Resource-constrained parallel machine model production scheduling method based on intensive strategy coordination
Through the resource-constrained parallel machine model scheduling method based on intensive strategies, combined with the field heuristic algorithm and model address intensive optimization algorithm, the information asymmetry problem of production and distribution in large-scale manufacturing mode in the home appliance industry is solved, and the information asymmetry problem of production and distribution in the large-scale manufacturing mode in the home appliance industry is improved. Productivity and reduce costs, ensuring continuity of models and addresses.
Patent Information
- Application Number
- CN202411233294.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-04
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-09-04
AI Technical Summary
In the home appliance industry, the traditional large-scale manufacturing model has high production costs due to variety diversification and rapid iteration, and the information asymmetry between production and distribution is difficult to coordinate, which increases the revenue pressure of enterprises, especially When facing the demand for small batch orders of multiple addresses and multiple varieties, the difficulty of scheduling for planners has increased significantly.
Using the resource-constrained parallel machine model scheduling method based on intensive strategies, obtain order data, use the domain-based heuristic algorithm to calculate the model scheduling solution, and use intelligent algorithms based on intensive optimization of model address to optimize the address Distribution and model production plans to improve production efficiency and reduce distribution costs.
It effectively solves the complex calculation problem under the complex relationship mapping between orders, addresses and models, reduces the difficulty of optimization of distribution and production scheduling, improves production efficiency, reduces distribution costs, and ensures model continuity under the conditions of variety. and address continuity.
Smart Images

Figure CN119090226A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data processing, and in particular to a resource-constrained parallel machine model scheduling method based on intensive strategy coordination. Background Art
[0002] In general, scale and cost advantages are important factors in the development of the home appliance industry. However, changes in the environment have put forward increasingly stringent requirements for the home appliance industry. Specifically, first of all, the market demand for small batches of multiple varieties has gradually replaced the original market demand for large batches of fewer varieties, making the traditional large-scale manufacturing model for inventory-oriented high production costs due to the diversification and rapid iteration of varieties. Secondly, in order to reduce the area of finished product warehouses, enterprises adopt the offline direct delivery model that directly sends the same order to the required address after completion. However, in the face of fragmented order demand and highly dispersed address demand, the large-scale manufacturing production model will cause the orders at the same address to be completed in a dispersed manner, thereby increasing distribution costs. Therefore, it is necessary to improve the current large-scale manufacturing model under the condition of variety diversification and combined with the intensive demand for production and distribution. However, most manufacturers use third-party logistics for distribution, resulting in information asymmetry between the producer and the distributor, making it difficult to jointly consider the problem of intensive production and distribution, thereby increasing the revenue pressure of the enterprise.
[0003] While taking into account both variety diversity and intensive production, considering intensive distribution at multiple addresses and formulating production scheduling plans will undoubtedly greatly increase the difficulty of production scheduling for planners. The decision-making dimensions, factors that need to be considered, and the amount of data required in the production scheduling process are growing exponentially, which increases the scheduling time for planners and the risk of omissions. Summary of the invention
[0004] The purpose of the present invention is to provide a resource-constrained parallel machine model scheduling method based on intensive strategy coordination. This method efficiently generates scientific and effective model scheduling solutions for resource-constrained parallel machine scenarios, and solves the difficulty of complex calculations under the complex relationship mapping between orders, addresses and models; it is necessary to consider the problem of solution space fragmentation caused by multiple constraints. At the same time, it proactively considers distribution, reduces the optimization difficulty of distribution scheduling, and achieves the purpose of improving production efficiency and reducing distribution costs.
[0005] A resource-constrained parallel machine model scheduling method based on intensive strategy coordination, comprising:
[0006] Get order data;
[0007] Calculate the model production schedule according to the order data using a domain-based heuristic algorithm;
[0008] An intelligent algorithm based on intensive optimization of model addresses is used to optimize the production scheduling plan of the model.
[0009] Preferably, the obtaining of order data comprises:
[0010] Obtain information on the number of machines in the factory, the models that the machines can produce, the machine start time, the time required for model production, the resources required for model production, and customer order information.
[0011] Preferably, the method of calculating the model production scheduling plan according to the order data using a domain-based heuristic algorithm includes:
[0012] Get order information, model information, resource information, work calendar information and configuration information;
[0013] Convert order information, model information, resource information, work calendar information and configuration information into integer data by establishing key-value pair matching;
[0014] Summarize the order information by month and model to obtain the model set that needs to be produced;
[0015] According to the data preprocessing process, the redundant data in the model information and resource information of the current version is eliminated;
[0016] Initialize the list of idle machines;
[0017] Determine whether the current flight is the first flight;
[0018] According to the current shift information, choose to continuously undertake the lower die strategy or the upper die strategy based on domain knowledge and constraints for the machine;
[0019] Output model production schedule.
[0020] Preferably, the method of selecting a continuous lower die strategy or an upper die strategy based on domain knowledge and constraints for the machine according to the current shift information includes:
[0021] If the current shift is the first shift, select the continuous die transfer strategy for the machine, including:
[0022] Obtain the model production schedule of the previous shift, and pre-allocate the model production schedule of the previous shift as the mold arrangement plan for the current shift;
[0023] According to the production schedule of the previous shift, determine whether the molds that have been loaded have reached the maximum number of continuous production days;
[0024] If the maximum number of continuous production days is reached, check the corresponding relationship between the remaining model demand and the current capacity. The current available capacity = the number of shifts that can be molded * the maximum number of molds of the current model * the UPH of the current model. If the number of machines with the current model cannot meet the minimum continuous production days limit when continuing to produce and the remaining demand can be completed within the monthly range, perform mold removal on one of the machines and check the current step again.
[0025] If the maximum continuous production days have not been reached, determine whether the mold that has been molded has reached the maximum continuous production days. If the maximum continuous production days have been reached, check whether the orders within the monthly range of the corresponding model of the mold have all been completed. If they have all been completed, perform the mold removal process. If they have not been completed, do not perform the mold removal process.
[0026] Check the door resources of the mold arrangement plan for the current shift. If the door resources are insufficient to support the production of the corresponding model of the mold, the mold will be processed. If the door resources are sufficient to support the production of the model, the mold will not be processed.
[0027] Perform capacity verification on the remaining model information to verify whether the demand can be met within the available capacity range of the model. If it cannot be met, the model with constraints is selected first, and then the mold with a larger number of upper molds is selected for lower mold;
[0028] Save a collection of completed die models.
[0029] Preferably, the method of selecting a continuous lower die strategy or an upper die strategy based on domain knowledge and constraints for the machine according to the current shift information includes:
[0030] If the current shift is not the first shift, select a die-up strategy based on domain knowledge and constraints, including:
[0031] Summarize the list of all idle machines in the current shift, and take the order of the order volume of the same model from high to low as the model priority;
[0032] Prioritize the models that are subject to model mutual exclusion and machine production constraints, and then select the models with higher model priority according to the model priority. After sorting the model set to be produced according to the model priority, the list of candidate models for the lower upper mold in the current shift is obtained;
[0033] The model set of the lower mold in the process of pre-allocation of the production plan for the previous shift is removed from the list of candidate models for the lower mold in the current shift;
[0034] Summarize the list of molds that have been loaded in the non-idle machines in the current shift, and remove the mutually exclusive models under the production models of the non-idle machines from the list of candidate models for loading in the current shift;
[0035] In the current shift, the list of candidate mold models is sorted from high to low. According to the remaining demand quantity of the order, the corresponding mold model is selected for mold processing. The maximum number of products that can be arranged = min (box mold production capacity, door resource supply capacity) / single mold daily production capacity;
[0036] Determine whether the number of mold types exceeds the maximum number of mold types. If so, reselect the model for mold processing and remove the model corresponding to the current mold from the list of mold candidate models for the current shift.
[0037] If the maximum number of mold types is not exceeded, determine whether the molds corresponding to the models are mutually exclusive. If the mold corresponding to the model to be newly molded is mutually exclusive with other molds that have been molded in the current shift, remove the model corresponding to the mold from the list of candidate models for molds to be molded in the current shift P, and reselect the model for mold processing;
[0038] If the molds corresponding to the models are not mutually exclusive, determine whether the current mold-mounting machine and mold have mold-mounting area restrictions. If so, reselect the model from the mold-mounting alternative model list for the current shift to perform mold-mounting processing;
[0039] If there is no restriction on the mold loading area, determine whether there is an idle machine in the current shift. If there is an idle machine in the current shift and the list of mold loading candidate models in the current shift has not been completely traversed, continue to select the model for mold loading;
[0040] If there is no idle machine in the current shift or the list of upper die alternative models in the current shift has been completely traversed, it will jump to the next shift and re-perform the previous shift to pre-allocate the production schedule and the upper die strategy based on domain knowledge and constraints.
[0041] Preferably, the optimization of the model production scheduling plan by using an intelligent algorithm based on model address intensive optimization includes:
[0042] Obtain order information, model information, resource information, work calendar information and previous model production schedule;
[0043] Initialize the number of iterations;
[0044] Summarize the order information by month, model and address to obtain the set of models and addresses that need to be produced;
[0045] According to the address intensive allocation algorithm, an address allocation sequence is obtained;
[0046] According to the address allocation sequence, the previous model scheduling plan, and the set of models and addresses to be produced, the address to be shipped for the number of models produced by each machine in each shift is determined and the information is summarized. At the same time, the set of models and addresses to be produced is updated;
[0047] Based on the joint optimization algorithm of model address, the model production schedule of the previous level is combined with the model address shipment summary information optimization to obtain a new model production schedule, and verify whether the resources can be met based on the resource information;
[0048] If resources are sufficient, output the optimized model production schedule.
[0049] Preferably, obtaining the address allocation sequence according to the address intensive allocation algorithm comprises:
[0050] Get the number of iterations, population size, and mutation probability configuration information;
[0051] Create a coding matrix by address and month through the configured population size, and randomly generate A different numbers of addresses by month to fill the matrix;
[0052] Calculate the fitness value and sort each chromosome S in the population according to the fitness value;
[0053] According to the roulette wheel method, two different chromosomes S1 and S2 are selected;
[0054] Randomly select a gene from S1 as the intersection point C1, record the row information where C1 is located, randomly select a gene in a row in S2 as the intersection point C2, and perform positional crossover on C1 and C2;
[0055] Check whether there is duplication or missing of coding sequence numbers in S1 and S2 after crossing;
[0056] If there are repeated or missing numbers in the coding sequence, a repeated number is randomly selected and replaced with the missing number;
[0057] Check whether the current situation meets the mutation conditions. If so, regenerate a new chromosome and insert it into the population, and remove those chromosomes with low fitness evaluation indicators.
[0058] Check whether the current situation has reached the upper limit of the number of iterations. If so, output the optimal result obtained by the current iteration;
[0059] If the upper limit of the number of iterations is not reached, the iterative optimization process will continue.
[0060] Preferably, decoding the chromosome using the address intensive optimization algorithm includes:
[0061] Extract the month information in the chromosome and the order information of each address in each month, and index each month one by one;
[0062] Get the model production plan for the current month, index the addresses in order according to the address sequence information, and get the current address information;
[0063] Summarize the order information of the current month and address by month, required quantity, model, and address, and sort them in descending order according to the number of models required to be produced at the same address;
[0064] Through the shift and opening hours information provided by the box foaming line work calendar, combined with the hourly production capacity information in the box mold information table, the number of units that can be scheduled for production = model UPH * address intensive time * number of molds on the machine corresponding to the model, the information on how many units of each model can be produced within the time range of the batch shipment is obtained, and the number of models that can be produced is sorted from large to small;
[0065] Traverse the collected information and select machines within the same address concentration time range for allocation in order according to the obtained information until all orders under the current address in the current month are allocated or there is no capacity to be allocated within the same address concentration time range. At the same time, update the information so that the occupied machines can no longer be allocated;
[0066] After the information is distributed, the next address and model are searched for and distributed;
[0067] After completing all address conveniences for the current month, continue looking for address allocation for the next month.
[0068] Preferably, the model-address joint optimization algorithm iteration strategy includes:
[0069] Get the number of iterations, model concentration importance, and decision step configuration information;
[0070] Calculate whether each machine has changed molds in each shift compared to the previous shift, and count the number of machines that have changed molds in each shift;
[0071] Count the number of addresses delivered in each shift, and calculate the degree of optimization of the model addresses encoded in each shift dimension based on the model concentration emphasis;
[0072] According to the framework solution of model scheduling and the mapping relationship of the address intensive allocation algorithm, the number of units produced in each address of the mold corresponding to each model is summarized, and the addresses are sorted according to the number of units produced;
[0073] Select the model, mold and address in order, reselect and allocate the molds and priority order corresponding to the model in the shift. Each time a mold is allocated, check whether the model allows simultaneous production, mold quantity limit, and machine production constraints. If the constraint check passes, use the corresponding mold. If the constraint check fails, select the next mold in the order for mold loading;
[0074] If it is possible to reduce mold switching and the number of addresses required can support single-shift allocation, the currently selected mold and address will be allocated;
[0075] If it is possible to reduce mold switching but the number of units required for the address is less than the number of units that the machine can produce in a shift, the currently selected mold, the remaining number of units to be produced at the current address, and the number of units required for the next sequential address in the current mold are allocated;
[0076] If it is possible to reduce mold switching but there is no mold corresponding to the required number of addresses that need to be allocated, select the next mold in the sequence for allocation.
[0077] Preferably, after optimizing the model production scheduling plan by using the intelligent algorithm based on model address intensive optimization, the optimized production scheduling plan is evaluated, specifically:
[0078] Based on the average number of days each address is assigned Number of model switching between adjacent shifts Conduct a comprehensive evaluation and obtain a comprehensive evaluation:
[0079]
[0080] Obtain the optimal model production scheduling plan under multiple iterative versions as the final production scheduling plan;
[0081] Among them, m represents month information, a represents address information, i represents shift information, j represents machine information, and p represents model information. Among them, inf2 represents the number of mold changes when the model address joint optimization algorithm is input, inf3 represents the average distribution days when the model address joint optimization algorithm is input, nf2 represents the number of mold changes of the solution result under the current optimization, and nf3 represents the average distribution days of the solution result under the current optimization.
[0082] The beneficial effects of the present invention are as follows: the core of the heuristic algorithm based on domain rules of the present invention is to configure relevant business parameters based on the actual production environment conditions, and to mold the molds required for the model shift by shift according to the work calendar through the business parameters, so as to meet the model continuity under the condition of ensuring variety diversity. The core of the intelligent algorithm based on intensive optimization of model addresses is to allocate addresses on the basis of the first part of the model scheduling plan and optimize the model and address through the intelligent algorithm, so as to improve the continuity of address to address. BRIEF DESCRIPTION OF THE DRAWINGS
[0083] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0084] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0085] Figure 1 A flow chart of a resource-constrained parallel machine model scheduling method based on intensive strategy coordination of the present invention;
[0086] Figure 2 It is a schematic diagram of the flow of the heuristic algorithm based on domain rules of the present invention;
[0087] Figure 3 It is a schematic diagram of the model address intensive optimization algorithm flow based on the intelligent algorithm of the present invention;
[0088] Figure 4 A schematic diagram of the decoding allocation strategy of the address intensive allocation algorithm of the present invention;
[0089] Figure 5 It is a schematic diagram of the iteration of the model-address joint optimization algorithm of the present invention. DETAILED DESCRIPTION
[0090] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.
[0091] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.
[0092] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.
[0093] While taking into account both variety diversity and intensive production, considering intensive distribution at multiple addresses and formulating production scheduling plans will undoubtedly greatly increase the difficulty of production scheduling for planners. The decision-making dimensions, factors that need to be considered, and the amount of data required in the production scheduling process are growing exponentially, which increases the scheduling time for planners and the risk of omissions.
[0094] The core of the heuristic algorithm based on domain rules of the present invention is to configure relevant business parameters based on the actual production environment conditions, and to mold the molds required by the model according to the work calendar shift by shift through the business parameters, so as to meet the model continuity under the condition of ensuring variety diversity. The core of the intelligent algorithm based on intensive optimization of model addresses is to allocate addresses on the basis of the first part of the model scheduling plan and optimize the model and address through intelligent algorithms, so as to improve the continuity of address to address.
[0095] A resource-constrained parallel machine model scheduling method based on intensive strategy coordination, comprising:
[0096] S100, obtaining order data;
[0097] S200, uses a domain-based heuristic algorithm to calculate model scheduling plans based on order data;
[0098] S300 uses an intelligent algorithm based on intensive optimization of model addresses to optimize model production scheduling.
[0099] Preferably, S100, obtaining order data includes:
[0100] Obtain information on the number of machines in the factory, the models that the machines can produce, the machine start time, the time required for model production, the resources required for model production, and customer order information.
[0101] Since the actual production data is configured by planners, there are situations where planners mismatch or miss match in the actual situation. For example, in order to meet market demand, the company launches a new model for sale, and the customer places an order for the new model, but the corresponding supply route cannot be found in the system. In order to ensure the efficiency and stability of the algorithm and shield the interference of abnormal data, the production data will go through the data preprocessing process before entering the algorithm.
[0102] Through the user's quick code configuration, we can obtain information such as the maximum and minimum continuous production days, the machine's production matrix, etc., and establish a key-value pair relationship for the data to achieve standardized processing of converting character information into integer information.
[0103] The products in the order information table are aggregated according to the dimension of month to obtain the model set N that needs to be produced.
[0104] According to the model set N that needs to be produced, search whether there is a corresponding relationship in the information table with the model field. If there is a single model, the data does not exist in the model-related information table, but there is data in the model set N produced, then delete the corresponding data to improve the efficiency of the algorithm data retrieval; if there is a single model, the data does not exist in the model-related information table, but there is data in the model set N produced, then transfer the order corresponding to the model to the new order set D, and feedback the relevant information to the user after completing the entire production scheduling process.
[0105] Preferably, S200, using a domain-based heuristic algorithm to calculate a model production scheduling plan according to order data includes:
[0106] Get order information, model information, resource information, work calendar information and configuration information;
[0107] Convert order information, model information, resource information, work calendar information and configuration information into integer data by establishing key-value pair matching;
[0108] Summarize the order information by month and model to obtain the model set that needs to be produced;
[0109] According to the data preprocessing process, the redundant data in the model information and resource information of the current version is eliminated;
[0110] Initialize the list of idle machines;
[0111] Determine whether the current flight is the first flight;
[0112] According to the current shift information, choose to continuously undertake the lower die strategy or the upper die strategy based on domain knowledge and constraints for the machine;
[0113] Output model production schedule.
[0114] In this module, the experience of production schedulers is summarized into a heuristic algorithm based on domain knowledge, and the model scheduling plan is obtained according to the algorithm flow of the heuristic algorithm. This can meet the variety of products and ensure production continuity, and avoid infeasible solutions caused by constraints. It is mainly divided into shift scheduling decision, continuous die-taking strategy, and die-up strategy based on domain knowledge and constraints. The relevant processes are as follows: Figure 2 shown.
[0115] Pseudo code of production intensive model scheduling algorithm based on domain knowledge:
[0116] Input: order information Order_Info, model information Fac_Info, resource information Res_Info, work calendar information Data_Info, configuration information Config_Info;
[0117] Convert the order information Order_Info, model information Fac_Info, resource information Res_Info, work calendar information Data_Info, and configuration information Config_Info into integer data by establishing key-value pair matching;
[0118] Summarize the order information Order_Info by month and model to obtain the model set N that needs to be produced;
[0119] for i=1:length(N)
[0120] According to the data preprocessing process, irrelevant data in the model information Fac_Info and resource information Res_Info in the current version are removed;
[0121] end for
[0122] for i=1:length(Data_Info)
[0123] Initialize the idle machine list idle_List;
[0124] If the current flight is the first flight
[0125] Inherit the mold information of the previous shift to the current shift, and according to the continuity of the mold placement strategy, place the mold on the machine, and update the idle machine list idle_List at the same time;
[0126] end if
[0127] According to the die-casting strategy based on domain knowledge and constraints, the idle machine list idle_List is die-casted;
[0128] end for
[0129] Output model scheduling scheme Mould_scheme1;
[0130] Preferably, the strategy of continuously taking over the lower die strategy or the upper die strategy based on domain knowledge and constraints for the machine is selected according to the current shift information, including:
[0131] If the current shift is the first shift, select the continuous die transfer strategy for the machine, including:
[0132] Obtain the model production schedule of the previous shift, and pre-allocate the model production schedule of the previous shift as the mold arrangement plan for the current shift;
[0133] According to the production schedule of the previous shift, determine whether the molds that have been loaded have reached the maximum number of continuous production days;
[0134] If the maximum number of continuous production days is reached, check the corresponding relationship between the remaining model demand and the current capacity. The current available capacity = the number of shifts that can be molded * the maximum number of molds of the current model * the UPH of the current model. If the number of machines with the current model cannot meet the minimum continuous production days limit when continuing to produce and the remaining demand can be completed within the monthly range, perform mold removal on one of the machines and check the current step again.
[0135] If the maximum continuous production days have not been reached, determine whether the mold that has been molded has reached the maximum continuous production days. If the maximum continuous production days have been reached, check whether the orders within the monthly range of the corresponding model of the mold have all been completed. If they have all been completed, perform the mold removal process. If they have not been completed, do not perform the mold removal process.
[0136] Check the door resources of the mold arrangement plan for the current shift. If the door resources are insufficient to support the production of the corresponding model of the mold, the mold will be processed. If the door resources are sufficient to support the production of the model, the mold will not be processed.
[0137] Perform capacity verification on the remaining model information to verify whether the demand can be met within the available capacity range of the model. If it cannot be met, the model with constraints is selected first, and then the mold with a larger number of upper molds is selected for lower mold;
[0138] Save a collection of completed die models.
[0139] Shift scheduling decision: This decision mainly determines whether the current shift can inherit the shift information. If it can inherit the shift information, it is the first shift. If it cannot inherit the shift information, it is not the first shift. If the current shift is the first shift, the upper mold strategy based on domain knowledge and constraints is executed. If the current shift is not the first shift, the previous shift scheduling information is obtained, and the current shift is pre-allocated according to the previous shift.
[0140] Preferably, the strategy of continuously taking over the lower die strategy or the upper die strategy based on domain knowledge and constraints for the machine is selected according to the current shift information, including:
[0141] If the current shift is not the first shift, select a die-up strategy based on domain knowledge and constraints, including:
[0142] Summarize the list of all idle machines in the current shift, and take the order of the order volume of the same model from high to low as the model priority;
[0143] Prioritize the models that are subject to model mutual exclusion and machine production constraints, and then select the models with higher model priority according to the model priority. After sorting the model set to be produced according to the model priority, the list of candidate models for the lower upper mold in the current shift is obtained;
[0144] The model set of the lower mold in the process of pre-allocation of the production plan for the previous shift is removed from the list of candidate models for the lower mold in the current shift;
[0145] Summarize the list of molds that have been loaded in the non-idle machines in the current shift, and remove the mutually exclusive models under the production models of the non-idle machines from the list of candidate models for loading in the current shift;
[0146] In the current shift, the list of candidate mold models is sorted from high to low. According to the remaining demand quantity of the order, the corresponding mold model is selected for mold processing. The maximum number of products that can be arranged = min (box mold production capacity, door resource supply capacity) / single mold daily production capacity;
[0147] Determine whether the number of mold types exceeds the maximum number of mold types. If so, reselect the model for mold processing and remove the model corresponding to the current mold from the list of mold candidate models for the current shift.
[0148] If the maximum number of mold types is not exceeded, determine whether the molds corresponding to the models are mutually exclusive. If the mold corresponding to the model to be newly molded is mutually exclusive with other molds that have been molded in the current shift, remove the model corresponding to the mold from the list of candidate models for molds to be molded in the current shift P, and reselect the model for mold processing;
[0149] If the molds corresponding to the models are not mutually exclusive, determine whether the current mold-mounting machine and mold have mold-mounting area restrictions. If so, reselect the model from the mold-mounting alternative model list for the current shift to perform mold-mounting processing;
[0150] If there is no restriction on the mold loading area, determine whether there is an idle machine in the current shift. If there is an idle machine in the current shift and the list of mold loading candidate models in the current shift has not been completely traversed, continue to select the model for mold loading;
[0151] If there is no idle machine in the current shift or the list of upper die alternative models in the current shift has been completely traversed, it will jump to the next shift and re-perform the previous shift to pre-allocate the production schedule and the upper die strategy based on domain knowledge and constraints.
[0152] Preferably, S300, optimizing the model production scheduling plan by using an intelligent algorithm based on model address intensive optimization includes:
[0153] Obtain order information, model information, resource information, work calendar information and previous model production schedule;
[0154] Initialize the number of iterations;
[0155] Summarize the order information by month, model and address to obtain the set of models and addresses that need to be produced;
[0156] According to the address intensive allocation algorithm, an address allocation sequence is obtained;
[0157] According to the address allocation sequence, the previous model scheduling plan, and the set of models and addresses to be produced, the address to be shipped for the number of models produced by each machine in each shift is determined and the information is summarized. At the same time, the set of models and addresses to be produced is updated;
[0158] Based on the joint optimization algorithm of model address, the model production schedule of the previous level is combined with the model address shipment summary information optimization to obtain a new model production schedule, and verify whether the resources can be met based on the resource information;
[0159] If resources are sufficient, output the optimized model production schedule.
[0160] Since the previous-level production scheduling plan takes into account the continuity of model production while ensuring variety diversity, ensuring the intensive completion of orders with the same address through local optimization based on the previous-level production scheduling plan can effectively improve the efficiency of the algorithm and the quality of the solution. The model-address intensive optimization algorithm based on the intelligent algorithm specifically establishes a mapping relationship between address and model based on the previous-level production scheduling plan, and then uses the mapping relationship to optimize the model and address intensive coordination of the model scheduling plan. Its core is mainly divided into the address intensive allocation algorithm and the model-address joint scheduling algorithm. The relevant processes are as follows: Figure 3 shown.
[0161] Model scheduling search algorithm for intensive production and distribution coordination:
[0162] Input: order information Order_Info, model information Fac_Info, resource information Res_Info, work calendar information Data_Info, and upper level model scheduling scheme Mould_scheme1;
[0163] Initialize the number of iterations count
[0164] Summarize the order information Order_Info by month, model and address to obtain the set N_add of models and addresses that need to be produced;
[0165] According to the address intensive allocation algorithm, an address allocation sequence M is obtained;
[0166] for i=1:length(M)
[0167] According to the address allocation sequence M, the previous model scheduling scheme Mould_scheme1 and the set of models and addresses to be produced N_add, determine the address to be shipped for the number of models produced by each machine in each shift and summarize the information Add_scheme, and update the set of models and addresses to be produced N_add at the same time;
[0168] end for
[0169] for i=1:length(count)
[0170] According to the joint optimization algorithm of model address, the model scheduling scheme Mould_scheme1 of the previous level is combined with the model address shipment summary information Add_scheme optimization to obtain the new model scheduling scheme Mould_scheme2, and check whether the resources can be met according to the resource information Res_Info;
[0171] end for
[0172] Output model scheduling scheme Mould_scheme2;
[0173] Preferably, according to the address intensive allocation algorithm, obtaining the address allocation sequence includes:
[0174] Get the number of iterations, population size, and mutation probability configuration information;
[0175] Create a coding matrix by address and month through the configured population size, and randomly generate A different numbers of addresses by month to fill the matrix;
[0176] Calculate the fitness value and sort each chromosome S in the population according to the fitness value;
[0177] According to the roulette wheel method, two different chromosomes S1 and S2 are selected;
[0178] Randomly select a gene from S1 as the intersection point C1, record the row information where C1 is located, randomly select a gene in a row in S2 as the intersection point C2, and perform positional crossover on C1 and C2;
[0179] Check whether there is duplication or missing of coding sequence numbers in S1 and S2 after crossing;
[0180] If there are repeated or missing numbers in the coding sequence, a repeated number is randomly selected and replaced with the missing number;
[0181] Check whether the current situation meets the mutation conditions. If so, regenerate a new chromosome and insert it into the population, and remove those chromosomes with low fitness evaluation indicators.
[0182] Check whether the current situation has reached the upper limit of the number of iterations. If so, output the optimal result obtained by the current iteration;
[0183] If the upper limit of the number of iterations is not reached, the iterative optimization process will continue.
[0184] Preferably, decoding the chromosome using the address intensive optimization algorithm includes:
[0185] Extract the month information in the chromosome and the order information of each address in each month, and index each month one by one;
[0186] Get the model production plan for the current month, index the addresses in order according to the address sequence information, and get the current address information;
[0187] Summarize the order information of the current month and address by month, required quantity, model, and address, and sort them in descending order according to the number of models required to be produced at the same address;
[0188] Through the shift and opening hours information provided by the box foaming line work calendar, combined with the hourly production capacity information in the box mold information table, the number of units that can be scheduled for production = model UPH * address intensive time * number of molds on the machine corresponding to the model, the information on how many units of each model can be produced within the time range of the batch shipment is obtained, and the number of models that can be produced is sorted from large to small;
[0189] Traverse the collected information and select machines within the same address concentration time range for allocation in order according to the obtained information until all orders under the current address in the current month are allocated or there is no capacity to be allocated within the same address concentration time range. At the same time, update the information so that the occupied machines can no longer be allocated;
[0190] After the information is distributed, the next address and model are searched for and distributed;
[0191] After completing all address conveniences for the current month, continue looking for address allocation for the next month.
[0192] For example, suppose address 1 requires 1.5X model 1, 0.5X model 2, and 2X model 3; address 2 requires 1.5X model 1 and 0.5X model 2. The allocation action is as follows: Figure 4 shown.
[0193] Preferably, reference Figure 5 ,The iterative strategies of the model address joint optimization algorithm include:
[0194] Get the number of iterations, model concentration importance, and decision step configuration information;
[0195] Calculate whether each machine has changed molds in each shift compared to the previous shift, and count the number of machines that have changed molds in each shift;
[0196] Count the number of addresses delivered in each shift, and calculate the degree of optimization of the model addresses encoded in each shift dimension based on the model concentration emphasis;
[0197] According to the framework solution of model scheduling and the mapping relationship of the address intensive allocation algorithm, the number of units produced in each address of the mold corresponding to each model is summarized, and the addresses are sorted according to the number of units produced;
[0198] Select the model, mold and address in order, reselect and allocate the molds and priority order corresponding to the model in the shift. Each time a mold is allocated, check whether the model allows simultaneous production, mold quantity limit, and machine production constraints. If the constraint check passes, use the corresponding mold. If the constraint check fails, select the next mold in the order for mold loading;
[0199] If it is possible to reduce mold switching and the number of addresses required can support single-shift allocation, the currently selected mold and address will be allocated;
[0200] If it is possible to reduce mold switching but the number of units required for the address is less than the number of units that the machine can produce in a shift, the currently selected mold, the remaining number of units to be produced at the current address, and the number of units required for the next sequential address in the current mold are allocated;
[0201] If it is possible to reduce mold switching but there is no mold corresponding to the required number of addresses that need to be allocated, select the next mold in the sequence for allocation.
[0202] Preferably, after optimizing the model production scheduling plan by using the intelligent algorithm based on model address intensive optimization, the optimized production scheduling plan is also evaluated, specifically:
[0203] Based on the average number of days each address is assigned Number of model switching between adjacent shifts Conduct a comprehensive evaluation and obtain a comprehensive evaluation:
[0204]
[0205] Obtain the optimal model production scheduling plan under multiple iterative versions as the final production scheduling plan;
[0206] Among them, m represents month information, a represents address information, i represents shift information, j represents machine information, and p represents model information. Among them, inf2 represents the number of mold changes when the model address joint optimization algorithm is input, inf3 represents the average distribution days when the model address joint optimization algorithm is input, nf2 represents the number of mold changes of the solution result under the current optimization, and nf3 represents the average distribution days of the solution result under the current optimization.
[0207] This module aims to evaluate the quality of the algorithm based on the iteration results, retain individuals with better fitness, thereby improving the algorithm optimization effect, and perform relevant robustness analysis to illustrate the robustness of the algorithm.
[0208] Table 1: Which models are produced in machine j during shift i.
[0209]
[0210]
[0211] The proposed allocation scheme for its address is shown in Table 2. The number before the comma in the table indicates the address of shipment, and the number after it indicates the number of units shipped. For example, the data [2,77] means that 77 refrigerators will be shipped to address 2.
[0212] Table 2: Proposed address allocation scheme
[0213]
[0214]
[0215] Robustness analysis of model production scheduling. Conduct robustness analysis on the model production scheduling plan. m represents the maximum value, f avg represents the average value, f dev Represents the deviation of the objective function. The data robustness analysis is shown in Table 3.
[0216] Table 3: Data robustness analysis
[0217]
[0218] The core of the heuristic algorithm based on domain rules of the present invention is to configure relevant business parameters based on the actual production environment conditions, and to mold the molds required by the model according to the work calendar shift by shift through the business parameters, so as to meet the model continuity under the condition of ensuring variety diversity. The core of the intelligent algorithm based on intensive optimization of model addresses is to allocate addresses on the basis of the first part of the model scheduling plan and optimize the model and address through intelligent algorithms, so as to improve the continuity of address to address.
[0219] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.
Claims
1. A resource-constrained parallel machine model scheduling method based on intensive strategy coordination, It is characterized in that include: Get order data; Calculate the model production schedule according to the order data using a domain-based heuristic algorithm; An intelligent algorithm based on intensive optimization of model addresses is used to optimize the production scheduling plan of the model.
2. A resource-constrained parallel machine model scheduling method based on intensive strategy coordination according to claim 1, It is characterized in that The obtaining of order data comprises: Obtain information on the number of machines in the factory, the models that the machines can produce, the time information of the machine opening hours, the time information required for model production, the resource information required for model production, and customer order information.
3. The resource-constrained parallel machine model scheduling method based on intensive strategy coordination according to claim 1, It is characterized in that The method of using a domain-based heuristic algorithm to calculate a model production scheduling plan according to the order data includes: Get order information, model information, resource information, work calendar information and configuration information; Convert order information, model information, resource information, work calendar information and configuration information into integer data by establishing key-value pair matching; Summarize the order information by month and model to obtain the model set that needs to be produced; According to the data preprocessing process, the redundant data in the model information and resource information of the current version is eliminated; Initialize the list of idle machines; Determine whether the current flight is the first flight; According to the current shift information, choose to continuously take over the lower die strategy or the upper die strategy based on domain knowledge and constraints for the machine; Output model production schedule.
4. The method for scheduling resource-constrained parallel machine orders based on a progressive locking strategy according to claim 3, It is characterized in that The method of selecting a continuous lower die strategy for the machine according to the current shift information or an upper die strategy based on domain knowledge and constraints includes: If the current shift is the first shift, select the continuous die transfer strategy for the machine, including: Obtain the model production schedule of the previous shift, and pre-allocate the model production schedule of the previous shift as the mold arrangement plan for the current shift; According to the production schedule of the previous shift, determine whether the molds that have been loaded have reached the maximum number of continuous production days; If the maximum number of continuous production days is reached, check the corresponding relationship between the remaining model demand and the current capacity. The current available capacity = the number of shifts that can be molded * the maximum number of molds of the current model * the UPH of the current model. If the number of machines with the current model cannot meet the minimum continuous production days limit when continuing to produce and the remaining demand can be completed within the monthly range, perform mold removal on one of the machines and check the current step again. If the maximum continuous production days have not been reached, determine whether the mold that has been molded has reached the maximum continuous production days. If the maximum continuous production days have been reached, check whether the orders within the monthly range of the corresponding model of the mold have all been completed. If they have all been completed, perform the mold removal process. If they have not been completed, do not perform the mold removal process. Check the door resources of the mold arrangement plan for the current shift. If the door resources are insufficient to support the production of the corresponding model of the mold, the mold will be processed. If the door resources are sufficient to support the production of the model, the mold will not be processed. Perform capacity verification on the remaining model information to verify whether the demand can be met within the available capacity range of the model. If it cannot be met, the model with constraints is selected first, and then the mold with a larger number of upper molds is selected for lower mold; Save a collection of completed die models.
5. The method for scheduling resource-constrained parallel machine orders based on a progressive locking strategy according to claim 3, It is characterized in that The method of selecting a continuous lower die strategy for the machine according to the current shift information or an upper die strategy based on domain knowledge and constraints includes: If the current shift is not the first shift, select a die-up strategy based on domain knowledge and constraints, including: Summarize the list of all idle machines in the current shift, and take the order of the order volume of the same model from high to low as the model priority; Prioritize the models that are subject to model mutual exclusion and machine production constraints, and then select the models with higher model priority according to the model priority. After sorting the model set to be produced according to the model priority, the list of candidate models for the lower upper mold in the current shift is obtained; The model set of the lower mold in the process of pre-allocation of the production plan for the previous shift is removed from the list of candidate models for the lower mold in the current shift; Summarize the list of molds that have been loaded in the non-idle machines in the current shift, and remove the mutually exclusive models under the production models of the non-idle machines from the list of candidate models for loading in the current shift; In the current shift, the list of candidate mold models is sorted from high to low. According to the remaining demand quantity of the order, the corresponding mold model is selected for mold processing. The maximum number of products that can be arranged = min (box mold production capacity, door resource supply capacity) / single mold daily production capacity; Determine whether the number of mold types exceeds the maximum number of mold types. If so, reselect the model for mold processing and remove the model corresponding to the current mold from the list of mold candidate models for the current shift. If the maximum number of mold types is not exceeded, determine whether the molds corresponding to the models are mutually exclusive. If the mold corresponding to the model to be newly molded is mutually exclusive with other molds that have been molded in the current shift, remove the model corresponding to the mold from the list of candidate models for molds to be molded in the current shift P, and reselect the model for mold processing; If the molds corresponding to the models are not mutually exclusive, determine whether the current mold-mounting machine and mold have mold-mounting area restrictions. If so, reselect the model from the mold-mounting alternative model list for the current shift to perform mold-mounting processing; If there is no restriction on the mold loading area, determine whether there is an idle machine in the current shift. If there is an idle machine in the current shift and the list of mold loading candidate models in the current shift has not been completely traversed, continue to select the model for mold loading; If there is no idle machine in the current shift or the list of upper die alternative models in the current shift has been completely traversed, it will jump to the next shift and re-perform the previous shift to pre-allocate the production schedule and the upper die strategy based on domain knowledge and constraints.
6. A resource-constrained parallel machine model scheduling method based on intensive strategy coordination according to claim 1, It is characterized in that The optimization of the model production scheduling plan by using an intelligent algorithm based on model address intensive optimization includes: Obtain order information, model information, resource information, work calendar information and previous model production schedule; Initialize the number of iterations; Summarize the order information by month, model and address to obtain the set of models and addresses that need to be produced; According to the address intensive allocation algorithm, an address allocation sequence is obtained; According to the address allocation sequence, the previous model scheduling plan, and the set of models and addresses to be produced, the address to be shipped for the number of models produced by each machine in each shift is determined and the information is summarized. At the same time, the set of models and addresses to be produced is updated; Based on the joint optimization algorithm of model address, the model production schedule of the previous level is combined with the model address shipment summary information optimization to obtain a new model production schedule, and verify whether the resources can be met based on the resource information; If resources are sufficient, output the optimized model production schedule.
7. A resource-constrained parallel machine model scheduling method based on intensive strategy coordination according to claim 6, It is characterized in that The method of obtaining the address allocation sequence according to the address intensive allocation algorithm includes: Get the number of iterations, population size, and mutation probability configuration information; Create a coding matrix by address and month through the configured population size, and randomly generate A different numbers of addresses by month to fill the matrix; Calculate the fitness value and sort each chromosome S in the population according to the fitness value; According to the roulette wheel method, two different chromosomes S1 and S2 are selected; Randomly select a gene from S1 as the intersection point C1, record the row information where C1 is located, randomly select a gene in a row in S2 as the intersection point C2, and perform positional crossover on C1 and C2; Check whether there is duplication or missing of coding sequence numbers in S1 and S2 after crossing; If there are repeated or missing numbers in the coding sequence, a repeated number is randomly selected and replaced with the missing number; Check whether the current situation meets the mutation conditions. If so, regenerate a new chromosome and insert it into the population, and remove those chromosomes with low fitness evaluation indicators. Check whether the current situation has reached the upper limit of the number of iterations. If so, output the optimal result obtained by the current iteration; If the upper limit of the number of iterations is not reached, the iterative optimization process will continue.
8. A resource-constrained parallel machine model scheduling method based on intensive strategy coordination according to claim 7, It is characterized in that The chromosome decoding using address intensive optimization algorithm includes: Extract the month information in the chromosome and the order information of each address in each month, and index each month one by one; Get the model production plan for the current month, index the addresses in order according to the address sequence information, and get the current address information; Summarize the order information of the current month and address by month, required quantity, model, and address, and sort them in descending order according to the number of models required to be produced at the same address; Through the shift and opening hours information provided by the box foaming line work calendar, combined with the hourly production capacity information in the box mold information table, the number of units that can be scheduled for production = model UPH * address intensive time * number of molds on the machine corresponding to the model, the information on how many units of each model can be produced within the time range of the batch shipment is obtained, and the number of models that can be produced is sorted from large to small; Traverse the collected information and select machines within the same address concentration time range for allocation in order according to the obtained information until all orders under the current address in the current month are allocated or there is no capacity to be allocated within the same address concentration time range. At the same time, update the information so that the occupied machines can no longer be allocated; After the information is distributed, the next address and model are searched for and distributed; After completing all address conveniences for the current month, continue looking for address allocation for the next month.
9. A resource-constrained parallel machine model scheduling method based on intensive strategy coordination according to claim 8, It is characterized in that The iteration strategies of the model-address joint optimization algorithm include: Get the number of iterations, model concentration importance, and decision step configuration information; Calculate whether each machine has changed molds in each shift compared to the previous shift, and count the number of machines that have changed molds in each shift; Count the number of addresses delivered in each shift, and calculate the degree of optimization of the model addresses encoded in each shift dimension based on the model concentration emphasis; According to the framework solution of model scheduling and the mapping relationship of the address intensive allocation algorithm, the number of units produced in each address of the mold corresponding to each model is summarized, and the addresses are sorted according to the number of units produced; Select the model, mold and address in order, reselect and allocate the molds and priority order corresponding to the model in the shift. Each time a mold is allocated, check whether the model allows simultaneous production, mold quantity limit, and machine production constraints. If the constraint check passes, use the corresponding mold. If the constraint check fails, select the next mold in the order for mold loading; If it is possible to reduce mold switching and the number of addresses required can support single-shift allocation, the currently selected mold and address will be allocated; If it is possible to reduce mold switching but the number of units required for the address is less than the number of units that the machine can produce in a shift, the currently selected mold, the remaining number of units to be produced at the current address, and the number of units required for the next sequential address in the current mold are allocated; If it is possible to reduce mold switching but there is no mold corresponding to the required number of addresses that need to be allocated, select the next mold in the sequence for allocation.
10. The resource-constrained parallel machine model scheduling method based on intensive strategy coordination according to claim 1, It is characterized in that After the model production scheduling scheme is optimized by using the intelligent algorithm based on model address intensive optimization, the optimized production scheduling scheme is evaluated, specifically: Based on the average number of days each address is assigned Number of model switching between adjacent shifts Conduct a comprehensive evaluation and obtain a comprehensive evaluation: Obtain the optimal model production scheduling plan under multiple iteration versions as the final production scheduling plan; Where m represents month information, a represents address information, i represents shift information, j represents machine information, and p represents model information. 2 Indicates the number of mold changes when inputting the model address joint optimization algorithm, inf 3 Indicates the average number of days allocated when inputting the model address joint optimization algorithm, nf 2 Indicates the number of mold changes of the solution result under the current optimization, nf 3 Indicates the average number of days for the solution results under the current optimization.
Citation Information
Patent Citations
Profile extrusion production scheduling method, readable storage medium and device
CN111882196A
Equipment production scheduling and intra-field logistics scheduling optimization method based on improved genetic algorithm
CN114971317A
Production scheduling method and apparatus for household appliance device, device, and storage medium
WO2023130709A1
Cited By
Foaming machine group production scheduling method under multi-target constraint
CN121433149A
