An edge computing driven light-weight hybrid optimization system
By leveraging the collaborative work of edge computing nodes and cloud computing centers, combined with lightweight heuristic algorithms and deep learning models, the real-time performance and resource utilization issues of the production planning and scheduling system were resolved. This enabled efficient, flexible, and multi-objective optimization of production scheduling, thereby improving the operational efficiency of enterprises.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- WUXI INSTITUTE OF TECHNOLOGY
- Filing Date
- 2026-02-09
- Publication Date
- 2026-05-29
AI Technical Summary
Existing production planning and scheduling systems suffer from insufficient real-time performance, large fluctuations in resource utilization, and weak ability to balance multiple objectives when facing complex manufacturing scenarios, creating a vicious cycle that leads to decreased production efficiency and increased operating costs.
A lightweight hybrid optimization system driven by edge computing is adopted. Real-time data acquisition and preliminary scheduling are achieved through edge computing nodes, task priority prediction and global optimization are performed in the cloud computing center, and data interfaces ensure data interaction between systems. By combining lightweight heuristic algorithms and deep learning models, rapid response and global optimal solution are achieved.
It improved the real-time performance of production scheduling and resource utilization, optimized costs, on-time performance and energy consumption, and enhanced production flexibility and corporate competitiveness.
Smart Images
Figure CN122111567A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial automation and intelligent manufacturing technology, specifically to a lightweight hybrid optimization system driven by edge computing. Background Technology
[0002] In the field of industrial automation and intelligent manufacturing technology, production planning and scheduling systems are a core component supporting the efficient operation of complex manufacturing scenarios such as petrochemicals, new energy, and shipbuilding. Their core function is to formulate scientific and reasonable production scheduling plans based on multi-dimensional information such as production order demand, real-time equipment status, and energy consumption costs, thereby achieving orderly production process advancement, optimized resource allocation, and multi-objective balance. As the manufacturing industry transforms towards intelligence and flexibility, dynamic events such as equipment failures and urgent order insertions occur frequently in production scenarios, placing higher demands on the real-time response capabilities, resource utilization efficiency, and multi-objective collaborative optimization capabilities of production planning and scheduling systems. However, existing production planning and scheduling systems and the technical solutions supporting their operation have gradually revealed many shortcomings in addressing these demands and are no longer adequate for the actual operational needs of complex manufacturing scenarios.
[0003] From the perspective of the technologies relied upon by current production planning and scheduling systems, traditional solutions generally suffer from three key limitations: First, insufficient real-time performance is a prominent issue. Existing systems mostly employ traditional optimization algorithms such as mixed integer programming. These algorithms have high computational complexity and are time-consuming. When facing equipment failures requiring urgent scheduling adjustments or the insertion of urgent orders necessitating the replanning of production tasks, it often takes a considerable amount of time to generate new feasible scheduling solutions. This cannot meet the needs of minute-level dynamic adjustments during production, easily leading to production interruptions due to response delays, or causing ineffective waste of resources such as equipment and raw materials. Second, resource utilization fluctuates significantly. Traditional systems are mostly designed based on static scheduling rules, unable to adjust task allocation according to dynamic information such as real-time equipment load and changes in order priority. This results in some equipment remaining idle for extended periods due to insufficient task allocation, leading to low utilization rates, while other equipment operates continuously under high load due to excessive task concentration. This not only easily triggers equipment failures and increases maintenance costs but also reduces the stability of the overall production process. Third, weak ability to balance multiple objectives. Existing algorithms are not designed to fully coordinate the relationship between cost control, on-time order delivery rate and energy consumption. This often results in the optimization of one objective at the expense of other objectives. For example, reducing production batches to lower inventory costs may lead to frequent equipment start-ups and shutdowns and a surge in energy consumption. Or, prioritizing urgent orders may disrupt the original production rhythm, causing delays in the delivery of non-urgent orders, ultimately affecting the overall production efficiency.
[0004] Furthermore, the aforementioned technical limitations are not isolated but interconnected, forming a vicious cycle that amplifies the performance deficiencies of existing systems: insufficient real-time performance prevents timely responses to dynamic events, exacerbating equipment load imbalances and making resource utilization fluctuations more pronounced; unstable resource utilization not only wastes resources but also affects production schedules, increasing the difficulty of order delivery control and weakening the effectiveness of multi-objective trade-offs; the weakness in multi-objective trade-off capabilities makes it difficult for the system to formulate adjustment plans that balance efficiency and cost when dealing with dynamic events, further extending response time and creating a vicious cycle of "poor real-time performance—resource waste—multi-objective imbalance." This vicious cycle ultimately leads to a significant decline in production efficiency, reduced customer satisfaction, and increased enterprise operating costs, failing to meet the core requirements of current complex manufacturing scenarios for the efficiency, flexibility, and dynamic adaptability of production planning and scheduling systems. There is an urgent need for a new technical solution that can overcome the limitations of traditional technologies and balance real-time response, resource optimization, and multi-objective equilibrium to support the performance upgrade of production planning and scheduling systems.
[0005] To address these issues, those skilled in the art propose a lightweight hybrid optimization system driven by edge computing. Summary of the Invention
[0006] To address the shortcomings of existing technologies, this invention provides a lightweight hybrid optimization system driven by edge computing, which solves the problems mentioned in the background section.
[0007] To achieve the above objectives, the present invention provides the following technical solution: a lightweight hybrid optimization system driven by edge computing, comprising edge computing nodes, a cloud computing center, and a data interface;
[0008] The edge computing nodes are used to perform real-time data acquisition, preliminary scheduling scheme generation, and event response operations to enable rapid processing of dynamic events on the production site.
[0009] The cloud computing center is used to perform task priority prediction and global optimization solutions to solve complex global problems in production planning and scheduling.
[0010] The data interface is used to enable data interaction between edge computing nodes, cloud computing centers and external production management systems, ensuring the smooth flow of production data between the systems.
[0011] Through the above technical solution, edge computing nodes enable real-time data acquisition from the production site, generation of preliminary scheduling plans, and event response, ensuring rapid processing of dynamic events. The cloud computing center is responsible for task priority prediction and global optimization, solving complex production planning and scheduling problems. Data interfaces ensure smooth data exchange between edge computing nodes, the cloud computing center, and external production management systems. The collaborative work of these three components makes production scheduling more intelligent and efficient, enabling rapid response to dynamic events, optimization of production plans, and improvement of production efficiency and resource utilization.
[0012] Preferably, the edge computing node includes a lightweight heuristic algorithm module, a real-time event response module, and a cloud collaboration module; the lightweight heuristic algorithm module integrates genetic algorithms and local search algorithms; the real-time event response module is used to receive equipment fault signals and emergency order insertion instructions, and trigger the lightweight heuristic algorithm module to recalculate the local scheduling scheme; the cloud collaboration module is used to encapsulate complex problems into a specified format and upload them to the cloud computing center through a preset communication protocol.
[0013] Through the aforementioned technical solution, the collaborative work of edge computing nodes, cloud computing centers, and data interfaces enables intelligent optimization and efficient management of production scheduling. Edge computing nodes utilize a lightweight heuristic algorithm module integrating genetic and local search algorithms to generate and optimize preliminary scheduling plans in real time, rapidly responding to dynamic events such as equipment failures and urgent orders, ensuring real-time decision-making and efficient operation on the production floor. Their cloud collaboration module uploads complex problems to the cloud, leveraging its powerful computing capabilities for in-depth optimization, further enhancing the global optimality of the scheduling plan. The cloud computing center focuses on task priority prediction and global optimization solutions, utilizing deep learning model modules and a global optimization engine module to provide precise guidance for production planning and scheduling, ensuring rational resource allocation and maximizing production efficiency. Data interfaces ensure smooth data interaction between systems, enabling real-time information sharing and collaboration, making the entire production scheduling process more intelligent and automated, effectively improving production flexibility and response speed, reducing operating costs, and enhancing the company's competitiveness in complex production environments.
[0014] Preferably, the cloud computing center includes a deep learning model module and a global optimization engine module; the deep learning model module uses an LSTM neural network, with input dimensions including historical order data, equipment status data, and energy cost data, and outputs task priority weights; the global optimization engine module uses a mixed integer programming algorithm, with inputs including complex problems uploaded by the cloud collaboration module and task priority weights output by the deep learning model module, and outputs the global optimal solution.
[0015] Through the above technical solutions, the cloud computing center achieves efficient integration of "precise prediction - global optimization" in production planning and scheduling by coordinating the deep learning model module and the global optimization engine module. The deep learning model module uses an LSTM neural network, which can accurately mine data correlation patterns based on three key inputs: historical order data, equipment status data, and energy cost data, to output scientific task priority weights, providing clear guidance for subsequent optimization. The global optimization engine module uses these task priority weights as the core basis, combined with the complex problems uploaded from edge computing nodes, to solve the system using a mixed integer programming algorithm, ultimately outputting the globally optimal solution. This ensures that the scheduling scheme meets actual production needs and achieves an optimal balance among multiple objectives such as cost, on-time performance, and energy consumption, effectively improving the rationality and overall coordination capabilities of production planning and scheduling.
[0016] Preferably, the objective function of the mixed integer programming algorithm in the global optimization engine module is set as MinimizeZ = α・delivery delay + β・equipment utilization rate − γ・energy cost, and the constraints include equipment capacity limit, process sequence constraint and emergency order priority rule, where α, β and γ are the task priority weights corresponding to delivery delay, equipment utilization rate and energy cost, respectively.
[0017] Through the above technical solution, the algorithm uses equipment capacity limitations, process sequence constraints, and emergency order priority rules as constraints to ensure that the globally optimal solution obtained not only meets the actual operating capacity and process technology requirements of the production equipment, but also prioritizes the response to emergency order demands. Finally, it generates a production scheduling scheme that takes into account feasibility, rationality, and economy, effectively solving the problems of multi-objective imbalance and scheme disconnect from actual production in traditional scheduling.
[0018] Preferably, the data interface includes an ERP / MES system interface module and a 5G real-time communication protocol module; the ERP / MES system interface module adopts the OPCUA protocol to read production order data and write scheduling results, and the production order data includes order quantity, order delivery date and raw material inventory data.
[0019] Through the above technical solution, the data interface, in conjunction with the ERP / MES system interface module and the 5G real-time communication protocol module, provides crucial support for the efficient collaboration between the optimization engine and the external production management system. Specifically, the ERP / MES system interface module, using the OPCUA protocol, can accurately read production order data, including order quantity, order delivery date, and raw material inventory data, providing comprehensive and accurate basic information for the optimization engine to generate scheduling plans. It can also promptly write the final scheduling results into the ERP / MES system, ensuring consistency between the production execution process and the scheduling plan. This module, in collaboration with the 5G real-time communication protocol module, further guarantees the stability and timeliness of data transmission, preventing the scheduling plan from becoming disconnected from actual production due to data interaction delays or missing information. This effectively achieves data connectivity between the optimization engine and the production management system, laying a data foundation for the dynamic adjustment and efficient execution of production planning and scheduling.
[0020] Preferably, the device fault signal received by the real-time event response module comes from an IoT sensor, which is used to collect data on the device's load rate, temperature, and energy consumption.
[0021] Through the above technical solution, the real-time event response module receives device fault signals from IoT sensors. Simultaneously, relying on the IoT sensors' ability to collect real-time data on device load rate, temperature, and energy consumption, it can, on the one hand, capture abnormal device states immediately. This provides accurate and timely fault information for quickly triggering a recalculation of the local scheduling scheme, preventing production interruptions or escalating risks due to undetected device faults. On the other hand, the device load rate, temperature, and energy consumption data collected by the IoT sensors also provide real-time data support for edge computing nodes to generate preliminary scheduling schemes and determine whether the device's operating status is normal. This ensures that the scheduling scheme aligns with the actual operating conditions of the equipment, further improving the rationality of production scheduling and the stability of equipment operation.
[0022] Preferably, the process of the lightweight heuristic algorithm module generating a preliminary scheduling scheme is as follows: the genetic algorithm generates candidate schemes based on the current order data, the local search algorithm optimizes the candidate schemes by specifying an exchange strategy, and finally outputs a preset number of preliminary scheduling schemes.
[0023] Through the above technical solution, the lightweight heuristic algorithm module generates candidate schemes through a genetic algorithm, optimizes them through a local search algorithm, and outputs a preset number of preliminary scheduling schemes. This process not only relies on the genetic algorithm to quickly generate multiple sets of candidate scheduling schemes with basic feasibility based on current order data, ensuring the diversity and efficiency of the preliminary schemes, but also uses a specified exchange strategy to specifically optimize local conflicts in the candidate schemes through a local search algorithm, improving the rationality and adaptability of the preliminary scheduling schemes. Finally, the preset number of preliminary scheduling schemes output can also provide multiple basic schemes for subsequent edge nodes to quickly respond to dynamic events or for cloud-based collaborative optimization, effectively balancing the efficiency and quality of preliminary scheduling and meeting the initial requirements of production scheduling for real-time performance and feasibility.
[0024] Preferably, after the global optimization engine module of the cloud computing center outputs the global optimal solution, it returns the global optimal solution to the edge computing node through the 5G real-time communication protocol module, and synchronously updates the production plan in the ERP system to ensure that the production plan is consistent with the optimization result.
[0025] Through the above technical solution, after the global optimization engine module of the cloud computing center outputs the global optimal solution, it quickly returns the optimal solution to the edge computing node with the help of the 5G real-time communication protocol module. At the same time, the production plan in the ERP system is updated synchronously. This process not only relies on the low latency characteristics of 5G technology to ensure the timeliness of the transmission of the global optimal solution to the edge node, ensuring that the edge node can quickly adjust the local scheduling execution according to the optimal solution, but also achieves seamless alignment between the production plan and the optimization result by synchronously updating the ERP system. This avoids the disconnect between production execution and scheduling plan due to information asynchrony, and ultimately ensures that the production process is strictly carried out in accordance with the global optimal scheduling plan, thereby improving the overall execution and accuracy of production scheduling.
[0026] Preferably, the edge computing nodes employ hardware devices with specified computing power to support real-time data acquisition and lightweight algorithm execution; the cloud computing center deploys instances with a specified number of CPU cores to meet the solution requirements of high-complexity mixed integer programming algorithms.
[0027] Through the above technical solutions, edge computing nodes employ hardware devices with specified computing power, which can fully meet the requirements of high efficiency in real-time data acquisition and smooth operation of lightweight algorithms. This ensures that the edge can quickly process real-time data from the production site and generate preliminary scheduling plans, avoiding data processing delays or algorithm lags due to insufficient hardware computing power. The cloud computing center deploys instances with a specified number of CPU cores, providing sufficient computing resources to support high-complexity mixed-integer programming algorithms. This ensures the efficiency and accuracy of the global optimal solution process, avoiding prolonged global optimization time due to insufficient hardware performance. Through precise matching of hardware performance and functional requirements, both provide a stable hardware foundation for the real-time response and global optimization capabilities of the optimization engine, ensuring the efficient progress of the entire production planning and scheduling process.
[0028] Preferably, the task priority weights output by the deep learning model module can be dynamically adjusted based on the device failure frequency and emergency order triggering signals fed back by the edge computing nodes. When an emergency order triggering signal exists, the task priority weight values related to the emergency order are increased by a preset ratio.
[0029] Through the above technical solution, the task priority weights output by the deep learning model module can be dynamically adjusted according to the equipment failure frequency and emergency order triggering signals fed back by the edge computing nodes. When an emergency order triggering signal is present, the corresponding weights are increased by a preset ratio. This design not only allows the task priority weights to adapt to the dynamic changes in the production site, but also allows the priority weights to be quickly increased when an emergency order occurs, ensuring that emergency orders are prioritized for scheduling and processing. This effectively avoids the problems of traditional static weight settings being unable to adapt to dynamic production scenarios and the lag in emergency order response, further improving the flexibility and targeting of production scheduling, and ensuring that the production process can still proceed efficiently when dealing with emergencies.
[0030] This invention provides a lightweight hybrid optimization system driven by edge computing. It has the following advantages:
[0031] 1. This invention utilizes a hybrid architecture that combines edge computing nodes and cloud computing centers, along with a lightweight heuristic algorithm to quickly generate preliminary scheduling schemes and a cloud-based global optimization engine to solve complex problems. This significantly improves the response speed of the production planning and scheduling system, effectively meets the real-time requirements for dynamic adjustments during production, avoids production interruptions or resource waste caused by system response delays, and ensures the smooth operation of the production process.
[0032] 2. This invention can dynamically adjust the scheduling target weights based on real-time feedback production data, reasonably balance equipment load distribution, avoid the problem of equipment being idle or overloaded under traditional static scheduling rules, and keep the equipment in a state of high efficiency, thereby improving the overall utilization rate of equipment resources and reducing resource waste caused by equipment idleness and frequent downtime maintenance costs caused by equipment overload.
[0033] 3. This invention predicts task priorities using a deep learning model and combines a mixed-integer programming algorithm to collaboratively optimize multiple objectives such as cost, on-time performance, and energy consumption. This solves the problem that existing algorithms have difficulty coordinating multiple objectives. While ensuring on-time order delivery, it effectively controls energy consumption and inventory costs in the production process, thereby improving overall production efficiency and enhancing the company's operational competitiveness in complex manufacturing scenarios. Attached Figure Description
[0034] Figure 1 This is the overall flowchart of the present invention;
[0035] Figure 2 This is a flowchart of the module interaction process of the present invention. Detailed Implementation
[0036] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0037] Please see the appendix Figure 1 - Appendix Figure 2 This invention provides a lightweight hybrid optimization system driven by edge computing, including edge computing nodes, cloud computing center and data interface;
[0038] Edge computing nodes are used to perform real-time data acquisition, preliminary scheduling scheme generation, and event response operations to enable rapid processing of dynamic events on the production site;
[0039] The edge computing node includes a lightweight heuristic algorithm module, a real-time event response module, and a cloud collaboration module. The lightweight heuristic algorithm module integrates genetic algorithms and local search algorithms. The real-time event response module is used to receive equipment fault signals and emergency order insertion instructions, and trigger the lightweight heuristic algorithm module to recalculate the local scheduling scheme. The cloud collaboration module is used to encapsulate complex problems into a specified format and upload them to the cloud computing center through a preset communication protocol.
[0040] Specifically, the lightweight heuristic algorithm module integrates genetic algorithms and local search algorithms to quickly generate preliminary scheduling schemes. The real-time event response module receives real-time information such as equipment fault signals and emergency order insertion instructions, and triggers the lightweight heuristic algorithm module to recalculate the local scheduling scheme, enabling timely response and processing of dynamic events. The cloud collaboration module encapsulates complex problems into a specified format and uploads them to the cloud computing center via a preset communication protocol. Leveraging the powerful computing capabilities of the cloud, further analysis and optimization are performed, thereby achieving collaborative work between edge computing and cloud computing, improving the overall system's intelligence and decision-making capabilities.
[0041] The cloud computing center is used to perform task priority prediction and global optimization solutions to solve complex global problems in production planning and scheduling.
[0042] The cloud computing center includes a deep learning model module and a global optimization engine module. The deep learning model module uses an LSTM neural network, with input dimensions including historical order data, equipment status data, and energy cost data, and outputs task priority weights. The global optimization engine module uses a mixed integer programming algorithm, with inputs including complex problems uploaded by the cloud collaboration module and task priority weights output by the deep learning model module, and outputs the global optimal solution.
[0043] Specifically, the deep learning model module and the global optimization engine module work together to solve complex global problems in production planning and scheduling. The deep learning model module uses an LSTM neural network to analyze historical order, equipment status, and energy cost data to generate task priority weights. These weights, along with the complex problem data, are input into the global optimization engine module, which uses a mixed-integer programming algorithm to calculate the optimal scheduling scheme, achieving global optimization and improving production planning efficiency and resource utilization.
[0044] The data interface is used to enable data interaction between edge computing nodes, cloud computing centers, and external production management systems, ensuring the smooth flow of production data between these systems. The data interface includes an ERP / MES system integration module and a 5G real-time communication protocol module. The ERP / MES system integration module uses the OPCUA protocol to read production order data and write scheduling results. The production order data includes order quantity, order delivery date, and raw material inventory data.
[0045] Specifically, the data interface, acting as a crucial bridge in the production system, utilizes the OPCUA protocol of the ERP / MES system interface module to read production order data and write scheduling results, ensuring bidirectional data flow between the production management system and computing nodes. Simultaneously, the 5G real-time communication protocol module, leveraging the 5G network, guarantees real-time and efficient data interaction between edge computing nodes, cloud computing centers, and the production management system, significantly reducing latency. The synergistic effect of these two components enables the data interface to precisely connect with production data logic, meeting the real-time response requirements of edge and cloud computing, driving the intelligent and automated upgrade of the production management system, and achieving closed-loop management and optimization of the entire production process.
[0046] The objective function of the mixed integer programming algorithm in the global optimization engine module is set as MinimizeZ = α・delivery delay + β・equipment utilization rate − γ・energy cost. The constraints include equipment capacity limit, process sequence constraint and emergency order priority rule, where α, β and γ are the task priority weights corresponding to delivery delay, equipment utilization rate and energy cost, respectively.
[0047] The real-time event response module receives device fault signals from IoT sensors, which collect data on device load rate, temperature, and energy consumption. The lightweight heuristic algorithm module generates a preliminary scheduling scheme as follows: a genetic algorithm generates candidate schemes based on current order data; a local search algorithm optimizes the candidate schemes using a specified exchange strategy; and finally, a preset number of preliminary scheduling schemes are output.
[0048] Specifically, IoT sensors continuously collect data such as device load rate, temperature, and energy consumption to monitor device status in real time. This data is transmitted to a real-time event response module, which is always on standby. Upon detecting a device fault signal or an urgent order insertion command, it immediately triggers a lightweight heuristic algorithm module. For example, if a sensor detects an abnormally high temperature in a device, potentially indicating an impending failure, the real-time event response module quickly captures this signal and activates the lightweight heuristic algorithm module to re-plan the scheduling scheme, striving to restore production order in the shortest possible time. The lightweight heuristic algorithm module employs a combination of genetic algorithms and local search algorithms. The genetic algorithm, based on current order data, simulates a biological evolution process to generate diverse candidate scheduling schemes. The local search algorithm focuses on these candidate schemes, using specific exchange strategies to fine-tune the schemes, eliminating unreasonable parts, strengthening advantageous aspects, and ultimately outputting a certain number of preliminary scheduling schemes.
[0049] After the global optimization engine module of the cloud computing center outputs the global optimal solution, it returns the global optimal solution to the edge computing node via the 5G real-time communication protocol module, and synchronously updates the production plan in the ERP system to ensure that the production plan is consistent with the optimization results. The edge computing node uses hardware devices with specified computing power to support real-time data acquisition and lightweight algorithm execution; the cloud computing center deploys instances with specified CPU cores to meet the solution requirements of high-complexity mixed integer programming algorithms.
[0050] Specifically, after the global optimization engine module of the cloud computing center outputs the globally optimal solution, it quickly transmits the result to the edge computing nodes using the 5G real-time communication protocol module. Simultaneously, the edge nodes update the production plan in the ERP system, ensuring consistency between the plan and the optimization results. The edge computing nodes are equipped with hardware devices with specified computing power, supporting real-time data acquisition and lightweight algorithm execution. The cloud computing center deploys instances with a specified number of CPU cores to meet the solution requirements of highly complex mixed-integer programming algorithms, ensuring the efficient operation of the entire system.
[0051] The task priority weights output by the deep learning model module can be dynamically adjusted based on the device failure frequency and emergency order triggering signals fed back by the edge computing nodes. When an emergency order triggering signal exists, the task priority weight values related to the emergency order are increased by a preset ratio.
[0052] Specifically, the deep learning model module dynamically adjusts task priority weights based on the frequency of device failures reported by edge computing nodes and emergency order trigger signals. When an emergency order is triggered, the priority weights of tasks related to that order are increased by a preset ratio to ensure that emergency tasks are processed first.
[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A lightweight hybrid optimization system driven by edge computing, characterized in that, This includes edge computing nodes, cloud computing centers, and data interfaces; The edge computing nodes are used to perform real-time data acquisition, preliminary scheduling scheme generation, and event response operations to enable rapid processing of dynamic events on the production site. The cloud computing center is used to perform task priority prediction and global optimization solutions to solve complex global problems in production planning and scheduling. The data interface is used to enable data interaction between edge computing nodes, cloud computing centers and external production management systems, ensuring the smooth flow of production data between the systems.
2. The edge computing-driven lightweight hybrid optimization system according to claim 1, characterized in that, The edge computing node includes a lightweight heuristic algorithm module, a real-time event response module, and a cloud collaboration module. The lightweight heuristic algorithm module integrates genetic algorithms and local search algorithms. The real-time event response module is used to receive equipment fault signals and emergency order insertion instructions, and trigger the lightweight heuristic algorithm module to recalculate the local scheduling scheme. The cloud collaboration module is used to encapsulate complex problems into a specified format and upload them to the cloud computing center through a preset communication protocol.
3. The edge computing-driven lightweight hybrid optimization system according to claim 1, characterized in that, The cloud computing center includes a deep learning model module and a global optimization engine module. The deep learning model module uses an LSTM neural network, with input dimensions including historical order data, equipment status data, and energy cost data, and outputs task priority weights. The global optimization engine module uses a mixed integer programming algorithm, with inputs including complex problems uploaded by the cloud collaboration module and task priority weights output by the deep learning model module, and outputs the global optimal solution.
4. The edge computing-driven lightweight hybrid optimization system according to claim 3, characterized in that, The objective function of the mixed integer programming algorithm in the global optimization engine module is set as MinimizeZ = α・delivery delay + β・equipment utilization rate − γ・energy cost. The constraints include equipment capacity limit, process sequence constraint and emergency order priority rule, where α, β and γ are the task priority weights corresponding to delivery delay, equipment utilization rate and energy cost, respectively.
5. The edge computing-driven lightweight hybrid optimization system according to claim 1, characterized in that, The data interface includes an ERP / MES system integration module and a 5G real-time communication protocol module; the ERP / MES system integration module adopts the OPCUA protocol to read production order data and write scheduling results, and the production order data includes order quantity, order delivery date and raw material inventory data.
6. The edge computing-driven lightweight hybrid optimization system according to claim 2, characterized in that, The device fault signal received by the real-time event response module comes from the IoT sensor, which is used to collect data on the device's load rate, temperature, and energy consumption.
7. The edge computing-driven lightweight hybrid optimization system according to claim 2, characterized in that, The lightweight heuristic algorithm module generates a preliminary scheduling scheme as follows: the genetic algorithm generates candidate schemes based on the current order data, the local search algorithm optimizes the candidate schemes through a specified exchange strategy, and finally outputs a preset number of preliminary scheduling schemes.
8. The edge computing-driven lightweight hybrid optimization system according to claim 3, characterized in that, After the global optimization engine module of the cloud computing center outputs the global optimal solution, it returns the global optimal solution to the edge computing node through the 5G real-time communication protocol module, and synchronously updates the production plan in the ERP system to ensure that the production plan is consistent with the optimization result.
9. The edge computing-driven lightweight hybrid optimization system according to claim 1, characterized in that, The edge computing nodes employ hardware devices with specified computing power to support real-time data acquisition and lightweight algorithm execution; the cloud computing center deploys instances with a specified number of CPU cores to meet the solution requirements of high-complexity mixed integer programming algorithms.
10. The edge computing-driven lightweight hybrid optimization system according to claim 3, characterized in that, The task priority weights output by the deep learning model module can be dynamically adjusted based on the device failure frequency and emergency order triggering signals fed back by the edge computing nodes. When an emergency order triggering signal exists, the task priority weight values related to the emergency order are increased by a preset ratio.