Multi-target collaborative optimization scheduling system and method based on artificial intelligence

Abstract

The invention provides a multi-target collaborative optimization scheduling system based on artificial intelligence, relates to the technical field of intelligent scheduling, and realizes automatic task classification, task association relationship modeling, resource optimization distribution and dynamic adjustment of a scheduling strategy through task analysis, target association analysis, scheduling strategy generation, real-time monitoring and a dynamic scheduling module. The system utilizes natural language processing, machine learning, graph neural network, reinforcement learning and other technologies to extract features from task information, analyze association between tasks and generate a scheduling strategy according to resource requirements, a real-time monitoring module tracks a task execution state, a dynamic scheduling module evaluates deviation influence through a causal inference model, and a task scheduling strategy is generated. The system is suitable for the electronic equipment, the instructions are stored through the storage medium, and efficient task scheduling and management are achieved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent scheduling technology, and more specifically, to a multi-objective collaborative optimization scheduling system and method based on artificial intelligence. Background Art

[0002] In modern complex systems, multi-objective collaborative optimization scheduling is a key technology to improve efficiency and resource utilization. With the development of artificial intelligence technology, scheduling methods based on artificial intelligence have gradually attracted attention. However, existing scheduling systems often have the following problems: First, traditional scheduling methods are difficult to meet the optimization needs of multiple objectives at the same time, resulting in overall low efficiency; second, the dynamic correlation between tasks is not considered enough, and it is impossible to effectively deal with the uncertainty in complex task environments; finally, there is a lack of real-time monitoring and dynamic adjustment mechanisms, and it is impossible to respond to deviations and changes in the task execution process in a timely manner.

[0003] Therefore, the existing scheduling system is difficult to meet the needs of multi-objective optimization at the same time, lacks consideration of the dynamic correlation between tasks, and lacks real-time monitoring and dynamic adjustment mechanisms. Summary of the invention

[0004] In order to overcome the problems that the existing scheduling system is difficult to simultaneously meet multi-objective optimization needs, lacks consideration of the dynamic correlation between tasks, and lacks real-time monitoring and dynamic adjustment mechanisms, the present invention designs a multi-objective collaborative optimization scheduling system and method based on artificial intelligence, which can effectively solve the above technical problems.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows: Multi-objective collaborative optimization scheduling system based on artificial intelligence, including task analysis module, target association analysis module, scheduling strategy generation module, real-time monitoring module and dynamic scheduling module; The task parsing module is used to obtain task information of multiple tasks, and automatically divide the tasks into multiple task categories according to the task information; the multiple task categories respectively contain multiple tasks; The target association analysis module is used to obtain target attributes of the multiple tasks respectively included in the multiple task categories, and analyze the association relationship between the multiple tasks according to the target attributes to obtain multiple task association sets; the number of the task association sets is related to the task category; The scheduling strategy generation module is used to obtain resource requirement information corresponding to the multiple tasks respectively, and plan the execution order and resource allocation of the multiple tasks according to the resource requirement information to generate a scheduling strategy; The real-time monitoring module is used to obtain the execution status information of the multiple tasks in real time, and analyze the execution status information to obtain task progress, resource usage and task execution deviation; The dynamic scheduling module is used to dynamically adjust the scheduling strategy according to the task progress, the resource usage and the task execution deviation to ensure multi-objective collaborative optimization.

[0006] Preferably, the task parsing module includes: A task information acquisition unit, used for acquiring the task information of the plurality of tasks, specifically acquiring detailed descriptions, priorities and deadline task information of the plurality of tasks from the task management system through a data interface; A feature extraction unit is used to extract key features from the task information and generate a task feature vector, specifically using natural language processing technology and machine learning algorithms to extract key features from the task information and generate a task feature vector; The category division unit is used to perform cluster analysis on the task feature vector to obtain multiple task categories. Specifically, a K-Means clustering algorithm is used to perform cluster analysis on the task feature vector to determine the task category.

[0007] Preferably, the target association analysis module includes: A target attribute acquisition unit, used to respectively acquire the target attributes of the plurality of tasks by using a data mining technique; An association relationship modeling unit, used to model the association relationship between the tasks according to the target attributes using a graph neural network to obtain a task association graph; The association set determining unit is used to determine multiple task association sets according to the task association graph, and specifically perform community discovery algorithm processing on the task association graph to obtain multiple task association sets.

[0008] Preferably, the scheduling strategy generation module includes: A resource demand acquisition unit, configured to acquire resource demand information corresponding to each of the plurality of tasks by using big data technology; the resource demand information includes manpower, material resources and time; A scheduling model building unit, used to build a scheduling model based on reinforcement learning; The strategy generating unit is used to input the task association set and the resource requirement information into the scheduling model to generate a scheduling strategy.

[0009] Preferably, the real-time monitoring module includes: A status information acquisition unit, used for acquiring the execution status information of the plurality of tasks in real time; and also for collecting and collating the execution status information in real time; A progress analysis unit, configured to analyze the execution status information to obtain the task progress of the plurality of tasks, specifically by comparing the actual completion steps of the tasks with the planned completion steps to calculate the task progress; A resource analysis unit, used to analyze the execution status information to obtain the resource usage, specifically count the actual usage and planned usage of resources, and analyze the resource usage; The deviation detection unit is used to detect the task execution deviation by comparing with the preset task execution standard, and specifically detect the task execution deviation by calculating the task progress deviation and the resource usage deviation.

[0010] Preferably, the dynamic scheduling module includes: An impact assessment unit, for assessing the impact on other tasks and overall objectives according to the task execution deviation; A strategy adjustment unit, used to adjust the scheduling strategy according to the impact assessment result; The execution unit is used to execute the adjusted scheduling strategy.

[0011] Preferably, the impact assessment unit uses a causal inference model to assess the impact on other tasks and the overall goal based on the task association set and the task execution deviation; The strategy adjustment unit generates an adjusted scheduling strategy by re-inputting the adjusted task information and resource information into the scheduling model; The execution unit specifically sends the adjusted scheduling strategy to the task execution system for execution.

[0012] The multi-objective collaborative optimization scheduling method based on artificial intelligence includes the following steps: Obtaining task information of multiple production tasks through the data interface of the factory production management system, wherein the task information includes a detailed description of the task, priority, deadline, required production equipment type, and process requirements; using natural language processing technology to extract key information in the task description, and using a machine learning algorithm to extract features of the task priority, deadline, and estimated production time to generate a task feature vector; using a K-Means clustering algorithm to perform cluster analysis on the task feature vector, and automatically dividing the task into multiple task categories, wherein the multiple task categories respectively contain multiple tasks, such as dividing the task into high-priority emergency tasks, routine production tasks, and quality inspection task categories; The target attributes of the multiple tasks respectively included in the multiple task categories are obtained by data mining technology, and the target attributes include the production process nodes, quality requirements, and dependencies with upstream and downstream processes of the tasks; the association relationship between the tasks is modeled according to the target attributes by using a graph neural network to obtain a task association graph, such as establishing associations between multiple tasks that require the same batch of raw materials, and establishing strong associations between tasks that are closely connected to the previous and next processes; the task association graph is processed by a community discovery algorithm to obtain multiple task association sets, the number of which is related to the task category, and each task association set represents a set of tasks with strong associations, such as a certain association set containing a series of tasks from raw material processing to semi-finished product assembly; The resource demand information corresponding to the multiple tasks is obtained through the factory production big data platform, and the resource demand information includes the type and quantity of production equipment, the type and amount of raw materials, the manpower demand and skill requirements and the production time window; a scheduling model based on reinforcement learning is constructed, and the deep Q network algorithm is used as an example. The intelligent agent of the model learns how to reasonably arrange the start sequence of tasks, allocate production equipment and human resources according to the category of the current task, the association set information and the available status of resources, so as to maximize the reward function, and the reward function includes multiple objectives such as on-time delivery rate, equipment utilization rate and production cost control; the task association set and the resource demand information are input into the scheduling model, and the execution sequence and resource allocation of the multiple tasks are planned to generate an initial scheduling strategy, such as giving priority to the use of key equipment for high-priority tasks and reasonably allocating skilled workers to tasks with high technical requirements; Acquire the execution status information of the multiple tasks in real time, collect equipment operation data, raw material consumption data, product quality inspection data and personnel operation data through the IoT sensor network in the factory workshop, and collect and organize the execution status information in real time; calculate the task progress by comparing the actual completion progress of the task with the planned completion progress, such as a task has completed half of the planned progress; count the actual usage and planned usage of resources, and analyze the resource usage; detect task execution deviation by calculating task progress deviation and resource usage deviation, such as a task is delayed due to equipment failure, and a subsequent task is suspended due to insufficient supply of raw materials; According to the task execution deviation, the causal inference model is used to evaluate the impact on other tasks and the overall goal. The delay of a key task may lead to a chain delay of multiple subsequent dependent tasks, affecting the delivery time of the entire production batch and increasing production costs. According to the impact assessment results, the adjusted task information and resource information are re-entered into the scheduling model to generate an adjusted scheduling strategy. In order to alleviate the impact of delays in key tasks, idle equipment support is reallocated, the start sequence of subsequent tasks is adjusted, and raw material procurement and replenishment are prioritized. The adjusted scheduling strategy is sent to the factory production execution system and executed to ensure multi-objective collaborative optimization.

[0013] An electronic device includes a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the program or instructions are executed by the processor, the steps of the multi-objective collaborative optimization scheduling method based on artificial intelligence as described above are implemented.

[0014] A readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the steps of the multi-objective collaborative optimization scheduling method based on artificial intelligence as described above are implemented.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention automatically divides tasks into multiple categories through the task parsing module, and models the correlation between tasks in combination with the target association analysis module, which can simultaneously consider the optimization needs of multiple targets and improve the overall scheduling efficiency and resource utilization; the graph neural network is used to model the correlation between tasks, and the task association set is determined in combination with the community discovery algorithm, which can effectively capture the dynamic correlation between tasks and adapt to the uncertainty in the complex task environment; the execution status information of the task is obtained through the real-time monitoring module, and the task progress, resource usage and task execution deviation are analyzed, and the dynamic scheduling module is combined to dynamically adjust according to the task execution deviation to ensure the real-time and adaptability of the scheduling strategy; in the dynamic scheduling module, the causal inference model is used to evaluate the impact of task execution deviation on other tasks and the overall goal, which can more accurately predict and respond to deviations and changes in the task execution process, and improve the scientificity and reliability of the scheduling strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived based on the provided drawings without paying any creative work.

[0017] Figure 1This is the structure diagram of the multi-objective collaborative optimization scheduling system based on artificial intelligence; Figure 2 This is a step-by-step diagram of the multi-objective collaborative optimization scheduling method based on artificial intelligence. DETAILED DESCRIPTION

[0018] The drawings are for illustrative purposes only and should not be construed as limiting the present patent; In order to better illustrate the present embodiment, some parts in the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product; It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0019] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0020] Example 1

[0021] Multi-objective collaborative optimization scheduling system based on artificial intelligence, please refer to Figure 1 , including task parsing module, target association analysis module, scheduling strategy generation module, real-time monitoring module and dynamic scheduling module; The task parsing module is used to obtain task information of multiple tasks, and automatically divide the tasks into multiple task categories according to the task information; the multiple task categories respectively contain multiple tasks; The target association analysis module is used to obtain target attributes of the multiple tasks respectively included in the multiple task categories, and analyze the association relationship between the multiple tasks according to the target attributes to obtain multiple task association sets; the number of the task association sets is related to the task category; The scheduling strategy generation module is used to obtain resource requirement information corresponding to the multiple tasks respectively, and plan the execution order and resource allocation of the multiple tasks according to the resource requirement information to generate a scheduling strategy; The real-time monitoring module is used to obtain the execution status information of the multiple tasks in real time, and analyze the execution status information to obtain task progress, resource usage and task execution deviation; The dynamic scheduling module is used to dynamically adjust the scheduling strategy according to the task progress, the resource usage and the task execution deviation to ensure multi-objective collaborative optimization.

[0022] The task parsing module comprises: A task information acquisition unit, used for acquiring the task information of the plurality of tasks, specifically acquiring detailed descriptions, priorities and deadline task information of the plurality of tasks from the task management system through a data interface; A feature extraction unit is used to extract key features from the task information and generate a task feature vector, specifically using natural language processing technology and machine learning algorithms to extract key features from the task information and generate a task feature vector; The category division unit is used to perform cluster analysis on the task feature vector to obtain multiple task categories. Specifically, a K-Means clustering algorithm is used to perform cluster analysis on the task feature vector to determine the task category.

[0023] The task management system sends detailed information about the factory's production tasks for the day to the scheduling system. This task information includes a specific description of each task, such as welding specific parts of a specific model of car body, assembling a certain component, priority (determined according to the model production plan and order delivery time, urgent order tasks have a high priority), deadline (the latest time point for completion of each task scheduled according to the production plan), etc. For example, one task is to weld the door part of the A model body, with a high priority and a deadline of 5 pm on the same day.

[0024] The feature extraction unit in the system uses natural language processing technology to process the task description and extract key feature words, such as "welding", "A model body", "door part", etc. At the same time, combined with the machine learning algorithm, the task priority, deadline and other information are converted into numerical or vector features together with the feature words to generate a task feature vector. For example, for the above welding task, the generated task feature vector may be [welding, A model body, door part, high priority, deadline 17:00].

[0025] The classification unit uses the K-Means clustering algorithm to perform cluster analysis on the feature vectors of all tasks. It is assumed that after cluster analysis, the factory's production tasks are divided into three categories: body welding task category, parts assembly task category, and vehicle inspection task category. Each category contains multiple tasks with similar characteristics. For example, the body welding task category contains tasks for welding different parts of the body, such as doors, hoods, trunks, etc., and the parts assembly task category contains tasks for assembling different parts assemblies, such as engine assemblies, seat assemblies, dashboard assemblies, etc.

[0026] The target association analysis module comprises: A target attribute acquisition unit, used to respectively acquire the target attributes of the plurality of tasks by using a data mining technique; An association relationship modeling unit, used to model the association relationship between the tasks according to the target attributes using a graph neural network to obtain a task association graph; The association set determining unit is used to determine multiple task association sets according to the task association graph, and specifically perform community discovery algorithm processing on the task association graph to obtain multiple task association sets.

[0027] The target attribute acquisition unit obtains the target attributes of each task from the production data acquisition system through data mining technology. For the body welding task, the target attributes may include welding quality indicators, such as weld strength, welding accuracy, welding time, required welding equipment type, etc.; for the parts assembly task, the target attributes may include assembly quality indicators, such as parts matching accuracy, assembly firmness, assembly time, required parts type and quantity, etc.; for the whole vehicle inspection task, the target attributes may include the pass rate of inspection items, inspection time, required inspection equipment type, etc.

[0028] The association relationship modeling unit uses graph neural networks to model the association relationships between tasks according to the target attributes of each task. For example, the welding quality of the body welding task may affect the assembly quality of the parts assembly task, and the assembly progress of the parts assembly task will affect the start time of the vehicle inspection task. These association relationships are represented in the graph neural network in the form of nodes representing tasks and edges representing association relationships to form a task association graph. The weights of the edges in the graph can be set according to the degree of association between tasks. The higher the degree of association, the greater the weight of the edge.

[0029] The association set determination unit processes the task association graph using a community discovery algorithm to determine multiple task association sets. Assume that after processing, three task association sets are determined: body welding-parts assembly association set, parts assembly-vehicle inspection association set, and body welding-vehicle inspection association set. The tasks in each task association set have a strong correlation and need to be considered collaboratively during the scheduling process.

[0030] The scheduling strategy generation module includes: A resource demand acquisition unit, configured to acquire resource demand information corresponding to each of the plurality of tasks by using big data technology; the resource demand information includes manpower, material resources and time; A scheduling model building unit, used to build a scheduling model based on reinforcement learning; The strategy generating unit is used to input the task association set and the resource requirement information into the scheduling model to generate a scheduling strategy.

[0031] The resource demand acquisition unit uses big data technology to obtain the resource demand information corresponding to each task from the factory's resource management system. The resource demand information includes manpower, such as the number of workers required, type of work, skill level, etc., material resources, such as the required production equipment, raw materials, parts, etc., and time, such as the expected task execution time, earliest start time, latest end time, etc. For example, the car body welding task requires 5 skilled welders, 3 welding robots, and a certain amount of welding materials, with an estimated execution time of 4 hours; the parts assembly task requires 10 assembly workers, 5 assembly equipment, and various parts and components, with an estimated execution time of 6 hours.

[0032] The scheduling model construction unit constructs a scheduling model based on reinforcement learning. The model takes task association sets and resource demand information as input, and continuously optimizes the generation process of scheduling strategies by learning historical scheduling data and reward mechanisms, such as giving positive rewards for completing tasks on time and high resource utilization, and giving negative rewards for task delays and resource waste.

[0033] The strategy generation unit inputs the task association set and resource requirement information into the scheduling model to generate a scheduling strategy. The scheduling strategy includes the execution order of each task and the resource allocation plan. For example, based on the current production situation and resource requirements, the scheduling strategy determines to execute high-priority tasks in the body welding task category first, then execute urgent tasks in the parts assembly task category, and finally execute tasks in the vehicle inspection task category. In terms of resource allocation, sufficient welders and welding robots are allocated to body welding tasks, and appropriate assembly workers and assembly equipment are allocated to parts assembly tasks to ensure that each task can be executed smoothly in the predetermined order and time requirements.

[0034] The real-time monitoring module comprises: A status information acquisition unit, used for acquiring the execution status information of the plurality of tasks in real time; and also for collecting and collating the execution status information in real time; A progress analysis unit, configured to analyze the execution status information to obtain the task progress of the plurality of tasks, specifically by comparing the actual completion steps of the tasks with the planned completion steps to calculate the task progress; A resource analysis unit, used to analyze the execution status information to obtain the resource usage, specifically count the actual usage and planned usage of resources, and analyze the resource usage; The deviation detection unit is used to detect the task execution deviation by comparing with the preset task execution standard, and specifically detect the task execution deviation by calculating the task progress deviation and the resource usage deviation.

[0035] The status information acquisition unit in the real-time monitoring module obtains the execution status information of each task by connecting to the factory's production data acquisition system in real time. This information includes the actual start time of the task, the amount of work completed, the current execution progress, and the actual use of resources, such as the actual number of workers used, equipment operating time, raw material consumption, etc. For example, the actual start time of a car body welding task is 9 am, 60% of the welding workload has been completed, and 4 welders and 2 welding robots are currently used.

[0036] The progress analysis unit analyzes the acquired execution status information, calculates the task progress of each task, and calculates the task progress percentage by comparing the actual completion steps of the task with the planned completion steps. For example, a certain component assembly task is planned to complete 10 assembly steps, and 7 steps have been completed so far, with a task progress of 70%.

[0037] The resource analysis unit analyzes the execution status information, counts the actual and planned resource usage, calculates the resource utilization rate, and analyzes the resource usage. For example, a vehicle inspection task plans to use three inspection equipment, but actually two are used, with a resource utilization rate of 66.7%; the planned usage time is three hours, but actually two hours have been used, and the resource utilization efficiency is relatively high.

[0038] The deviation detection unit detects the execution deviation of each task by comparing it with the preset task execution standards, such as task deadline, quality standard, resource usage limit, etc. For example, the deadline for a car body welding task is 5 pm, but the actual execution progress is slow and it is expected that it cannot be completed on time, so there is a task progress deviation; the actual quality inspection result of a component assembly task does not meet the quality standard, so there is a quality deviation; the actual resource usage of a whole vehicle inspection task exceeds the planned usage, so there is a resource usage deviation.

[0039] The dynamic scheduling module includes: An impact assessment unit, for assessing the impact on other tasks and overall objectives according to the task execution deviation; A strategy adjustment unit, used to adjust the scheduling strategy according to the impact assessment result; The execution unit is used to execute the adjusted scheduling strategy.

[0040] The impact assessment unit uses a causal inference model to assess the impact on other tasks and the overall goal based on the task association set and the task execution deviation; The strategy adjustment unit generates an adjusted scheduling strategy by re-inputting the adjusted task information and resource information into the scheduling model; The execution unit specifically sends the adjusted scheduling strategy to the task execution system for execution.

[0041] The impact assessment unit in the dynamic scheduling module evaluates the impact of the deviation on other tasks and the overall goal based on the task execution deviation, using the causal inference model and combining the task association set information. For example, the progress deviation of the body welding task may cause the start time of the subsequent parts assembly task to be delayed, thereby affecting the progress of the vehicle inspection task, and ultimately may cause delays in the vehicle delivery time; the quality deviation of the parts assembly task may require rework, increase resource input, and affect the quality and progress of the subsequent vehicle inspection task.

[0042] The strategy adjustment unit adjusts the scheduling strategy based on the impact assessment results. For the progress deviation of the body welding task, the resource allocation can be adjusted, the number of welders or welding robots can be increased, or the task execution order can be adjusted to prioritize the execution of tasks on the critical path; for the quality deviation of the parts assembly task, the task execution process can be adjusted, the quality inspection link can be strengthened, or skilled assembly workers can be reallocated for rework; for the resource use deviation, the resource use plan can be optimized, resources can be reasonably allocated, and resource waste can be avoided.

[0043] The execution unit sends the adjusted scheduling strategy to the factory's production execution system. The production execution system adjusts and controls each production link according to the new scheduling strategy to ensure multi-objective collaborative optimization. For example, the production execution system rearranges the work tasks of workers and equipment according to the adjusted scheduling strategy and adjusts the production plan so that each task can be smoothly executed according to the new scheduling strategy, ultimately achieving efficient, coordinated and optimized production scheduling in automobile manufacturing plants.

[0044] Example 2

[0045] Multi-objective collaborative optimization scheduling method based on artificial intelligence, please refer to Figure 2 , including the following steps: Obtaining task information of multiple production tasks through the data interface of the factory production management system, wherein the task information includes a detailed description of the task, priority, deadline, required production equipment type, and process requirements; using natural language processing technology to extract key information in the task description, and using a machine learning algorithm to extract features of the task priority, deadline, and estimated production time to generate a task feature vector; using a K-Means clustering algorithm to perform cluster analysis on the task feature vector, and automatically dividing the task into multiple task categories, wherein the multiple task categories respectively contain multiple tasks, such as dividing the task into high-priority emergency tasks, routine production tasks, and quality inspection task categories; The target attributes of the multiple tasks respectively included in the multiple task categories are obtained by data mining technology, and the target attributes include the production process nodes, quality requirements, and dependencies with upstream and downstream processes of the tasks; the association relationship between the tasks is modeled according to the target attributes by using a graph neural network to obtain a task association graph, such as establishing associations between multiple tasks that require the same batch of raw materials, and establishing strong associations between tasks that are closely connected to the previous and next processes; the task association graph is processed by a community discovery algorithm to obtain multiple task association sets, the number of which is related to the task category, and each task association set represents a set of tasks with strong associations, such as a certain association set containing a series of tasks from raw material processing to semi-finished product assembly; The resource demand information corresponding to the multiple tasks is obtained through the factory production big data platform, and the resource demand information includes the type and quantity of production equipment, the type and amount of raw materials, the manpower demand and skill requirements and the production time window; a scheduling model based on reinforcement learning is constructed, and the deep Q network algorithm is used as an example. The intelligent agent of the model learns how to reasonably arrange the start sequence of tasks, allocate production equipment and human resources according to the category of the current task, the association set information and the available status of resources, so as to maximize the reward function, and the reward function includes multiple objectives such as on-time delivery rate, equipment utilization rate and production cost control; the task association set and the resource demand information are input into the scheduling model, and the execution sequence and resource allocation of the multiple tasks are planned to generate an initial scheduling strategy, such as giving priority to the use of key equipment for high-priority tasks and reasonably allocating skilled workers to tasks with high technical requirements; Acquire the execution status information of the multiple tasks in real time, collect equipment operation data, raw material consumption data, product quality inspection data and personnel operation data through the IoT sensor network in the factory workshop, and collect and organize the execution status information in real time; calculate the task progress by comparing the actual completion progress of the task with the planned completion progress, such as a task has completed half of the planned progress; count the actual usage and planned usage of resources, and analyze the resource usage; detect task execution deviation by calculating task progress deviation and resource usage deviation, such as a task is delayed due to equipment failure, and a subsequent task is suspended due to insufficient supply of raw materials; According to the task execution deviation, the causal inference model is used to evaluate the impact on other tasks and the overall goal. The delay of a key task may lead to a chain delay of multiple subsequent dependent tasks, affecting the delivery time of the entire production batch and increasing production costs. According to the impact assessment results, the adjusted task information and resource information are re-entered into the scheduling model to generate an adjusted scheduling strategy. In order to alleviate the impact of delays in key tasks, idle equipment support is reallocated, the start sequence of subsequent tasks is adjusted, and raw material procurement and replenishment are prioritized. The adjusted scheduling strategy is sent to the factory production execution system and executed to ensure multi-objective collaborative optimization.

[0046] An electronic device includes a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the program or instructions are executed by the processor, the steps of the multi-objective collaborative optimization scheduling method based on artificial intelligence as described above are implemented.

[0047] A readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the steps of the multi-objective collaborative optimization scheduling method based on artificial intelligence as described above are implemented.

[0048] In the specific implementation, the task information of multiple production tasks is obtained through the data interface of the factory production management system (MES). The task information includes a detailed description of the task, such as the model, specification, priority (high, medium, low), deadline, required type of production equipment, such as lathes, milling machines, testing equipment and process requirements, such as machining accuracy and surface treatment requirements. For example, Task 1 is to produce high-precision engine parts, with a high priority and a deadline of within 24 hours. A high-precision lathe is required and the process requirement is an accuracy of ±0.01mm.

[0049] Natural language processing technology is used to extract key information in the task description, such as component model, processing technology keywords, etc., and machine learning algorithms are used to extract features of task priority, deadline, and estimated production time to generate a task feature vector. For example, the priority is encoded as high = 2, medium = 1, and low = 0. The deadline is converted into a time window value, and the estimated production time is normalized. The K-Means clustering algorithm is used to perform cluster analysis on the task feature vector, and the tasks are automatically divided into multiple task categories, such as high-priority emergency tasks, routine production tasks, and quality inspection task categories. The feature vector of task 1 is (2, 0.8, 0.6), and after cluster analysis, it is classified as a high-priority emergency task category.

[0050] The target attributes of the multiple tasks respectively included in the multiple task categories are obtained through data mining technology, and the target attributes include the production process nodes of the tasks, such as raw material processing, semi-finished product assembly, finished product testing, quality requirements, such as accuracy, surface roughness, and dependencies with upstream and downstream processes. For example, the output of a task is the input of another task. The target attributes of Task 1 include production process nodes of raw material processing and semi-finished product assembly, the quality requirement is accuracy of ±0.01mm, and the dependency is that semi-finished product assembly can only be carried out after raw material processing is completed.

[0051] A graph neural network (GNN) is used to model the association between the tasks according to the target attributes to obtain a task association graph. For example, multiple tasks that require the same batch of raw materials are associated, and strong associations are established between tasks that are closely connected to the previous and next processes. Task 1 and Task 2 are associated because they use the same batch of raw materials, and the output of Task 1 is the input of Task 3. Therefore, a strong association is established between Task 1 and Task 3.

[0052] The task association graph is processed by a community discovery algorithm to obtain multiple task association sets. The number of the task association sets is related to the task category. Each task association set represents a set of tasks with strong associations. A task association set includes a series of tasks from raw material processing to semi-finished product assembly, such as task 1, task 2 and task 3 forming an association set.

[0053] The resource demand information corresponding to the multiple tasks is obtained through the factory production big data platform. The resource demand information includes the type and quantity of production equipment, the type and usage of raw materials, manpower requirements and skill requirements, and the production time window. Task 1 requires a high-precision lathe. The raw material is steel, the usage is 100kg, and it requires a skilled worker. The production time window is 8 hours.

[0054] A scheduling model based on reinforcement learning is constructed. Taking the deep Q network (DQN) algorithm as an example, the intelligent agent of the model learns how to reasonably arrange the start order of tasks and allocate resources such as production equipment and manpower according to the category of the current task, the associated set information and the available status of resources, so as to maximize the reward function. The reward function includes multiple objectives such as on-time delivery rate, equipment utilization rate and production cost control. The reward function is designed with a weight of 0.5 for on-time delivery rate, a weight of 0.3 for equipment utilization rate and a weight of 0.2 for production cost control.

[0055] The task association set and the resource requirement information are input into the scheduling model, the execution order and resource allocation of the multiple tasks are planned, and an initial scheduling strategy is generated. The initial scheduling strategy is to give priority to the use of key equipment high-precision lathe for high-priority task 1, reasonably allocate skilled workers to task 1 with high technical requirements, and arrange tasks 2 and 3 to start in sequence after task 1 is completed.

[0056] The execution status information of the multiple tasks is obtained in real time. The equipment operation data, raw material consumption data, product quality inspection data and personnel operation data are collected through the Internet of Things sensor network in the factory workshop, and the execution status information is collected and sorted in real time. The equipment operation data shows that the high-precision lathe is operating normally, the raw material consumption data shows that 50 kg of steel has been consumed, the product quality inspection data shows that the processing accuracy of parts for Task 1 meets the requirements, and the personnel operation data shows that skilled workers have completed half of Task 1. The task progress is calculated by comparing the actual completion progress of the task with the planned completion progress. For example, Task 1 has completed half of the planned progress. The planned progress of Task 1 is 100%, the actual progress is 50%, and the progress deviation is -50%. The actual and planned usage of resources are counted to analyze the resource usage. For example, the planned usage time of the high-precision lathe is 8 hours, and the actual usage is 4 hours; the planned steel usage is 100kg, and the actual consumption is 50kg. By calculating the task progress deviation and resource usage deviation, the task execution deviation is detected. For example, a task is delayed due to equipment failure, and the subsequent tasks are suspended due to insufficient supply of a certain raw material. Task 1 is delayed for 2 hours due to equipment failure, and the progress deviation is -25%; the raw material steel supply is insufficient, causing Task 2 to be suspended waiting for raw materials to be replenished.

[0057] According to the task execution deviation, the causal inference model is used to evaluate the impact on other tasks and the overall goal. For example, the delay of a key task may lead to a chain delay of multiple subsequent dependent tasks, affecting the delivery time of the entire production batch and increasing production costs. The delay of Task 1 causes Task 2 and Task 3 to be unable to start on time, affecting the delivery time of the entire production batch, and it is estimated that the production cost will increase by 10%. According to the impact assessment results, the adjusted task information and resource information are re-entered into the scheduling model to generate an adjusted scheduling strategy. For example, in order to alleviate the impact of the delay of key tasks, idle equipment support is reallocated, the start sequence of subsequent tasks is adjusted, and raw material procurement and replenishment are prioritized. The adjusted scheduling strategy is to assign idle ordinary lathes to Task 2, adjust the start sequence of Task 2 in advance, and prioritize the purchase of raw materials to replenish steel to ensure that Task 2 and Task 3 can be started as soon as possible. The adjusted scheduling strategy is sent to the factory production execution system (MES) to execute the adjusted strategy to ensure multi-objective collaborative optimization. According to the adjusted scheduling strategy, the MES system notifies the equipment operator to adjust the equipment allocation and guides the logistics personnel to allocate raw materials to ensure that Task 2 and Task 3 can be smoothly promoted according to the adjusted strategy.

[0058] The electronic device includes a processor and a memory. The memory stores programs or instructions that can be run on the processor. When the program or instructions are executed by the processor, the steps of the above-mentioned multi-objective collaborative optimization scheduling method based on artificial intelligence are implemented. The processor can be the CPU of a computer, and the memory can be a hard disk or memory. The program or instructions contain codes corresponding to a task parsing module, a target association analysis module, a scheduling strategy generation module, a real-time monitoring module and a dynamic scheduling module. The processor completes the automated scheduling and optimization of common factory tasks by executing these codes.

[0059] A readable storage medium such as a CD, a USB flash drive or a hard disk stores a program or instruction thereon. When the program or instruction is executed by the processor, the steps of the multi-objective collaborative optimization scheduling method based on artificial intelligence are implemented. The program or instruction contains codes of algorithms such as natural language processing, machine learning, graph neural networks and reinforcement learning, as well as control logic codes of modules such as task parsing, target association analysis, scheduling strategy generation, real-time monitoring and dynamic scheduling. When these codes are loaded into the computer's memory and executed by the CPU, multi-objective collaborative optimization scheduling of factory tasks can be achieved.

[0060] The same or similar reference numerals correspond to the same or similar components; The terms used in the drawings to describe positional relationships are only used for illustrative purposes and should not be construed as limiting this patent; Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not limitations on the implementation methods of the present invention. For ordinary technicians in the relevant field, other different forms of changes or modifications can be made on the basis of the above description. It is not necessary and impossible to list all the implementation methods here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. The multi-objective collaborative optimization scheduling system based on artificial intelligence is characterized by: It includes task analysis module, target association analysis module, scheduling strategy generation module, real-time monitoring module and dynamic scheduling module; The task parsing module is used to obtain task information of multiple tasks, and automatically divide the tasks into multiple task categories according to the task information; the multiple task categories respectively contain multiple tasks; The target association analysis module is used to obtain target attributes of the multiple tasks respectively included in the multiple task categories, and analyze the association relationship between the multiple tasks according to the target attributes to obtain multiple task association sets; the number of the task association sets is related to the task category; The scheduling strategy generation module is used to obtain resource requirement information corresponding to the multiple tasks respectively, and plan the execution order and resource allocation of the multiple tasks according to the resource requirement information to generate a scheduling strategy; The real-time monitoring module is used to obtain the execution status information of the multiple tasks in real time, and analyze the execution status information to obtain task progress, resource usage and task execution deviation; The dynamic scheduling module is used to dynamically adjust the scheduling strategy according to the task progress, the resource usage and the task execution deviation to ensure multi-objective collaborative optimization.

2. The multi-objective collaborative optimization scheduling system based on artificial intelligence according to claim 1 is characterized in that: The task parsing module comprises: A task information acquisition unit, used for acquiring the task information of the plurality of tasks, specifically acquiring detailed descriptions, priorities and deadline task information of the plurality of tasks from the task management system through a data interface; A feature extraction unit is used to extract key features from the task information and generate a task feature vector, specifically using natural language processing technology and machine learning algorithms to extract key features from the task information and generate a task feature vector; The category division unit is used to perform cluster analysis on the task feature vector to obtain multiple task categories. Specifically, a K-Means clustering algorithm is used to perform cluster analysis on the task feature vector to determine the task category.

3. The multi-objective collaborative optimization scheduling system based on artificial intelligence according to claim 2 is characterized in that: The target association analysis module comprises: A target attribute acquisition unit, used to respectively acquire the target attributes of the plurality of tasks by using a data mining technique; An association relationship modeling unit, used to model the association relationship between the tasks according to the target attributes using a graph neural network to obtain a task association graph; The association set determining unit is used to determine multiple task association sets according to the task association graph, and specifically perform community discovery algorithm processing on the task association graph to obtain multiple task association sets.

4. The multi-objective collaborative optimization scheduling system based on artificial intelligence according to claim 3 is characterized in that: The scheduling strategy generation module includes: A resource demand acquisition unit, configured to acquire resource demand information corresponding to each of the plurality of tasks by using big data technology; the resource demand information includes manpower, material resources and time; A scheduling model building unit, used to build a scheduling model based on reinforcement learning; The strategy generating unit is used to input the task association set and the resource requirement information into the scheduling model to generate a scheduling strategy.

5. The multi-objective collaborative optimization scheduling system based on artificial intelligence according to claim 1 is characterized in that: The real-time monitoring module comprises: A status information acquisition unit, used for acquiring the execution status information of the plurality of tasks in real time; and also for collecting and collating the execution status information in real time; A progress analysis unit, configured to analyze the execution status information to obtain the task progress of the plurality of tasks, specifically by comparing the actual completion steps of the tasks with the planned completion steps to calculate the task progress; A resource analysis unit, used to analyze the execution status information to obtain the resource usage, specifically count the actual usage and planned usage of resources, and analyze the resource usage; The deviation detection unit is used to detect the task execution deviation by comparing with the preset task execution standard, and specifically detect the task execution deviation by calculating the task progress deviation and the resource usage deviation.

6. The multi-objective collaborative optimization scheduling system based on artificial intelligence according to claim 5 is characterized in that: The dynamic scheduling module includes: An impact assessment unit, for assessing the impact on other tasks and overall objectives according to the task execution deviation; A strategy adjustment unit, used to adjust the scheduling strategy according to the impact assessment result; The execution unit is used to execute the adjusted scheduling strategy.

7. The multi-objective collaborative optimization scheduling system based on artificial intelligence according to claim 6 is characterized in that: The impact assessment unit uses a causal inference model to assess the impact on other tasks and the overall goal based on the task association set and the task execution deviation; The strategy adjustment unit generates an adjusted scheduling strategy by re-inputting the adjusted task information and resource information into the scheduling model; The execution unit specifically sends the adjusted scheduling strategy to the task execution system for execution.

8. A multi-objective collaborative optimization scheduling method based on artificial intelligence, used to implement the multi-objective collaborative optimization scheduling system based on artificial intelligence according to any one of claims 1 to 7, characterized in that: The following steps are involved: Obtaining task information of multiple production tasks through the data interface of the factory production management system, wherein the task information includes a detailed description of the task, priority, deadline, required production equipment type, and process requirements; using natural language processing technology to extract key information in the task description, and using a machine learning algorithm to extract features of the task priority, deadline, and estimated production time to generate a task feature vector; using a K-Means clustering algorithm to perform cluster analysis on the task feature vector, and automatically dividing the task into multiple task categories, wherein the multiple task categories respectively contain multiple tasks, such as dividing the task into high-priority emergency tasks, routine production tasks, and quality inspection task categories; Acquire target attributes of the multiple tasks respectively included in the multiple task categories by using data mining technology, wherein the target attributes include production process nodes, quality requirements, and dependencies with upstream and downstream processes of the tasks; The graph neural network is used to model the association relationship between the tasks according to the target attributes to obtain a task association graph, such as establishing associations between multiple tasks that require the same batch of raw materials, and establishing strong associations between tasks that are closely connected to the previous and next processes; the task association graph is processed by a community discovery algorithm to obtain multiple task association sets, the number of which is related to the task category, and each task association set represents a set of tasks with strong associations, such as a certain association set containing a series of tasks from raw material processing to semi-finished product assembly; The resource demand information corresponding to the multiple tasks is obtained through the factory production big data platform, and the resource demand information includes the type and quantity of production equipment, the type and amount of raw materials, the manpower demand and skill requirements and the production time window; a scheduling model based on reinforcement learning is constructed, and the deep Q network algorithm is used as an example. The intelligent agent of the model learns how to reasonably arrange the start sequence of tasks, allocate production equipment and human resources according to the category of the current task, the association set information and the available status of resources, so as to maximize the reward function, and the reward function includes multiple objectives such as on-time delivery rate, equipment utilization rate and production cost control; the task association set and the resource demand information are input into the scheduling model, and the execution sequence and resource allocation of the multiple tasks are planned to generate an initial scheduling strategy, such as giving priority to the use of key equipment for high-priority tasks and reasonably allocating skilled workers to tasks with high technical requirements; Acquire the execution status information of the multiple tasks in real time, collect equipment operation data, raw material consumption data, product quality inspection data and personnel operation data through the IoT sensor network in the factory workshop, and collect and organize the execution status information in real time; calculate the task progress by comparing the actual completion progress of the task with the planned completion progress, such as a task has completed half of the planned progress; count the actual usage and planned usage of resources, and analyze the resource usage; detect task execution deviation by calculating task progress deviation and resource usage deviation, such as a task is delayed due to equipment failure, and a subsequent task is suspended due to insufficient supply of raw materials; According to the task execution deviation, the causal inference model is used to evaluate the impact on other tasks and the overall goal. The delay of a key task may lead to a chain delay of multiple subsequent dependent tasks, affecting the delivery time of the entire production batch and increasing production costs. According to the impact assessment results, the adjusted task information and resource information are re-entered into the scheduling model to generate an adjusted scheduling strategy. In order to alleviate the impact of delays in key tasks, idle equipment support is reallocated, the start sequence of subsequent tasks is adjusted, and raw material procurement and replenishment are prioritized. The adjusted scheduling strategy is sent to the factory production execution system and executed to ensure multi-objective collaborative optimization.

9. An electronic device, characterized in that: It includes a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the steps of the multi-objective collaborative optimization scheduling method based on artificial intelligence as claimed in claim 8 are implemented.

10. A readable storage medium, characterized in that: The readable storage medium stores programs or instructions, and when the programs or instructions are executed by the processor, the steps of the multi-objective collaborative optimization scheduling method based on artificial intelligence as claimed in claim 8 are implemented.

Images