An intelligent scheduling method for die casting resources of automobile parts in a cloud manufacturing environment

Abstract

The application provides a kind of cloud manufacturing environment under automobile parts die casting resource intelligent scheduling method, and the technical field is artificial intelligence and intelligent manufacturing.It is characterized by: the method combines the production characteristics of automobile parts die casting under cloud manufacturing environment, and carries out the task decomposition of automobile parts die casting order project and under cloud manufacturing environment, and classifies die casting resources;Establish a die casting resource scheduling model, determine the objective function and constraint condition of optimization scheduling;And using deep policy gradient (DDPG) algorithm to solve the die casting resource scheduling problem of automobile parts.The application is widely used in automobile parts die casting enterprises and vehicle manufacturing enterprises, can meet the real-time scheduling needs of automobile parts die casting resources, and can effectively solve the difficulty of automobile parts die casting resource scheduling problem under current cloud manufacturing environment.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence and intelligent manufacturing technology, specifically relating to an intelligent scheduling method for die-casting resources of automotive parts in a cloud manufacturing environment. Background Technology

[0002] With the integration of new technologies such as cloud computing, the Internet of Things, and big data with advanced manufacturing, the production methods of the manufacturing industry are developing towards networking and service orientation. Against this backdrop, Academician Li Bohu's team proposed "cloud manufacturing" in 2010, aiming to integrate various discrete manufacturing resources into virtualized service units for centralized management and operation via the network. Scheduling problems in the cloud manufacturing environment are key to production planning and control. With the development of artificial intelligence, scholars at home and abroad have focused on machine learning methods and applied them to resource scheduling problems in the cloud manufacturing environment, achieving significant results. Die casting, due to its high efficiency and precision, has become an important manufacturing method in the automotive industry. Automotive die castings are diverse, mainly classified by material as aluminum alloy, magnesium alloy, and zinc alloy die castings, primarily involving power systems, chassis, body and structural parts, interior, and electrical systems. In terms of product categories, die castings also include shells with high airtightness requirements, load-bearing structural parts with high strength requirements, large thin-walled parts with very strict forming processes, and special components with high thermal conductivity / high temperature resistance. This diversity results in the high complexity and strong constraints of die casting production itself. To improve the efficiency of die-casting machines and reduce capacity loss, efficient and rational scheduling and optimization control are crucial. Current research on die-casting resource scheduling largely focuses on single production lines or individual factories, and the solution methods mostly employ traditional heuristic algorithms. However, the scheduling of automotive parts die-casting resources in a cloud manufacturing environment is characterized by numerous resources, diverse products, significant process differences, complex constraints, and high production line flexibility requirements, posing greater challenges compared to other industries. Existing scheduling methods are insufficient to handle such a complex production environment. Therefore, a new method is urgently needed to achieve rapid response and adaptive scheduling for automotive parts die-casting production in a cloud manufacturing environment. Summary of the Invention

[0003] To address the challenge of existing scheduling methods being unsuitable for the personalized scenarios of intelligent scheduling of automotive parts die-casting resources in a cloud manufacturing environment, this invention fully considers the characteristics of the die-casting process and provides an intelligent scheduling method for automotive parts die-casting resources in a cloud manufacturing environment. The technical solution of this invention is as follows:

[0004] Step S10: The die-casting resource scheduling problem within a single workshop is reduced to a mixed flow workshop scheduling problem, based on this design assumption.

[0005] Step S20: Based on the characteristics of automotive parts die casting in the cloud manufacturing environment and the information uploaded to the cloud platform by resource demanders, decompose automotive parts die casting order projects and tasks in the cloud manufacturing environment;

[0006] Step S30: Based on the information uploaded to the cloud platform by the resource provider, classify the die-casting resources of automotive parts in the cloud manufacturing environment;

[0007] Step S40: Based on the die casting order project and task decomposition in step S20 and the die casting resource classification in step S30, establish an automotive parts die casting resource scheduling model in the cloud manufacturing environment, and determine the objective function and constraints for optimal scheduling.

[0008] Step S50: Establish a deep reinforcement learning model based on DDPG, determine the state space, action space and reward function of the agent, construct the network structure of the agent, and train the DDPG model through the agent's dynamic interaction mechanism;

[0009] Step S60: Based on the proposed assumptions and the constructed die-casting resource scheduling model, the trained DDPG model is used to optimize the scheduling of die-casting resources for automotive parts.

[0010] Furthermore, in step S10, the specific assumptions are as follows:

[0011] 1) All die-casting resources are available in the initial stage.

[0012] 2) A resource can execute subtasks under different order tasks, but it cannot execute multiple subtasks at the same time. The next subtask can only be executed after the current subtask is completed.

[0013] 3) Subtasks are the smallest units that cannot be further decomposed in die casting resource scheduling.

[0014] 4) There can be waiting time between subtasks, and die-casting resources can also accommodate a certain period of idle time.

[0015] 5) Different subtasks under the same order task are executed according to the die-casting process and cannot be executed in parallel.

[0016] 6) Subtasks under the same priority level have the same scheduling priority.

[0017] 7) Different order tasks under the same order project can be assigned to different die casting companies for processing. This will result in outbound logistics time and logistics costs between the company and the customer. Outbound logistics time depends on transportation distance and transportation speed, while outbound logistics costs depend on unit distance transportation cost and transportation distance.

[0018] Furthermore, in step S20, the breakdown of automotive parts die-casting order projects and tasks in the cloud manufacturing environment is as follows:

[0019] Based on the characteristics of automotive parts die-casting manufacturing in a cloud manufacturing environment, die-cast parts are divided into powertrain and chassis structural parts, body structural parts, and functional parts. An order project set is defined. have One order item, of which order items Include Different types of die castings Define the classification function Determine the category of each part:

[0020]

[0021] in, , , For category identifiers; .

[0022] Based on the classification results, order items It is broken down into three independent resource pool task packages, each containing several order tasks, namely:

[0023]

[0024]

[0025] Furthermore, resource pool task packages Order tasks within The expression is as follows:

[0026]

[0027] in, For order item identifier; For order task identifier; The order task structure includes the specific processes required for the order task and the number of subtasks; The required metrics for order tasks, among which Indicates quality requirements. Indicate cost requirements, Indicates the project timeline requirements; Total workload for the order task; The current geographical location of the order task; For already scheduled subtasks; This represents the priority level of the order task.

[0028] Furthermore, the order task The process flow is broken down into sub-tasks corresponding to each step:

[0029] .

[0030] Furthermore, in step S30, the die-casting resources for automotive parts in the cloud manufacturing environment are categorized as follows:

[0031] Based on the three categories of automotive die-cast parts, three virtual die-casting resource pools are defined:

[0032] Powertrain and chassis structural component resource pool This corresponds to the set of die-casting resources for producing powertrain and chassis structural components; and the resource pool for body structural components. This corresponds to the set of die-casting resources for producing body structural parts; the functional component resource pool. This corresponds to the set of die-casting resources used in the production of functional components; Inside Die casting resources.

[0033] The complete resource set consisting of all die casting resources The union of the three resource pools mentioned above is:

[0034]

[0035] in, For die casting resource identifiers, , The quantity of die-casting resources within a single enterprise; This is an identifier for die-casting companies. The existence of overlap between the various resource pools indicates that some die-casting resources have the flexibility to serve multiple types of order tasks.

[0036] Based on the types of parts that can be processed from die-casting resources, a classification function is defined. Used to transfer die-casting resources Assigned to one or more resource pools, the specific classification criteria are as follows:

[0037] Conditions:

[0038] in, For the vacuum die-casting capability of resources, , This indicates that the resource lacks vacuum die-casting capability. This indicates that resources can be provided for vacuum die casting as needed. This indicates that the resource must have vacuum die-casting capability; For the heat treatment capacity of resources, , This indicates that the resource has no heat treatment capability. This indicates that resources can be provided for heat treatment as needed. This indicates that the resource must have heat treatment capabilities; The collection of materials that can be used from resources; This refers to the set of materials required for powertrain and chassis structural components. For the dimensions of die-cast parts that can be machined from resources; This refers to the size range required for die-casting resources for powertrain and chassis structural components.

[0039] Conditions:

[0040] in, The set of materials required for vehicle body structural components; The minimum machining dimensions required for die-casting of vehicle body structural components; For the ability to precisely control resources; This refers to the minimum precision control capability required for die-casting resources of vehicle body structural components.

[0041] Conditions:

[0042] in, The set of materials required for functional components; The minimum machining dimension required for die-casting of functional components; This refers to the minimum precision control capability required for die-casting resources of functional components.

[0043] Furthermore, in step S40, the specific scheduling model for automotive parts die-casting resources in the cloud manufacturing environment is as follows:

[0044] The objective function of the scheduling model is:

[0045]

[0046]

[0047]

[0048]

[0049]

[0050]

[0051]

[0052]

[0053] in, The optimization objective is to optimize the global scheduling. to These are the weighting coefficients; To improve the average quality of die casting manufacturing services; This is the subtask identifier; For scheduled subtasks A collection arranged in the order of processes; For a binary decision variable, if The Middle Subtasks are assigned to The first in the enterprise For each resource processing, ,otherwise ; The service quality that die-casting resources can achieve; Number of subtasks; Number of times the service type for die casting resources can be switched; For a binary decision variable, if the first... In the first order task The type of die-casting resource service required for each sub-task With the In the first order task The type of die-casting resource service required for each sub-task If they are the same, then ,otherwise ; This represents the maximum completion time for die casting. For order tasks The completion time of the last subtask in the process; This refers to the time for logistics and transportation from the factory to the destination. For transportation distance; For transportation speed; Cost of die casting services; Processing cost for subtasks; This refers to the unit distance transportation cost for outbound logistics. The unit time processing cost of die-casting resources; The workload of subtasks; To improve the processing speed of die-casting resources; Cost per unit distance for AGV delivery; Indicates the first The workpiece for each sub-task is delivered to the... The required transportation distance for each resource.

[0054] The scheduling model has the following constraints:

[0055] Die casting resource constraints: for each type of workpiece have Each process can only be completed by one die-casting resource; each order task For the processing of a specific workpiece, each subtask This corresponds to a single processing step.

[0056] Process sequence constraints: The execution of subtasks must strictly follow the inherent process route of automotive parts die casting; for the same workpiece consecutive subtasks and The time requirement is as follows:

[0057]

[0058] in, Subtasks Completion time on the allocated resources; Subtasks Start time; For subtasks arrive The necessary transition time mainly includes the time for switching between die-casting resource service types and the time for AGV transportation within the plant.

[0059] Schedule constraints: for any order task Maximum completion time The deadline for the order must not be exceeded. ,Right now: .

[0060] Service quality constraints: Execute order tasks Average service quality The service quality must not be lower than the minimum service quality required for the order. ,Right now: .

[0061] Service cost constraint: completing the order task Die casting service cost The cost must not exceed the cost requirement of the order. ,Right now: .

[0062] Furthermore, in step S50, the representation method of the agent's state space is as follows:

[0063] state space of an agent It consists of the state of each die-casting resource, its corresponding sub-task, and the corresponding logistics resources. Among them, Represents the resource state matrix, Represents the subtask state matrix. This represents the logistics state matrix, and these matrices together constitute the agent's perception of the environment. Resource state matrix. It provides a comprehensive overview of currently available resources and a subtask state matrix. Detailed records of subtask information and logistics status matrix were provided. It covers various types of data related to logistics.

[0064] Furthermore, in step S50, the representation method of the agent's action space is as follows:

[0065] 1) Action Type

[0066] The agent's actions are reflected in resource scheduling decisions, particularly in selecting a suitable resource to serve eligible subtasks. This process considers factors such as the subtask's priority, resource availability, processing speed, processing cost, service quality, and logistical status. The agent can determine the next pending order and its subtasks, and can also dynamically adjust its actions based on the urgency of the order and the required resources.

[0067] 2) Status input and action output

[0068] The agent selects the optimal action based on the current environmental state. At each decision-making moment, the agent observes the environmental state matrix, including the resource state matrix, sub-task state matrix, and logistics state matrix. Then, the agent evaluates the value of different actions through its network model and selects the action that maximizes the expected value.

[0069] Furthermore, in step S50, the agent's reward function is as follows:

[0070] Design a hybrid reward function that includes step-by-step rewards and a final reward based on the objective function in the mathematical model.

[0071] Step-by-step reward function :

[0072]

[0073] in, to These are the weighting coefficients for each indicator; The waiting time for a subtask from its schedulable time to its current assigned time; If the scheduling violates the constraint indicator function, then ,otherwise .

[0074] Design the final reward function based on the complete scheduling results. :

[0075]

[0076] in, to Corresponding weight coefficients in the scheduling model; A penalty weight for constraint violation is specifically introduced to ensure that the agent prioritizes satisfying all constraints. This is a global constraint violation indicator function; if the entire scheduling scheme violates any global constraints, then... ,otherwise .

[0077] The total reward is obtained by summing the discounted amounts of the step-by-step rewards and adding the final reward. :

[0078]

[0079] in, This is a discount factor used to weigh the importance of near-term rewards against long-term rewards; This represents the total number of steps in a scheduling cycle.

[0080] Furthermore, in step S50, the specific neural network structure of the MLP-based deep reinforcement learning model is as follows:

[0081] The agent's network structure comprises two neural network models: an Actor network and a Critic network. The Actor network updates the policy based on the expected value, while the Critic network outputs the current value to the Actor network for calculating the expected reward. Environment state. From matrix group Characterization:

[0082]

[0083] in, This is a matrix flattening function that converts a two-dimensional state matrix into a one-dimensional input vector acceptable to the agent network. This is a vector concatenation function used to integrate state information from different sources.

[0084] Actor network receives Output action probability distribution :

[0085]

[0086]

[0087]

[0088]

[0089]

[0090] in, to This represents the activation vector of each hidden layer in the Actor network; For activation function, ; To activate the function, ; to These are the weights of the Actor network; to For the bias of the Actor network; This is the output of the last linear layer; It's an action mask, at the time step. Dynamically generated based on constraints, infeasible resource allocation actions are marked as... This ensures that the agent explores only within the feasible solution space.

[0091] The forward propagation process of the Critic network is represented as follows:

[0092]

[0093]

[0094]

[0095]

[0096] to This represents the activation vector of each hidden layer in the Critic network; to These are the weights of the Critic network; to The bias of the Critic network; The expected discounted cumulative return that can be obtained by following the current strategy.

[0097] Furthermore, in step S50, the dynamic interaction mechanism of the intelligent agent is as follows:

[0098] Step S501, Order item decomposition: Refresh the die casting order item set, select order items one by one according to priority, and decompose the order items into three independent resource pool task packages based on the three categories of automotive die casting parts, until all order items are traversed.

[0099] Step S502, Order task decomposition: Select each order task in each resource pool task package according to priority, and decompose it into several sub-tasks according to its inherent process flow, until all order tasks are traversed.

[0100] Step S503, Die-casting resource scheduling: Match resources in the virtual die-casting resource pool to sub-tasks according to priority, generate a scheduling plan, and continue until all sub-tasks and die-casting resources are traversed;

[0101] Step S504: Refresh the status of all die-casting resources and update the system time;

[0102] Step S505: Determine whether all order items are completed. If not, return to step S501 and repeat. Attached Figure Description

[0103] Figure 1 This is an example of an automotive parts die-casting order project and task breakdown in a cloud manufacturing environment;

[0104] Figure 2 This is an example of resource classification for die-casting of automotive parts in a cloud manufacturing environment;

[0105] Figure 3 This is a flowchart of the resource scheduling process for die-casting of automotive parts in a cloud manufacturing environment;

[0106] Figure 4 Here is an example of the state space of an intelligent agent;

[0107] Figure 5 It is a comparison of the training performance of reinforcement learning algorithm models with different depths;

[0108] Figure 6 This is a comparison of the solution performance of reinforcement learning algorithm models with different depths. Detailed Implementation

[0109] The present invention will now be described in further detail with reference to the accompanying drawings. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0110] Example 1: This invention proposes an intelligent scheduling method for die-casting resources of automotive parts in a cloud manufacturing environment. The specific steps include:

[0111] Step S10: The customer publishes the order project to the cloud platform. The intelligent agent breaks down the order project into multiple order tasks based on the types of parts required by the customer and categorizes them into different resource pool task packages, i.e. , Each order task can be broken down into multiple sub-tasks based on the processing technology of the die-cast parts, i.e. The cloud platform selects a resource pool and matches corresponding die-casting resources for each subtask based on the characteristics and status of resources in the three virtual die-casting resource pools. Each subtask corresponds to one resource. Finally, the corresponding scheduling plan is distributed to the relevant resource-owning enterprises for processing, while collecting relevant production data for updating the virtual die-casting resource pools and training the agent. Examples of die-casting order project and task decomposition, and die-casting resource classification are shown below. Figure 1 , Figure 2 As shown, the resource scheduling process for automotive parts die casting in a cloud manufacturing environment is as follows: Figure 3 As shown.

[0112] Step S20: Based on step S10, maximize the average die-casting manufacturing service quality. Minimize the number of times die casting resource service types are switched. Minimize the maximum completion time of die casting and minimize die casting service costs To optimize the die casting resource scheduling model, a die casting resource scheduling model was constructed.

[0113] Step S30: Construct the agent's state space. Through the resource state matrix, the agent can understand resource availability and limitations, and make reasonable resource allocation decisions. In the resource state matrix... middle, The characteristics of each die-casting resource reflect any time step Real-time status of each die-casting resource. Indicates the resource's operating status. Indicates the location of the resource. Indicates the next available time of the resource. For processing speed, The processing cost per unit time. For service quality, Assuming it's a service type. Based on the characteristics of die-casting resources, construct a resource state matrix, such as... Figure 4 As shown in Figure a, the number of rows in the matrix represents the number of features of the die-casting resources, and the number of columns in the matrix represents the quantity of the die-casting resources.

[0114] The subtask state matrix helps the agent understand the urgency and complexity of the current order task. (The subtask state matrix...) middle, The characteristics of each subtask reflect any time step The real-time status of each subtask. Indicates priority. Indicates quality requirements. Indicate cost requirements, Indicates the construction period requirements. Indicates the workload of the subtask. Indicates the type of die-casting resource service required by the subtask. Indicates the current state of the subtask. The current system time. This represents the cumulative cost percentage. Based on the characteristics of the subtasks, a subtask state matrix is ​​constructed, such as... Figure 4 As shown in b. The number of rows in the matrix represents the number of features of the subtasks, and the number of columns in the matrix represents the number of subtasks.

[0115] The logistics state matrix is ​​of great significance for agents to evaluate logistics efficiency and optimize supply chain management. middle, The characteristics of logistics reflect any time step Real-time status of logistics. Indicates the type of logistics transportation from the factory. Indicates the distance of logistics transportation from the factory. Indicates the speed of logistics transportation from the factory. This indicates the unit distance transportation cost in the outbound logistics process. Indicates the status of AGVs in the factory's logistics system. Indicates the transportation distance of AGVs in the factory logistics. This indicates the unit distance delivery cost of AGVs in the factory's logistics system. This indicates the delivery speed of AGVs in the factory's logistics. Based on logistics characteristics, a logistics status matrix is ​​constructed, such as... Figure 4 As shown in c.

[0116] Step S40: Import the die casting production data of a certain automotive parts into the die casting resource scheduling model. This case dataset contains 47 die casting resources from 20 companies; a total of 63 order tasks and 315 sub-tasks were completed.

[0117] Step S50: Perform model training. To verify the advantages of the DDPG model in training, it is compared with AC, DQN, and DDQN models. The training results are as follows: Figure 5 As shown, with increasing iteration count, the performance of the DDQN model gradually falls below that of the AC and DQN models, while the DQN model consistently outperforms the AC model and exhibits a more stable training trajectory. Compared to the other three models, the DDPG model demonstrates better convergence in its training trajectory. The model training parameters were set as follows: 10,000 iterations, a discount factor of 0.7, and an Actor network learning rate of 10. -5 The Critic network has a learning rate of 5×10⁻⁶. -4 The hidden layer contains 512 neurons; the batch size is 512; and the soft update parameter is 0.006.

[0118] Step S60: Perform model testing. To verify the solution performance of the DDPG model, the solution reward and solution time of the DDPG, AC, DDQN, and DQN models are compared. The results are as follows: Figure 6 As shown in the figure. The results demonstrate that the intelligent scheduling method proposed in this invention can effectively describe the state of die-casting resources and order tasks, exhibits good solution performance, and can provide a scheduling scheme that meets the requirements. This aligns with the practical requirements of intelligent scheduling of automotive parts die-casting resources in a cloud manufacturing environment.

Claims

1. A method for intelligent scheduling of die-casting resources for automotive parts in a cloud manufacturing environment, characterized in that, The steps of this method are as follows: Step S10: The die-casting resource scheduling problem within a single workshop is reduced to a mixed flow workshop scheduling problem, based on this design assumption. Step S20: Based on the characteristics of automotive parts die casting in the cloud manufacturing environment and the information uploaded to the cloud platform by resource demanders, decompose automotive parts die casting order projects and tasks in the cloud manufacturing environment; Step S30: Based on the information uploaded to the cloud platform by the resource provider, classify the die-casting resources of automotive parts in the cloud manufacturing environment; Step S40: Based on the die casting order project and task decomposition in step S20 and the die casting resource classification in step S30, establish an automotive parts die casting resource scheduling model in the cloud manufacturing environment, and determine the objective function and constraints for optimizing the scheduling. Step S50: Establish a deep reinforcement learning model based on DDPG, determine the state space, action space and reward function of the agent, construct the network structure of the agent, and train the DDPG model through the agent's dynamic interaction mechanism; Step S60: Based on the proposed assumptions and the constructed die-casting resource scheduling model, the trained DDPG model is used to optimize the scheduling of die-casting resources for automotive parts. Furthermore, in step S20, the breakdown of automotive parts die-casting order projects and tasks in the cloud manufacturing environment is as follows: Based on the characteristics of automotive parts die casting manufacturing in a cloud manufacturing environment, die castings are divided into powertrain and chassis structural parts, body structural parts, and functional parts; an order project set is set up. have One order item, of which order items Include Different types of die castings Define the classification function Determine the category of each part: ； in, , , For category identifiers; ; Based on the classification results, order items It is broken down into three independent resource pool task packages, each containing several order tasks, namely: ； ； Resource Pool Task Pack Order tasks within The expression is as follows: ； in, For order item identifier; For order task identifier; The order task structure includes the specific processes required for the order task and the number of subtasks; The required metrics for order tasks, among which Indicates quality requirements. Indicate cost requirements, Indicates the project timeline requirements; Total workload for the order task; The current geographical location of the order task; For already scheduled subtasks; The priority level number to which the order task belongs; Order task The process flow is broken down into sub-tasks corresponding to each step: ； Furthermore, in step S50, the dynamic interaction mechanism of the intelligent agent is as follows: Step S501, Order item decomposition: Refresh the die casting order item set, select order items one by one according to priority, and decompose the order items into three independent resource pool task packages based on the three categories of automotive die casting parts, until all order items are traversed. Step S502, Order task decomposition: Select each order task in each resource pool task package according to priority, and decompose it into several sub-tasks according to its inherent process flow, until all order tasks are traversed. Step S503, Die-casting resource scheduling: Match resources in the virtual die-casting resource pool to sub-tasks according to priority, generate a scheduling plan, and continue until all sub-tasks and die-casting resources are traversed; Step S504: Refresh the status of all die-casting resources and update the system time; Step S505: Determine whether all order items are completed. If not, return to step S501 and repeat.

2. The intelligent scheduling method for die-casting resources of automotive parts in a cloud manufacturing environment according to claim 1, characterized in that, The specific assumptions made in step S10 are as follows: 1) All die-casting resources are available in the initial stage; 2) A resource can execute subtasks under different order tasks, but it cannot execute multiple subtasks at the same time. The next subtask can only be executed after the current subtask is completed. 3) The subtask is the smallest unit that cannot be further decomposed in the die-casting resource scheduling; 4) There can be waiting time between subtasks, and die-casting resources can also accommodate a certain period of idle time; 5) Different subtasks under the same order task are executed according to the die-casting process and cannot be executed in parallel; 6) Subtasks at the same priority level have the same scheduling priority; 7) Different order tasks under the same order project can be assigned to different die casting companies for processing. This will result in outbound logistics time and logistics costs between the company and the customer. Outbound logistics time depends on transportation distance and transportation speed, while outbound logistics costs depend on unit distance transportation cost and transportation distance.

3. The intelligent scheduling method for die-casting resources of automotive parts in a cloud manufacturing environment according to claim 1, characterized in that, In step S30, the classification of automotive parts die-casting resources in the cloud manufacturing environment is as follows: Based on the three categories of automotive die-cast parts, three virtual die-casting resource pools are defined: Powertrain and chassis structural component resource pool This corresponds to the set of die-casting resources for producing powertrain and chassis structural components; and the resource pool for body structural components. This corresponds to the set of die-casting resources for producing body structural parts; the functional component resource pool. This corresponds to the set of die-casting resources used in the production of functional components; Inside One die-casting resource; The complete resource set consisting of all die casting resources The union of the three resource pools mentioned above is: ； in, For die casting resource identifiers, , The quantity of die-casting resources within a single enterprise; This is an identifier for die-casting companies. The existence of overlap between the various resource pools indicates that some die-casting resources have the flexibility to serve multiple types of order tasks. Based on the types of parts that can be processed from die-casting resources, a classification function is defined. Used to transfer die-casting resources Assigned to one or more resource pools, the specific classification criteria are as follows: Conditions: ; in, For the vacuum die-casting capability of resources, , This indicates that the resource lacks vacuum die-casting capability. This indicates that resources can be provided for vacuum die casting as needed. This indicates that the resource must have vacuum die-casting capability; For the heat treatment capacity of resources, , This indicates that the resource has no heat treatment capability. This indicates that resources can be provided for heat treatment as needed. This indicates that the resource must have heat treatment capabilities; The collection of materials that can be used from resources; This refers to the set of materials required for powertrain and chassis structural components. For the dimensions of die-cast parts that can be machined from resources; The size range required for die-casting resources of powertrain and chassis structural components; Conditions: ; in, The set of materials required for vehicle body structural components; The minimum machining dimensions required for die-casting of vehicle body structural components; For the ability to precisely control resources; Minimum precision control capability required for die-casting resources of vehicle body structural components; Conditions: ; in, The set of materials required for functional components; The minimum machining dimension required for die-casting of functional components; This refers to the minimum precision control capability required for die-casting resources of functional components.

4. The intelligent scheduling method for die-casting resources of automotive parts in a cloud manufacturing environment according to claim 1, characterized in that, The objective function of the automotive parts die-casting resource scheduling model in the cloud manufacturing environment (step S40) is as follows: ； ； ； ； ； ； ； ； in, The optimization objective is to optimize the global scheduling. to These are the weighting coefficients; To improve the average quality of die casting manufacturing services; This is the subtask identifier; For scheduled subtasks A collection arranged in the order of processes; Given a binary decision variable, if The Middle Subtasks are assigned to The first in the enterprise For each resource processing, ,otherwise ; The service quality that die-casting resources can achieve; Number of subtasks; Number of times the service type for die casting resources can be switched; For a binary decision variable, if the first... In the first order task The type of die-casting resource service required for each sub-task With the In the first order task The type of die-casting resource service required for each sub-task If they are the same, then ,otherwise ; This represents the maximum completion time for die casting. For order tasks The completion time of the last subtask in the process; This refers to the time for logistics and transportation from the factory to the destination. For transportation distance; For transportation speed; Cost of die casting services; Processing cost for subtasks; This refers to the unit distance transportation cost for outbound logistics. The unit time processing cost of die-casting resources; The workload of subtasks; To improve the processing speed of die-casting resources; Cost per unit distance for AGV delivery; Indicates the first The workpiece for each sub-task is delivered to the... The required transportation distance for each resource.

5. The intelligent scheduling method for die-casting resources of automotive parts in a cloud manufacturing environment according to claim 1, characterized in that, The specific constraints of the automotive parts die-casting resource scheduling model in step S40 of the cloud manufacturing environment are as follows: Die casting resource constraints: for each type of workpiece have Each process can only be completed by one die-casting resource; each order task For the processing of a specific workpiece, each subtask Processing corresponding to one step; Process sequence constraints: The execution of subtasks must strictly follow the inherent process route of automotive parts die casting; for the same workpiece consecutive subtasks and The time requirement is as follows: ； in, Subtasks Completion time on the allocated resources; Subtasks Start time; For subtasks arrive The necessary transition time mainly includes the time for switching between die-casting resource service types and the time for AGV transportation within the plant; Schedule constraints: for any order task Maximum completion time The deadline for the task must not exceed the specified timeframe. ,Right now: ; Service quality constraints: Execute order tasks Average service quality The service quality must not be lower than the minimum service quality required for the order. ,Right now: ; Service cost constraint: completing the order task Die casting service cost The cost must not exceed the cost requirement of the order. ,Right now: .

Images