Information learning method and system considering unauthorized slab matching
By constructing an information learning model and a tabu search optimization algorithm, the problem of matching slabs without mandate was solved, improving the solution efficiency and quality, reducing production and inventory costs, and enhancing resource utilization and production management.
Patent Information
- Application Number
- CN202511155788.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-18
- Publication Date
- 2025-11-21
AI Technical Summary
Existing technologies are insufficient to effectively solve the problem of matching slabs without a contract, resulting in low resource utilization, high inventory costs, and low efficiency and poor solution quality of commonly used optimization methods, which are difficult to meet actual needs.
By employing an information learning approach, a mathematical model is constructed and combined with information entropy and entropy value from information theory. The algorithm for matching slabs without mandate is optimized through tabu search and fast evaluation mechanisms. The information learning algorithm guides the search process, thereby improving the quality and efficiency of the solution.
It significantly improves the solution quality and efficiency of unmanaged slab matching problems, reduces production and inventory costs, shortens production cycles, and enhances resource allocation efficiency and production management capabilities.
Smart Images
Figure CN120996148A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of computing intelligence and production planning optimization, in particular to an information learning method and system considering non-commissioned slab matching. BACKGROUND
[0002] The production of silicon steel involves multiple stages of processes, namely ironmaking, steelmaking-continuous casting, hot rolling and cold rolling. Companies adopt the strategy of production on order, and the production plan is closely dependent on customer orders. Due to the characteristics of batch production and the uncertainty of the production process, the slabs coming out of the steelmaking-continuous casting may not be completely matched with the orders. Therefore, it is inevitable to produce open order slabs, i.e. non-commissioned slabs, which do not belong to any order. The unnecessary non-commissioned slabs will increase the inventory cost and hinder the production process. Therefore, enterprises need to continuously optimize the matching strategy between slabs and orders, reduce the generation of non-commissioned slabs, and ensure the coordination and efficiency of the production process.
[0003] By introducing the Internet of Things unique identification technology, the slabs can be uniquely coded and tracked, and accurate matching with orders can be realized. Combined with industrial data analysis means, the production process, inventory information and order demand can be analyzed in real time to provide decision basis for optimizing the matching strategy. The manufacturing execution system (MES system) is used to coordinate and schedule each link of the production process to ensure the efficient advancement of the production task.
[0004] Only by effectively controlling the number of non-commissioned slabs, can enterprises improve resource utilization, reduce inventory and production costs, and thus achieve the dual goals of improving production efficiency and economic benefits. In the production management of modern steel enterprises, there are many challenges such as low utilization rate of non-matching slab resources and low completion rate of customer contracts on time. In order to improve production efficiency and customer satisfaction, an efficient non-commissioned slab matching optimization method is urgently needed, which can quickly respond to the complex and large-scale order matching demand in reality. This method should ensure the matching quality while taking into account the computing efficiency, so as to effectively reduce the production and inventory costs, shorten the overall production cycle, and improve the resource allocation efficiency and on-site scheduling level. By optimizing the matching process, it not only helps to improve the production management ability of enterprises, but also ultimately realizes the steady improvement of enterprise benefits, enhances market competitiveness and customer service ability.
[0005] The non-commissioned slab matching problem has important application value, but it is a typical NP-hard problem when considering cost factors, and it is difficult to solve. The commonly used optimization methods such as genetic algorithm and ant colony algorithm have the problems of low solving efficiency and low solution quality, which are difficult to meet the actual demand. The information learning method proposed in the present application can extract useful information from the historical optimal solution to guide the search process, significantly improve the solving quality and efficiency, and provide a more efficient and reliable technical means for solving complex and large-scale matching problems. SUMMARY
[0006] In view of the defects of the prior art, the present application provides an information learning method and system considering no-entrust slab matching, which has the advantages of effectively obtaining high-quality solutions to provide reliable slab matching schemes for managers and the like, and solves the above technical problems.
[0007] To achieve the above object, the present application provides the following technical scheme: an information learning method considering no-entrust slab matching, comprising the following steps: S1, a mathematical model of a no-entrust slab entrust problem is constructed, and constraints are set for the established model, and the specific expression is as follows: s.t. wherein, represents a slab set, represents an order set, , , represents a slab, represents an order, represents an allocation cutting cost of each slab and each order, represents a cutting cost of two slabs from the same mother plate, being allocated to different orders, represents the sum of the allocation cutting cost and the same mother plate cutting cost, represents a minimum value, represents the weight limit of each order, represents the maximum processing capacity of each order, represents that the slab is allocated to the order , 0 represents that the slab is not allocated to the order , represents any, represents the processing capacity of the slab , represents whether the slab and the slab from the same mother plate are both allocated to the same order for processing, wherein =1 represents that the slab is allocated to the order , 0 represents that the slab Not assigned to an order ; S2: Construct an information learning-based, delegate-free slab matching algorithm to solve the slab matching problem in step S1.
[0008] As a preferred technical solution of the present invention, step S2 specifically includes the following steps: S2.1: Encoding: Agent-based task allocation encoding; S2.2: Set the relevant parameters in the information learning-based no-delegation slab matching algorithm. The relevant parameters include population size, similarity threshold, tabu length, and tabu search depth. S2.3: Initialize the population based on the allocation cost; if an infeasible solution is generated, repair it. S2.4: Initialize the taboo list; S2.5: Learn information based on the population's encoding and calculate the population similarity. ; S2.6: Determine the compressible encoding length based on similarity. Therefore, the size of the sampling space required to produce a new solution can be calculated. ; S2.7: Randomly select a solution from the population and determine the population similarity. Has the set similarity threshold been reached? ; If the desired solution is achieved, all slabs assigned to the same order in the selected solution will be treated as a whole and exchanged with slabs from other orders. Each whole exchange will be evaluated, and a whole exchange will be selected based on the roulette wheel to generate a new solution. If the target is not reached, the entropy of each coding bit is calculated based on the population coding, and then the entropy of each coding position is sorted from largest to smallest. The coding bit with the largest entropy is selected. Each encoding bit represents the selected solution. A new solution is generated by roulette wheel sampling of one of the encoded bits, while the encoding of the remaining encoded bits remains unchanged; S2.8: Perform tabu search optimization on the newly generated solution in S2.7 based on a fast evaluation mechanism; S2.9: Evaluate the objective function Determine whether the new solution meets the population management conditions. If it does, replace the solution selected from the population. If it does not, discard the new solution and keep the population unchanged. S2.10: Select the optimal solution from the population, retaining the objective function and its encoding; S2.11: Determine whether the set termination condition has been met. If the algorithm termination condition has not been met, return to S2.5 to continue iterating.
[0009] As a preferred embodiment of the present invention, the population similarity calculation formula in S2.5 is as follows: in, This refers to the actual entropy value of the current population. The maximum entropy value of the current population. Indicates the similarity of the current population. Represents the relative rate of entropy.
[0010] As a preferred embodiment of the present invention, the actual entropy value of the current population The expression is as follows: in, Represents a specific slab in the set of slabs , Represents a specific order in the order set. , In the current population Board matching The probability of an order. For all of the current population The board is allocated Conditional probability of an order; The maximum entropy value of the current population The calculation formula is as follows: in, That is the total number of orders. Represents a logarithmic function.
[0011] As a preferred embodiment of the present invention, the specific steps of the tabu search optimization based on the fast evaluation mechanism in S2.8 are as follows: S2.8.1: The encoding initialization based on S2.1 is a tabu search matrix based on fast evaluation, and its expression is as follows; in, This represents the initial tabu search matrix based on fast evaluation. and Let these represent the number of rows and columns of the probability matrix, respectively. The size is consistent with the size of the slab assembly. The size is the same as the size of the order set. After a rapid evaluation, the slab... Assigned to orders The cost difference between the original allocation plan and the subsequent one slab Assigned to order The extent of the taboo surrounding this action; S2.8.2: Based on the matrix Choose one in At the same time, Find the minimum local optimization solution, recalculate the objective function value, and update the matrix. Specifically: Order Equal to the forbidden length, all other non-zero values The value is reduced by one accordingly; Calculate all values in the new matrix using a fast evaluation method. The search depth is incremented by one, and the search exits after reaching the given search depth, outputting the code of the optimal solution for the objective function during the search process.
[0012] The present invention also provides an information learning system considering unmanaged slab matching, and based on the above-mentioned information learning method considering unmanaged slab matching, includes: A module is established to build a mathematical model for the slab-less matching problem that considers minimizing the total matching cost as the optimization objective. The optimization module is used to optimize the optimization objective in the above model using an evolutionary algorithm based on information learning.
[0013] Compared with the prior art, the present invention provides an information learning method and system that considers matching of slabs without commission, which has the following beneficial effects: This invention opens up new avenues for optimizing the uncommissioned slab matching problem by utilizing the theory of redundancy in information theory. It proposes an information learning algorithm based on information learning to compress the sampling space and solve the problem. The main components include: inventing a global search strategy based on information learning; the invented information learning method relies on information source redundancy and information entropy in information theory; specifically, the information learning algorithm determines the sampling size based on the similarity of the decoded data in the population, and then evaluates the information entropy of the decoded data to determine which positions to sample in the parent generation. This information learning mechanism is expected to make the search more focused and efficient. This information learning optimization algorithm can obtain higher quality solutions than other advanced algorithms within an acceptable CPU time, further reducing the cost for enterprises in uncommissioned slab matching. Attached Figure Description
[0014] Figure 1 This is a flowchart of the overall algorithm of the present invention.
[0015] Figure 2 This is a diagram illustrating the encoding method of the present invention.
[0016] Figure 3 This is a schematic diagram of the repair operation of the present invention.
[0017] Figure 4 An overall exchange operation of the present application is schematically shown. DETAILED DESCRIPTION
[0018] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.
[0019] Please refer to Figure 1 Figure 4 The present application provides the following technical solutions: An optimization method for information learning without a billet matching commission, comprising the following steps: S1, a mathematical model of a no-commission billet commission problem is constructed, and constraints are set for the established model, and the specific expression is as follows: s.t. (constraints) The formula indicates that each billet can be assigned to only one order, and each billet can be assigned only once The formula indicates that the billet assigned to an order cannot exceed the capacity of each order; The formula indicates that the billet assigned to an order cannot exceed the processing capacity of each order; The formula indicates that the value range of the decision variable is only 0 and 1; Wherein, represents a set of billets, represents a set of orders, , , represents a billet, represents an order, represents an assignment cutting cost of each billet matched with each order, represents a cutting cost of two billets from the same mother plate assigned to different orders, represents the sum of the assignment cutting cost and the same mother plate cutting cost, represents a minimum value, represents a weight limit of each order, represents a maximum processing capacity of each order, represents a billet Assigned to orders , 0 indicates slab Not assigned to an order , Indicates any, Indicates slab Processing capacity Slabs originating from the same motherboard and slab Are they all assigned to the same order? To handle, among which =1 indicates slab Assigned to orders , 0 indicates slab Not assigned to an order ; S2. Construct an information-based learning-based algorithm for assigning slabs without mandates to solve the assignment problem in step S1. This includes the following steps: S2.1 Encoding. Based on the problem characteristics, an agent-based integer vector π is used to represent the solution of a task assignment agent. For example, π=[1,3,4,4,2,1] means that slab 1 and slab 6 are assigned to order 1, slab 2 is assigned to order 3, slab 4 and slab 5 are assigned to order 4, and slab 5 is assigned to order 2. The encoding diagram is as follows. Figure 2 As shown; S2.2 Set the relevant parameters in the information learning-based no-delegation slab matching algorithm. The relevant parameters include population size, similarity threshold, tabu length, and tabu search depth. S2.3. The population is initialized based on the allocation cost. First, each slab is selected using a roulette wheel based on the allocation cost of different orders, and the slab is allocated to the appropriate order. Then, a repair function is used to correct individuals that violate the constraints. Finally, a relatively high-quality initial population is created. S2.6 Determine the compressible encoding length based on similarity. Therefore, the size of the sampling space required to produce a new solution can be calculated. As a population continues to evolve, the similarity within the population... The number will gradually increase, and the required sampling space will decrease accordingly; S2.7 First, randomly select a solution from the population, and then determine whether the population similarity reaches the set similarity threshold. If the condition is met, all slabs allocated to the same order in the selected solution are treated as a whole and exchanged with slabs from other orders. Each whole exchange is evaluated, and a new solution is generated by selecting a whole exchange based on the roulette wheel. The whole exchange operation is as follows:Figure 4 As shown, if not reached, the entropy of each coding bit is calculated according to the population coding, and then the entropy of each coding position is sorted from large to small, and the coding bit with larger entropy is selected The roulette sampling is performed on the selected coding bit of the solution to generate a new solution, and the coding of the remaining coding bits remains unchanged. S2.8, the newly generated solution in S2.7 is optimized by the fast evaluation-based tabu search; S2.9, the objective function is evaluated, and it is judged whether the new solution meets the population management condition, if it meets the condition, the selected solution from the population is replaced, if it does not meet the condition, the new solution is abandoned, and the population remains unchanged. S2.10, the optimal solution in the population is selected, and the objective function and the coding are reserved. S2.11, it is judged whether the termination condition is reached, if not, the algorithm returns to S2.5 for iteration.
[0020] Further, the repairing operation of the individual violating the constraint mentioned in step S2.3, since the individual violating the constraint, i.e. the infeasible solution, may appear in the process of initializing the population according to the allocated cost, and the infeasible solution obviously does not meet the requirement, therefore, the infeasible solution needs to be repaired to convert the infeasible solution into a feasible solution. According to the characteristics of the problem, the main idea is to balance the idle capacity of each order, i.e. the constraint of each order, so as to convert the infeasible solution into a feasible solution. In the present application, only the order violating the constraint (i.e. the constraint is less than 0) is considered, and the related slab in these orders is moved to the order with the largest idle capacity. The idle capacity of each order and the total number of orders violating the constraint are dynamically updated. The above operation is repeated until a feasible solution is generated. The specific repairing operation is shown in Figure 3 .
[0021] Further, the information learning method in step S2.5 is based on the residual degree calculation in information theory, according to the coding characteristics of the solution in the current population, the residual degree, i.e. the population similarity, is calculated, and the population similarity calculation formula in step S2.5 is as follows: Wherein, the actual entropy value of the current population, the maximum entropy value of the current population, the similarity of the current population, the relative rate of the entropy.
[0022] Further, the actual entropy value calculation formula of the population in the population similarity calculation formula is as follows: in, Represents a specific slab in the set of slabs , Represents a specific order in the order set. , In the current population Board matching The probability of an order. For all of the current population The board is allocated Conditional probability of an order.
[0023] Furthermore, the formula for calculating the maximum entropy of the population in the population similarity calculation formula is as follows: in, That is the total number of orders. It is a logarithmic function.
[0024] Furthermore, the formula for calculating the entropy encoded in step S2.7 is as follows: in, Represents unallocated slabs , Representing orders, In the population The entropy value of the encoded bits, Unallocated slabs in the representative population Assigned to orders The probability of.
[0025] Furthermore, the overall exchange in step S2.8 is an optimization operation on the population encoding based on the magnitude of population similarity; in addition, the specific steps of the tabu search based on fast search are as follows: S2.8.1, The encoding initialization based on S2.1 is based on the tabu search matrix of fast evaluation, and its expression is as follows; in, This represents the initial tabu search matrix based on fast evaluation. and Let these represent the number of rows and columns of the probability matrix, respectively. The size is consistent with the size of the slab assembly. The size is the same as the size of the order set. After a rapid evaluation, the slab... Assigned to orders The cost difference between the original allocation plan and the subsequent one. slab Assigned to orders The size of the tabu list; S2.8.2: According to the matrix Select one in At the same time, The minimum local optimization scheme, and recalculate the objective function value and update the matrix Specifically, let Equal to the tabu length, and the rest of the non-zero Value is reduced by one, and after reaching the given search depth, the optimal scheme code of the objective function in the search process is output.
[0026] In order to verify the effectiveness of the algorithm proposed in the application, the information learning-based non-delegation slab matching algorithm (ILSF), the algorithm is compared with four mainstream advanced algorithms DPLS, IDE, HBMO and NGHS for 46 examples of large scale (which cannot be solved by the solver). First, ILSF uses CPU time 5000s as the termination condition, is written in Matlab, and is run on a PC with Intel Core-i5 3.4 ghz processor and 16 gb memory.
[0027] ILSF is independently run 20 times for each instance. The result report has three indicators, namely BV, AV and SD. BV, AV and SD are the best value, average value and standard deviation. We notice that the value of the most known solution (BKS) of all instances is obtained by the four comparison algorithms (DPLS, IDE, HBMO, NGHS). At the same time, we mark the BV of the instance with an asterisk (*) to represent the strictly optimal target value in the comparison algorithm, which also corresponds to the newly discovered upper bound.
[0028] The experimental results are shown in Table 1: Table 1 ILSF results on 46 difficult instances The BKS of the 46 challenging instances is obtained by the four reference algorithms. For the BKS indicator, ILSF achieves 39 better, 2 same. There are 6 worse results. This finding shows that the information learning mechanism enables the ILSF algorithm to more efficiently utilize the given computational budget and more effectively find high-quality solutions.
[0029] The application also realizes the above-mentioned non-delegation slab matching optimization method through the information learning optimization system for non-delegation slab matching, which comprises: The establishment module is used to establish a mathematical model of the non-slab matching problem considering the minimization of the cost by taking the minimization of the total matching cost as the optimization objective; An optimization module is configured to optimize the optimization objective in the model using an information learning based evolutionary algorithm.
[0030] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be made without departing from the spirit and scope of the present application, which is defined by the following claims and their equivalents.
Claims
1. An information learning method considering uncommissioned slab matching, characterized in that: Includes the following steps: S1. Construct a mathematical model for the problem of slab commissioning without commissioning, and set constraints on the established model. The specific expression is as follows: st in, Represents a set of slabs. Represents a set of orders. , , Indicates slab, Indicates an order. This represents the allocated cutting cost for each slab matched with each order. Indicates two slabs from the same mother plate Cutting costs allocated to different orders This represents the sum of the allocated cutting cost and the cost of cutting the same motherboard. This represents the minimum value. This indicates that there is a weight limit for each order. This indicates the maximum processing capacity for each order. Indicates slab Assigned to orders , 0 indicates slab Not assigned to an order , Indicates any, Indicates slab Processing capacity Slabs originating from the same motherboard and slab Are they all assigned to the same order? To handle, =1 indicates slab Assigned to order , 0 indicates slab Not assigned to an order ; S2: Construct an information learning-based, delegate-free slab matching algorithm to solve the slab matching problem in step S1.
2. The information learning method considering uncommissioned slab matching according to claim 1, characterized in that: S2 specifically includes the following steps: S2.1: Encoding: Agent-based task allocation encoding; S2.2: Set the relevant parameters in the information learning-based no-delegation slab matching algorithm. The relevant parameters include population size, similarity threshold, tabu length, and tabu search depth. S2.3: Initialize the population according to the allocation cost. If an infeasible solution is generated, it needs to be repaired. S2.4: Initialize the taboo list; S2.5: Learn information based on the population's encoding and calculate the population similarity. ; S2.6: Determine the compressible encoding length based on similarity. Therefore, the size of the sampling space required to produce a new solution can be calculated. ; in, Represents a set of orders; S2.7: Randomly select a solution from the population and determine the population similarity. Has the set similarity threshold been reached? ; If the desired solution is achieved, all slabs assigned to the same order in the selected solution will be treated as a whole and exchanged with slabs from other orders. Each whole exchange will be evaluated, and a whole exchange will be selected based on the roulette wheel to generate a new solution. If the target is not reached, the entropy of each coding bit is calculated based on the population coding, and then the entropy of each coding position is sorted from largest to smallest. The bit with the highest entropy is selected. Each encoding bit represents the selected solution. A new solution is generated by roulette wheel sampling of one of the encoded bits, while the encoding of the remaining encoded bits remains unchanged; S2.8: Perform tabu search optimization on the newly generated solution in S2.7 based on a fast evaluation mechanism; S2.9: Evaluate the objective function Determine whether the new solution meets the population management conditions. If it does, replace the solution selected from the population. If it does not, discard the new solution and keep the population unchanged. S2.10: Select the optimal solution from the population, retaining the objective function and its encoding; S2.11: Determine whether the set termination condition has been met. If the algorithm termination condition has not been met, return to S2.5 to continue iterating.
3. The information learning method considering uncommissioned slab matching according to claim 2, characterized in that: The formula for calculating population similarity in S2.5 is as follows: in, This refers to the actual entropy value of the current population. The maximum entropy value of the current population. Indicates the similarity of the current population. Represents the relative rate of entropy.
4. The information learning method considering uncommissioned slab matching according to claim 3, characterized in that: The actual entropy value of the current population The expression is as follows: in, Represents a specific slab in the set of slabs , Represents a specific order in the order set. , In the current population Board matching The probability of an order. For all of the current population The board is allocated Conditional probability of an order. and Both represent summation; The maximum entropy value of the current population The calculation formula is as follows: in, That is the total number of orders. Represents a logarithmic function.
5. The information learning method considering uncommissioned slab matching according to claim 2, characterized in that: The specific steps of the tabu search optimization based on the fast evaluation mechanism in S2.8 are as follows: S2.8.1: The encoding initialization based on S2.1 is a tabu search matrix based on fast evaluation, and its expression is as follows; in, This represents the initial tabu search matrix based on fast evaluation. and Let these represent the number of rows and columns of the probability matrix, respectively. The size is consistent with the size of the slab assembly. The size is the same as the size of the order set. After a rapid evaluation, the slab... Assigned to order The cost difference between the original allocation plan and the subsequent one slab Assigned to order The extent of the taboo surrounding this action; S2.8.2: Based on the matrix Choose one in At the same time, Find the minimum local optimization solution, recalculate the objective function value, and update the matrix. Specifically: Order Equal to the forbidden length, all other non-zero values The value is reduced by one accordingly; Calculate all values in the new matrix using a fast evaluation method. The search depth is incremented by one, and the search exits after reaching the given search depth, outputting the code of the optimal solution for the objective function during the search process.
6. An information learning system considering unmanaged slab matching, based on the information learning method considering unmanaged slab matching as described in any one of claims 1-5, characterized in that: include: A module is established to build a mathematical model for the slab-less matching problem that considers minimizing the total matching cost as the optimization objective. The optimization module is used to optimize the optimization objective in the above model using an evolutionary algorithm based on information learning.