogeneous devices such as PLCs, DCSs, field instruments, and video surveillance, supporting the acquisition of structured data (numerical and status) and unstructured data (images and sound); the acquired raw data is format-converted to a standard format, and key features are extracted from the raw data to reduce data complexity.
[0044] Furthermore, some of the original data were manually labeled to create a training dataset. The labeling included anomaly types and skill parameter ratings. The preprocessed data was then divided into training, validation, and test sets to ensure that the datasets were representative and consistent. These three datasets will be used for subsequent model training and evaluation.
[0045] S2: Based on preprocessed data and multilayer perceptron algorithms, an edge prediction model is constructed and trained to achieve real-time analysis and prediction of industrial data.
[0046] Specifically, the Multilayer Perceptron (MLP) algorithm has multiple input layers, multiple hidden layers, and one output layer. Each hidden layer contains multiple neurons, each connected to all neurons in the previous layer, and each connection has a weight. MLPs are trained using backpropagation, adjusting the connection weights with known input and output data so that the model can learn and approximate the target function. After training, MLPs can be used for tasks such as classification and regression. Its advantages include the ability to model non-linear data, strong fitting ability, and applicability to various data types. However, MLPs also have some drawbacks, such as high sensitivity to data quality and quantization parameters, and the potential for overfitting during training.
[0047] Furthermore, to address the needs of real-time analysis and prediction, an edge prediction model is constructed using a multilayer perceptron algorithm. The edge prediction model is trained using a training set, and its weights and biases are adjusted using backpropagation. The model is then evaluated using a test set, and its hyperparameters are fine-tuned based on the evaluation results to improve performance. Hyperparameters include the learning rate, the number of hidden layer nodes, and the activation function. The trained edge prediction model is then deployed to an edge computing device to acquire new industrial data. Through pre-inference and prediction, corresponding analysis results and predicted values are obtained.
[0048] Furthermore, the relevant formulas for the edge prediction model are as follows:
[0049]
[0050] Where Y is the edge prediction model, g is the activation function, and N is the edge prediction model. L Let W be the layer number of layer L. (i) Let a be the weight matrix of the i-th layer. (i-1) For the output of the (i-1)th layer, b (i-1) is the bias vector of the (i-1)th layer.
[0051] Specifically, after receiving real-time data streams from field devices, the edge prediction model immediately analyzes and predicts this new data. Through feedforward computation, the model quickly outputs analysis results, such as operational status assessment, anomaly detection, and quality warnings. For results that meet normal expectations, no further action is required at the edge; however, if potential anomalies or performance deviations are detected, the model will trigger a real-time warning mechanism. Simultaneously, based on the type, location, and impact of the problems identified in the analysis, the intelligent decision engine will provide comprehensive evaluation and suggested response strategies, which may include alarm prompts, parameter adjustments, equipment switching, and other options. The decision plan will be directly sent to the control execution system, completing real-time adjustments to the production process within milliseconds, ensuring timely correction of problems and dynamic optimization of the production line. This closed-loop decision response is rapid and reliable, greatly improving production efficiency and operational quality. Furthermore, the edge will continuously feed back new patterns and rules discovered during the analysis to the cloud for knowledge base updates, constantly strengthening the model's intelligent analysis capabilities.
[0052] S3: Based on real-time analysis results, output control decisions and optimized scheduling at the edge side in real time, adjust the equipment operating status, and output dynamic adjustment strategies.
[0053] Specifically, the relevant formulas for control decisions are as follows:
[0054] u * =argmin[σ(Y,Y * )+λΩ(u)],u∈U
[0055] subject_to g(x,u)≤0,h(x,u)=0
[0056] Among them, u * Let U be the optimal control decision vector, and σ(Y,Y) be the set of control strategies. * ) represents the difference between the predicted target and the actual output, λ is the regularization coefficient, Ω(u) is the penalty term for the constrained optimization problem, g(x,u) is the state variable inequality constraint function, and h(x,u) is the state variable equality constraint function.
[0057] Furthermore, the latency sensitivity of control decisions is classified based on the complexity of the operating state; the specific formula for the complexity of the operating state is as follows:
[0058]
[0059] Where C represents the complexity of the running state, n represents the number of factors, and w i For the weights of each factor, l i The values for each factor.
[0060] Furthermore, when the latency sensitivity of control decisions is Level 1, execution is performed directly at the edge, reducing data transmission and processing time and improving response speed. When the latency sensitivity of control decisions is Level 2, in-depth analysis of industrial data is conducted, and the analyzed data is transmitted to the cloud for processing. When the latency sensitivity of control decisions is Level 3, data is sent to the central server for analysis, and the analysis results are returned to local edge nodes for processing. Tasks that cannot be processed by local edge nodes are broken down into several sub-tasks, which are executed in a distributed manner through collaboration between edge nodes. The calculation results of each node are summarized to obtain the final analysis results. The results are then sent back to the corresponding edge nodes through cloud analysis to guide edge-side operations, output control decisions, and optimize scheduling.
[0061] It should be noted that Level 1 sensitivity means the control system is highly sensitive to latency and requires decision-making and execution to be completed within milliseconds; Level 2 sensitivity means the control system is moderately sensitive to latency and requires decision-making and execution to be within a few to tens of milliseconds; Level 3 sensitivity means the control system is low sensitive to latency and requires decision-making and execution to be within a few to hundreds of milliseconds.
[0062] Furthermore, the specific formula for the dynamic adjustment strategy is as follows:
[0063]
[0064] Where, x t+1 Let x be the state variable at time t+1. tLet be the state variable at time t. Let ξ be the optimal control decision at time t. t Let f be the external disturbance at time t, and f be the state transition function.
[0065] S4: Input the dynamic adjustment strategy into the edge prediction model for prediction, update the edge prediction model parameters according to the prediction results, and send them back to each edge node to realize closed-loop intelligent edge operation and processing.
[0066] Specifically, the formula for the prediction result is as follows:
[0067]
[0068] Among them, Y t+k The prediction result at time t+k is... κ(x) represents the transitive relationship between state variables and control decisions. t+k () is used to correct the prediction results.
[0069] Furthermore, when the prediction error is greater than or equal to the threshold ε, the edge prediction model parameters are updated. If the prediction result is good, the edge prediction model is sent back to each edge node, and small batches of parameters are updated at regular intervals. If the prediction result is average, the edge prediction model is adjusted, and parameters are updated in batches. If the prediction result is poor, the edge prediction model is rebuilt immediately. If the prediction confidence is low, the edge prediction model is assigned a high priority. If it is a core edge node, the updated edge prediction model is sent to this node first. If the network quality is good, the entire model is updated. If the network quality is poor, only key parameters are updated. When the prediction error is less than the threshold ε, the edge prediction model is sent back to each edge node directly, realizing closed-loop intelligent edge operation and processing.
[0070] Furthermore, the specific formula for the threshold is as follows:
[0071]
[0072] Where ε is the threshold, α is the weighting factor, and y r For the actual value, y p The predicted value is n, and the number of factors is n.
[0073] Furthermore, this embodiment also provides an edge computing device processing system for the Industrial Internet, including: a collection module for collecting industrial data in real time through edge-side sensors and preprocessing it; a construction module for constructing and training an edge prediction model based on the preprocessed data and a multilayer perceptron algorithm to achieve real-time analysis and prediction of industrial data; an adjustment module for outputting control decisions and optimized scheduling in real time at the edge side based on the real-time analysis results, adjusting the equipment operating status, and outputting dynamic adjustment strategies; and a processing module for inputting the dynamic adjustment strategies into the edge prediction model for prediction, updating the edge prediction model parameters based on the prediction results, and sending them back to each edge node to achieve closed-loop intelligent edge operation processing.
[0074] This embodiment also provides a computer device applicable to the edge computing device processing method for the Industrial Internet, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the edge computing device processing method for the Industrial Internet as proposed in the above embodiment.
[0075] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.
[0076] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program performs the following steps: real-time acquisition of industrial data via edge-side sensors and preprocessing; construction and training of an edge prediction model based on the preprocessed data and a multilayer sensor algorithm to achieve real-time analysis and prediction of industrial data; outputting control decisions and optimized scheduling at the edge side in real time according to the real-time analysis results, adjusting the equipment operating status, and outputting dynamic adjustment strategies; inputting the dynamic adjustment strategies into the edge prediction model for prediction, updating the edge prediction model parameters according to the prediction results, and sending them back to each edge node to achieve closed-loop intelligent edge operation processing.
[0077] In summary, this invention achieves intelligent production processes by real-time acquisition, analysis, and prediction of industrial data, and real-time output of control decisions and optimized scheduling at the edge. Real-time acquisition and preprocessing of data through edge sensors ensures data reliability and accuracy. An edge prediction model based on a multilayer sensor algorithm enables real-time analysis and prediction of industrial data, identifying potential problems in advance. Real-time output of control decisions and optimized scheduling adjusts equipment operating states, improving production efficiency and quality. Simultaneously, closed-loop intelligent edge operation processing achieves production optimization and automated control, promoting intelligent, efficient, and sustainable development of the production process.
[0078] Example 2
[0079] Referring to Table 1, which is the second embodiment of the present invention, this embodiment provides a processing method for an edge computing device for the Industrial Internet. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.
[0080] Specifically, as shown in Table 1, data from four experiments are presented, including equipment status, product processing status, environmental monitoring results, and real-time analysis results. Regarding equipment status and product processing status, the data performance of this invention is similar to that of existing technologies; even under abnormal conditions, changes in equipment status and product processing status are accurately recorded. Similarly, in terms of environmental monitoring, the data performance of this invention is also similar to that of existing technologies, effectively recording changes in environmental parameters.
[0081] Table 1 Comparison between the present invention and prior art
[0082] Experiment number Equipment status Product Processing Environmental monitoring Real-time analysis results Existing technology analysis Analysis of this invention Experiment 1 85 92 78 Normal operation Abnormal equipment failure Normal operation Experiment 2 78 85 72 Abnormal equipment failure Abnormal equipment failure Abnormal equipment failure Experiment 3 90 88 80 Normal operation Normal operation Normal operation Experiment 4 82 75 70 Abnormal equipment failure normal Abnormal equipment failure
[0083] Furthermore, the real-time analysis results demonstrate the differences between the present invention and existing technologies in identifying anomalies during the production process. The present invention can more accurately identify anomalies, such as equipment malfunctions and production abnormalities, while existing technologies suffer from certain misjudgments or delays. In summary, through comparative analysis of experimental data, the advantages of the present invention in real-time analysis capabilities are clearly demonstrated, enabling more accurate and timely identification of anomalies in the production process, thereby improving the efficiency and quality of production.
[0084] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A processing method for an edge computing device used in the Industrial Internet, characterized in that: include, Industrial data is collected in real time and preprocessed using edge-side sensors; Based on preprocessed data and multilayer perceptron algorithms, an edge prediction model is constructed and trained to achieve real-time analysis and prediction of industrial data. Based on real-time analysis results, control decisions and optimized scheduling are output in real time at the edge, the operating status of equipment is adjusted, and dynamic adjustment strategies are output. The dynamic adjustment strategy is input into the edge prediction model for prediction. The parameters of the edge prediction model are updated based on the prediction results and sent back to each edge node to achieve closed-loop intelligent edge operation and processing.
2. The edge computing device processing method for the Industrial Internet as described in claim 1, characterized in that: The industrial data includes equipment operating status data, product processing data, and environmental monitoring data; the control decision-making process includes the following steps: Classify the latency sensitivity of control decisions based on the complexity of the operational state; The specific formula for the complexity of the operating state is as follows: Where C represents the complexity of the running state, n represents the number of factors, and w i For the weights of each factor, l i The values for each factor; When the latency sensitivity of control decisions is Level 1, they are executed directly at the edge, reducing data transmission and processing time and improving response speed. When the delay sensitivity of control decisions is level two, the industrial data is analyzed in depth and the analyzed data is transmitted to the cloud for processing. When the delay sensitivity of the control decision is level three, the data is sent to the central server for analysis, and the analysis results are returned to the local edge nodes for processing. Tasks that cannot be handled by local edge nodes are broken down into several sub-tasks, and the sub-tasks are executed in a distributed manner through cooperation between edge nodes. The calculation results of each node will be summarized to obtain the final analysis results; The results are analyzed in the cloud and sent back to the corresponding edge nodes to guide the operation on the edge side, output control decisions and optimize scheduling.
3. The edge computing device processing method for the Industrial Internet as described in claim 2, characterized in that: The relevant formulas for the control decision are as follows: you * =arg min[σ(Y,Y * )+λΩ(u)],u∈U subject_to g(x,u)≤0,h(x,u)=0 Among them, u * Let U be the optimal control decision vector, and σ(Y,Y) be the set of control strategies. * ) represents the difference between the predicted target and the actual output, λ is the regularization coefficient, Ω(u) is the penalty term for the constrained optimization problem, g(x,u) is the state variable inequality constraint function, and h(x,u) is the state variable equality constraint function. The specific formula for the dynamic adjustment strategy is as follows: Where, x t+1 Let x be the state variable at time t+1. t Let be the state variable at time t. Let ξ be the optimal control decision at time t. t Let f be the external disturbance at time t, and f be the state transition function.
4. The edge computing device processing method for the Industrial Internet as described in claim 1, characterized in that: The edge prediction model includes the following steps: To meet the needs of real-time analysis and prediction, an edge prediction model is constructed using a multilayer perceptron algorithm; The edge prediction model is trained using a training set, and the weights and biases of the edge prediction model are adjusted using the backpropagation algorithm. The edge prediction model is evaluated using a test set, and the model is then tuned based on the evaluation results. The hyperparameters of the model are adjusted to improve its performance. The hyperparameters include the learning rate, the number of hidden layer nodes, and the activation function; The trained edge prediction model is deployed to edge computing devices to acquire new industrial data. Through pre-inference and prediction, the corresponding analysis results and prediction values are obtained.
5. The edge computing device processing method for the Industrial Internet as described in claim 4, characterized in that: The relevant formulas for the edge prediction model are as follows: Where Y is the edge prediction model, g is the activation function, and N is the edge prediction model. L Let W be the layer number of layer L. (i) Let a be the weight matrix of the i-th layer. (i -1) For the output of the (i-1)th layer, b (i-1) This is the bias vector for the (i-1)th layer; The specific formula for the prediction result is as follows: Among them, Y t+k The prediction result at time t+k is... κ(x) represents the transitive relationship between state variables and control decisions. t+k () is used to correct the prediction results.
6. The edge computing device processing method for the Industrial Internet as described in claim 5, characterized in that: The prediction results include, When the prediction error is greater than or equal to the threshold ε, the edge prediction model parameters are updated. If the prediction result is good, the edge prediction model will be sent back to each edge node, and small batch parameters will be updated at regular intervals. If the prediction result is average, adjust the edge prediction model and update the parameters in batches. If the prediction result is poor, the edge prediction model should be rebuilt immediately. If the prediction confidence is low, assign a high priority to the edge prediction model; If it is a core edge node, the updated edge prediction model will be distributed to this node first. If the network quality is good, then update the model as a whole; If the network quality is poor, only update the key parameters; When the prediction error is less than the threshold ε, the edge prediction model is directly sent back to each edge node to achieve closed-loop intelligent edge operation and processing. The specific formula for the threshold is as follows: Where ε is the threshold, α is the weighting factor, and y r For the actual value, y p The predicted value is n, and the number of factors is n.
7. The edge computing device processing method for the Industrial Internet as described in claim 2, characterized in that: The first level of sensitivity refers to the control system's high sensitivity to delays, requiring decisions to be made and executed within milliseconds. The second-level sensitivity refers to the control system being generally sensitive to delay, requiring the decision-making and execution time range to be from a few milliseconds to tens of milliseconds; the third-level sensitivity refers to the control system being less sensitive to delay, requiring the decision-making and execution time range to be from tens of milliseconds to hundreds of milliseconds.
8. An edge computing device processing system for the Industrial Internet, based on the edge computing device processing method for the Industrial Internet according to any one of claims 1 to 7, characterized in that: include, The data collection module is used to collect industrial data in real time via edge-side sensors and perform preprocessing. The module builds and trains an edge prediction model based on preprocessed data and a multilayer perceptron algorithm to enable real-time analysis and prediction of industrial data. The adjustment module is used to output control decisions and optimized scheduling in real time at the edge side based on real-time analysis results, adjust the equipment operating status, and output dynamic adjustment strategies. The processing module is used to input the dynamic adjustment strategy into the edge prediction model for prediction, update the edge prediction model parameters according to the prediction results, and send them back to each edge node to realize closed-loop intelligent edge operation processing.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the edge computing device processing method for the Industrial Internet as described in any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the edge computing device processing method for the Industrial Internet as described in any one of claims 1 to 7.