A self-governed decentralized intelligent factory architecture for complex manufacturing environments

Abstract

The application relates to a self-governed decentralized intelligent factory architecture for a complex manufacturing environment, which comprises: an atomic node, which is the most basic self-governed unit of a self-governed decentralized system, can independently extract information from a data field and internally process, and actively sends processing results and other internal information to the data field in a broadcast mode, and the sent information circulates in the data field; a group data field, which realizes information sharing of the atomic node; a group management node, which is the manager of the whole group; and an edge node, which is the terminal intelligent control unit of the intelligent self-governed decentralized system. Due to decentralization of the system, the fault tolerance and flexibility of the system are improved, so that the complex manufacturing environment is homogenized, self-governed controllable and self-governed coordinatable, and the system has excellent online fault tolerance, online expansion and online maintenance, and the control demand of continuous change and development is improved by deconstructing the complexity of the system.

Description

Technical Field

[0001] This invention relates to a self-regulating distributed architecture design for intelligent factories that incorporates edge computing for complex manufacturing environments, specifically a self-regulating distributed intelligent factory system architecture for complex manufacturing environments. Background Technology

[0002] The essence of intelligent manufacturing is to achieve horizontal integration across enterprise value networks, vertical integration across different levels such as equipment, control, and management, and end-to-end integration throughout the entire product lifecycle through the deep integration of next-generation information technology and advanced manufacturing technology. With the increasing decentralization, scale, and complexity of manufacturing environments, grasping the overall system and controlling the entire system becomes extremely difficult. Furthermore, the problem of controlling the entire system from a single centralized point makes the system less resilient to failures. Therefore, we break away from the traditional centralized or distributed system architecture and propose a novel intelligent system framework for complex manufacturing environments—an Intelligent Autonomous Decentralized System (E-ADS) incorporating edge nodes. This ensures that each subsystem or node has self-management capabilities while not interfering with the affairs of other subsystems or nodes, and is also not subject to interference from other subsystems or nodes, coordinating its work with other subsystems or nodes.

[0003] In autonomous distributed systems, the system's functions are distributed across various subsystems; distribution and autonomy are two key terms. Due to this decentralization, the system's fault tolerance and flexibility are improved. Fault tolerance means that even if one part of the system fails, other parts can still operate autonomously, and the overall stability and availability of the system remain unaffected. Flexibility means that subsystems can be added, deleted, and replaced online according to changes in the overall functional requirements of the system, thereby achieving homogenization, autonomous controllability, and autonomous coordination of complex manufacturing environments, as well as excellent online fault tolerance, online expansion, and online maintainability. By deconstructing the system's complexity, it enhances the control capabilities to meet constantly changing and evolving needs. Summary of the Invention

[0004] In response to the decentralization, large-scale operation, and complexity of complex manufacturing environments, the technical solution adopted by this invention to achieve the above objectives is a self-regulating decentralized intelligent factory system architecture for complex manufacturing environments.

[0005] The technical solution adopted by this invention to achieve the above objectives is: a self-regulating decentralized smart factory architecture for complex manufacturing environments, comprising:

[0006] Edge nodes are used to receive and process real-time detection information from the device. When the real-time detection information is determined to be normal operating data, the edge nodes store the real-time detection information and periodically transmit it to the corresponding atomic nodes through passive transmission mode. When the real-time detection information is determined to be abnormal operating data, the edge nodes activate active transmission mode to transmit the real-time detection information to the corresponding atomic nodes.

[0007] Atomic nodes are used to extract and process information from the data domain and broadcast the information containing the processing results back to the data domain so that the sent information circulates in the data field.

[0008] The data domain is used for information exchange with atomic nodes and management nodes, enabling information sharing among atomic nodes;

[0009] The management node is used to manage each atomic node through data interaction with the data field.

[0010] The atomic node is used to represent a subsystem within the group, including:

[0011] The autonomous control processor is used to process the information obtained by the application module and send the information containing the processing results to the data domain in a data broadcasting mode without a destination address.

[0012] The application module is associated with the data domain and is driven by data in the data domain; it extracts information from the data domain and sends it to the autonomous control processor.

[0013] The autonomous control processor includes:

[0014] The perceptron, with sensitivity Se, receives and processes real-time detection information from edge nodes or other atomic nodes, providing the processed information to the decision generator.

[0015] The decision-maker, possessing autonomy (Au) and initiative (Ac), outputs instructions to the feedback unit based on the information input from the sensor.

[0016] A feedback mechanism, which has a reactive Ac, receives instructions from the decision-maker and generates output actions that act on edge nodes or other atomic nodes.

[0017] The management node is used to construct a model expressing the functional collaboration and coordination relationships between subsystems, and to optimize the subsystem variables within the model to obtain the optimal subsystem S. i The device control variables are sent to the corresponding subsystem S via the data domain. i Atomic nodes.

[0018] The process of constructing a model that expresses the functional collaboration and coordination relationships between subsystems includes the following steps:

[0019] A=(Au, Ac, Se, So, Re, Mo)

[0020] Where A is the atomic node characteristic quantification matrix; Au is autonomy; Re is initiative; Se is sensitivity; Ac is reactivity; Mo is mobility; and So is sociality.

[0021] The group system S = (S1, S2, ..., S9) has the following system state variable matrix representation:

[0022]

[0023] i = 1, 2, ..., 9

[0024]

[0025] J = f(J1, J2, ..., J9)

[0026] Where i is the group subsystem number, A i For subsystem S i Characteristic quantization matrix; A i =[Au i Ac i Se i So i Re i Mo i ] is the subsystem S i The feature quantization matrix, i.e., the threshold of the characteristic state variable; For subsystem S i The state variable matrix, where z is the number of control variables; For subsystem S i The control parameter matrix, i.e., the threshold values ​​of each control variable within an atomic node; u z For subsystem S i Control variables; Let J1, J2, ..., J9 be the evaluation function matrix of the group system S, where J1, J2, ..., J9 represent the multivariate evaluation functions of each atomic node, and f represents the group-level evaluation function.

[0027] The process involves optimizing the variables of the subsystems within the model to obtain the optimal subsystem S. i The device control variables are sent to the corresponding subsystem S via the data domain. i The atomic nodes include the following steps:

[0028] For subsystem S i When entering an abnormal state, i.e., when real-time detection information shows abnormal operating conditions, the group-level evaluation function J is globally optimized through quadratic programming, and the optimized control variables are sent to the corresponding subsystem S. i Atomic nodes.

[0029] A method for implementing an autonomous, decentralized smart factory architecture for complex manufacturing environments includes the following steps:

[0030] Edge nodes receive and process real-time detection information from devices. When the real-time detection information is determined to be normal operating data, the edge nodes store the real-time detection information and periodically transmit it to their corresponding atomic nodes through passive transmission mode. When the real-time detection information is determined to be abnormal operating data, the edge nodes activate active transmission mode to transmit the real-time detection information to their corresponding atomic nodes.

[0031] Atomic nodes extract and process information from the data domain, and broadcast the information containing the processing results back to the data domain so that the sent information circulates in the data field.

[0032] Information exchange between the data domain and atomic nodes, as well as management nodes, enables information sharing among atomic nodes;

[0033] The management node manages each atomic node by interacting with the data field.

[0034] The management node constructs a model expressing the functional collaboration and coordination relationships between subsystems, and optimizes the model by finding the subsystem variables to obtain the optimal subsystem S. i The device control variables are sent to the corresponding subsystem S via the data domain. i Atomic nodes.

[0035] The present invention has the following beneficial effects and advantages:

[0036] (1) Intelligent autonomous distributed systems are multi-agent systems with distributed artificial intelligence and autonomous control and autonomous coordination capabilities;

[0037] (2) Each unit (atomic node) can freely enter and leave the consortium data field, thus meeting the needs of online expansion, fault tolerance and online maintenance;

[0038] (3) Add edge computing nodes to the end of each unit (atomic node) to further decentralize specific computing tasks, thereby maximizing operating efficiency and reducing costs;

[0039] (4) The autonomy of each unit (atomic node) is restricted by the agreement or protocol, which has strong security and controllability;

[0040] (5) Each unit (atomic node) can satisfy: equality, locality and self-containment, and can achieve self-control and self-coordination. Attached Figure Description

[0041] Figure 1 Organizational structure diagram of Midea Refrigeration Appliances Group;

[0042] Figure 2 This is a schematic diagram of the self-regulating decentralized smart factory architecture of the Midea Refrigeration Appliances Group.

[0043] Figure 3 A schematic diagram of the atomic node construction;

[0044] Figure 4 Schematic diagram of data domain network structure. Detailed Implementation

[0045] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0046] This invention relates to a self-regulating decentralized intelligent factory architecture for complex manufacturing environments, comprising: atomic nodes, which are the most basic self-regulating units in the self-regulating decentralized system, capable of independently extracting information from the data domain and performing internal processing, while also proactively broadcasting processing results and other internal information to the data domain, with the emitted information circulating within the data domain; a group-based data domain, which is the logical space for information propagation in the self-regulating decentralized system, enabling information sharing among atomic nodes; a group-based management node, which is the manager of the entire group, possessing the functions and characteristics of atomic nodes, and also responsible for the overall management, coordination, optimization, and control of the entire group system; and edge nodes, which are the final intelligent control units of the intelligent self-regulating decentralized system, sinking some application functions to the edge of the access network, providing storage, computing, and networking functions, and enabling independent operation and edge-to-edge collaboration within the empowered scope. Due to the decentralization of the system, this invention improves the system's fault tolerance and flexibility, thereby achieving homogenization, self-regulating controllability and self-regulating coordination in complex manufacturing environments, as well as excellent online fault tolerance, online expansion, and online maintainability. Furthermore, by deconstructing the complexity of the system, it enhances the control capabilities to meet constantly changing and evolving needs. Among them, complex manufacturing environment refers to the need to complete not only vertical integration across different levels such as equipment, control and management within the group, but also horizontal integration across the value network of subsidiaries and end-to-end integration throughout the product lifecycle.

[0047] Intelligent Autonomous Decentralized Systems (E-ADS) can be considered as being developed based on the combination of Edge Intelligence Systems (EAT) and Autonomous Decentralized Systems (ADS). As shown below:

[0048] EAI+ADS→E-ADS

[0049] In the formula, EAT represents edge intelligence system; + represents integration; ADS represents autonomous decentralized system; → represents development; and E-ADS represents intelligent autonomous decentralized system.

[0050] Based on the characteristics of the target system, a group-based intelligent autonomous distributed system (GE-ADS) is adopted. The system includes atomic nodes (Atom), a group-based data field (Group-Data Field), a group-based management node (Group Manager Atom), and edge nodes (Edge). Edge nodes primarily receive and process real-time information from intelligent control units at the network's end, such as data from intelligent sensors (pressure, temperature, flow rate, etc.) and parameters from intelligent actuators (valve, robotic arm, etc.). Edge nodes store this information in real-time and make intelligent decisions. When the information is determined to be normal operating data, the data is stored on the edge node and periodically transmitted to the atomic nodes via passive transmission mode. Only when the edge node determines that an abnormal operating condition has occurred will it activate active transmission mode to transmit data to the atomic nodes. This ensures the smooth flow of the data field while maintaining stable operation. Information between atomic nodes is shared through the group-based data field, and the group-based management node performs real-time data analysis, i.e., optimal operating point optimization, to optimize the allocation of resources such as raw materials, energy, and manpower within the group.

[0051] An atomic node is a structure containing the management functions and information necessary to ensure the self-regulation, controllability, and coordination of each subsystem. It is the most basic self-regulating unit in a self-regulating decentralized system. Physically, it corresponds to a computer or intelligent gateway. It can independently extract information from the data domain and process it internally, while also proactively broadcasting the processing results and other internal information back to the data domain. The transmitted information circulates within the data domain. There is no direct coupling between nodes; they are only interested in the content of the information in the data domain and do not need to know where the information comes from, effectively reducing the overall complexity of the system. Atomic nodes all include an Autonomous Control Processor (ACP) and an Application Programming Module (APM).

[0052] a. The Autonomous Control Processor (ACP) is responsible for managing the operation of the atomic nodes themselves and coordinating with other ACPs. In an autonomous distributed system, both data broadcasting without destination addresses and data-driven operations are implemented by the ACP.

[0053] b. Application Modules (APMs) can be installed on any computer without needing to know the installation locations of other application software modules, because all nodes and even modules are only associated with the data domain and driven by the data in the data domain. Furthermore, application modules can be moved from one computer to another without notifying other computers of their location change, thus enabling application module mobility.

[0054] The Group Data Field (GIN) is the logical space for information propagation in a self-regulating distributed system. Physically, it's analogous to a network or storage device, enabling information sharing among atomic nodes. It's a Group Intranet (GIN) supported by a Group Information Base (GIB). All atomic nodes actively send information to the data field and simultaneously retrieve information from it as needed to complete the functions of their internal modules. The portion of the data field extending into the atomic nodes is called the node data field, where the data required by the node's user application modules flows.

[0055] The Group Manager Atom is the manager of the entire group. In addition to having the functions and characteristics of an atomic node, it is also responsible for the overall management, coordination, optimization and control of the entire group system.

[0056] An edge node is a subsystem built at the edge of an atomic node network, serving as the final intelligent control unit in an intelligent, autonomous, distributed system. It offloads some application functions to the edge of the access network, providing storage, computing, and networking capabilities, enabling independent operation and edge-to-edge collaboration within its authorized scope. Physically, it corresponds to edge gateways, edge controllers, and intelligent sensors. This approach maximizes the freedom of atomic nodes and data domains.

[0057] Midea Refrigeration Appliances Group comprises nine subsystems: Refrigeration Research Institute, Washing Machine Business Unit, Compressor Business Unit, Refrigerator Business Unit, Domestic Residential Air Conditioner Business Unit, Overseas Residential Air Conditioner Business Unit, China Marketing Headquarters, International Business Unit, and Central Air Conditioning Business Unit. With the increasing decentralization, scale, and complexity of manufacturing environments, traditional centralized or distributed architectures are no longer sufficient to effectively control the entire group-level system. Furthermore, due to the coupling between subsystems, factories, and workshops, traditional architectures are ineffective in coordinating and optimizing horizontally across different stages.

[0058] To address the challenges of such a large, sophisticated, and high-speed system and achieve perfect synergy among people, materials, resources, and energy, we have broken away from the existing system architecture and proposed a new intelligent system framework for Midea Refrigeration and Home Appliances Group—the Group-E-ADS (Group-E-ADS) system, which introduces edge nodes. This framework ensures that each subsystem or node has self-management capabilities without interfering with the affairs of other subsystems or nodes, and can coordinate with other subsystems. This enables vertical integration across different levels of the group's subsystems, including equipment, control, and management layers; horizontal integration across the value networks of the subsystems within the group; and end-to-end integration throughout the entire product lifecycle.

[0059] The specific steps are as follows:

[0060] (1) Deconstruction of the Group System

[0061] Midea Refrigeration Appliances Group is divided into nine subsystems (atomic nodes) based on its Refrigeration Research Institute, Washing Machine Business Unit, Compressor Business Unit, Refrigerator Business Unit, Domestic Residential Air Conditioner Business Unit, Overseas Residential Air Conditioner Business Unit, China Marketing Headquarters, International Business Unit, and Central Air Conditioning Business Unit. Figure 1 As shown.

[0062] The Washing Machine Business Unit is divided into 10 units: Electronics Company, Twin-tub Company, Fully Automatic Company, Drum Washing Machine Company, Quality Management Department, Supply Chain Management Department, Technology R&D Center, Financial Management Department, Securities Business Unit, and Operations and Human Resources Department.

[0063] The Compressor Business Unit is divided into eight units: Meizhi Center, Technology R&D Center, Manufacturing Center, Marketing Center, Quality Management Department, Supply Chain Management Department, Financial Management Department, and Operations and Human Resources Department.

[0064] The Refrigerator Division is divided into five units: Freezer Product Center, Technology R&D Center, Manufacturing Center, Financial Management Department, and Operations and Human Resources Department.

[0065] The Domestic Business Unit for Residential Air Conditioning is divided into five units: Technology R&D Center, Manufacturing Center, Product Strategy Department, Financial Management Department, and Operations and Human Resources Department.

[0066] The Overseas Business Division for Residential Air Conditioning is divided into four units: Technology R&D Center, Manufacturing Center, Quality and Product Management Department, and Business Management Department.

[0067] China Marketing Headquarters: Divided into 11 units including Central Air Conditioning Sales Department, Washing Machine Sales Department, Refrigerator Sales Department, Air Conditioning Sales Department, After-sales Management Department, Planning and Logistics Department, Public Relations Department, Channel Development Department, Brand Marketing Department, Financial Management Department, and Operations and Human Resources Department;

[0068] International Business Division: Divided into 16 units including Strategic Customers Department, Central Air Conditioning Business Department, Washing Machine Business Department, Refrigerator Business Department, HVAC Products Department, Planning and Logistics Department, Marketing Department, Vietnam Company, Middle East and Africa Region, Asia Pacific Region, Eastern Europe Region, Western Europe Region, Greater Latin America Region, North America Region, Financial Management Department, and Operations and Human Resources Department.

[0069] The Central Air Conditioning Business Unit is divided into 10 units: Assembly Workshop, Automation Products Company, Industrial Washing Machine Products Company, Water Heater Products Company, Multi-Split Air Conditioner Products Company, Supply Chain Management Department, Chongqing Department, Quality Management Department, Financial Management Department, and Operations and Human Resources Department.

[0070] (2) System Architecture Design

[0071] Based on the characteristics of Midea Refrigeration Appliance Group and drawing on the concept of ADS (Advanced Device Management System), a group-style intelligent autonomous decentralized system (Group-E-ADS) incorporating edge nodes was established, such as... Figure 2 As shown, it includes:

[0072] The Group Manager Atom is the manager of the entire Midea Refrigeration and Home Appliance Group, and is a human-like intelligent entity or multi-intelligent entity system.

[0073] A Group Data Field is a Group Intranet (GIN) supported by a Group Information Base (GIB).

[0074] The atomic nodes are planned to be the nine subsystems of the group-style intelligent autonomous decentralized system, which are divided into nine subsystems: Refrigeration Research Institute, Washing Machine Business Unit, Compressor Business Unit, Refrigerator Business Unit, Domestic Residential Air Conditioning Business Unit, Overseas Residential Air Conditioning Business Unit, China Marketing Headquarters, International Business Unit, and Central Air Conditioning Business Unit.

[0075] Edge nodes are intelligent control units such as edge gateways, edge controllers, and smart sensors established at the network edges of the nine atomic nodes, including the Refrigeration Research Institute and the Washing Machine Division.

[0076] Design concept:

[0077] Group-based intelligent self-regulating decentralized systems possess distributed artificial intelligence and limited self-regulation and decentralization.

[0078] Each atomic node in the group is managed by a group-style management node and has limited autonomy and independence;

[0079] The operation and management within the group are achieved through a group-based data domain, which combines a "centralized-decentralized" approach. That is, group management is centralized, while the operation of each meta-entity is decentralized.

[0080] Each atomic node can satisfy the following: equality, locality, and restraint, which are different from the group-style management intelligence.

[0081] The design of a group-based intelligent self-regulatory decentralized system can draw inspiration from the group's operations.

[0082] Intelligent autonomous decentralized systems not only possess the autonomous control and coordination of autonomous decentralized systems, but also exhibit human-like distributed artificial intelligence, such as initiative, mobility, sensitivity, responsiveness, and sociality.

[0083] (3) Atomic node construction

[0084] The system consists of subsystems (atomic nodes) that share functions. In such a system, subsystems are driven by external input information or information received from other subsystems. The processing results of a subsystem are passed to other subsystems, which then perform similar processing in sequence. This driving condition also includes cases of logical AND and OR operations with multiple inputs. Therefore, through coordination among the subsystems, the overall function of the system can be achieved.

[0085] Atomic node structure, see Figure 3 .

[0086] A decision maker is a decision generator with autonomy (Au) and initiative (Ac) that outputs instructions to a feedback unit based on subjective judgment and information input from sensors.

[0087] A sensor, which has sensitivity Se, receives input stimulus information from the external environment or other intelligent agents and provides the processed information to the decision generator.

[0088] An Actuator, which is reactive (Ac), receives instructions from a decision generator and produces output actions that act on the external environment or other atomic nodes.

[0089] For example, the atomic node sensor of the compressor business unit receives production scheduling information (stimulus information) from atomic nodes of the refrigerator business unit, home air conditioning business unit, central air conditioning business unit, etc., and after the decision-maker performs internal optimization of the atomic node, it issues the optimal production scheduling instruction to the production workshop through the feedback device.

[0090] (4) System Modeling

[0091] In intelligent autonomous decentralized systems, a model is established to represent the functional collaboration and coordination relationships between subsystems, and driving conditions are added:

[0092] A=(Au, Ac, Se, So, Re, Mo)

[0093] Where A is the atomic node characteristic quantization matrix; Au is autonomy; Re is initiative; Se is sensitivity; Ac is reactivity; Mo is mobility; and So is sociality. The characteristic quantization matrix parameters need to be quantized empirically.

[0094] The group system S = (S1, S2, ..., S9) has the following system state variable matrix representation:

[0095]

[0096] i = 1, 2, ..., 9

[0097]

[0098] J = f(J1, J2, ..., J9)

[0099] Where i is the group subsystem number, A i For subsystem S i Characteristic quantization matrix; A i =[Au i Ac i Se i So i Re i Mo i ] is the subsystem S i The feature quantification matrix, namely the threshold values ​​of state variables such as autonomy, initiative, and sensitivity; For subsystem S i The state variable matrix corresponds to the feature quantization matrix; z is the control variable number; For subsystem S i The control system parameter matrix, i.e., the threshold values ​​of each control variable; u z For subsystem S i Control variables, such as valves, robotic arms, and rotational speed; Let S be the evaluation function matrix of system S.

[0100] Self-regulation and controllability: For any subsystem S i When entering an abnormal state, quadratic programming is used to apply the multivariable function J of the subsystem. i The system performs global variable optimization to achieve internal controllability of subsystems. This means that if any subsystem is in an abnormal state, it does not affect the self-management and functional operation of other subsystems, and other subsystems can arbitrarily control its scope of responsibility. For example, the atomic node sensor of the compressor division receives production scheduling information from atomic nodes of the refrigerator division, residential air conditioning division, and central air conditioning division. Through quadratic programming, it performs global variable optimization on the multivariate production scheduling function value of its own subsystem and issues optimization instructions to the production workshop.

[0101] Self-regulating and coordinating property: For any subsystem S i When entering an abnormal state, the system optimizes its evaluation function J-value to achieve coordination between subsystems. This means that even if one subsystem is in an abnormal state, other subsystems can still complete their tasks and operate collaboratively. For example, when an atomic node in the compressor division experiences abnormal operating conditions, such as reduced compressor production or shutdown, the group-level atomic node uses quadratic programming to optimize global variables such as raw materials, energy, and labor resources within the group using a multivariate evaluation function, and then issues optimization instructions to all relevant atomic nodes.

[0102] (5) Group-style data domain construction

[0103] A dual-ring gateway architecture is adopted, with each gateway node having four network interfaces. There are 10 gateway nodes in the network, forming 20 different network segments. These 20 different network segments can form different data domains or multicast groups. Heterogeneous systems can connect to different network segments, achieving heterogeneous system integration through gateway node links, such as... Figure 4 As shown.

[0104] (6) Edge nodes

[0105] The "proximity" characteristic of edge nodes determines their ability to react quickly, reduce latency caused by network transmission, and enable real-time computing. In particular, they can perform independent localized operations when the subsystem network fails.

[0106] Edge nodes enable data acquisition, local data processing, local data storage, and execution of corresponding operations based on control commands. By being located close to or embedded in field devices such as sensors, instruments, robots, and machine tools, they support real-time intelligent interconnection and intelligent applications for field equipment. Depending on their function, edge nodes can be categorized into physical units such as intelligent sensors and intelligent actuators.

Claims

1. A self-regulating decentralized smart factory architecture for complex manufacturing environments, characterized in that, include: Edge nodes are used to receive and process real-time detection information from the device. When the real-time detection information is determined to be normal operating data, the edge nodes store the real-time detection information and periodically transmit it to the corresponding atomic nodes through passive transmission mode. When the real-time detection information is determined to be abnormal operating data, the edge nodes activate active transmission mode to transmit the real-time detection information to the corresponding atomic nodes. Atomic nodes are used to extract and process information from the data domain and broadcast the information containing the processing results back to the data domain so that the sent information circulates in the data field. The data domain is used for information exchange with atomic nodes and management nodes, enabling atomic nodes to share information. 、 Management nodes are used to manage each atomic node through data interaction with the data field; The management node is used to construct a model expressing the functional collaboration and coordination relationships between subsystems, and to optimize the subsystem variables within the model to obtain the optimal subsystem. The device control variables are transmitted via the data domain to the corresponding subsystem. Atomic nodes.

2. The autonomous decentralized smart factory architecture for complex manufacturing environments according to claim 1, characterized in that, The atomic node is used to represent a subsystem within the group, including: The autonomous control processor is used to process the information obtained by the application module and send the information containing the processing results to the data domain in a data broadcasting mode without a destination address. The application module is associated with the data domain and is driven by data in the data domain; it extracts information from the data domain and sends it to the autonomous control processor.

3. The autonomous decentralized smart factory architecture for complex manufacturing environments according to claim 1 or 2, characterized in that, The autonomous control processor includes: The perceptron, with sensitivity Se, receives and processes real-time detection information from edge nodes or other atomic nodes, providing the processed information to the decision generator. The decision-maker, possessing autonomy (Au) and initiative (Ac), outputs instructions to the feedback unit based on the information input from the sensor. A feedback mechanism, which has a reactive Ac, receives instructions from the decision-maker and generates output actions that act on edge nodes or other atomic nodes.

4. The autonomous decentralized smart factory architecture for complex manufacturing environments according to claim 1, characterized in that, The process of constructing a model that expresses the functional collaboration and coordination relationships between subsystems includes the following steps: ； in, Quantization matrix for atomic node characteristics; For self-discipline; For initiative; Sensitivity; It is reactive; For mobility; For social purposes; Group System The system state variable matrix is ​​represented as follows: ； ； ； ； in, Number the group's subsystems. For subsystem Characteristic quantization matrix; For subsystem The feature quantization matrix, i.e., the threshold of the characteristic state variable; For subsystem The state variable matrix, To control the number of variables; For subsystem The control parameter matrix, i.e., the threshold values ​​of each control variable within an atomic node; For subsystem Control variables; For the group system The evaluation function matrix, Let f represent the multivariate evaluation function for each atomic node, and let f represent the group-level evaluation function.

5. The autonomous decentralized smart factory architecture for complex manufacturing environments according to claim 1, characterized in that, The process involves optimizing the variables of the subsystems within the model to obtain the optimal subsystem. The device control variables are transmitted via the data domain to the corresponding subsystem. The atomic nodes include the following steps: For subsystems Entering an abnormal state, i.e., real-time detection information shows abnormal operating conditions, a quadratic programming approach is used to evaluate the group-level evaluation function. The values ​​are globally optimized, and the optimized control variables are sent to the corresponding subsystems. Atomic nodes.

6. A method for implementing an autonomous decentralized smart factory architecture for complex manufacturing environments, wherein the method is applied to the autonomous decentralized smart factory architecture for complex manufacturing environments described in claim 1, characterized in that, Includes the following steps: Edge nodes receive and process real-time detection information from devices. When the real-time detection information is determined to be normal operating data, the edge nodes store the real-time detection information and periodically transmit it to their corresponding atomic nodes through passive transmission mode. When the real-time detection information is determined to be abnormal operating data, the edge nodes activate active transmission mode to transmit the real-time detection information to their corresponding atomic nodes. Atomic nodes extract and process information from the data domain, and broadcast the information containing the processing results back to the data domain so that the sent information circulates in the data field. Information exchange between the data domain and atomic nodes, as well as management nodes, enables information sharing among atomic nodes; The management node manages each atomic node by interacting with the data field.

7. The method for implementing an autonomous decentralized smart factory architecture for complex manufacturing environments according to claim 6, characterized in that, The management node constructs a model expressing the functional collaboration and coordination relationships between subsystems, and optimizes the model using subsystem variables to obtain the optimal subsystem. The device control variables are transmitted via the data domain to the corresponding subsystem. Atomic nodes.

Images