Multi-objective production planning optimization method and system for digital factories

Abstract

The present invention relates to the technical field of digital factory production management, and discloses a multi-objective production planning optimization method and system for digital factories. The method uses a production task decomposition module to decompose enterprise orders into multiple subtasks and extract relevant information; a resource dynamic assessment module collects data in real time to generate a resource status matrix; a multi-objective optimization algorithm module constructs a model and uses an improved non-dominated sorting genetic algorithm to solve it; a dynamic priority allocation module adjusts subtask priorities based on real-time data; and a conflict resolution unit resolves resource allocation conflicts. This method can accurately decompose tasks, evaluate resources in real time, collaboratively optimize multiple objectives, flexibly respond to anomalies, and efficiently resolve conflicts, comprehensively improving the production efficiency and benefits of digital factories and effectively resolving complex issues in the production planning and optimization process of digital factories.

Description

Technical Field

[0001] The present invention relates to the technical field of digital factory production management, and in particular to a multi-objective production plan optimization method and system for digital factories. Background Art

[0002] With the digital transformation of the manufacturing industry, digital factories have demonstrated significant advantages in production efficiency and quality control, but they also face the complex challenge of optimizing production plans. Under traditional production models, production planning often relies on empirical experience or simple calculations, which are difficult to adapt to the complex and ever-changing production environment of digital factories. First, the diversity and complexity of production tasks have increased. The increasing variety of business orders and demand for personalized product customization have made the decomposition and scheduling of production tasks extremely complex. For example, in automobile manufacturing, different models have diverse configurations, and component production tasks are intertwined. Failure to accurately decompose tasks and consider process constraints can easily lead to production disruptions and extended production cycles.

[0003] Secondly, resource management becomes more difficult. Digital factories feature a wide variety of production equipment, dynamically changing material inventories, and complex human resource structures. Equipment operating conditions are unstable and subject to malfunction at any time. Material supply is affected by market fluctuations, making precise inventory control difficult. Human resource skills and work efficiency vary widely, and there are restrictions on working hours. For example, if an electronics manufacturer fails to adjust production plans promptly due to sudden equipment failures during production, this can lead to material backlogs and order delays. Furthermore, failure to properly allocate human resources can result in some employees being overloaded and others being idle, reducing overall production efficiency.

[0004] Furthermore, the problem of multi-objective conflicts in the production process is prominent. Companies typically pursue multiple production goals, such as shortening production cycles, reducing energy costs, and improving equipment utilization, but these goals often constrain each other. In chemical production, increasing equipment speed to shorten production cycles can lead to significant increases in energy consumption, as well as uneven equipment loads and accelerated equipment wear and tear.

[0005] Furthermore, traditional production planning optimization methods struggle to adapt to real-time changes in the production process. When production anomalies occur, such as raw material quality issues or urgent order changes, traditional methods are unable to quickly adjust production plans and lack dynamic adaptability. In the apparel manufacturing industry, if fabric quality fails, traditional production plans struggle to adapt quickly and effectively, leading to production stagnation and increased costs.

[0006] Existing production planning optimization technologies face numerous shortcomings when addressing these challenges of digital factories. Some methods focus solely on single-objective optimization and fail to comprehensively balance multiple production objectives. Some algorithms exhibit high computational complexity and low solution efficiency, making them inefficient in meeting the real-time demands of production. Furthermore, some methods fail to monitor and respond promptly to dynamic resource changes, leading to a disconnect between production plans and actual conditions. Summary of the Invention

[0007] The purpose of the present invention is to provide a multi-objective production planning optimization method and system for digital factories to solve the problems raised in the above background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a multi-objective production planning optimization system for a digital factory, the system comprising:

[0009] The production task decomposition module is used to receive enterprise order data and decompose it into multiple independent production subtasks. Each subtask contains a task number i. , k is a positive integer greater than 2, and the process constraint conditions and resource requirement type set of each subtask are extracted;

[0010] The resource dynamic assessment module is used to collect real-time operating status data of production equipment, material inventory characteristic values, and human resource availability parameters in the digital factory to generate a real-time resource status matrix;

[0011] Multi-objective optimization algorithm module, which is used to build a multi-objective optimization model based on the decomposed subtask set and resource status matrix. The multi-objectives include minimizing the production cycle, minimizing energy consumption costs, and maximizing equipment load balance;

[0012] Dynamic priority allocation module, which is used to calculate the dynamic priority weight of each subtask based on the real-time updated production abnormal event data and equipment failure rate prediction value, and adjust the subtask execution sequence;

[0013] Conflict resolution unit, used to detect resource allocation conflicts and reset the time windows of conflicting tasks using a relaxation constraint algorithm;

[0014] The method for constructing the multi-objective optimization model is:

[0015] Define the decision variables as the equipment allocation plan and time scheduling sequence of subtasks;

[0016] Establish a set of objective functions, including total production cycle function, total energy consumption function and equipment load variance function;

[0017] Introducing process constraints, resource capacity limitations, and task dependencies as constraints;

[0018] An improved non-dominated sorting genetic algorithm is used to iteratively solve the multi-objective optimization model and generate a non-dominated solution set.

[0019] Preferably, the specific method for generating the real-time resource status matrix in the resource dynamic evaluation module is:

[0020] Collect real-time idle time windows, remaining maintenance cycles, and energy efficiency parameters of production equipment to build equipment resource vectors;

[0021] Collect material inventory, replenishment cycle, and quality attenuation coefficient to construct a material resource vector;

[0022] Collect personnel skills matching, working time limit and task processing efficiency to build human resource vector;

[0023] Equipment, material and human resource vectors are aggregated by time slices to form a three-dimensional resource status matrix.

[0024] Preferably, the specific optimization steps of the improved non-dominated sorting genetic algorithm include:

[0025] When initializing the population, a heuristic rule based on process constraints is used to generate feasible solutions;

[0026] Introducing the device affinity factor in the crossover operation, giving priority to the allocation scheme with a high historical frequency of device collaboration for gene exchange;

[0027] Dynamic mutation probability is used in the mutation operation, and the mutation intensity is adaptively adjusted according to the degree of population convergence;

[0028] The entropy weight method is used to assign weights to the non-dominated solution set and screen the Pareto frontier optimal solution.

[0029] Preferably, the specific method for calculating the dynamic priority weight in the dynamic priority allocation module is:

[0030] Get the basic priority coefficient of each subtask from the local database ;

[0031] Real-time collection of equipment failure prediction values and order urgency parameters , j represents the device number;

[0032] Calculate the real-time priority weight of subtask i :

[0033]

[0034] in represents the dependency of subtask i on equipment j, and m represents the total number of production equipment in the digital factory.

[0035] Preferably, the specific steps of adopting the relaxed constraint algorithm in the conflict resolution unit include:

[0036] Identify conflicting task groups with overlapping resource occupancy and extract the process paths and time windows of the conflicting tasks;

[0037] Calculate the upper limit of delay time and cost penalty coefficient of each task;

[0038] A conflict resolution objective function is constructed, with minimizing the total delay penalty as the optimization direction, and the branch and bound algorithm is used to solve the optimal time window adjustment solution.

[0039] Preferably, the calculation method for maximizing the device load balancing degree is:

[0040] Statistics of the total working time of each device in the scheduling cycle and idle time ;

[0041] Calculate device load balancing :

[0042]

[0043] in, is the average working time of the equipment, is the average idle time of the device, and the denominator is the load difference value normalized by the total time dimension.

[0044] Preferably, the method for extracting the material inventory characteristic value includes:

[0045] Obtain the batch code and storage time stamp of the material through the RFID sensor;

[0046] Combined with environmental temperature and humidity monitoring data, calculate the real-time quality decay rate of materials :

[0047]

[0048] in is the reference attenuation coefficient, For real-time humidity, For ideal humidity, is the adjustment parameter.

[0049] Preferably, the digital method of the process constraints is:

[0050] Analyze the process sequence and equipment dependency in CAD process drawings and generate a directed acyclic graph;

[0051] Map the nodes in the graph to subtasks, and the edge weights represent the time cost of task switching;

[0052] The initial feasible scheduling sequence is generated by the topological sorting algorithm.

[0053] Preferably, the optimization method for minimizing energy consumption cost includes:

[0054] Collect the power curve of the equipment at different load rates and fit the segmented line performance consumption function;

[0055] Introducing time-of-use electricity price parameters to convert energy consumption costs into time-dependent weighted functions;

[0056] Lagrange multipliers are embedded in the improved non-dominated sorting genetic algorithm to dynamically relax the energy consumption constraint.

[0057] The present invention also includes a multi-objective production planning optimization method for a digital factory, comprising the following steps:

[0058] Step 1: Use the production task decomposition module to receive enterprise order data and decompose it into multiple independent production subtasks. Each subtask contains a task number i. , k is a positive integer greater than 2, and the process constraint conditions and resource requirement type set of each subtask are extracted;

[0059] Step 2: The resource dynamic assessment module collects real-time data on the operating status of production equipment, material inventory characteristics, and human resource availability parameters in the digital factory to generate a real-time resource status matrix.

[0060] Step 3: A multi-objective optimization algorithm module is used to construct a multi-objective optimization model based on the decomposed subtask set and resource status matrix. The decision variables are defined as the equipment allocation plan and time scheduling sequence of the subtasks. A set of objective functions is established, including the total production cycle function, the total energy consumption function, and the equipment load variance function. Process constraints, resource capacity limitations, and task dependencies are introduced as constraints. An improved non-dominated sorting genetic algorithm is used to iteratively solve the multi-objective optimization model and generate a non-dominated solution set. The optimization objectives include minimizing the production cycle, minimizing the energy consumption cost, and maximizing the equipment load balance.

[0061] Step 4: Use the dynamic priority allocation module to calculate the dynamic priority weight of each subtask based on the real-time updated production abnormal event data and equipment failure rate prediction value, and adjust the subtask execution sequence;

[0062] Step 5: Detect resource allocation conflicts with the help of the conflict resolution unit and use the relaxation constraint algorithm to reset the time windows of the conflicting tasks.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] The production task decomposition module of the present invention can accurately decompose enterprise order data into multiple independent production subtasks and extract the corresponding process constraints and resource requirement type sets. This makes production task scheduling more organized and enables targeted planning based on the characteristics and requirements of different subtasks. Taking machinery manufacturing companies as an example, large-scale equipment manufacturing orders can be carefully decomposed into subtasks such as parts processing, component assembly, and overall commissioning. The process path and resource requirements of each subtask are clearly defined, avoiding confusion in task scheduling, improving production accuracy and efficiency, and ensuring the orderly progress of the entire production process.

[0065] The dynamic resource assessment module generates a real-time resource status matrix by collecting real-time data on production equipment operating status, material inventory characteristics, and human resource availability parameters. This function enables companies to monitor resource dynamics in real time and adjust production plans promptly. For example, in the electronics manufacturing process, when the inventory of a key material falls below a safe threshold, the system can quickly detect and adjust the production sequence, prioritizing other tasks with lower reliance on the material. This prevents production stoppages caused by material shortages, improves resource utilization efficiency, and reduces inventory backlogs and waste.

[0066] The multi-objective optimization algorithm module constructs a multi-objective optimization model that comprehensively considers minimizing production cycle time, minimizing energy costs, and maximizing equipment load balance. In actual production, for example, using this optimization model to rationally schedule equipment operating hours and production tasks at a steel production company, not only has the product production cycle been shortened, order delivery times reduced, and energy costs lowered, but the loads across equipment have also been more balanced, extending equipment lifespans and reducing maintenance costs. This multi-objective collaborative optimization approach effectively balances the conflicts between different production objectives within a company and enhances its overall competitiveness.

[0067] The dynamic priority allocation module calculates the dynamic priority weights of each subtask based on real-time production anomaly data and predicted equipment failure rates, and adjusts the subtask execution sequence. If a critical piece of equipment experiences a potential failure during the production process, the system automatically increases the priority of critical subtasks dependent on that equipment, pre-schedules production, and adjusts the order of other tasks to ensure continuous and stable production. This enables companies to respond quickly to emergencies, ensuring smooth production and minimizing losses caused by production anomalies.

[0068] The conflict resolution unit uses a relaxed constraint algorithm to detect resource allocation conflicts and reset the time windows for conflicting tasks. In complex production environments, resource competition is inevitable, such as when multiple tasks simultaneously require a specific piece of equipment. This unit quickly identifies conflicts and constructs a conflict resolution objective function by calculating the upper limit of each task's delay and the cost penalty coefficient. Using a branch-and-bound algorithm, it finds the optimal time window adjustment solution, resolving resource allocation conflicts and ensuring the orderly execution of production plans, thereby improving the feasibility and stability of production plans.

[0069] From the precise extraction of material inventory characteristic values ​​to the digital processing of process constraints, to the optimization of energy costs and the rational construction of human resource vectors, this invention comprehensively optimizes every detail of the production process. For example, by accurately calculating the real-time quality decay rate of materials, the order of use is rationally arranged to ensure product quality; utilizing the optimization steps of an improved non-dominated sorting genetic algorithm improves the efficiency and accuracy of solving multi-objective optimization models; and by real-time collection of employee physiological indicators, adjusting human resource performance parameters, achieving efficient utilization of human resources. These detailed optimizations comprehensively enhance the overall production efficiency and benefits of the digital factory, creating greater value for the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 This is a working principle diagram of the multi-objective production planning optimization system of the present invention;

[0071] Figure 2 This is the optimization flow chart of the improved non-dominated sorting genetic algorithm;

[0072] Figure 3 Flowchart of the relaxed constraint algorithm used for the conflict resolution unit. DETAILED DESCRIPTION

[0073] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0074] See also Figure 1-3 The present invention provides a multi-objective production planning optimization system for a digital factory, the system comprising:

[0075] Production task decomposition module: After receiving the enterprise order data, it decomposes it into multiple independent production subtasks, and each subtask is given a unique task number i. , k is a positive integer greater than 2. At the same time, the process constraints and resource requirement types for each subtask are extracted. For example, if a company receives an order to produce a batch of automotive parts, this module will break the order down into multiple subtasks, such as casting, machining, and assembly, based on the parts' production process. It will also clearly define the process sequence requirements for each subtask (process constraints) and the types of equipment, materials, and human resources required (resource requirement types).

[0076] Dynamic Resource Assessment Module: This module collects real-time data on the operational status of production equipment, material inventory characteristics, and human resource availability parameters within the digital factory, generating a real-time resource status matrix. This process provides accurate resource information for subsequent production planning, ensuring the feasibility of the plan.

[0077] Multi-objective optimization algorithm module: Based on the decomposed subtask set and resource status matrix, a multi-objective optimization model is constructed. The multi-objectives include minimizing the production cycle, minimizing energy consumption costs, and maximizing equipment load balance. By constructing a reasonable model and selecting an effective algorithm, an optimal production plan that meets multiple objectives is found. The specific construction method is as follows: the decision variables are defined as the equipment allocation plan and time scheduling sequence of the subtasks; a set of objective functions is established, including the total production cycle function, the total energy consumption function, and the equipment load variance function; process constraints, resource capacity limits, and task dependencies are introduced as constraints; and an improved non-dominated sorting genetic algorithm is used to iteratively solve the multi-objective optimization model to generate a non-dominated solution set.

[0078] Dynamic Priority Assignment Module: Based on real-time updated production anomaly data and equipment failure rate predictions, it calculates the dynamic priority weights of each subtask and adjusts the subtask execution sequence. This allows important or urgent tasks to be prioritized in the face of unexpected production situations, ensuring smooth production.

[0079] Conflict Resolution Unit: Detects resource allocation conflicts and uses a relaxed constraint algorithm to reset the time windows of conflicting tasks. When multiple subtasks compete for the same resource, this unit uses a reasonable algorithm to resolve the conflict, ensuring the orderly execution of the production plan.

[0080] The implementation of the present invention will be further described below with reference to Examples 1 to 5.

[0081] Example 1:

[0082] This embodiment describes in detail the specific implementation of the resource dynamic assessment module to generate a real-time resource status matrix, which is used to accurately obtain and integrate the real-time status information of various resources in the digital factory.

[0083] For production equipment, real-time idle time windows, remaining maintenance cycles, and energy efficiency parameters are collected. For example, for a CNC machine tool, the device's sensors provide real-time information about the time interval during which the machine is not in use (the real-time idle time window). Simultaneously, the remaining time until the next maintenance (the remaining maintenance cycle) and the energy efficiency data for the machine in different operating modes (energy efficiency parameters) are obtained from the equipment management system. This data is used to construct a device resource vector, which comprehensively reflects the machine's usable time, maintenance requirements, and energy consumption characteristics.

[0084] For material resources, collect the material inventory, supply cycle and quality decay coefficient. Obtain the current inventory quantity (inventory) of the material through the warehouse management system, and obtain the material supply cycle information from the supplier. At the same time, combine the environmental temperature and humidity monitoring data to calculate the real-time quality decay rate of the material. , the calculation formula is:

[0085]

[0086] in is the reference attenuation coefficient, For real-time humidity, For ideal humidity, To adjust the parameters, a material resource vector is constructed based on these data to accurately reflect the quantity, supply timeliness and quality changes of the materials.

[0087] In terms of human resources, we collect information on employee skill matching, work hour limits, and task processing efficiency. Through the employee skill management system, we determine the degree to which employee skills match the required skills for each subtask (personnel skill matching). We then determine employee work hour limits based on labor laws and company regulations, while also recording employee efficiency data from previous similar tasks. This constructs a human resource vector that comprehensively reflects an employee's ability to complete a task, time limits, and efficiency level.

[0088] Finally, the equipment, material, and human resource vectors are aggregated by time slice to form a three-dimensional resource status matrix. For example, taking one hour as a time slice, the equipment, material, and human resource vector information collected within that time slice is integrated to generate a three-dimensional matrix. This matrix is ​​dynamically updated along the time dimension, providing the multi-objective optimization algorithm module with real-time, comprehensive resource status information, enabling precise production planning based on actual resource conditions.

[0089] Example 2:

[0090] When initializing the population, heuristic rules based on process constraints are used to generate feasible solutions. For example, in the case of automotive parts production, when formulating the initial population for the production plan, the algorithm prioritizes task combinations that adhere to the process sequence specified in the process drawings, such as casting, machining, and assembly. Furthermore, the algorithm considers the process requirements of the equipment, such as the fact that certain high-precision machining tasks can only be performed on specific high-precision equipment. This ensures that the generated initial solution meets the process constraints, avoids infeasible production plan scenarios, and improves the quality of the initial solution and the speed of convergence.

[0091] By introducing a device affinity factor into crossover operations, allocation schemes with a high historical frequency of device collaboration are prioritized for gene exchange. For example, in a digital factory, devices A and B have collaborated frequently to complete specific production tasks in the past, and their historical frequency of collaboration is high. During a crossover operation, if a gene fragment involving device allocation contains allocation information for both devices A and B, and this combination has a higher device affinity factor than other gene fragments, then this combination is prioritized for gene exchange. This fully leverages the collaborative experience between devices, improves the quality of generated new solutions, makes production plans more rational in terms of device collaboration, and reduces the time and cost losses associated with device switching.

[0092] The mutation operation uses a dynamic mutation probability, adaptively adjusting the mutation intensity based on the population's convergence. In the early stages of the algorithm's iterations, when the population diversity is high and convergence is slow, the mutation probability is appropriately increased to encourage the algorithm to explore a wider range of solutions and avoid being trapped in a local optimum. As iterations progress, the population gradually converges. If the population converges too quickly, potentially leading to premature convergence, the mutation probability can be reduced, allowing the algorithm to conduct a refined search near the currently optimal solution, further optimizing solution quality. For example, by setting a convergence metric, such as when the change in the optimal solution over several consecutive generations is less than a certain threshold, the population is considered to be converging, and the mutation probability can be adjusted accordingly.

[0093] The entropy weighting method is used to assign weights to the non-dominated solution set and screen for optimal solutions on the Pareto front. The entropy weighting method is an objective weighting method that determines the weight of each objective by calculating the degree of difference between different solutions in the non-dominated solution set. For example, for the three objectives of production cycle time, energy cost, and equipment load balance, the entropy weighting method calculates their respective weights based on their distribution in the non-dominated solution set. Each solution in the non-dominated solution set is then comprehensively evaluated based on these weights, and the solution with the best overall score is selected as the optimal Pareto front solution. This ensures that the resulting production plan achieves a better balance between multiple objectives.

[0094] Example 3:

[0095] This embodiment introduces in detail the specific process of calculating the dynamic priority weight by the dynamic priority allocation module. Its function is to reasonably adjust the priority of each subtask according to the real-time situation in the production process, ensure that the production plan can respond to abnormal events and equipment failure risks in a timely manner, and ensure efficient production.

[0096] First, obtain the basic priority coefficient of each subtask from the local database For example, in an electronics production project, the database pre-assigns higher basic priority coefficients for the production of key components, due to their significant impact on the quality and performance of the entire product. Meanwhile, auxiliary subtasks, such as preparing packaging materials, have relatively lower basic priority coefficients. These basic priority coefficients are pre-assigned based on factors such as the subtask's importance and its contribution to the overall product value.

[0097] Next, collect equipment failure prediction values ​​in real time and order urgency parameters , where j represents the device number. Through the equipment failure monitoring system and data analysis model, the predicted probability of each device failing in the future (equipment failure prediction value ). At the same time, the order urgency parameter is calculated based on factors such as the customer's order delivery time requirements and order amount. For example, if the delivery time of an order is approaching and the order amount is large, then the corresponding order urgency parameter It is higher.

[0098] Then, calculate the real-time priority weight of subtask i , the calculation formula is:

[0099] in represents the dependency of subtask i on equipment j, and m represents the total number of production equipment in the digital factory. For example, in a certain machinery manufacturing factory, subtask i is a processing task for a key component, which is highly dependent on equipment j for high-precision processing. Therefore, the dependency of subtask i on equipment j is Higher. Assume that the failure prediction value of device j is obtained by calculation is 0.2, the basic priority coefficient of subtask i The order urgency parameter is 0.8. is 1.5, and subtask i only depends on device j, then the real-time priority weight of subtask i According to the calculated real-time priority weight, the execution sequence of subtasks is dynamically adjusted, and subtasks with high priority weights are given priority for production, thereby improving the flexibility of production plans and the ability to respond to emergencies.

[0100] Example 4:

[0101] During the conflicting task group identification phase, real-time resource allocation is monitored to identify conflicting task groups with overlapping resource usage. The process paths and time windows for the conflicting tasks are then extracted. For example, in a garment production workshop, Task A and Task B both require the use of the same sewing machine during the same time period, constituting a conflicting task group. The entire process path for Task A and Task B, from raw material preparation to finished product completion, is also captured, along with the planned start and end time windows for each task.

[0102] Calculate the maximum delay time and cost penalty coefficient for each task. For each conflicting task, determine the maximum delay time based on its process characteristics and the time requirements of subsequent tasks. For example, if Task A is delayed for more than 3 hours, it will affect the timely completion of multiple subsequent tasks, so the maximum delay time for Task A is 3 hours. The cost penalty coefficient is determined based on the impact of task delays on production efficiency, cost increases, etc. For example, delayed production of Task A will lead to increased raw material backlog costs and increased equipment idle costs. After comprehensively considering these factors, the cost penalty coefficient of Task A is determined to be a fine of 500 yuan for every hour of delay.

[0103] A conflict resolution objective function is constructed, with minimizing the total delay penalty as the optimization direction. A branch-and-bound algorithm is used to find the optimal time window adjustment solution. The branch-and-bound algorithm searches for all possible time window adjustment solutions, expanding the solution space by continuous branching. It then prunes branches based on the objective function value, discarding branches that are unlikely to yield the optimal solution, ultimately finding a time window adjustment solution that minimizes the total delay penalty. For example, suppose the conflicting task group consists of tasks A, B, and C. The algorithm calculates that delaying the start time of task A by 1 hour, leaving task B's time window unchanged, and starting task C 0.5 hours earlier can minimize the total delay penalty. This solution is then used to reset the time windows of the conflicting tasks, resolving the resource allocation conflict.

[0104] Example 5:

[0105] In terms of maximizing equipment load balancing, statistics are collected on the total working time of each equipment during the scheduling cycle. and idle time For example, in a one-month scheduling cycle, the total working time of device 1 is 150 hours and the idle time is 50 hours; the total working time of device 2 is 130 hours and the idle time is 70 hours, etc. Then calculate the device load balance degree B, the calculation formula is:

[0106] in, is the average working time of the equipment, is the average idle time of the device, and the denominator normalizes the load difference value by the total time dimension. Assume that a digital factory has 3 devices, and the working time of device 1 is Hours, the working hours of device 2 hours, the working time of device 3 hours, average working hours hours; the idle time of device 1 Hours, the idle time of device 2 hours, the idle time of device 3 hours, average idle time Hour.

[0107] The device load balance By improving the load balancing of equipment, you can reduce excessive use or idleness of equipment, extend the service life of equipment, and improve the overall utilization of equipment.

[0108] In terms of extracting material inventory feature values, RFID sensors are used to obtain the batch code and storage time stamp of the material. For example, in a food processing plant, RFID sensors can quickly and accurately obtain the coding information of each batch of raw materials and the storage time of the batch of raw materials. Combined with the ambient temperature and humidity monitoring data, the real-time quality decay rate of the material is calculated. By accurately obtaining the characteristic values ​​of material inventory, we can better understand the changes in material quality, arrange production reasonably, and avoid affecting product quality due to the use of deteriorated materials.

[0109] To digitize process constraints, the process sequence and equipment dependencies in CAD process drawings are analyzed to generate a directed acyclic graph. For example, in a mechanical parts processing process drawing, the sequence of roughing, semi-finishing, and finishing processes is clearly specified, along with the equipment required for each process. This information is converted into a directed acyclic graph, where nodes represent subtasks and edge weights represent the time cost of task switching. A topological sorting algorithm is then used to generate an initial feasible scheduling sequence, ensuring that production tasks are arranged according to a reasonable process sequence and improving production fluidity.

[0110] To minimize energy costs, we collect power curves from equipment at different load rates and fit a piecewise linear performance function. For example, for an injection molding machine, we collect power data at different injection volumes (corresponding to different load rates) and mathematically fit a piecewise linear performance function. We also introduce time-of-use electricity pricing parameters to convert energy costs into a time-dependent weighted function. Assuming that electricity prices are higher during peak daytime hours and lower during off-peak evening hours, we weight energy costs based on these prices. We embed Lagrange multipliers in an improved non-dominated sorting genetic algorithm to dynamically relax energy constraints, minimizing energy costs while meeting production needs.

[0111] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0112] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multi-objective production planning optimization system for digital factories, characterized by: The system comprises: The production task decomposition module is used to receive enterprise order data and decompose it into multiple independent production subtasks. Each subtask contains a task number i. , k is a positive integer greater than 2, and the process constraint conditions and resource requirement type set of each subtask are extracted; The resource dynamic assessment module is used to collect real-time operating status data of production equipment, material inventory characteristic values, and human resource availability parameters in the digital factory to generate a real-time resource status matrix; Multi-objective optimization algorithm module, which is used to build a multi-objective optimization model based on the decomposed subtask set and resource status matrix. The multi-objectives include minimizing the production cycle, minimizing energy consumption costs, and maximizing equipment load balance; Dynamic priority allocation module, which is used to calculate the dynamic priority weight of each subtask based on the real-time updated production abnormal event data and equipment failure rate prediction value, and adjust the subtask execution sequence; Conflict resolution unit, used to detect resource allocation conflicts and reset the time windows of conflicting tasks using a relaxation constraint algorithm; The method for constructing the multi-objective optimization model is: Define the decision variables as the equipment allocation plan and time scheduling sequence of subtasks; Establish a set of objective functions, including total production cycle function, total energy consumption function and equipment load variance function; Introducing process constraints, resource capacity limitations, and task dependencies as constraints; An improved non-dominated sorting genetic algorithm is used to iteratively solve the multi-objective optimization model and generate a non-dominated solution set. The specific optimization steps of the improved non-dominated sorting genetic algorithm include: When initializing the population, a heuristic rule based on process constraints is used to generate feasible solutions; Introducing the device affinity factor in the crossover operation, giving priority to the allocation scheme with a high historical frequency of device collaboration for gene exchange; Dynamic mutation probability is used in the mutation operation, and the mutation intensity is adaptively adjusted according to the degree of population convergence; Use the entropy weight method to assign weights to the non-dominated solution set and screen the Pareto frontier optimal solution; The specific method for calculating the dynamic priority weight in the dynamic priority allocation module is: Get the basic priority coefficient of each subtask from the local database ; Real-time collection of equipment failure prediction values and order urgency parameters , j represents the device number; Calculate the real-time priority weight of subtask i : in represents the dependency of subtask i on equipment j, and m represents the total number of production equipment in the digital factory.

2. The system according to claim 1, wherein: The specific method for generating the real-time resource status matrix in the resource dynamic evaluation module is: Collect real-time idle time windows, remaining maintenance cycles, and energy efficiency parameters of production equipment to build equipment resource vectors; Collect material inventory, replenishment cycle, and quality attenuation coefficient to construct a material resource vector; Collect personnel skills matching, working time limit and task processing efficiency to build human resource vector; Equipment, material and human resource vectors are aggregated by time slices to form a three-dimensional resource status matrix.

3. The system according to claim 1, wherein: The specific steps of adopting the relaxed constraint algorithm in the conflict resolution unit include: Identify conflicting task groups with overlapping resource occupancy and extract the process paths and time windows of the conflicting tasks; Calculate the upper limit of delay time and cost penalty coefficient of each task; A conflict resolution objective function is constructed, with minimizing the total delay penalty as the optimization direction, and the branch and bound algorithm is used to solve the optimal time window adjustment solution.

4. The system according to claim 3, characterized in that The calculation method for maximizing the device load balancing degree is: Statistics of the total working time of each device in the scheduling cycle and idle time ; Calculate device load balancing : in, is the average working time of the equipment, is the average idle time of the device, and the denominator is the load difference value normalized by the total time dimension.

5. The system according to claim 4, characterized in that The method for extracting the material inventory characteristic value includes: The batch code and storage time stamp of the material are obtained through the RFID sensor.

6. The system according to claim 5, characterized in that The digital method of the process constraints is: Analyze the process sequence and equipment dependency in CAD process drawings and generate a directed acyclic graph; Map the nodes in the graph to subtasks, and the edge weights represent the time cost of task switching; The initial feasible scheduling sequence is generated by the topological sorting algorithm.

7. The system according to claim 6, characterized in that The optimization method for minimizing energy consumption cost includes: Collect the power curve of the equipment at different load rates and fit the segmented line performance consumption function; Introducing time-of-use electricity price parameters to convert energy consumption costs into time-dependent weighted functions; Lagrange multipliers are embedded in the improved non-dominated sorting genetic algorithm to dynamically relax the energy consumption constraint.

8. A multi-objective production planning optimization method for digital factories, characterized by: The following steps are involved: Step 1: Use the production task decomposition module to receive enterprise order data and decompose it into multiple independent production subtasks. Each subtask contains a task number i. , k is a positive integer greater than 2, and the process constraint conditions and resource requirement type set of each subtask are extracted; Step 2: The resource dynamic assessment module collects real-time data on the operating status of production equipment, material inventory characteristics, and human resource availability parameters in the digital factory to generate a real-time resource status matrix. Step 3: A multi-objective optimization algorithm module is used to construct a multi-objective optimization model based on the decomposed subtask set and resource status matrix. The decision variables are defined as the equipment allocation plan and time scheduling sequence of the subtasks. A set of objective functions is established, including the total production cycle function, the total energy consumption function, and the equipment load variance function. Process constraints, resource capacity limitations, and task dependencies are introduced as constraints. An improved non-dominated sorting genetic algorithm is used to iteratively solve the multi-objective optimization model and generate a non-dominated solution set. The optimization objectives include minimizing the production cycle, minimizing the energy consumption cost, and maximizing the equipment load balance. Step 4: Use the dynamic priority allocation module to calculate the dynamic priority weight of each subtask based on the real-time updated production abnormal event data and equipment failure rate prediction value, and adjust the subtask execution sequence; Step 5: Detect resource allocation conflicts with the help of the conflict resolution unit and use the relaxation constraint algorithm to reset the time windows of the conflicting tasks; The specific optimization steps of the improved non-dominated sorting genetic algorithm include: When initializing the population, a heuristic rule based on process constraints is used to generate feasible solutions; Introducing the device affinity factor in the crossover operation, giving priority to the allocation scheme with a high historical frequency of device collaboration for gene exchange; Dynamic mutation probability is used in the mutation operation, and the mutation intensity is adaptively adjusted according to the degree of population convergence; Use the entropy weight method to assign weights to the non-dominated solution set and screen the Pareto frontier optimal solution; The specific method for calculating the dynamic priority weight in the dynamic priority allocation module is: Get the basic priority coefficient of each subtask from the local database ; Real-time collection of equipment failure prediction values and order urgency parameters , j represents the device number; Calculate the real-time priority weight of subtask i : in represents the dependency of subtask i on equipment j, and m represents the total number of production equipment in the digital factory.

Images