Intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning
By employing an intelligent scheduling method that combines multi-dimensional pre-decision-making and dynamic fine-tuning, the problems of time-consuming, labor-intensive, and resource-wasting traditional manual scheduling have been solved. This has enabled efficient and reliable intelligent scheduling for steel coil production, improving the quality of production planning and resource utilization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI PINJIAN INTELLIGENT TECH CO LTD
- Filing Date
- 2025-07-10
- Publication Date
- 2026-04-28
AI Technical Summary
Traditional manual scheduling is time-consuming and labor-intensive in steel coil production. It lacks global optimization capabilities, is prone to resource idleness and waste, and is prone to errors under complex processes, leading to increased production uncertainty.
An intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning is adopted. Through data preprocessing, batch partitioning, MIP model optimization and strategy decision-making, the optimal arrangement of steel coils is achieved, including material set filtering, classification, roll number screening and process interval insertion, to ensure the accuracy and efficiency of production planning.
It significantly reduces the time and errors of manual scheduling, improves the quality and reliability of production planning, optimizes resource utilization and production efficiency, and adapts to the needs of complex processes.
Smart Images
Figure CN121032025B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent scheduling technology, specifically an intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning. Background Technology
[0002] Steel coil scheduling is a crucial and highly complex step in the steel coil production process. The scheduling process involves multiple intricate technological steps, each requiring precise timing and resource allocation, especially when generating the production sequence plan, which necessitates comprehensive consideration of multiple dimensions such as steel coil resources and production concentration. In the early stages of steel coil production or when production tasks are relatively simple, manual scheduling can usually balance the various dimensions of the production sequence plan effectively. However, with constantly changing production demands and increasingly complex technological requirements, the limitations of traditional manual scheduling have become increasingly apparent. Manual scheduling typically consumes a significant amount of time, leading to decreased production efficiency. Furthermore, manual scheduling lacks global optimization capabilities, making it difficult to effectively coordinate resource allocation between steps, resulting in resource idleness and waste. Due to over-reliance on experience, manual scheduling is prone to human error in increasingly complex processes, further increasing the uncertainty of the production process. Therefore, combining advanced algorithmic models to address the problems existing in traditional scheduling methods and promoting the intelligentization of steel coil production has significant practical implications. Summary of the Invention
[0003] The purpose of this invention is to provide a solution to the problems raised in the prior art.
[0004] To achieve the above objectives, the present invention provides the following technical solution: an intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning, the method comprising the following steps:
[0005] Step S1: Obtain the basic attributes of the steel coils to be produced, preprocess the data to generate data for intelligent scheduling, filter the data that does not need to be scheduled according to the scheduling requirements, and output the material set.
[0006] Step S2: Receive the output material set, divide the material set into batches according to the steel type, initially determine the order of the scheduling batch list, classify the material set according to the scheduling batch list, and divide the material set into main scheduling materials, bounce materials and insertable materials.
[0007] Step S3: Based on the output material set, exclusion list of rollers, and number of selectable rollers, filter out the roller numbers with the largest material quantity in the material set;
[0008] Step S4: Based on the quantity of main materials, determine the scheduling strategy, establish the MIP model, handle batch switching, handle material insertion during process intervals, and perform post-processing on the scheduled plans.
[0009] Furthermore, the process of outputting the material set includes:
[0010] Step S1.1: Clean and mark the initial data of the steel coils in the warehouse to obtain the order width, order thickness, available work rolls, weight, welding group fields and their corresponding values required for scheduling;
[0011] Step S1.2: Based on the stage, order, and process, filter out materials that do not need to be scheduled from all steel coil data, and output the remaining data as a set of materials to be scheduled, thus realizing intelligent data filtering under personalized needs.
[0012] Further steps to determine the order of the schedule set batch list include:
[0013] Step S2.1: Set the default batching order based on manual experience, and use the batching category of the connecting volumes as the first batch in the batching list;
[0014] Step S2.2: For batches with production time constraints, check if all materials have arrived. If not, remove them. If they have, compare the thickness d of the connecting coil with the thickness a of the steel coil in the batch. If d > a, add them to the batch list in order. Otherwise, treat the steel coil in the batch as the first batch. This prioritizes the processing of time-limited batches to ensure the timely execution of the production plan.
[0015] Step S2.3: Traverse the remaining batches according to the default batch order, and put them into the batch list according to the total tonnage of the materials. Output the batch list as the batch list for subsequent scheduling.
[0016] Further steps in classifying the material set include:
[0017] Step S2.4: According to the process of the order, classify the material set output in step S1.2, set a width threshold, sort the materials in the material set whose width is greater than the threshold by width from widest to narrowest and classify them as main materials, sort the materials in the material set whose width is less than the threshold by width from narrowest to widest and classify them as bounce materials, and classify other batch materials that can be mixed with the current batch as insertable materials if the process allows, so as to adapt to the optimal scheduling under the condition of different roll widths after the completion of the previous batch.
[0018] Furthermore, the process of selecting roller numbers includes:
[0019] Step S3.1: Based on the marked available working rollers for the material, take the set of available working rollers for the input material and generate a unique set of available working rollers.
[0020] Step S3.2: Traverse all available working rollers, count the materials that can be processed with the current traversal number in the input material set, and record them in the database. The primary key is the roller number, and the value is the corresponding material quantity.
[0021] Step S3.3: Select the n roller numbers with the most corresponding materials and output them as a set {1,2,3,...,n}. Based on the material set required for production scheduling, obtain all non-repeating available work rollers. During the production process, the most suitable work rollers can be accurately selected according to material properties and process requirements to improve work efficiency.
[0022] Furthermore, the process of deciding on scheduling strategies includes:
[0023] Step S4.1: Record the width of the steel coil at the end of the previous batch as width A. Set a width judgment threshold. If A is greater than the threshold, execute the main arrangement strategy, arrange production from widest to narrowest, and connect the virtual coil at the end. If A is less than the threshold, execute the reverse jump strategy, arrange production from narrowest to widest, and connect the widest coil of the main batch that will be continued at the end. This maximizes the working efficiency and material utilization between the two batches.
[0024] Furthermore, the process of establishing a MIP model includes:
[0025] Step S4.2: By default, the MIP model is built according to the main sorting strategy. The main sorting materials and inserted materials are optimized and sorted. The goal is to output a material sorting scheme that minimizes the total connection weight.
[0026] The optimization of its connection relationship can be represented by the following mathematical model:
[0027]
[0028] The uniqueness constraint of its connection method is as follows:
[0029]
[0030] The following constraints prevent repeated connections:
[0031]
[0032] The virtual volume of its main sorting model must have the following constraints selected:
[0033]
[0034] The real volume of its bounce model must have the following constraints selected:
[0035]
[0036] The main production constraints of the model are as follows: Define a set of coil serial numbers WS that must be strictly sorted from wide to narrow, j∈WS and i≠j, u j Defined as the sequence number of steel coil j in the main layout model output plan, a ijDefined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j. This constraint strictly ensures that the steel coils are sorted from widest to narrowest.
[0037]
[0038] The production constraints of its bounce model are as follows: Define a set TS of steel coil serial numbers that must be strictly sorted from narrow to wide, j∈TS and i≠j, u j Defined as the sequence number of steel coil j in the output plan of the bounce model, a ij Defined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j. This constraint strictly ensures that the steel coils are sorted from narrowest to widest.
[0039]
[0040] The above G is defined as the set of all coil numbers, i and j are defined as coil serial numbers, r is defined as the actual coil serial number, and i, j, o, r, v ∈ G and i ≠ j; x ij Defined as a variable of 0 or 1, used to determine whether coil i and coil j are connected; w ij The connection weight between steel coil i and steel coil j is used to measure the degree of fit between them; u i Define the sequence number of steel coil i in the current model output plan, u j Define the sequence number of steel coil j in the current model output plan; define N as the total number of steel coils; a ij Defined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j; y i Defined as a variable of 0 or 1, used to determine whether steel coil i is selected into the plan, y j Defined as a variable of 0 or 1, used to determine whether steel coil j is selected for the plan; y v With y r Both are defined as 1 to ensure that the virtual volume v of the main row model and the real volume r of the bounce model must be selected; M is defined as a maximum value;
[0041] The above model and constraints strictly guarantee that steel coils are uniquely and non-repeatablely sorted according to width under different scheduling strategies.
[0042] Furthermore, the process of handling batch switching includes:
[0043] Step S4.3: By statistically analyzing the material quantity of each batch in the latest material set, define the upper limit tonnage, total tonnage, and remaining order tonnage of the current batch for a single production run, and define the lower limit tonnage and total tonnage of the next batch for a single interval.
[0044] Step S4.4: Set the production upper limit gap threshold. Assuming that the difference between the output of each batch and the upper limit of a single production is less than this threshold, calculate the future batch number = (remaining order tonnage in the current production cycle - total tonnage) / upper limit tonnage of a single production. The current batch number = total tonnage / upper limit tonnage of a single production. Set the production lower limit gap threshold. Assuming that when the next batch is used as an interval material, the difference between the output of each batch and the lower limit of a single production is less than this threshold, calculate the maximum number of times it can be used as a process interval = total tonnage / lower limit tonnage of a single interval.
[0045] Step S4.5: Compare the ratio of the number of production runs of the current batch to the next batch. Based on the experience value determined by the planning personnel based on their production experience, decide on the conversion between batches. When the remaining material in the current batch is not much, switch batches in a timely manner and prioritize the production of other batches to achieve a significant improvement in work efficiency.
[0046] Furthermore, the process of inserting materials into the processing interval includes:
[0047] Step S4.6: Based on the ratio of the number of times the current batch can be produced to the next batch as stated in step S4.5, the current batch has reached the upper limit of single production scheduling. According to the ratio of the remaining inventory of the current batch material to the batch material of the process interval in the latest material set, the MIP model of the master scheduling strategy is called to incorporate the batch material of the process interval into the latest plan.
[0048] Furthermore, the post-processing of the scheduled plans includes:
[0049] Step S4.7: Obtain the latest plan calculated by MIP. Obtain all steel coils that are not in the current batch from the latest material set. Iterate through each coil in the plan, calculate the average specifications of the coils before and after it, and find a steel coil in the material set that uses the same roller, meets the process rules, and is closer to the average specifications than the original steel coil to replace it. Put the original steel coil into the material set and merge the plan after material replacement into the scheduled plan.
[0050] Compared with the prior art, the beneficial effects of the present invention are:
[0051] 1. Data Preprocessing and Multi-Dimensional Screening. This invention employs a systematic approach to preprocess the scattered steel coil data and selects suitable scheduling materials through three dimensions: "batch production, centralized roll usage, and scheduling strategy." This effectively solves the problem of difficulty in screening and making decisions when faced with a large amount of steel coil data in manual scheduling, significantly reducing time and effort.
[0052] 2. Differentiated scheduling strategies. This invention divides the scheduling process into two strategies: "main scheduling" and "reverse scheduling," which correspond to wide-to-narrow and narrow-to-wide arrangements, respectively. This effectively simulates the logic of manual scheduling, making the scheduling plan more in line with human expectations.
[0053] 3. Optimal Arrangement. Under different scheduling strategies, this invention establishes and solves a MIP model that conforms to the process rules to achieve the optimal arrangement of steel coils, avoiding errors in the work plan due to the complexity of scheduling rules, and significantly improving the quality and reliability of the plan. Attached Figure Description
[0054] Figure 1 This is the overall flowchart of the intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning of the present invention;
[0055] Figure 2 for Figure 1 Flowchart of the decision logic for the batch list of medium-sized scheduling sets;
[0056] Figure 3 for Figure 1 Flowchart of the decision logic for the middle roller number decision module;
[0057] Figure 4 for Figure 1 Flowchart of the decision logic for material insertion during intermediate process intervals;
[0058] Figure 5 for Figure 1 CIMC batch switching decision logic flowchart;
[0059] Figure 6 for Figure 1 Flowchart of the decision-making logic for post-planning processing;
[0060] Figure 7 for Figure 1 Diagram of production scheduling using the main scheduling / reverse scheduling strategy. Detailed Implementation
[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0062] Example: Figures 1-7 As shown, this invention provides a technical solution: an intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning, the method comprising the following steps:
[0063] Step S1: Obtain the basic attributes of the steel coils to be produced, preprocess the data to generate data for intelligent scheduling, filter the data that does not need to be scheduled according to the scheduling requirements, and output the material set.
[0064] Step S2: Receive the output material set, divide the material set into batches according to the steel type, initially determine the order of the scheduling batch list, classify the material set according to the scheduling batch list, and divide the material set into main scheduling materials, bounce materials and insertable materials.
[0065] Step S3: Based on the output material set, exclusion list of rollers, and number of selectable rollers, filter out the roller numbers with the largest material quantity in the material set;
[0066] Step S4: Based on the quantity of main materials, determine the scheduling strategy, establish the MIP model, handle batch switching, handle material insertion during process intervals, and perform post-processing on the scheduled plans.
[0067] Furthermore, the process of outputting the material set includes:
[0068] Step S1.1: Clean and mark the initial data of the steel coils in the warehouse to obtain the order width, order thickness, available work rolls, weight, welding group fields and their corresponding values required for scheduling;
[0069] Step S1.2: Based on the stage, order, and process, filter out materials that do not need to be scheduled from all steel coil data, and output the remaining data as a set of materials to be scheduled, thus realizing intelligent data filtering under personalized needs.
[0070] Further steps to determine the order of the schedule set batch list include:
[0071] Step S2.1: Set the default batching order based on manual experience, and use the batching category of the connecting volumes as the first batch in the batching list;
[0072] Step S2.2: For batches with production time constraints, check if all materials have arrived. If not, remove them. If they have, compare the thickness d of the connecting coil with the thickness a of the steel coil in the batch. If d > a, add them to the batch list in order. Otherwise, treat the steel coil in the batch as the first batch. This prioritizes the processing of time-limited batches to ensure the timely execution of the production plan.
[0073] Step S2.3: Traverse the remaining batches according to the default batch order, and put them into the batch list according to the total tonnage of the materials. Output the batch list as the batch list for subsequent scheduling.
[0074] Further steps in classifying the material set include:
[0075] Step S2.4: According to the process of the order, classify the material set output in step S1.2, set a width threshold, sort the materials in the material set whose width is greater than the threshold by width from widest to narrowest and classify them as main materials, sort the materials in the material set whose width is less than the threshold by width from narrowest to widest and classify them as bounce materials, and classify other batch materials that can be mixed with the current batch as insertable materials if the process allows, so as to adapt to the optimal scheduling under the condition of different roll widths after the completion of the previous batch.
[0076] Furthermore, the process of selecting roller numbers includes:
[0077] Step S3.1: Based on the marked available working rollers for the material, take the set of available working rollers for the input material and generate a unique set of available working rollers.
[0078] Step S3.2: Traverse all available working rollers, count the materials that can be processed with the current traversal number in the input material set, and record them in the database. The primary key is the roller number, and the value is the corresponding material quantity.
[0079] Step S3.3: Select the n roller numbers with the most corresponding materials and output them as a set {1,2,3,...,n}. Based on the material set required for production scheduling, obtain all non-repeating available work rollers. During the production process, the most suitable work rollers can be accurately selected according to material properties and process requirements to improve work efficiency.
[0080] Furthermore, the process of deciding on scheduling strategies includes:
[0081] Step S4.1: Record the width of the steel coil at the end of the previous batch as width A. Set a width judgment threshold. If A is greater than the threshold, execute the main arrangement strategy, arrange production from widest to narrowest, and connect the virtual coil at the end. If A is less than the threshold, execute the reverse jump strategy, arrange production from narrowest to widest, and connect the widest coil of the main batch that will be continued at the end. This maximizes the working efficiency and material utilization between the two batches.
[0082] Furthermore, the process of establishing a MIP model includes:
[0083] Step S4.2: By default, the MIP model is built according to the main sorting strategy. The main sorting materials and inserted materials are optimized and sorted. The goal is to output a material sorting scheme that minimizes the total connection weight.
[0084] The optimization of its connection relationship can be represented by the following mathematical model:
[0085]
[0086] The uniqueness constraint of its connection method is as follows:
[0087]
[0088] The following constraints prevent repeated connections:
[0089]
[0090] The virtual volume of its main sorting model must have the following constraints selected:
[0091]
[0092] The real volume of its bounce model must have the following constraints selected:
[0093]
[0094] The main production constraints of the model are as follows: Define a set of coil serial numbers WS that must be strictly sorted from wide to narrow, j∈WS and i≠j, u j Defined as the sequence number of steel coil j in the main layout model output plan, a ij Defined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j. This constraint strictly ensures that the steel coils are sorted from widest to narrowest.
[0095]
[0096] The production constraints of its bounce model are as follows: Define a set TS of steel coil serial numbers that must be strictly sorted from narrow to wide, j∈TS and i≠j, u j Defined as the sequence number of steel coil j in the output plan of the bounce model, a ij Defined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j. This constraint strictly ensures that the steel coils are sorted from narrowest to widest.
[0097]
[0098] The above G is defined as the set of all coil numbers, i and j are defined as coil serial numbers, r is defined as the actual coil serial number, and i, j, o, r, v ∈ G and i ≠ j; x ij Defined as a variable of 0 or 1, used to determine whether coil i and coil j are connected; w ij The connection weight between steel coil i and steel coil j is used to measure the degree of fit between them; u i Define the sequence number of steel coil i in the current model output plan, u j Define the sequence number of steel coil j in the current model output plan; define N as the total number of steel coils; a ij Defined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j; y i Defined as a variable of 0 or 1, used to determine whether steel coil i is selected into the plan, y jDefined as a variable of 0 or 1, used to determine whether steel coil j is selected for the plan; y v With y r Both are defined as 1 to ensure that the virtual volume v of the main row model and the real volume r of the bounce model must be selected; M is defined as a maximum value;
[0099] The above model and constraints strictly guarantee that steel coils are uniquely and non-repeatablely sorted according to width under different scheduling strategies.
[0100] Furthermore, the process of handling batch switching includes:
[0101] Step S4.3: By statistically analyzing the material quantity of each batch in the latest material set, define the upper limit tonnage, total tonnage, and remaining order tonnage of the current batch for a single production run, and define the lower limit tonnage and total tonnage of the next batch for a single interval.
[0102] Step S4.4: Set the production upper limit gap threshold. Assuming that the difference between the output of each batch and the upper limit of a single production is less than this threshold, calculate the future batch number = (remaining order tonnage in the current production cycle - total tonnage) / upper limit tonnage of a single production. The current batch number = total tonnage / upper limit tonnage of a single production. Set the production lower limit gap threshold. Assuming that when the next batch is used as an interval material, the difference between the output of each batch and the lower limit of a single production is less than this threshold, calculate the maximum number of times it can be used as a process interval = total tonnage / lower limit tonnage of a single interval.
[0103] Step S4.5: Compare the ratio of the number of production runs of the current batch to the next batch. Based on the experience value determined by the planning personnel based on their production experience, decide on the conversion between batches. When the remaining material in the current batch is not much, switch batches in a timely manner and prioritize the production of other batches to achieve a significant improvement in work efficiency.
[0104] Furthermore, the process of inserting materials into the processing interval includes:
[0105] Step S4.6: Based on the ratio of the number of times the current batch can be produced to the next batch as stated in step S4.5, the current batch has reached the upper limit of single production scheduling. According to the ratio of the remaining inventory of the current batch material to the batch material of the process interval in the latest material set, the MIP model of the master scheduling strategy is called to incorporate the batch material of the process interval into the latest plan.
[0106] Furthermore, the post-processing of the scheduled plans includes:
[0107] Step S4.7: Obtain the latest plan calculated by MIP. Obtain all steel coils that are not in the current batch from the latest material set. Iterate through each coil in the plan, calculate the average specifications of the coils before and after it, and find a steel coil in the material set that uses the same roller, meets the process rules, and is closer to the average specifications than the original steel coil to replace it. Put the original steel coil into the material set and merge the plan after material replacement into the scheduled plan.
[0108] In a specific implementation scenario, during a production scheduling process, the basic attributes of the steel coils to be scheduled are obtained, preprocessed to generate data for intelligent scheduling, and output as a material set M. Material set M is divided into five batches (A, B, C, D, and E) based on the process and steel type. Batch B uses a special process, requires process intervals, and is prioritized for production; batch C requires production at a specific time point; batch E serves as the process interval for batch B; batches A and D have no special requirements. Actual production proceeds in a fixed batch order, therefore, a decision needs to be made regarding the production sequence of the batches.
[0109] After batch division is completed, the factory's production experts evaluate the batches and provide a default batch order list (batch_list). Considering the continuity of steel coil production, the last coil of the previous implemented production plan is used as the connecting coil for the current plan, and the batch type of the connecting coil is used as the first batch in the scheduled batch list (selected_batch_list).
[0110] For batch C, which needs to be produced at a specific time, check whether the current time meets its production requirements. If not, batch C is removed from the batch_list and is no longer an option for this production schedule. If it does meet the requirements, batch C is added to the selected_batch_list, and its order is determined: assuming thick_1 represents the maximum thickness of the connecting batch C, to reduce specification fluctuations during batch switching, based on the thickness d of the connecting roll, if d > thick_1, batch C is added to the batch list in order; otherwise, batch C is inserted as the first item in the batch list.
[0111] Batch production requires meeting a certain weight limit. Therefore, for batches not selected in the selected_batch_list, the batches that meet weight_1 will be added to the selected_batch_list according to the default batch order batch_list.
[0112] In actual scheduling, the first batch in selected_batch_list is selected by default as the main production batch for the remaining steps.
[0113] Even within the same batch and scheduling strategy, steel coil production may utilize different work rolls. To reduce production costs and complexity associated with changing work rolls, a roll number decision module determines the work rolls for the current batch, concentrating materials requiring the same rolls in the scheduling process as much as possible. The roll number decision module selects the final set of work rolls based on the input material set and the parameters `excluded_roll_list` (a list of work roll numbers not to be considered) and `roll_num` (the number of output roll numbers).
[0114] Based on the available working rolls of the marked materials, a set of available working fields for the input materials is obtained to generate a unique set of available working rolls, and the roll numbers in excluded_roll_list are removed.
[0115] Iterate through all available working rollers, count the materials that can be processed using the currently traversed roller number in the input material set, and record them as a dict, with the roller number as the primary key and the corresponding material quantity as the value.
[0116] Select the rolls with the most corresponding material quantities (roll_num) and output them as a set {roll_1, roll_2, ...}.
[0117] The production of some materials requires interval processing. The production of batch B requires inserting at least y tons of batch E material every x tons as a process interval to ensure the quality of batch B. Therefore, when scheduling batch B, it is necessary to determine whether the already scheduled materials meet the conditions for inserting process interval materials. If not, skip this step; if they do, proceed with inserting the process interval materials.
[0118] Based on the existing schedule or the latest schedule calculated by the MIP model, determine whether the current batch is batch B that requires a process interval, and whether the production output has reached the upper limit. If yes, insert the process interval material; otherwise, exit the current step.
[0119] The insertion amount of materials in the process interval is not fixed. Therefore, under the premise of meeting its lower limit, the appropriate insertion amount will be calculated based on the ratio of the remaining batch B materials in the latest material set M to the batch E materials in the process interval.
[0120] After determining the process material insertion quantity, the MIP model of the master scheduling strategy is called to arrange the order of process interval materials and incorporate them into the already scheduled plan or the latest plan calculated by the MIP model.
[0121] After screening based on three dimensions—production scheduling strategy, batching, and roll usage—if the main production material P is sufficient, the main production scheduling strategy is implemented. If the material does not meet the requirements of the main production scheduling strategy, such as insufficient roll count, then other roll numbers under the current batch must be selected for production scheduling. Since the main production scheduling strategy generally proceeds from wide to narrow, this is usually located at a relatively narrow width. Therefore, a rebound strategy is first used to widen the width.
[0122] Special batch B requires priority scheduling, but as scheduling progresses and the set of materials to be scheduled remains unchanged, its priority will change. When there is little material remaining in batch B, but a large amount of batch B material will be received into the warehouse later, it can be transferred to other batches for scheduling. Therefore, the existing batches and the batch list selected_batch_list need to be adjusted.
[0123] Switching between batches primarily occurs between a specific batch B and other batches. Decisions are made by statistically analyzing the material quantities of each batch in the latest material set M. Batch B is defined as having a single-production upper limit of tonnage B_s_weight, a total tonnage B_weight, and remaining order tonnage B_os_weight within the current production cycle; batch E has a single-interval lower limit of tonnage E_s_weight and a total tonnage E_weight.
[0124] Assuming that batch B's production scheduling is close to the upper limit of a single production run, calculate the possible number of batches B in the future: B_batch = (B_os_weight - B_weight) / B_s_weight; the current number of batches that can be scheduled is: B_batch_now = B_weight / B_s_weight; assuming that when batch E is used as an interval material, its production scheduling is close to the lower limit of a single production run, calculate the maximum number of times batch E can currently be used as a process interval: E_batch_now = E_weight / E_s_weight.
[0125] If the current batch is B, compare the ratio of the number of times batch B can be produced to the number of times batch E can be produced. Based on empirical values H1 and H2, decide on the conversion between batches. These empirical values are determined by the planners based on production experience. The pseudocode is shown below:
[0126] If B_batch_now / E_batch_now > H1, continue batching B;
[0127] If B_batch_now > 1, then continue with batch B;
[0128] If: B_batch / (E_batch_now-(B_batch_now / E_batch_now)>H2, then continue batching B;
[0129] else: Transfer to other batches;
[0130] If the current batch is another type, the decision to switch between batches is made based on the number of batches that can be set in batch B and the empirical value H3. The pseudocode is as follows:
[0131] If B_batch_now > 1, then convert batch B;
[0132] If B_batch_now / E_batch_now > H3, then convert batch B;
[0133] else: Continue with the current batch;
[0134] Scheduling strategies are divided into primary scheduling strategies and bounce strategies. For example... Figure 7 As shown, the MIP model corresponding to the main sorting strategy arranges production from wide to narrow, with a virtual roll at the end; the MIP model corresponding to the bounce strategy arranges production from narrow to wide, with the widest roll of the main sorting to be continued at the end. In the actual sorting result of each strategy, the last roll will serve as the starting roll for the next strategy. The main sorting strategy calls the corresponding MIP model to optimize the sorting of the main sorting material P and the inserted material I, with the goal of outputting a material sorting scheme that minimizes the total connection weight.
[0135] The optimization of its connection relationship can be represented by the following mathematical model:
[0136]
[0137] The uniqueness constraint of its connection method is as follows:
[0138]
[0139] The following constraints prevent repeated connections:
[0140]
[0141] The virtual volume of its main sorting model must have the following constraints selected:
[0142]
[0143] The real volume of its bounce model must have the following constraints selected:
[0144]
[0145] The main production constraints of the model are as follows: Define a set of coil serial numbers WS that must be strictly sorted from wide to narrow, j∈WS and i≠j, u j Defined as the sequence number of steel coil j in the main layout model output plan, a ij Defined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j. This constraint strictly ensures that the steel coils are sorted from widest to narrowest.
[0146]
[0147] The production constraints of its bounce model are as follows: Define a set TS of steel coil serial numbers that must be strictly sorted from narrow to wide, j∈TS and i≠j, u jDefined as the sequence number of steel coil j in the output plan of the bounce model, a ij Defined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j. This constraint strictly ensures that the steel coils are sorted from narrowest to widest.
[0148]
[0149] The above G is defined as the set of all coil numbers, i and j are defined as coil serial numbers, r is defined as the actual coil serial number, and i, j, o, r, v ∈ G and i ≠ j; x ij Defined as a variable of 0 or 1, used to determine whether coil i and coil j are connected; w ij The connection weight between steel coil i and steel coil j is used to measure the degree of fit between them; u i Define the sequence number of steel coil i in the current model output plan, u j Define the sequence number of steel coil j in the current model output plan; define N as the total number of steel coils; a ij Defined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j; y i Defined as a variable of 0 or 1, used to determine whether steel coil i is selected into the plan, y j Defined as a variable of 0 or 1, used to determine whether steel coil j is selected for the plan; y v With y r Both are defined as 1 to ensure that the virtual volume v of the main row model and the real volume r of the bounce model must be selected; M is defined as a maximum value.
[0150] Given a set of steel coils G = {1, 2, 3, 4, 5}, corresponding to five steel coils, the final output should be the arrangement order of these five steel coils.
[0151] The widths of the five steel coils are as follows:
[0152] Steel coils 1-100 (mm);
[0153] Steel coils 2-200 (mm);
[0154] Steel coils 3-400 (mm);
[0155] Steel coil 4-230 (mm);
[0156] Steel coils 5-250 (mm);
[0157] Where, for the steel coil t∈WS = {1, 5, 3}, in the current model output plan, the requirement of width from wide to narrow must be met, that is, the width of any steel coil arranged before the steel coil t must be greater than or equal to the width of the steel coil t.
[0158] Regarding the uniqueness constraint of the connection method:
[0159]
[0160] In the current model output plan, there is only one way to connect any two steel coils. Assuming that the next steel coil after steel coil 1 is steel coil 3, then steel coil 1 cannot be immediately followed by steel coils 2, 4, or 5, ensuring the uniqueness of the final arrangement order.
[0161] For the constraint that connections cannot be repeated:
[0162]
[0163] This formula is a classic constraint in graph theory in operations research (it has been theoretically proven and does not need to be proven again). This formula restricts the occurrence of "cycles" in the model, that is, steel coil 1 is followed by steel coil 3, and steel coil 3 is followed by steel coil 1. This does not conform to the reality. This formula can be used to restrict the final arrangement order to be non-repeating.
[0164] For the production constraints of the master-slot model:
[0165]
[0166] Production constraints of the bounce model:
[0167]
[0168] The above two equations are explained in two parts: left and right.
[0169] Right side of the equation: y i / y j This indicates whether coil i and coil j are included in the plan output by the model. If y i / y j If there exists one or more values equal to 0, meaning that coil i or coil j is not included in the output plan of the model, then the right side of the equation is a maximum value, and the constraint is invalid; if y i / y j Both are 1, the right side of the expression = 0, and the constraint takes effect.
[0170] Left side of the equation: a ij As a pre-calculated value, if coil j is wider than coil i, then a ij =1; otherwise a ij =0. u i / u j This represents the sequence number of coil i / coil j in the current model output plan.
[0171] If both coils i and j are selected into the main layout model output plan, then the right side of the equation = 0. Based on the widths of the previous five coils, coil 1 (100mm) is placed at position 2 in the main layout model output plan as coil i, i.e., u i=2, steel coil 2 (200mm) is placed in position 3 in the main layout model output plan as steel coil j, i.e., u j =3,
[0172] The numerical value of this expression is as follows:
[0173] The constraint (3-2)*1<= 0 is not valid, and the current order does not meet the requirements. Because j∈WS, for a steel coil j with a width of 200mm, the steel coils preceding it must have a width greater than or equal to 200mm, while the steel coil i preceding it currently has a width of 100mm.
[0174] If the output plan meets the constraints, then coil j should be placed before coil i. The order of coil i and coil j should be swapped to make u i =3, u j If the value is 2, then the corresponding constraint values are as follows:
[0175] (2-3)*1<=0, the constraint holds;
[0176] If both coils i and j are selected into the rebound model output plan, then the right side of the equation = 0. Based on the widths of the previous five coils, coil 5 (250mm) is placed at position 3 in the rebound model output plan as coil i, i.e., u i =3, steel coil 3 (400mm) is placed in position 1 of the rebound model output plan as steel coil j, i.e., u j =1,
[0177] The numerical value of this expression is as follows:
[0178] The constraint (3-1)*1<=0 is not valid, and the current order does not meet the requirements. Because j∈TS, a steel coil j with a width of 400mm requires that the width of the steel coil following it be greater than or equal to 400mm, but now the width of the steel coil i following it is 250mm.
[0179] If the output plan meets the constraints, then coil j should be placed after coil i. The order of coil i and coil j should be swapped to make u i =1, u j If the value is 3, then the corresponding constraint values are as follows:
[0180] (1-3)*1<=0, the constraint holds.
[0181] If the MIP model has a solution, obtain the MIP calculation result, determine whether the current batch is a special batch B and has reached the single production scheduling limit, and if so, insert the interval material; otherwise, skip this step. If the MIP model has no solution, it is considered that the main material under the roller screening of the current batch cannot be scheduled, and the bounce strategy is entered, calling the corresponding MIP model to sort the bounce material P.
[0182] The bounce plan prioritizes minimizing material placement, quickly widening the width to the widest coil width of the connecting main batch. If the MIP model for the bounce strategy has a solution, material replacement optimization is performed. The latest plan calculated by MIP is obtained, and all coils not in the current batch are retrieved from the latest material set M to form SM. Each coil in the plan is traversed, calculating the average specification of its preceding and following coils. A coil with the same roll, meeting the process rules, and closer to the average specification than the original coil is found in the material set SM for replacement. The original coil is then added to the material set SM. The material-replaced plan is merged into the already scheduled plan, making the specification transition of the already scheduled plan smoother.
[0183] If the MIP model of the bounce strategy has no solution, it is considered that the current batch cannot continue to be scheduled, and the batch is removed from the selected_batch_list and the next batch scheduling is carried out; if the selected_batch_list is empty, it is considered that all batch scheduling is completed and the scheduling is exited.
[0184] If the scheduled time has reached the preset duration, the scheduling ends; otherwise, the scheduling of the current batch continues.
[0185] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.
Claims
1. An intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning, characterized in that, The method includes the following steps: Step S1: Obtain the basic attributes of the steel coils to be produced, preprocess the data to generate data for intelligent scheduling, filter the data that does not need to be scheduled according to the scheduling requirements, and output the material set. Step S2: Receive the material set output in step S1, divide the material set into batches according to the steel type, preliminarily determine the order of the scheduling batch list, classify the material set according to the scheduling batch list, and divide the material set into main scheduling materials, rebound materials and insertable materials. Step S3: Using the material set, exclusion list of rollers and number of selectable rollers output in step S1, filter out the roller numbers with the largest material quantity in the material set; Step S4: Based on the quantity of main materials determined in step S2, decide on the scheduling strategy, establish the MIP model, handle batch switching, handle material insertion during process intervals, and perform post-processing on the scheduled plans. The decision scheduling strategy includes: Step S4.1: Record the width of the last coil of the previous batch as width A. Set a width judgment threshold. If A is greater than the threshold, execute the main rowing strategy, arrange production from widest to narrowest, and connect the virtual coil at the end. If A is less than the threshold, execute the reverse jump strategy, arrange production from narrowest to widest, and connect the widest coil of the main row that will be continued at the end. The process of establishing the MIP model includes: Step S4.2: By default, the MIP model is built according to the main sorting strategy. The main sorting materials and inserted materials are optimized and sorted. The goal is to output a material sorting scheme that minimizes the total connection weight. The optimization of its connection relationship can be represented by the following mathematical model: ； The uniqueness constraint of its connection method is as follows: ； The following constraints prevent repeated connections: ； The virtual volume of its main sorting model must have the following constraints selected: ； The real volume of its bounce model must have the following constraints selected: ； The main production constraints of the model are as follows: Define a set of coil serial numbers WS that must be strictly sorted from wide to narrow, j∈WS and i≠j, u j Defined as the sequence number of steel coil j in the main layout model output plan, a ij Defined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j. This constraint strictly ensures that the steel coils are sorted from widest to narrowest. ； Its anti-boost model production constraints are as follows: Define a set TS of steel coil serial numbers that must be strictly sorted from narrow to wide, j∈TS and i≠j, u j Defined as the sequence number of steel coil j in the output plan of the bounce model, a ij Defined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j. This constraint strictly ensures that the steel coils are sorted from narrowest to widest. ； The above G is defined as the set of all coil numbers, i and j are defined as coil serial numbers, r is defined as the actual coil serial number, and i, j, o, r, v ∈ G and i ≠ j; x ij Defined as a variable of 0 or 1, used to determine whether coil i and coil j are connected; w ij The connection weight between steel coil i and steel coil j is used to measure the degree of fit between them; u i Define the sequence number of steel coil i in the current model output plan, u j Define the sequence number of steel coil j in the current model output plan; define N as the total number of steel coils; a ij Defined as a variable of 0 or 1, used to compare the widths of steel coil i and steel coil j; y i Defined as a variable of 0 or 1, used to determine whether steel coil i is selected into the plan, y j Defined as a variable of 0 or 1, used to determine whether steel coil j is selected for the plan; y v With y r Both are defined as 1 to ensure that the virtual volume v of the main row model and the real volume r of the bounce model must be selected; M is defined as a maximum value.
2. The intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning according to claim 1, characterized in that, The process of outputting the material set in step S1 includes: Step S1.1: Clean and mark the initial data of the steel coils in the warehouse to obtain the order width, order thickness, available work rolls, weight, welding group fields and their corresponding values required for scheduling; Step S1.2: Based on the stage, order, and process, filter out materials that do not need to be scheduled from all steel coil data, and output the remaining data as a set of materials to be scheduled.
3. The intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning according to claim 1, characterized in that, The step of determining the order of the scheduling batch list in step S2 includes: Step S2.1: Set the default batching order based on manual experience, and use the batching category of the connecting volumes as the first batch in the batching list; Step S2.2: For batches with production time constraints, check if all materials have arrived. If not, remove them. If so, compare the thickness d of the connecting coil with the thickness a of the steel coil in the batch. If d > a, add them to the batch list in order. Otherwise, treat the steel coil in the batch as the first batch. Step S2.3: Traverse the remaining batches according to the default batch order, and put them into the batch list according to the total tonnage of the materials. Output the batch list as the batch list for subsequent scheduling.
4. The intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning according to claim 2, characterized in that, The step of classifying the material set in step S2 includes: Step S2.4: According to the process of the order, classify the material set output in step S1.2, set a width threshold, sort the materials in the material set whose width is greater than the threshold by width from widest to narrowest and classify them as main materials, sort the materials in the material set whose width is less than the threshold by width from narrowest to widest and classify them as bounce materials, and classify other batch materials that can be mixed with the current batch as insertable materials if the process allows.
5. The intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning according to claim 1, characterized in that, The process of selecting roller numbers in step S3 includes: Step S3.1: Based on the marked available working rollers for the material, take the set of available working rollers for the input material and generate a unique set of available working rollers. Step S3.2: Traverse all available work rollers, count the materials that can be processed with the current traversal number in the input material set, and record them in the database. The primary key is the roller number, and the value is the corresponding material quantity. Step S3.3: Select the n roller numbers with the most corresponding materials and output them as a set {1,2,3,...,n}. Based on the material set required for production scheduling, obtain all non-repeating available work rollers.
6. The intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning according to claim 1, characterized in that, The process of batch switching in step S4 includes: Step S4.3: By statistically analyzing the material quantity of each batch in the latest material set, define the upper limit tonnage, total tonnage, and remaining order tonnage of the current batch for a single production run, and define the lower limit tonnage and total tonnage of the next batch for a single interval. Step S4.4: Set the production upper limit gap threshold. Assuming that the difference between the output of each batch and the upper limit of a single production is less than this threshold, calculate the future batch number = (remaining order tonnage in the current production cycle - total tonnage) / upper limit tonnage of a single production. The current batch number = total tonnage / upper limit tonnage of a single production. Set the production lower limit gap threshold. Assuming that when the next batch is used as an interval material, the difference between the output of each batch and the lower limit of a single production is less than this threshold, calculate the maximum number of times it can be used as a process interval = total tonnage / lower limit tonnage of a single interval. Step S4.5: Compare the ratio of the number of times the current batch can be produced to the next batch, and decide on the conversion between batches based on the experience value determined by the planning personnel based on their production experience.
7. The intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning according to claim 6, characterized in that, The process of inserting materials into the processing interval in step S4 includes: Step S4.6: Based on the ratio of the number of times the current batch can be produced to the next batch as stated in step S4.5, the current batch has reached the upper limit of single production scheduling. According to the ratio of the remaining inventory of the current batch material to the batch material of the process interval in the latest material set, the MIP model of the master scheduling strategy is called to incorporate the batch material of the process interval into the latest plan.
8. The intelligent scheduling method based on multi-dimensional scale pre-decision and dynamic fine-tuning according to claim 1, characterized in that, The post-processing of the scheduled plan in step S4 includes: Step S4.7: Obtain the latest plan calculated by MIP. Obtain all steel coils that are not in the current batch from the latest material set. Iterate through each coil in the plan, calculate the average specifications of the coils before and after it, and find a steel coil in the material set that uses the same roller, meets the process rules, and is closer to the average specifications than the original steel coil to replace it. Put the original steel coil into the material set and merge the plan after material replacement into the scheduled plan.
Citation Information
Patent Citations
Personalized scheduling method for cold rolling galvanizing unit
CN119130059A
Controlling operation of energy-consuming devices
US20170003043A1