A knowledge-based multi-stage casting and rolling collaborative scheduling method and system

Abstract

This invention relates to a knowledge-based multi-stroke casting and rolling collaborative scheduling method and system, belonging to the field of metallurgical automation technology. The method includes data preparation; pre-sorting / slab clustering, utilizing the slab insertion constraint knowledge set in the knowledge base as a constraint condition during merging, performing agglomerative hierarchical clustering algorithm to cluster slabs and output slab groups; subsequent steps include main material group division, hot-roll material selection, merging main material groups, merging hot-roll materials and main materials, selection of finishing materials, and KPI calculation. The overall scheduling time is reduced from 20+ minutes to 7 seconds, achieving one-click intelligent multi-stroke automatic scheduling, effectively improving KPI indicators such as rolling hot charging rate and rolling mileage, significantly reducing the workload of planners while providing more stable scheduling quality, and improving the automation level of steel plants.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of metallurgical automation, and in particular to a multi-pass casting-rolling collaborative scheduling method and system based on a knowledge base. BACKGROUND

[0002] At present, the production process of steel mostly adopts processes such as "ironmaking-steelmaking-continuous casting-rolling- finishing", and the processes are closely connected in sequence, and the process production requirements are harsh and complex. The continuous casting process is to smelt pig iron and scrap steel at high temperature to remove impurities and obtain high-purity billets, and the hot rolling in the rolling process is to continuously roll the billets produced by continuous casting into various specifications and shapes of steel. At present, hot rolling mostly adopts mixed rolling, that is, cold charging and hot charging (direct charging) mixed rolling to form a same rolling unit.

[0003] The quality of the rolling plan scheduling of the continuous casting and hot rolling production line will directly affect the rolling line capacity, smooth logistics, product quality, roll life, etc., and the rolling plan has the characteristics of multiple targets, complex process constraints, multiple disturbances, large scale, etc., and is the core content and key technology of rolling production management.

[0004] At present, the hot rolling rolling plan of most steel plants is manually specified by a planner. Since the planner needs to face multiple production plan coordination, multiple pouring plan, thousands of slab arrangement, complex process constraints, and multiple upstream and downstream capacity constraints, the operation amount is extremely large, and the scheduling experience of different planners is different, the scheduling quality and production indicators will also fluctuate, which is not conducive to the stability, standardization and fine management of the plant production, and the production efficiency is low. At present, some researches on hot rolling automatic scheduling mostly use heuristic algorithms, which have the problems of slow speed, poor algorithm KPI effect, unsatisfactory process rules, limited input scale, etc., and can only arrange a single rolling pass, without considering the casting-rolling collaboration. SUMMARY

[0005] In view of the above analysis, the present application aims to provide a multi-pass casting-rolling collaborative scheduling method and system based on a knowledge base, which has more stable scheduling quality, improves the automation level of the steel plant, and improves the production efficiency.

[0006] In one aspect, the present application provides a multi-pass casting-rolling collaborative scheduling method based on a knowledge base, comprising the following steps:

[0007] S1: data preparation;

[0008] S2: pre-sorting / slab grouping, using the slab insertion constraint knowledge set of the knowledge base as the limiting condition during merging, performing the agglomerative hierarchical clustering algorithm, clustering the slabs and outputting the slab groups;

[0009] S3: Main Material Group Division. Based on the main material group division constraint knowledge set in the knowledge base, the clustered slab groups are divided into main material groups, and the main material groups are given priority scores according to information such as slab quantity, casting start time, and steel grade priority. Then, based on the main material group division constraint knowledge set in the knowledge base, the main material groups are merged according to the scores to form multiple candidate main material group sets. Finally, the candidate slab groups are divided into multiple rolling strokes, with each main material group set corresponding to one rolling stroke. The division objective is to maximize rolling mileage and hot charging rate.

[0010] S4: Selection of ironing roller material. For each main material group set in the main material division result, traverse all unused slabs. Based on the ironing roller material constraint knowledge set of the knowledge base, determine whether the slab can be used as the ironing roller material of the current main material set. If so, put it into the ironing roller material set corresponding to the main material set and select the available ironing roller material.

[0011] S5: Merge the main material groups. Merge adjacent slabs in the main material group based on the slab insertion constraint knowledge set in the knowledge base. If the merger fails, iterate through unused slabs. Then select suitable transition materials for splicing based on the slab insertion constraint knowledge set in the knowledge base. If the splicing fails, delete the next main material group. Repeat the above steps to form a continuous main rolling material group.

[0012] S6: Combine the ironing roller material and the main material.

[0013] S7: Selection of finishing material; If the rolling mileage is insufficient, use the Monte Carlo tree search algorithm to select finishing material to supplement it, thereby improving the rolling mileage performance;

[0014] S8: KPI Calculation. Calculates the KPIs for the rolling plan, organizes and fills in the rolling results, and returns them to the front end for display.

[0015] Furthermore, the data preparation is used to read slab data, steel grade configuration information, and scheduling status information from various heterogeneous underlying storage systems.

[0016] Furthermore, the agglomerative hierarchical clustering algorithm first merges clusters with the closest similarity that are allowed to be placed together according to process requirements. Secondly, the termination condition is that no new clusters are merged. Finally, no final pruning step is needed to determine the final clusters.

[0017] Furthermore, the similarity includes target width, target thickness, steel type, and whether it is a hot billet.

[0018] Furthermore, the main material group partitioning constraint knowledge set includes casting sequence information, rolling rhythm, and production process constraints.

[0019] Furthermore, the slab insertion constraint knowledge set includes steel grade, furnace temperature, furnace time, slab width, slab type, slab length, and slab attribute jumps.

[0020] Furthermore, the knowledge set of constraints on the hot rolling material includes steel grade, slab width, slab type, and slab length.

[0021] Furthermore, the KPI calculation involves calculating core technical indicators, including rolling mileage, hot charging rate, and machine time rate.

[0022] On the other hand, the present invention provides a multi-stroke casting and rolling collaborative scheduling system based on a knowledge base, including the multi-stroke casting and rolling collaborative scheduling method based on a knowledge base as described in the present invention. The system includes a data preparation module, a slab aggregation module, a hot roll material solver module, a tailing material solver module, a main material division module, a main material merging module, a result organization module, a KPI calculation module, and a knowledge base.

[0023] Furthermore, the data preparation module is used to read slab data, steel grade configuration information, scheduling status, and other information from various heterogeneous underlying storage systems.

[0024] The knowledge base module is an abstraction of the expert system, which stores the scheduling experience of scheduling experts using a production-based knowledge representation method.

[0025] Slab Clustering Module: Since there are thousands of slabs to be scheduled, exploring the scheduling plan directly at the particle size of a single slab would greatly increase the solution time. Therefore, we have innovatively introduced a slab clustering function in this module. Based on the scheduling rules, similar slabs are clustered together. When selecting main materials or finishing materials, selection is based on the group form. While ensuring the effect, the solution time can be greatly reduced.

[0026] The heat-pressing roller material solver module is used to search for the best heat-pressing roller material solution based on the main material specifications and a depth-first search algorithm.

[0027] Finishing material solver module: Used to search for the best finishing material solution based on the main material specifications and the Monte Carlo tree search algorithm to make the rolling length as long as possible;

[0028] Main material division module: It is used to divide thousands of input slabs into multiple rolling strokes based on factors such as upstream casting information, rolling rhythm, and production process limitations, with the goal of maximizing rolling mileage and hot charging rate;

[0029] Main material combination module: This module selects appropriate transition materials to smoothly transition the selected main material groups while meeting the production process requirements;

[0030] Results organization module: Based on the final scheduling results, fill in basic information such as rolling time, rolling mileage, and billet properties for downstream display;

[0031] KPI calculation module: Used to calculate the core technical indicators of the rolling process, and to quantify the scheduling effect of the rolling plan, including rolling mileage, hot charging rate, and machine time rate.

[0032] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0033] 1. This invention provides a multi-stroke casting and rolling collaborative scheduling method. Based on a knowledge base, it adds a clustering process based on an improved agglomerative hierarchical clustering algorithm. After clustering, the slab properties within the group are similar. When selecting main materials, the group is used as the unit, reducing the problem action space from thousands to less than hundreds, and reducing the overall scheduling time from 20+ minutes to 7 seconds. When dividing the main materials, casting and rolling are comprehensively considered to maximize the rolling rhythm and hot charging rate. At the same time, the Monte Carlo tree search algorithm is introduced to improve the rolling mileage, improve KPI performance, and ensure the stability of scheduling quality.

[0034] 2. This invention proposes a knowledge-based multi-stroke casting and rolling collaborative scheduling system. This system is based on a knowledge-based expert system, improved condensed hierarchical clustering, Monte Carlo tree search, and other algorithms. It comprehensively considers various factors such as scheduling process and casting sequence, and realizes one-click intelligent multi-stroke automatic scheduling. It effectively improves the rolling hot charging rate, rolling mileage and other KPI indicators, greatly reduces the workload of planners, has more stable scheduling quality, and improves the automation level of steel plants.

[0035] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0036] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0037] Figure 1 The flowchart shows a knowledge-based multi-stage casting and rolling collaborative scheduling method.

[0038] Figure 2 A flowchart of the Monte Carlo tree search algorithm;

[0039] Figure 3 This is a flowchart of the knowledge base execution process.

[0040] Figure 4 This is a diagram of a knowledge-based multi-stage casting and rolling collaborative scheduling system. Detailed Implementation

[0041] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which constitute a part of the present invention and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0042] In the steel production industry, there are numerous processes and complex operations, especially in the rolling stage, which is characterized by multiple objectives, complex process constraints, numerous disturbances, and large scale. Rolling schedules are often formulated manually by planners. Planners face numerous challenges, including coordinating multiple production plans, diverse casting schedules, handling thousands of slabs, complex process constraints, and fluctuating capacity constraints from upstream and downstream suppliers. This results in an extremely large workload, and the varying scheduling experience among planners leads to fluctuations in scheduling quality and production indicators, hindering the stability, standardization, and refined management of factory production. Existing research on automated hot rolling scheduling primarily relies on heuristic algorithms, which can only schedule single rolling cycles and do not consider casting-rolling coordination. Furthermore, the scheduling quality is poor, the automation level of the steel plant is low, and production efficiency is low.

[0043] Therefore, this invention provides a knowledge-based multi-stroke casting and rolling collaborative scheduling method, comprising the following steps:

[0044] S1: Data preparation;

[0045] S2: Pre-sorting / slab clustering. The slab insertion constraint knowledge set in the knowledge base is used as a constraint condition during merging. An agglomerative hierarchical clustering algorithm is used to cluster the slabs and output slab groups.

[0046] S3: Main Material Group Division. Based on the main material group division constraint knowledge set in the knowledge base, the clustered slab groups are divided into main material groups, and the main material groups are given priority scores according to information such as slab quantity, casting start time, and steel grade priority. Then, based on the main material group division constraint knowledge set in the knowledge base, the main material groups are merged according to the scores to form multiple candidate main material group sets. Finally, the candidate slab groups are divided into multiple rolling strokes, with each main material group set corresponding to one rolling stroke. The division objective is to maximize rolling mileage and hot charging rate.

[0047] S4: Selection of ironing roller material. For each main material group set in the main material division result, traverse all unused slabs. Based on the ironing roller material constraint knowledge set of the knowledge base, determine whether the slab can be used as the ironing roller material of the current main material set. If so, put it into the ironing roller material set corresponding to the main material set and select the available ironing roller material.

[0048] S5: Merge the main material groups. Merge adjacent slabs in the main material group based on the slab insertion constraint knowledge set in the knowledge base. If the merger fails, iterate through unused slabs. Then select suitable transition materials for splicing based on the slab insertion constraint knowledge set in the knowledge base. If the splicing fails, delete the next main material group. Repeat the above steps to form a continuous main rolling material group.

[0049] S6: Combine the ironing roller material and the main material.

[0050] S7: Selection of finishing material; If the rolling mileage is insufficient, use the Monte Carlo tree search algorithm to select finishing material to supplement it, thereby improving the rolling mileage performance;

[0051] S8: KPI Calculation. Calculates the KPIs for the rolling plan, organizes and fills in the rolling results, and returns them to the front end for display.

[0052] Compared with existing technologies, this invention is based on a knowledge base and adds a clustering process based on an improved agglomerative hierarchical clustering algorithm. After clustering, the slabs within the group have similar properties. When selecting main materials, the group is used as the unit, which reduces the problem action space from thousands to less than hundreds, and the overall scheduling time is reduced from 20+ minutes to 7 seconds. When dividing the main materials, casting and rolling are considered together to maximize the rolling rhythm and hot charging rate. At the same time, the Monte Carlo tree search algorithm is introduced to improve rolling mileage, improve KPI performance, and ensure the stability of scheduling quality.

[0053] In this invention, the rolling production process includes: obtaining slabs of various specifications through casting production. Since the number of slabs is relatively large, the slabs are classified into multiple main material groups. The main material groups are further divided into different priorities based on information such as the number of slabs, casting start time, and steel grade priority. Appropriate hot roll materials and transition materials are selected and combined with the main materials. If the rolling mileage is insufficient, finishing materials need to be added to finally complete the entire rolling process.

[0054] It should be noted that the definition of terms in this invention is as follows:

[0055] Rolling Stroke: The daily hot rolling plan can be divided into multiple rolling strokes. Each rolling stroke corresponds to the lifespan of a finishing roll, which means that each replacement of the finishing roll corresponds to the start of a rolling stroke.

[0056] Main material: The main rolled slab varieties in one rolling stroke (finishing roll service).

[0057] Preheating roller material: Material used for preheating rollers.

[0058] Transition material: In order to meet process constraints and ensure stable production of coils with high surface requirements or thin specifications, the transition material is required from hot-rolled material to main rolled material or main rolled material of different specifications.

[0059] Rolling process KPI: Data indicators for evaluating the quality of the rolling process.

[0060] Knowledge base: A knowledge base, also called a knowledge base system, is a system for acquiring, transforming, interpreting, storing, retrieving, and creating knowledge. It is a knowledge management tool that can simulate domain experts in solving various complex problems.

[0061] Specifically, the data preparation is used to read slab data, steel grade configuration information, scheduling status and other information from various heterogeneous underlying storage systems.

[0062] It should be noted that data preparation is an important step in this invention, which includes two parts: data collection and storage. First, slab inventory information, casting information, steel type information, and various rule configurations are collected and then stored in the underlying storage system for easy retrieval later.

[0063] Specifically, the agglomerative hierarchical clustering algorithm first merges clusters with the closest similarity that are allowed to be grouped together according to process requirements. Secondly, the termination condition is that no new clusters are merged. Finally, no final pruning step is needed to determine the final clusters.

[0064] Specifically, similarity includes target width, target thickness, steel type, and whether it is a hot billet.

[0065] Specifically, the slab insertion constraint knowledge set in the knowledge base includes steel grade, tapping temperature, tapping time, slab width, slab type, slab length, and slab attribute jumps.

[0066] It should be noted that hierarchical clustering is an unsupervised clustering method based on a hierarchical structure. It iteratively merges highly similar sample points into the same cluster, forming a tree-like clustering structure. Hierarchical clustering has advantages such as easy definition of distance and rule similarity and no need to predetermine the number of clusters. Agglomerative hierarchical clustering is a branch of hierarchical clustering algorithms. It adopts a bottom-up strategy, clustering similar slabs together based on production scheduling rules. When selecting main materials or finishing materials, selection is based on groups, which can significantly reduce solution time while ensuring effectiveness.

[0067] Slab Clustering: Generally, the total number of candidate slabs can reach several thousand. If the scheduling plan is searched directly at the granularity of a single slab, the solution time will be greatly increased. This invention introduces a slab clustering method, which uses an improved agglomerative hierarchical clustering algorithm combined with a knowledge base to cluster similar slabs together. When selecting main materials or finishing materials, the slab group is used as the basis for selection. This can significantly reduce the solution time while ensuring the scheduling effect.

[0068] The original agglomerative hierarchical clustering algorithm starts with each sample as an independent cluster, then iteratively merges clusters with the closest similarity (such as Euclidean distance, cosine similarity, etc.), repeating this process until only one cluster remains, forming a clustering tree where each node represents a cluster or sub-cluster. Then, based on actual needs, pruning is used to select the final number of clusters.

[0069] In the application scenarios of this invention, some slabs are not allowed to be rolled together due to process requirements. The slab clustering of this invention first merges the clusters with the closest similarity that are allowed to be rolled together according to process requirements. Secondly, the termination condition is that no new clusters are merged. Finally, no final pruning step is needed to determine the final cluster.

[0070] Specifically, this invention uses the Euclidean distance of attributes such as target width, target thickness, steel type, and whether it is a hot billet as the similarity, and uses the slab insertion constraint knowledge set of the knowledge base as the restriction condition during merging, to perform the above-mentioned agglomerative hierarchical clustering algorithm.

[0071] Specifically, the knowledge set for defining the main material group includes casting information, rolling rhythm, and production process constraints.

[0072] Specifically, the slab insertion constraint knowledge set includes constraint rules for steel grade, tapping temperature, tapping time, slab width, slab type, slab length, and slab attribute jumps.

[0073] Specifically, the knowledge set of constraints on hot rolling materials includes limitations such as steel grade, slab width, slab type, and slab length.

[0074] Specifically, the selection of finishing material; if the rolling mileage is insufficient, the Monte Carlo tree search algorithm is used to select finishing material to supplement it, thereby improving the rolling mileage performance.

[0075] It should be noted that the specific steps / methods of the Monte Carlo tree search algorithm are as follows:

[0076] Reference Figure 2 Initialize the root node to the current rolling schedule and set the root node as the current node.

[0077] Algorithm begins:

[0078] First, initialize the root node, set the current node's state to the initial schedule (rolling plan), set the parent node to empty, Q value to 0, N value to 0, set the loop count to 0, and set the root node to the current node.

[0079] Determine if it is a terminal node: If yes, the algorithm terminates and outputs the root node status, i.e., the rolling plan. If no, continue to the next step.

[0080] Loop count control: Determine if the loop count is less than the preset iteration count (IterCount). If it is, the node has been fully explored. Then, find the child node with the highest score from the current node's child nodes based on formula (1) (v in the formula represents the current node, Q(v) represents taking the Q value of node v, N(v) represents taking the N value of node v, and v' is any child node of v). Update the schedule to the current schedule and repeat step 2. If not, stop the search and continue to the next step.

[0081]

[0082] The subsequent execution steps include selection, expansion, simulation, and propagation.

[0083] Determine if it is a terminal node: If yes, proceed to the simulation step; if no, continue to the next step; Determine if the current node has been fully explored: If yes, the node has been fully explored, then find the child node with the highest score from the current node's child nodes based on Formula 1 as the current node; if no, proceed to the expansion step.

[0084] Randomly select an unexplored slab s that meets the knowledge base K constraint; add s to the schedule, create a new node with the new schedule as the state, the new node's Q value and N value are 0, and this new node becomes a child node of the current node; then proceed to the simulation step;

[0085] Calculate the distances between all available candidate slabs and the end of the rolling stroke, and use the simulation reward as the current node; set the current node to V1; then proceed to the propagation step;

[0086] Set Q = Q + reward for V1, and set N = N + 1 for V1;

[0087] Determine if V1 is the root node. If it is, increment the loop count by 1 and proceed to the "loop count is less than IterCount" check step. If not, set V1 as the parent node of V1 and then proceed to the "set V1's Q = Q + reward, set V1's N = N + 1" step.

[0088] Algorithm termination: When a certain condition is met (such as reaching the iteration limit or finding the optimal solution), the algorithm terminates and outputs the final root node state.

[0089] Specifically, KPI calculation involves calculating and filling core technical indicators, including rolling mileage, hot charging rate, and machine time rate.

[0090] It should be noted that KPI calculation is the final output result, which can reflect or be used to evaluate the quality of rolling plans and production scheduling.

[0091] Specifically, in this invention, the knowledge base stores the experience of production scheduling experts in a production knowledge representation method. The knowledge base is categorized, such as slab insertion constraint knowledge set, main material merging constraint knowledge set, hot rolling material constraint knowledge set, and finishing material selection rule set.

[0092] The knowledge base execution process is as follows: Figure 3 As shown, the specific steps include: inputting the name of the knowledge set and the set of facts required by the knowledge set; selecting the knowledge set based on the name of the knowledge set; traversing each piece of knowledge in the knowledge set; inputting the set of facts and deciding whether to execute action Q based on condition P; and finally returning the result of the action execution.

[0093] Its representation also adopts the production rule notation, as detailed below.

[0094] Knowledge-generating knowledge representation method: IFPTHENQ;

[0095] Fact-based knowledge representation: (object, attribute, value).

[0096] Based on a knowledge base, this invention also provides a system for a multi-stage casting and rolling collaborative scheduling method, including a data preparation module, a slab aggregation module, a hot roll material solver module, a tailing material solver module, a main material division module, a main material merging module, a result organization module, a KPI calculation module, and a knowledge base.

[0097] Data preparation module: used to read slab data, steel grade configuration information, scheduling status and other information from various heterogeneous underlying storage systems;

[0098] The knowledge base module is an abstraction of the expert system, which stores the scheduling experience of scheduling experts using a production-based knowledge representation method.

[0099] Slab Clustering Module: Since there are thousands of slabs to be scheduled, exploring the scheduling plan directly at the particle size of a single slab would greatly increase the solution time. Therefore, we have innovatively introduced a slab clustering function in this module. Based on the scheduling rules, similar slabs are clustered together. When selecting main materials or finishing materials, selection is based on the group form. While ensuring the effect, the solution time can be greatly reduced.

[0100] The heat-pressing roller material solver module is used to search for the best heat-pressing roller material solution based on the main material specifications and a depth-first search algorithm.

[0101] Finishing material solver module: Used to search for the best finishing material solution based on the main material specifications and the Monte Carlo tree search algorithm to make the rolling length as long as possible;

[0102] Main material division module: It is used to divide thousands of input slabs into multiple rolling strokes based on factors such as upstream casting information, rolling rhythm, and production process limitations, with the goal of maximizing rolling mileage and hot charging rate;

[0103] Main material combination module: This module selects appropriate transition materials to smoothly transition the selected main material groups while meeting the production process requirements;

[0104] Results organization module: Based on the final scheduling results, fill in basic information such as rolling time, rolling mileage, and billet properties for downstream display;

[0105] KPI calculation module: Used to calculate the core technical indicators of the rolling process, and to quantify the scheduling effect of the rolling plan, including rolling mileage, hot charging rate, machine time rate, etc.

[0106] Compared with existing technologies, this invention proposes a knowledge-based multi-stroke casting and rolling collaborative scheduling system. This system is based on a knowledge-based expert system, improved condensed hierarchical clustering, Monte Carlo tree search, and other algorithms. It comprehensively considers various factors such as scheduling process and casting sequence, and realizes one-click intelligent multi-stroke automatic scheduling. It effectively improves the rolling hot charging rate, rolling mileage and other KPI indicators, greatly reduces the workload of planners, has more stable scheduling quality, and improves the automation level of steel plants.

[0107] Specifically, the knowledge base module includes a slab insertion constraint knowledge set, a main material merging constraint knowledge set, a hot rolling material constraint knowledge set, and a finishing material selection rule set.

[0108] The multi-stroke casting and rolling collaborative scheduling system provided by this invention first requires data collection and storage through a data preparation module. The data is stored in the slab module in the knowledge base and object model and used as subsequent limiting conditions and calls.

[0109] Then, the slab clustering module uses the slab insertion constraint knowledge set in the knowledge base as the constraint condition for merging, performs agglomerative hierarchical clustering algorithm, clusters the slabs and outputs slab groups, and stores the data in the slab group module in the object model.

[0110] Then, through the main material segmentation module, the clustered slab groups are divided into main material groups, and priority is assigned.

[0111] Next, through the hot rolling material solver module and based on the hot rolling material constraint knowledge set of the knowledge base, it is determined whether the slab can be used as the hot rolling material of the current main material set. If so, it is placed into the hot rolling material set corresponding to the main material set, and the available hot rolling material is selected.

[0112] Then, through the main material merging module, adjacent main materials in the main material set are merged based on the slab insertion constraint knowledge set in the knowledge base, and suitable transition materials are selected for splicing.

[0113] Furthermore, the tailings solver module is used to select tailings to supplement the main material and ensure that the entire rolling mileage meets the requirements.

[0114] Finally, the results are displayed through the results organization module, the KPI calculation module, and the scheduling status module in the object model, which facilitates the scheduler or operator to adjust the overall rolling schedule in real time.

[0115] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A knowledge base based multi-pass rolling co-scheduling method, characterized in that, The method comprises the following steps: S1: data preparation; the data preparation is used for reading slab data, steel grade configuration information and scheduling state information from a plurality of heterogeneous underlying storage systems; S2: pre-sorting / slab grouping, using a slab insertion constraint knowledge set of a knowledge base as a limiting condition during merging, performing a hierarchical clustering algorithm by a clustering method, clustering slabs and outputting slab groups; S3: main material group division, based on a main material group division constraint knowledge set of the knowledge base, dividing the clustered slab groups into main material groups, and performing priority scoring on the main material groups according to the number of slabs, pouring time, steel grade priority; based on the main material group division constraint knowledge set of the knowledge base, merging the main material groups according to the scoring scores, and forming a plurality of candidate main material group sets, and finally dividing the candidate slab groups into a plurality of rolling processes, each main material group set corresponding to a rolling process, and the division target is to maximize the rolling mileage and the hot charging rate; S4: selection of a hot rolling material, for each main material group set in the main material division result, all unused slabs are traversed, based on a hot rolling material constraint knowledge set of the knowledge base, whether the slab can be used as the hot rolling material of the current main material set, if yes, the slab is put into the hot rolling material set corresponding to the main material set, and the available hot rolling material is selected; S5: merging of main material groups, adjacent slabs in the main material set are merged based on a slab insertion constraint knowledge set of the knowledge base, if the merging fails, the unused slabs are traversed; then, a suitable transition material is selected based on the slab insertion constraint knowledge set of the knowledge base to splice, if the splicing fails, the last main material group is deleted, the above steps are repeatedly executed, and a continuous main rolling material group is formed; S6: merging of hot rolling materials and main materials, S7: selection of a finishing material; if the rolling mileage is not enough, a Monte Carlo tree search algorithm is used to select a finishing material to supplement, so as to improve the rolling mileage performance; S8: KPI calculation, KPI of the rolling plan is calculated, and a rolling result is organized and filled, and is returned to the front end for display.

2. The knowledge base based multi-pass rolling co-scheduling method according to claim 1, wherein, The hierarchical clustering algorithm first merges clusters with the closest similarity and process requirements that allow them to be placed together during merging, the termination condition is that no new cluster is merged, and finally, a final pruning step is not required to determine the final clustering cluster.

3. The knowledge base based multi-pass rolling co-scheduling method according to claim 2, wherein, The similarity includes target width, target thickness, steel grade and whether it is a hot slab.

4. The knowledge base based multi-pass rolling co-scheduling method according to claim 1, wherein, The main material group division constraint knowledge set includes pouring information, rolling rhythm and production process limitation.

5. The knowledge base based multi-pass rolling co-scheduling method according to claim 1, wherein, The slab insertion constraint knowledge set includes steel grade, tapping temperature, tapping time, slab width, slab type, slab length and slab attribute jump.

6. The knowledge base based multi-pass rolling co-scheduling method according to claim 1, wherein, The hot rolling material constraint knowledge set includes steel grade, slab width, slab type and slab length.

7. The knowledge base based multi-pass rolling co-scheduling method according to claim 1, wherein, The KPI calculation is to calculate the filling core technology index, including rolling mileage, hot charging rate and machine time rate.

8. A knowledge base based multi-pass rolling co-scheduling system, characterized by, The system comprises a data preparation module, a slab grouping module, a hot rolling material solver module, a finishing material solver module, a main material division module, a main material merging module, a result organization module, a KPI calculation module and a knowledge base.

9. The knowledge base based multi-pass rolling co-scheduling system of claim 8, wherein, The data preparation module is used for reading slab data, steel grade configuration information and scheduling state information from a plurality of heterogeneous underlying storage systems; Knowledge base module: is the expert system to abstract, to production knowledge representation method storage scheduling expert various scheduling experience; Slab clustering module: because the number of slabs to be scheduled is in the thousands, if directly exploring the scheduling plan with single slab granularity will greatly increase the solving time, so in this module, the slab clustering function is introduced, similar slabs are clustered together based on scheduling rules, and the main material or the end material is selected in the form of a group as the basis, which can greatly reduce the solving time while ensuring the effect; Roller material solver module: used to search the best roller material scheme based on the depth-first search algorithm according to the main material specification; End material solver module: used to search the best end material scheme based on the Monte Carlo tree search algorithm according to the main material specification, so that the rolling length is as long as possible; Main material division module: used to divide thousands of input slabs into multiple rolling processes according to the upstream casting information, rolling rhythm, and production process restrictions, with the goal of maximizing rolling mileage and hot charging rate; Main material combination module: this module selects appropriate transition materials to smoothly transition between selected main material groups under the condition of meeting the production process; Result organization module: according to the final scheduling result, fill in the rolling time, rolling mileage, and slab properties for downstream display; KPI calculation module: used to calculate the core technical indicators of the rolling process, which is used to quantify the scheduling effect of the rolling plan, including rolling mileage, hot charging rate, and machine time rate.

Images