Industrial automation control method and system

Abstract

The present invention discloses an industrial automation control method and system, which relates to the field of data processing technology, including collecting production equipment operation data, building a digital twin model and forming a virtual production network; obtaining the resource status data of each factory in the virtual production network in real time, dynamically generating the optimal cross-factory scheduling plan, simulating the scheduling plan in the virtual production network for simulation verification and adjusting the plan; deploying the scheduling plan to the physical factory, controlling the equipment to perform tasks through the automatic control system, and summarizing the production data and simulation results for storage and backup. The present invention significantly improves the accuracy and efficiency of industrial automation control, can process equipment operation data more efficiently, accurately simulate the dynamic behavior of equipment, and generate the optimal scheduling plan across factories in real time, and form a closed-loop optimization mechanism through simulation verification and feedback adjustment, so as to realize the dynamic adaptability and global optimality of the scheduling plan, thereby greatly improving the resource utilization efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an industrial automation control method and system. Background Art

[0002] With the widespread application of the concept of Industry 4.0, industrial automation control technology has developed rapidly. Modern industrial production is moving towards a high degree of digitalization, intelligence and integration, and advanced technologies including industrial Internet of Things (IoT), digital twin technology and artificial intelligence (AI) have emerged. However, the application of these technologies in actual industrial scenarios still faces many challenges. For example, existing industrial automation control methods usually lack the ability to refine the processing and dynamic modeling of equipment status data, resulting in insufficient adaptability and efficiency of scheduling schemes in complex production environments. Although some technologies have achieved basic perception of equipment status through IoT and CAD modeling, there is still much room for improvement in the real-time and global optimization capabilities of task allocation between devices. In addition, due to the diversity and dynamic nature of physical equipment operation data, traditional methods are difficult to efficiently process and utilize these data, resulting in an imbalance in resource scheduling and task allocation, limiting the further improvement of production efficiency. Therefore, how to fully combine IoT, digital twins and intelligent optimization technologies to build a more efficient and dynamic industrial automation control system has become a core issue in the current technological development. Summary of the invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, the present invention provides an industrial automation control method to solve how to fully combine IoT, digital twins and intelligent optimization technologies to build a more efficient and dynamic industrial automation control system, which has become a core issue in current technological development.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides an industrial automation control method, which comprises:

[0007] Collect production equipment operation data, build a digital twin model and form a virtual production network;

[0008] Obtain resource status data of each factory in the virtual production network in real time, dynamically generate the optimal cross-factory scheduling plan, simulate the scheduling plan in the virtual production network for simulation verification and adjust the plan;

[0009] Deploy the scheduling plan to the physical factory, control the equipment to perform tasks through the automated control system, and summarize the production data and simulation results for storage and backup.

[0010] As a preferred solution of the industrial automation control method described in the present invention, wherein: the collection of production equipment operation data, construction of a digital twin model and formation of a virtual production network refers to deploying loT sensors on each production equipment, collecting the temperature, pressure and vibration data of the equipment in real time, classifying the data by equipment number, generating a unique ID for each equipment, normalizing the data after removing outliers using the triple standard deviation method, uploading the pre-processed data to the central data platform using the MQTT industrial protocol, and storing them as a data flow table classified by equipment number, where each record contains the equipment number, data category and timestamp;

[0011] Use CAD software to build the equipment geometry model and clarify the key components and connection relationships of the equipment;

[0012] Extract the temperature, pressure and vibration data of the equipment from the data stream table as variables to dynamically drive the state change of the geometric model;

[0013] Use finite element analysis to simulate the heat conduction path during equipment operation, mark high temperature areas and generate distribution maps, and use state space models to describe the dynamic behavior of equipment;

[0014] Based on dynamic status mapping, resource status indicators such as equipment load rate, resource utilization rate and order priority are extracted;

[0015] The equipment load rate is calculated based on the ratio of the current task volume to the total capacity of the equipment. The resource utilization rate is calculated by counting the ratio of the task usage time of each equipment to the total available time. The order priority is calculated based on the importance of the order in the historical scheduling table.

[0016] Construct a virtual production network, take the geometric model of each factory as a node and construct a node set H, construct the task allocation relationship between devices as an edge set O, the node contains real-time resource status and historical operation data, define the task allocation relationship in the virtual production network, and form an interaction graph G between nodes;

[0017] Establish a weight relationship for task allocation and describe the task priorities between devices;

[0018] Use the Industrial Internet Protocol to synchronize data between the physical factory and the virtual production network in both directions.

[0019] As a preferred solution of the industrial automation control method of the present invention, the resource status data of each factory in the virtual production network is obtained in real time, and the optimal cross-factory scheduling scheme is dynamically generated to obtain the resource status data of each factory from the virtual production network at specified intervals, and the data groups of each factory are combined according to the equipment number ID to obtain the data set corresponding to each factory, and each data set is cleaned and the integrity is tested, and the missing data is supplemented by the time series interpolation method, and then the data is standardized;

[0020] The resource status data includes equipment load rate, resource utilization rate and order priority;

[0021] Convert the preprocessed numerical data of load rate and resource utilization directly into cloud model representation ;

[0022] Transform order priority into cloud model representation through fuzzy language model ;

[0023] Integrate all cloud model data to form a resource status matrix S, where each row represents a task and each column represents the resource status;

[0024] Convert the resource state matrix into the initial state of the DDPG network and build a dual-objective Critic network and , initialize the Actor network;

[0025] Collect historical task scheduling records, combine them with the current state matrix S, generate scheduling samples, and build empirical data based on the scheduling samples;

[0026] The experience data includes the state ,action and reward value ;

[0027] According to the reward value Perform recursive experience pool grouping, set thresholds z and x, and z>x, if the reward value is greater than the threshold z, it is a high priority group. If the reward value Less than or equal to the threshold z and greater than the threshold x, it is a medium priority group. If the reward value If it is less than the threshold x, it is a low priority group;

[0028] Dynamically adjust the experience sampling probability according to the TD error, give priority to extracting samples from the high priority group for training, and construct the sample data contained in different training batches according to the priority of the experience pool group;

[0029] According to the sample data, use the dual-objective Critic network and Calculate the target value;

[0030] Define the loss function of the Critic network as the mean error between the target value and the current value, and use the Adam optimizer to update the parameters of the Critic network;

[0031] Use the Actor network to generate the optimal scheduling action based on the current state;

[0032] The goal of optimizing the Actor network is to maximize the Q value given by the Critic network;

[0033] Update the Actor network parameters through gradient descent and use the soft update strategy to update the target network parameters;

[0034] Use the critic network weights and task results to optimize the comprehensive reward function;

[0035] Reward evaluation for actions output by the Actor network;

[0036] Set a threshold l. If the reward value is lower than the threshold l, return to the Actor network for retraining and adjust the scheduling strategy.

[0037] Integrate multiple groups of scheduling plans output by the Actor network. Each group of plans contains performance indicators. The performance indicators are standardized into dimensionless values. A hierarchy analysis method is used to assign weights to each indicator. The weighted performance indicator values ​​are obtained according to the assigned weights and a weighted matrix is ​​generated.

[0038] Define the positive ideal solution as the optimal value of each indicator, and define the negative ideal solution as the worst value of each indicator;

[0039] Calculate the distance between the positive and negative ideal solutions, and calculate the closeness coefficient based on the distance between the ideal solutions;

[0040] The scheduling plans are sorted in descending order according to the proximity coefficient, the scheduling plan with the largest proximity coefficient is selected as the optimal plan, and a scheduling list is generated according to the optimal plan;

[0041] The scheduling list includes task numbers, factory numbers, equipment numbers and time schedules.

[0042] As a preferred solution of the industrial automation control method of the present invention, wherein: the simulation of the scheduling scheme in the virtual production network for simulation verification refers to importing the generated scheduling list into the virtual production network, simulating the execution process of the scheduling scheme in the virtual production network, dynamically updating the equipment operation status, task execution progress and resource consumption of each task and recording them, and calculating the total simulation delay D according to the task planning time and the actual simulation completion time;

[0043] The real-time load rate of the device is calculated based on the ratio of the task occupancy time and the total available time of the device in the simulation, and a threshold P is set. If the device load rate is greater than the threshold P, it is a bottleneck device.

[0044] Extract the number, actual completion time and delay time of the delayed task in the simulation process and associate them with the equipment number, summarize the bottleneck equipment and delayed task information, and form a feedback data table;

[0045] The feedback data table includes task number, device number, delay time and device load rate.

[0046] As a preferred solution of the industrial automation control method of the present invention, wherein: the scheme adjustment refers to traversing all non-bottleneck devices according to the bottleneck device information recorded in the feedback data, calculating the real-time load rate, selecting the device with the lowest load rate as the substitute device, and preferentially allocating the tasks of the bottleneck device to the substitute device;

[0047] Extract the number and delay time of the delayed task, adjust the task priority incrementally according to the task delay time, reorder the tasks from high to low according to the new priority, and update the task execution order.

[0048] As a preferred solution of the industrial automation control method described in the present invention, wherein: the deployment of the scheduling plan to the physical factory and the control of the equipment to perform tasks through the automated control system refers to delivering the optimized scheduling plan to the automated control system of the physical factory in a standardized format, the control system parses the scheduling plan, initializes the equipment required for the task, checks the resources required for the task, and according to the scheduling plan, the control system gradually starts the equipment according to the task priority, sets the task execution parameters to control the equipment to perform the task.

[0049] As a preferred solution of the industrial automation control method described in the present invention, the storage and backup of the summarized production data and simulation results refers to collecting the production data and real-time simulation results after executing the task, sorting them according to timestamps and storing them in a database, synchronously backing up the stored data regularly, regularly checking the security and integrity of the stored data and backup data, and generating a test report.

[0050] In a second aspect, the present invention provides an industrial automation control system, comprising:

[0051] Data acquisition module, used to collect production equipment data in real time and perform preprocessing;

[0052] The construction module is used to build a digital twin model based on equipment operation data and interconnect the digital twin models of factory equipment to form a virtual production network;

[0053] The scheduling optimization module is used to regularly collect factory resource status data and generate the optimal scheduling plan using the DDPG network;

[0054] The adjustment module is used to simulate the execution process of the scheduling scheme in the virtual production network and adjust the scheduling strategy according to the simulation feedback;

[0055] A deployment module, used to deploy the optimized scheduling scheme to the automation control system of the physical plant;

[0056] The data storage module is used to summarize production data and simulation results, store them in the database and back them up regularly.

[0057] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the industrial automation control method as described in the first aspect of the present invention is implemented.

[0058] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the industrial automation control method as described in the first aspect of the present invention is implemented.

[0059] The beneficial effects of the present invention are as follows: the present invention significantly improves the accuracy and efficiency of industrial automation control by introducing IoT data-driven digital twin modeling technology, dynamic interactive analysis of virtual production networks, and scheduling optimization mechanism based on deep reinforcement learning. It can process equipment operation data more efficiently, accurately simulate equipment dynamic behavior, and generate optimal scheduling plans across factories in real time. At the same time, a closed-loop optimization mechanism is formed through simulation verification and feedback adjustment, which realizes the dynamic adaptability and global optimality of the scheduling plan, thereby greatly improving resource utilization efficiency, reducing task delays, and enhancing the flexibility and reliability of the production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0061] Figure 1 This is a flow chart of the industrial automation control method in Example 1.

[0062] Figure 2 This is a structural diagram of the industrial automation control system in Example 1. DETAILED DESCRIPTION

[0063] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0064] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0065] 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 term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0066] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides an industrial automation control method, comprising the following steps:

[0067] S1. Collect production equipment operation data, build a digital twin model and form a virtual production network;

[0068] Specifically, collecting the operation data of production equipment, building a digital twin model and forming a virtual production network means deploying loT sensors on each production equipment to collect the temperature, pressure and vibration data of the equipment in real time. By deploying IoT sensors on production equipment, key parameters (temperature, pressure, vibration) can be collected in real time to achieve all-round monitoring of the equipment operation status. Combined with data number classification and triple standard deviation method to clean up outliers, the quality of data can be effectively improved, providing a reliable foundation for subsequent modeling and analysis. Data is classified by equipment number, and a unique ID is generated for each device. The data is normalized after removing outliers using the triple standard deviation method. The pre-processed data is uploaded to the central data platform using the MQTT industrial protocol, and stored as a data flow table classified by equipment number. Each record contains the equipment number, data category and timestamp.

[0069] Use CAD software to build a geometric model of the equipment, clarify the key components of the equipment (such as bearings, gears and sleeves) and the connection relationship (mark the physical assembly method and motion constraints between the components). The geometric model can accurately depict the key components and connection relationships of the equipment, and provide a physical basis for mapping and simulating the dynamic state of the equipment. In particular, with the support of finite element analysis, it can identify the thermal conduction path and high stress area during equipment operation, helping to predict the risk of equipment failure;

[0070] Extract the temperature, pressure and vibration data of the equipment from the data stream table as variables to dynamically drive the state change of the geometric model;

[0071] Finite element analysis is used to simulate the heat conduction path of the equipment during operation, mark the high-temperature area and generate a distribution map, and the state space model is used to describe the dynamic behavior of the equipment. Finite element analysis can capture the complex physical characteristics of the equipment, such as heat and stress distribution, and the state space model converts the dynamic behavior of the equipment into a quantifiable mathematical expression. This combination ensures the accuracy of the physical model and improves the real-time performance of dynamic behavior modeling.

[0072] Based on dynamic status mapping, resource status indicators such as equipment load rate, resource utilization rate and order priority are extracted;

[0073] The equipment load rate is calculated based on the ratio of the current task volume to the total equipment capacity. The resource utilization rate is calculated by counting the ratio of the task usage time of each device to the total available time. The order priority is calculated based on the importance ranking of the orders in the historical scheduling table. By calculating the equipment load rate and resource utilization rate, the imbalance of resource usage can be identified. The order priority helps optimize the task execution sequence and reduce task delays. These indicators are combined to construct a resource status matrix to provide a quantitative basis for scheduling optimization.

[0074] Construct a virtual production network, take the geometric model of each factory as a node and construct a node set H, construct the task allocation relationship between devices as an edge set O, and the node contains real-time resource status and historical operation data. The virtual production network integrates the digital twin models of all factories and constructs a node interaction graph based on task allocation. The priority relationship between devices is described by the task allocation weight, which supports the global optimization and resource sharing of complex production tasks. The task allocation relationship is defined in the virtual production network to form an interaction graph G between nodes:

[0075] ,

[0076] In the formula, H is the node set and O is the edge set;

[0077] Establish the weight relationship of task allocation and describe the task priority between devices:

[0078] ,

[0079] In the formula, is the weight of task allocation from device h to device u, is the amount of resource transfer, The task allocation weight is optimized by balancing the resource transmission amount and transmission time, accurately describing the task allocation priority between devices, and helping to generate a more efficient task scheduling solution;

[0080] The Industrial Internet Protocol is used for two-way synchronization of data between physical factories and virtual production networks. The Industrial Internet Protocol supports real-time synchronization of data between physical factories and virtual production networks, ensuring that the virtual environment can accurately reflect the actual equipment status, and at the same time efficiently transmit the optimized scheduling plan to the physical factory for execution.

[0081] Through the deep integration of IoT technology, digital twin modeling, virtual production network and industrial Internet protocol, the intelligent and dynamic optimization capabilities of industrial automation control have been significantly improved. In terms of data collection and processing, the real-time and reliability of equipment operation data are ensured. In terms of digital twin modeling, accurate simulation of equipment status is achieved through geometric modeling, dynamic state mapping and finite element analysis. In the virtual production network, the task relationship interaction diagram of nodes and edges is constructed to support the optimization solution of complex scheduling problems. Finally, the two-way synchronization of physical factories and virtual production networks is achieved through the industrial Internet protocol, forming a closed-loop optimization process.

[0082] Furthermore, the resource status data of each factory in the virtual production network is obtained in real time, and the optimal cross-factory scheduling plan is dynamically generated. The resource status data of each factory is obtained from the virtual production network at specified intervals, and the data of each factory is grouped according to the equipment ID to obtain the data set corresponding to each factory. The data set is cleaned and the integrity is tested. The missing data is supplemented by the time series interpolation method, and the data is standardized;

[0083] The resource status data includes equipment load rate, resource utilization rate and order priority;

[0084] Convert the preprocessed numerical data of load rate and resource utilization directly into cloud model representation :

[0085] ,

[0086] In the formula, b is the current value of the numeric data, is the uncertainty coefficient, which indicates the possible error range of numerical data and is obtained based on statistical analysis. It is the volatility coefficient, which indicates the dynamic fluctuations that may occur in numerical data over time and is determined by the statistical results of the fluctuation range in historical data;

[0087] Transform order priority into cloud model representation through fuzzy language model :

[0088] ,

[0089] In the formula, is the expected value, indicating the core value of the priority. is the entropy value, indicating the uncertainty of the priority, is the super entropy value, which indicates the dynamic volatility of priority. is the task delay time, and are the influence coefficients, which represent the impact of delay on entropy and volatility, respectively, and are set through historical statistical analysis;

[0090] Integrate all cloud model data to form a resource status matrix S, where each row represents a task and each column represents the resource status (such as load rate, utilization rate, and priority);

[0091] The cloud model can quantify the data volatility and error range in complex production scenarios, overcoming the lack of adaptability of traditional linear models to dynamic data. The cloud model represents order priority, combined with historical delays and dynamic fluctuations, making the scheduling strategy more responsive to changes.

[0092] Convert the resource state matrix into the initial state of the DDPG network and build a dual-objective Critic network and , initialize the Actor network;

[0093] The resource state matrix provides a comprehensive and dynamic environment representation for the DDPG network input. The combination with reinforcement learning effectively improves the stability and applicability of the scheduling strategy. The dual-objective critic network optimizes order delay and equipment energy consumption at the same time, significantly improving the overall performance of the scheduling solution.

[0094] Collect historical task scheduling records, combine them with the current state matrix S, generate scheduling samples, and build empirical data based on the scheduling samples;

[0095] The experience data includes the state ,action and reward value ;

[0096] The status Refers to the rows of the resource status matrix, actions Refers to the equipment allocation, time arrangement, and reward value in historical data Refers to the actual feedback value of historical scheduling;

[0097] According to the reward value Recursive experience pool grouping is performed, and thresholds z and x are set by statistical analysis of historical state data, and z>x. If the reward value is greater than the threshold z, it is a high priority group. If the reward value Less than or equal to the threshold z and greater than the threshold x, it is a medium priority group. If the reward value If it is less than the threshold x, it is a low priority group;

[0098] Dynamically adjust the experience sampling probability according to the TD error, give priority to extracting samples from the high priority group for training, and construct the sample data contained in different training batches according to the priority of the experience pool group;

[0099] The priority experience replay mechanism accelerates model convergence and enhances scheduling accuracy through dynamic grouping and sample priority sorting. It combines the resource state matrix input network with experience replay to dynamically optimize the scheduling strategy and achieve high-precision scheduling in a rapidly changing production environment.

[0100] According to the sample data, use the dual-objective Critic network and Calculate the target value:

[0101] ,

[0102] In the formula, is the target value of the Critic network, indicating the current state and actions Expected return, is the reward value, is the discount factor, which indicates the importance weight of future returns, and is the output of the dual-objective Critic network, estimating the value of the current state and action respectively. is the next state, It is the next action;

[0103] Define the loss function of the Critic network as the mean error between the target value and the current value, and use the Adam optimizer to update the parameters of the Critic network;

[0104] Use the Actor network to generate the optimal scheduling action based on the current state:

[0105] ,

[0106] In the formula, It is the action generated by the Actor network, including task number, device allocation, and execution time. is the strategy function of the Actor network, N(t) is the exploration noise used to expand the search space, and in this embodiment it is Gaussian noise;

[0107] Through nonlinear mapping and exploration noise, the Actor network generates accurate task numbers, device allocations, and execution times, avoiding the regularization limitations of traditional methods. Exploration noise (such as Gaussian noise) expands the search space and increases the possibility of scheduling strategy innovation.

[0108] The strategy function is expressed as: ,in, is the parameter set of the Actor network, It is a nonlinear mapping implemented by a neural network, outputting action values;

[0109] The goal of optimizing the Actor network is to maximize the Q value given by the Critic network:

[0110] ,

[0111] In the formula, It’s a strategy The objective function is to use the strategy in all sampled states. The comprehensive value that can be obtained, N is the number of samples in the experience pool, is the importance weight of the i-th sample, which is used to highlight the key samples and is determined by prioritizing experience playback. Is the current state Next, according to the strategy Selected Action The cumulative expected reward of is the regularization coefficient, which is determined by experimental tuning. is a regularization term used to constrain the complexity of the policy function, usually the L2 norm of the Actor network parameters;

[0112] The goal is to maximize the Q value of the critic network to ensure that the scheduling action still has a high comprehensive value in complex scenarios;

[0113] Update the Actor network parameters through gradient descent and use the soft update strategy to update the target network parameters;

[0114] Use the Critic network weights and task results to optimize the comprehensive reward function:

[0115] ,

[0116] In the formula, is the reward value, is the order delay cost, is the energy consumption of the equipment, is the equipment load factor, , and is the weight coefficient, which is set by particle swarm optimization method, is the load rate mapping scale factor, which is set based on experimental tuning;

[0117] Combined with multiple objectives such as order delay, equipment energy consumption and load rate, the weight coefficient is flexibly tuned to meet the optimization needs of different production scenarios. The weight coefficient is set through particle swarm optimization to adapt to different task priorities and resource status changes, improve the versatility of the scheduling strategy, and the optimized reward function provides a better performance balance in a multi-task environment, reducing resource waste while improving task completion efficiency.

[0118] Reward evaluation for actions output by the Actor network:

[0119] ,

[0120] In the formula, A represents the current state Take action The cumulative rewards you can get, is the reward value, is the discount factor, is in the next state Take action The cumulative rewards that can be obtained;

[0121] The threshold l is set by clustering the historical scheduling data. If the reward value is lower than the threshold l, the Actor network is returned to retrain and the scheduling strategy is adjusted.

[0122] Integrate multiple groups of scheduling plans output by the Actor network. Each group of plans contains performance indicators (such as order delay cost, equipment energy consumption, and equipment load rate). Standardize the performance indicators into dimensionless values. Use the hierarchical analysis method to assign weights to each indicator. According to the assigned weights, the weighted performance indicator values ​​are obtained and a weighted matrix is ​​generated.

[0123] Define the positive ideal solution as the optimal value of each indicator, and define the negative ideal solution as the worst value of each indicator. Define positive and negative ideal solutions, combine standardized performance indicators, and quickly select the optimal scheduling solution to avoid suboptimal problems in resource allocation;

[0124] Calculate the distance between the positive and negative ideal solutions, calculate the closeness coefficient based on the distance of the ideal solution, and scientifically evaluate multiple groups of scheduling plans through hierarchical analysis method and closeness coefficient calculation to reduce interference from human subjective factors;

[0125] The scheduling plans are sorted in descending order according to the proximity coefficient, the scheduling plan with the largest proximity coefficient is selected as the optimal plan, and a scheduling list is generated according to the optimal plan;

[0126] The scheduling list includes task numbers, factory numbers, equipment numbers and time schedules.

[0127] By constructing a resource state matrix based on a cloud model, it is possible to quantify data fluctuations and uncertainties in complex environments, providing high-quality input for subsequent scheduling optimization. At the same time, the reinforcement learning framework combining the dual-objective Critic network and the Actor network achieves comprehensive optimization of order delays and equipment energy consumption, significantly improving the scientificity and accuracy of the scheduling strategy. In the generation and optimization stage of scheduling schemes, the present invention organically integrates multi-dimensional indicators through the analytic hierarchy process and the calculation of the proximity coefficient, providing a solid guarantee for the global optimization of the final scheduling scheme. The closed-loop optimization mechanism of simulation verification and feedback adjustment enables the scheduling strategy to continuously adapt to the dynamic changes of the production environment and ensure the consistency of the actual execution effect with the planning goals. The final generated scheduling list is directly transmitted to the automation control system, realizing an end-to-end industrial automation control process.

[0128] S2. Obtain resource status data of each factory in the virtual production network in real time, dynamically generate the optimal cross-factory scheduling plan, simulate the scheduling plan in the virtual production network for simulation verification and adjust the plan;

[0129] Specifically, simulating the scheduling scheme in the virtual production network for simulation verification means importing the generated scheduling list into the virtual production network, simulating the execution process of the scheduling scheme in the virtual production network, dynamically updating the equipment operation status, task execution progress and resource consumption of each task and recording them, and calculating the total simulation delay D according to the task planning time and the actual simulation completion time:

[0130] ,

[0131] Where m is the total number of tasks, is the actual completion time of task j, is the planned completion time of task j;

[0132] The real-time load rate of the device is calculated based on the ratio of the task occupancy time and the total available time of the device in the simulation. The threshold P is set based on industry experience. If the device load rate is greater than the threshold P, it is a bottleneck device.

[0133] Extract the number, actual completion time and delay time of the delayed task in the simulation process and associate them with the equipment number, summarize the bottleneck equipment and delayed task information, and form a feedback data table;

[0134] The feedback data table includes task number, device number, delay time and device load rate.

[0135] The simulation verification method based on the virtual production network comprehensively reflects the operation performance of the scheduling scheme by dynamically updating the task status and equipment load rate. By calculating the total simulation delay and identifying the bottleneck equipment, the scheduling scheme can be quantitatively evaluated and key problems can be located. The generated feedback data table provides data support for the improvement of the scheduling strategy, forming a closed-loop optimization mechanism, which significantly improves the utilization efficiency of production resources and the accuracy of task scheduling. This method can not only reduce the task delay rate, but also effectively reduce the risk of equipment downtime caused by overload, thereby greatly improving the flexibility and reliability of industrial production.

[0136] Furthermore, adjusting the plan means traversing all non-bottleneck devices according to the bottleneck device information recorded in the feedback data, calculating the real-time load rate, selecting the device with the lowest load rate as the substitute device, and allocating the tasks of the bottleneck device to the substitute device in priority;

[0137] Extract the number and delay time of the delayed task, adjust the task priority incrementally according to the task delay time, reorder the tasks from high to low according to the new priority, and update the task execution order.

[0138] Through the scheme adjustment method of the present invention, a dynamic closed-loop optimization mechanism is formed in the execution of production tasks. The tasks of bottleneck equipment are preferentially allocated to low-load equipment, which significantly alleviates the problem of uneven resource allocation and reduces task delays. At the same time, the task sorting is optimized in combination with priority adjustment, which improves the scheduling efficiency and the timeliness of order delivery. The calculation of real-time load rate provides a scientific basis for global task optimization, comprehensively improves the flexibility, reliability and production efficiency of the system, can meet the needs of complex production environments in modern industries, and provides strong technical support for scenarios of multi-device collaboration and resource sharing.

[0139] S3, deploy the scheduling plan to the physical factory, control the equipment to perform tasks through the automated control system, and summarize the production data and simulation results for storage and backup;

[0140] Specifically, deploying the scheduling plan to the physical factory and controlling the equipment to perform tasks through the automated control system means delivering the optimized scheduling plan in a standardized format (XML format or JSON format) to the automated control system (including PLC and SCADA systems) of the physical factory. The control system parses the scheduling plan, initializes the equipment required for the task, and checks the resources required for the task (such as energy consumption and materials). According to the scheduling plan, the control system gradually starts the equipment according to the task priority, sets the task execution parameters (such as speed, pressure, power, etc.) to control the equipment to perform the task.

[0141] The optimized scheduling plan is transmitted to the automation control system of the physical factory in XML or JSON format. This standardized format ensures the consistency and portability of data, facilitates the compatibility and parsing of different types of control systems, ensures the reliability of scheduling data transmission, avoids information loss or misunderstanding in format conversion, adapts to multi-system interoperability in complex industrial environments, and improves the flexibility of system deployment. The automation control system receives and parses the scheduling plan, identifies the equipment allocation and resource requirements (such as energy consumption and materials) of the task. The task initialization process includes checking the equipment availability, rationality of resource allocation and task preparation, ensuring that the status of resources and equipment meets the standards before the task is executed, avoiding task interruption due to equipment failure or resource shortage, improving the efficiency of task initialization, and shortening the idling time of equipment. According to the scheduling plan, the automation control system sets execution parameters (such as speed, pressure, and power) for each task. These parameters can be dynamically adjusted to adapt to production needs or respond to abnormal situations, achieve precise equipment control, and improve the stability of the production process and product quality. The control system gradually starts the relevant equipment and executes the task according to the task priority. This process avoids the power overload and resource competition problems that may be caused by the concurrent startup of equipment, prevents the impact of instantaneous high energy consumption on the power grid, and improves system stability. The automated control system monitors the operating status of equipment in real time through the sensor network, records task progress and abnormal events. The monitoring data is uploaded to the central database, providing a basis for subsequent analysis and optimization, ensuring the transparency and controllability of the production process, and timely discovering and solving potential problems.

[0142] Furthermore, aggregating production data and simulation results for storage and backup means collecting production data and real-time simulation results after executing a task, sorting them according to timestamps and storing them in a database, synchronously performing regular backup of the stored data, regularly performing security and integrity checks on the stored data and backup data, and generating a test report.

[0143] Through the summary storage, regular backup, and security and integrity testing of production data and real-time simulation results, efficient management and reliable protection of industrial data are achieved. The storage method sorted by timestamp improves the logic and traceability of data, while security and integrity testing ensures the reliability and risk resistance of data. The generated test report not only provides a comprehensive system status assessment tool, but also provides important support for the audit and compliance of enterprise data management. This method effectively improves the stability and efficiency of the production system, provides strong data protection and optimization support for industrial automation control, and is an important technical means to promote the construction of intelligent factories.

[0144] This embodiment also provides an industrial automation control system, including:

[0145] Data acquisition module, used to collect production equipment data in real time and perform preprocessing;

[0146] The construction module is used to build a digital twin model based on equipment operation data and interconnect the digital twin models of factory equipment to form a virtual production network;

[0147] The scheduling optimization module is used to regularly collect factory resource status data and generate the optimal scheduling plan using the DDPG network;

[0148] The adjustment module is used to simulate the execution process of the scheduling scheme in the virtual production network and adjust the scheduling strategy according to the simulation feedback;

[0149] A deployment module, used to deploy the optimized scheduling scheme to the automation control system of the physical plant;

[0150] The data storage module is used to summarize production data and simulation results, store them in the database and back them up regularly.

[0151] This embodiment also provides a computer device suitable for the case of an industrial automation control method, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the industrial automation control method proposed in the above embodiment.

[0152] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0153] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the industrial automation control method proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0154] In summary, the present invention significantly improves the accuracy and efficiency of industrial automation control by introducing IoT data-driven digital twin modeling technology, dynamic interaction analysis of virtual production networks, and scheduling optimization mechanism based on deep reinforcement learning. It can process equipment operation data more efficiently, accurately simulate equipment dynamic behavior, and generate optimal scheduling plans across factories in real time. At the same time, a closed-loop optimization mechanism is formed through simulation verification and feedback adjustment, which realizes the dynamic adaptability and global optimality of the scheduling plan, thereby greatly improving resource utilization efficiency, reducing task delays, and enhancing the flexibility and reliability of the production process.

[0155] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An industrial automation control method, characterized in that: include: Collect production equipment operation data, build a digital twin model and form a virtual production network; Obtain resource status data of each factory in the virtual production network in real time, dynamically generate the optimal cross-factory scheduling plan, simulate the scheduling plan in the virtual production network for simulation verification and adjust the plan; Deploy the scheduling plan to the physical factory, control the equipment to perform tasks through the automated control system, and summarize the production data and simulation results for storage and backup; The resource status data of each factory in the virtual production network is obtained in real time, and the optimal cross-factory scheduling plan is dynamically generated. The resource status data of each factory is obtained from the virtual production network at specified intervals, and the data groups of each factory are combined according to the equipment number ID to obtain the data set corresponding to each factory, and each data set is cleaned and tested for integrity, and the missing data is supplemented by the time series interpolation method, and then the data is standardized; The resource status data includes equipment load rate, resource utilization rate and order priority; Convert the preprocessed numerical data of load rate and resource utilization directly into cloud model representation ; Transform order priority into cloud model representation through fuzzy language model ; Integrate all cloud model data to form a resource status matrix S, where each row represents a task and each column represents the resource status; Convert the resource state matrix into the initial state of the DDPG network and build a dual-objective Critic network and , initialize the Actor network; Collect historical task scheduling records, combine them with the current state matrix S, generate scheduling samples, and build empirical data based on the scheduling samples; The experience data includes the state ,action and reward value ; According to the reward value Perform recursive experience pool grouping, set thresholds z and x, and z>x, if the reward value is greater than the threshold z, it is a high priority group. If the reward value Less than or equal to the threshold z and greater than the threshold x, it is a medium priority group. If the reward value If it is less than the threshold x, it is a low priority group; Dynamically adjust the experience sampling probability according to the TD error, give priority to extracting samples from the high priority group for training, and construct the sample data contained in different training batches according to the priority of the experience pool group; According to the sample data, use the dual-objective Critic network and Calculate the target value; Define the loss function of the Critic network as the mean error between the target value and the current value, and use the Adam optimizer to update the parameters of the Critic network; Use the Actor network to generate the optimal scheduling action based on the current state; The goal of optimizing the Actor network is to maximize the Q value given by the Critic network; Update the Actor network parameters through gradient descent and use the soft update strategy to update the target network parameters; Use the critic network weights and task results to optimize the comprehensive reward function; Reward evaluation for actions output by the Actor network; Set a threshold l. If the reward value is lower than the threshold l, return to the Actor network for retraining and adjust the scheduling strategy. Integrate multiple groups of scheduling plans output by the Actor network. Each group of plans contains performance indicators. The performance indicators are standardized into dimensionless values. A hierarchy analysis method is used to assign weights to each indicator. The weighted performance indicator values ​​are obtained according to the assigned weights and a weighted matrix is ​​generated. Define the positive ideal solution as the optimal value of each indicator, and define the negative ideal solution as the worst value of each indicator; Calculate the distance between the positive and negative ideal solutions, and calculate the closeness coefficient based on the distance between the ideal solutions; The scheduling plans are sorted in descending order according to the proximity coefficient, the scheduling plan with the largest proximity coefficient is selected as the optimal plan, and a scheduling list is generated according to the optimal plan; The scheduling list includes task numbers, factory numbers, equipment numbers and time schedules.

2. The industrial automation control method according to claim 1, characterized in that: The collecting of production equipment operation data, building a digital twin model and forming a virtual production network refers to deploying loT sensors on each production equipment, collecting the temperature, pressure and vibration data of the equipment in real time, classifying the data by equipment number, generating a unique ID for each equipment, normalizing the data after removing outliers using the triple standard deviation method, uploading the pre-processed data to the central data platform using the MQTT industrial protocol, and storing them as data flow tables classified by equipment number, where each record contains the equipment number, data category and timestamp; Use CAD software to build the equipment geometry model and clarify the key components and connection relationships of the equipment; Extract the temperature, pressure and vibration data of the equipment from the data stream table as variables to dynamically drive the state change of the geometric model; Use finite element analysis to simulate the heat conduction path during equipment operation, mark high temperature areas and generate distribution maps, and use state space models to describe the dynamic behavior of equipment; Based on dynamic status mapping, resource status indicators such as equipment load rate, resource utilization rate and order priority are extracted; The equipment load rate is calculated based on the ratio of the current task volume to the total capacity of the equipment. The resource utilization rate is calculated by counting the ratio of the task usage time of each equipment to the total available time. The order priority is calculated based on the importance of the order in the historical scheduling table. Construct a virtual production network, take the geometric model of each factory as a node and construct a node set H, construct the task allocation relationship between devices as an edge set O, the node contains real-time resource status and historical operation data, define the task allocation relationship in the virtual production network, and form an interaction graph G between nodes; Establish a weight relationship for task allocation and describe the task priorities between devices; Use the Industrial Internet Protocol to synchronize data between the physical factory and the virtual production network in both directions.

3. The industrial automation control method according to claim 2, characterized in that: The simulation verification of the scheduling scheme in the virtual production network refers to importing the generated scheduling list into the virtual production network, simulating the execution process of the scheduling scheme in the virtual production network, dynamically updating and recording the equipment operation status, task execution progress and resource consumption of each task, and calculating the total simulation delay D according to the task planning time and the actual simulation completion time; The real-time load rate of the device is calculated based on the ratio of the task occupancy time and the total available time of the device in the simulation, and a threshold P is set. If the device load rate is greater than the threshold P, it is a bottleneck device. Extract the number, actual completion time and delay time of the delayed task in the simulation process and associate them with the equipment number, summarize the bottleneck equipment and delayed task information, and form a feedback data table; The feedback data table includes task number, device number, delay time and device load rate.

4. The industrial automation control method according to claim 3, characterized in that: The scheme adjustment refers to traversing all non-bottleneck devices according to the bottleneck device information recorded in the feedback data, calculating the real-time load rate, selecting the device with the lowest load rate as the replacement device, and allocating the tasks of the bottleneck device to the replacement device in priority; Extract the number and delay time of the delayed task, adjust the task priority incrementally according to the task delay time, reorder the tasks from high to low according to the new priority, and update the task execution order.

5. The industrial automation control method according to claim 4, characterized in that: The deployment of the scheduling plan to the physical factory and controlling the equipment to perform tasks through the automated control system refers to delivering the optimized scheduling plan to the automated control system of the physical factory in a standardized format. The control system parses the scheduling plan, initializes the equipment required for the task, checks the resources required for the task, and according to the scheduling plan, the control system gradually starts the equipment according to the task priority, sets the task execution parameters, and controls the equipment to perform the task.

6. The industrial automation control method according to claim 5, characterized in that: The aggregating of production data and simulation results for storage and backup refers to collecting the production data and real-time simulation results after executing the task, sorting them according to timestamps and storing them in the database, synchronously backing up the stored data regularly, regularly checking the security and integrity of the stored data and backup data and generating a test report.

7. An industrial automation control system, based on the industrial automation control method according to any one of claims 1 to 6, characterized in that: include, Data acquisition module, used to collect production equipment data in real time and perform preprocessing; The construction module is used to build a digital twin model based on equipment operation data and interconnect the digital twin models of factory equipment to form a virtual production network; The scheduling optimization module is used to regularly collect factory resource status data and generate the optimal scheduling plan using the DDPG network; The adjustment module is used to simulate the execution process of the scheduling scheme in the virtual production network and adjust the scheduling strategy according to the simulation feedback; A deployment module, used to deploy the optimized scheduling scheme to the automation control system of the physical plant; The data storage module is used to summarize production data and simulation results, store them in the database and back them up regularly.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the industrial automation control method described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the industrial automation control method described in any one of claims 1 to 6 are implemented.

Images