Metallurgy reservoir area stack position decision-making method based on multi-objective optimization
Through the multi-objective optimization of stacking position decision-making method in metallurgical reservoir area, the problem of hindering logistics efficiency in the back stacking operation in the reservoir area of large steel mills is solved, and the effect of reducing the stacking volume and improving the efficiency of the sky truck is achieved.
Patent Information
- Application Number
- CN202411144071.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-20
- Publication Date
- 2024-11-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Due to the large number of layers and unreasonable distribution of stacking positions in large steel mills, a large number of back-up stacking operations are caused, which reduces the effective utilization rate of logistics equipment and affects the efficiency of slab entry and exit.
The stacking decision-making method of metallurgical warehouse area based on multi-objective optimization is adopted. By obtaining the slab properties and the trolley properties, it is divided into three types of stacking phenomena to calculate the number of stacking times, and the improved NSGA-II algorithm is used for multi-objective optimization. Optimize the stacking position and scheduling of the slabs to reduce the amount of stacking and shorten the feed length of the slabs on the rollers.
Effectively reduce the amount of stacking in the warehouse area, reduce the degree of irregular stacking positions, reduce logistics waiting time, improve the working efficiency of the sky truck, and improve the efficiency of slab outage.
Smart Images

Figure CN118966972A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of metallurgical storage area logistics scheduling, and in particular to a metallurgical storage area stacking position decision method based on multi-objective optimization. Background Art
[0002] Warehouse area optimization management refers to optimizing the position of slabs in the warehouse area, i.e., the stack, so as to minimize the use of operating resources. Minimum use of operating resources means achieving the highest space utilization and the most efficient overhead crane operation while satisfying the stack storage constraints, including the shortest moving distance, the fastest arrival at the designated stack, and reducing stacking.
[0003] At present, the layout of the slab warehouse of some large steel mills has many layers. The unreasonable allocation of the stacking positions of the incoming slabs will lead to a large number of ineffective stacking operations, reducing the effective utilization rate of logistics equipment. In addition, the unreasonable scheduling of the overhead crane will affect the efficiency of the slab in and out, thereby limiting the production rhythm. Therefore, the optimal stacking position allocation and reasonable overhead crane scheduling in the slab warehouse are of great significance.
[0004] In traditional stacking operations, the obstructing slab needs to be moved to other stacks by overhead crane equipment. If the obstructing slab moves to another stack where the target slab is located, a secondary stacking will occur, and the overhead crane needs to be moved in another operation time cycle. A large number of studies have shown that stacking operations are the key factor hindering the logistics efficiency of warehouses. Therefore, a stacking position decision method that can reasonably optimize the allocation of stacking positions to reduce or even avoid stacking operations is urgently needed to solve this problem. Summary of the invention
[0005] 1. Technical issues to be resolved
[0006] In view of the deficiencies of the prior art, the present invention provides a stack position decision method for a metallurgical storage area based on multi-objective optimization, which solves the problem in the prior art that the stacking operation will hinder the logistics efficiency of the storage area.
[0007] (II) Technical solution
[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:
[0009] The metallurgical storage area stacking position decision method based on multi-objective optimization of the present invention comprises:
[0010] Obtain the type of steel, steel temperature, length, temperature, width of the slab, and the current material distribution in the warehouse;
[0011] Simplify the abstract constraints of the warehouse area, establish constraints for the allocation of stacking positions, and establish constraints for the operation of the overhead crane; define the warehouse stacking position of the slab and the selected overhead crane operation instructions for executing the warehouse operation as decision variables, and define the parameters related to the properties of the warehoused slab, the properties of the overhead crane, and the properties of the stacking position in the warehouse as auxiliary variables;
[0012] The stacking phenomenon is divided into three categories, and the corresponding stacking times are calculated respectively;
[0013] The stacking phenomenon and the corresponding stacking times are calculated as follows:
[0014] when That is, if there is a slab k that has been taken away earlier in the lower slab of the same stack as slab i, slab i will be moved to other stacks in advance. Since there is no secondary stacking, the number of stacking times N for slab i is i =0;
[0015] when and That is, no slab in the lower layer of slab i is removed first, but a slab on the upper layer of slab i is removed first. When it is the turn of slab i, the number of slabs on its upper layer changes, resulting in a change in the number of stacking times. At this time, N i =D i -max(Q ik -D k )-1;
[0016] when and That is, there is no priority in the plan number of slabs in the upper and lower layers of the same stack of slab i. At this time, the number of slabs i stacked down is the initial number of slabs in the upper layer, that is, N i =D i ;
[0017] The number of times N the slab i is stacked is obtained i , the calculation formula for the number of stacking times is:
[0018]
[0019] Carry out multi-objective optimization of slab stacking model to minimize the amount of slab stacking required by hot rolling production plan when it is out of the warehouse;
[0020] The NSGA-II algorithm is used to improve the problems existing in the practical application of the NSGA algorithm. For the multi-objective optimization problem of minimization, the dominance relationship of the solution is redefined. The NSGA-II algorithm sorts all individuals in the population according to the fast non-dominated sorting, and continues to sort in its dominance solution until all individuals have a level.
[0021] The calculation steps of the NSGA-II algorithm are:
[0022] a. For each individual in the population, calculate its dominant solution set S p and the number n dominated p ;
[0023] b. The number n that will be dominated p= 0 are assigned to the F1 set with the highest pareo level;
[0024] c. For the individuals in the F1 set, do n for all the individuals in their dominant solution set p =n p -1, for n p =0 individuals are classified into the F2 set;
[0025] d. Repeat step c to complete the classification of F3, F4, ..., and the classification ends when all individuals in the population have completed the classification;
[0026] The casting slab needs to enter the slab yard, and the optimization is carried out with the goal of reducing the amount of stacking when the slab is taken away and shortening the feeding length of the slab on the roller when entering the slab yard;
[0027] The specific character string encoding form of reducing the stacking amount when taking away the slab and shortening the feeding length of the slab on the roller when entering the slab yard is specifically:
[0028] Input = {slab number, length, width, temperature, hot rolling schedule number, billet picking sequence number, storage roller number, storage sequence number} N*8
[0029] Where N represents the number of slabs to be optimized;
[0030] The crossover operation is achieved by exchanging the genetic information of random length at the head of two populations and regenerating the genetic information at the tail.
[0031] The adaptive formula is improved to make it suitable for dual-objective operation.
[0032] (III) Beneficial effects
[0033] The present invention provides a stacking position decision method for metallurgical storage areas based on multi-objective optimization. It has the following beneficial effects:
[0034] The present invention establishes an intelligent model for slab storage stacking decision with the goals of reducing the amount of stacking in the warehouse, reducing the irregularity of stacking positions, reducing logistics waiting time, and improving the working efficiency of the overhead crane. In combination with the problem characteristics and empirical knowledge, a multi-objective intelligent optimization algorithm based on the improved NSGA-II is designed to efficiently solve the problem, thereby solving the problem in the prior art that the stacking operation will hinder the logistics efficiency of the warehouse area. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 It is the information flow diagram of the in-and-out model in the present invention;
[0036] Figure 2 This is a schematic diagram of the layout of the slab warehouse in the present invention;
[0037] Figure 3 This is a framework diagram of the slab storage optimization model in the present invention;
[0038] Figure 4 A classification diagram for calculating the number of stacking failures in the present invention;
[0039] Figure 5 It is a ten-bit binary code diagram in the present invention;
[0040] Figure 6 It is the eight-bit floating point number coding diagram in the present invention;
[0041] Figure 7 It is a statistical diagram of the information of the slabs entering the warehouse in the present invention;
[0042] Figure 8 It is a stacking position allocation diagram of the slabs to be stored in the present invention;
[0043] Fig. 9 This is a flow chart of population initialization in the present invention;
[0044] Fig.10 is a single-point exchange graph in the present invention;
[0045] Fig.11 is the interval exchange graph in the present invention;
[0046] Fig.12 It is a single point insertion diagram in the present invention;
[0047] Fig.13 It is the flip diagram in the present invention;
[0048] Fig.14 It is an interval insertion diagram in the present invention;
[0049] Fig.15 It is a cross operation diagram in the present invention;
[0050] Fig.16 It is a variation operation diagram in the present invention;
[0051] Fig.17 This is the adaptive effect diagram of the present invention. DETAILED DESCRIPTION
[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0053] Example
[0054] Reservoir area optimization management can be divided into Figure 1Inbound planning management optimization and outbound planning management optimization are shown.
[0055] Figure 2 The layout of the slab storage area of a large steel plant is shown. There are three spans, and each span has two overhead cranes for cooperation. There are two continuous casting machines on the left side of the storage area. There are two horizontal transfer rollers in the storage area that pass through the three spans to transport slabs in or out of the storage. In addition, there are cross-span vehicles in the storage area to realize the transportation of slabs between different spans. The maximum number of slabs that can be stacked per stack is 13 layers, and the specific stacking specifications of the width difference, length difference and temperature difference of two adjacent layers of slabs cannot be too different.
[0056] The obstructing slab in the stacking operation needs to be moved to other stacks by the overhead crane. If the obstructing slab is moved to another stack where the target slab is located, a secondary stacking will occur, and the overhead crane needs to remove it in another operation time cycle. A large number of studies have shown that stacking operations are the key factor hindering the logistics efficiency of the warehouse area, so it is necessary to reasonably optimize the allocation of stack positions to reduce or even avoid stacking operations.
[0057] This embodiment refers to the actual warehouse layout of a slab warehouse in a domestic steel plant and the physical state information of the overhead crane operation. According to the received production data information, including the continuous casting billet discharge plan and the hot rolling billet supply plan information, the goal is to reduce the amount of stacking when the hot rolling billet is discharged and optimize the distribution of the billets in the warehouse area. It is expected to output the optimal warehouse billet stack allocation instructions and overhead crane operation scheduling instructions within a certain period of time, and establish a slab warehouse storage optimization model. The model framework is as follows: Figure 3 shown.
[0058] The stacking decision problem requires determining the most suitable position for the slab to ensure that the slab can meet the requirements of fast, reliable and low consumption when it is out of the warehouse. It is affected by many factors, including the material, specifications and current distribution of slabs in the warehouse.
[0059] The present invention provides a stacking position decision method for a metallurgical storage area based on multi-objective optimization, comprising:
[0060] Obtain the type of steel, steel temperature, length, temperature, width of the slab and the distribution of the slabs in the warehouse;
[0061] Simplify the abstract constraints on the warehouse area, establish constraints on the allocation of stacking positions, and establish constraints on overhead crane operations. Among them:
[0062] The constraints for stack allocation specifically include:
[0063] Requirements for setting up storage areas for incoming slabs: DHCR, a direct hot-charged continuous casting slab, enters the heating furnace directly through a roller table, HCR, a hot-charged continuous casting slab, is transported from the roller table to the heating furnace by a crane, and CCR, a cold-charged continuous casting slab, is transported to the stack by a crane, and no further stacking is allowed on the stack;
[0064] Stacks that need to be inspected for quality reasons will be blocked, which means that no more stacking is allowed on these stacks;
[0065] Slabs of the same specification whose length, width and thickness are within the tolerance range specified in the rolling regulations, and whose material, steel grade and hardness code are the same, are stacked together;
[0066] During storage, no slabs were withdrawn from the allocated stack and no slabs from other parts of the yard entered the allocated stack.
[0067] The specific operating constraints of the overhead crane include:
[0068] Overhead cranes in the same span cannot cross each other, and there must be a minimum safety distance between any two adjacent overhead cranes;
[0069] When the slab is transferred on the roller table, the overhead crane lifts it and starts the operation;
[0070] The speed of the overhead crane remains constant during the movement;
[0071] The accidental factor of equipment failure is not taken into account in the model.
[0072] In order to accurately describe the stacking position decision process in mathematical form, the present invention defines the following parameters and variables.
[0073] (1) Model parameter symbols
[0074] According to the production plan of the hot rolling production line downstream of the slab warehouse, the outbound time of each slab in the current warehouse and the upstream continuous casting plan that will be put into the warehouse can be calculated, and the outbound sequence of each slab can be obtained, so that the outbound sequence of the slabs can be used as the parameter variable of the model:
[0075] Represents the slab collection in the warehouse;
[0076] Represents time / overhead crane operation phase;
[0077] (x, y, z)∈(X, Y, Z) represents the stack position (row, column, number of layers) of the slab;
[0078] represents the Skytrain;
[0079] b i represents the set of obstacle slabs for slab i;
[0080] Pcitb represents the starting position of slab i operated by overhead crane c in stage t;
[0081]
[0082] m srepresents the slab at the bottom of the target slab set in stack s;
[0083]
[0084] N i represents the number of stacking times of slab i;
[0085] D i represents the number of slabs on the upper layer of slab i. If slab i is at the top of the stack, then D i =0;
[0086] I is represents the set of slabs in the same stack s as slab i;
[0087] P ct represents the position of overhead travelling crane c at time t;
[0088] A represents the maximum number of billets that can be lifted by the overhead crane in a single operation, A=1;
[0089] N s represents the maximum number of slabs that can be placed in stack s, N s =13;
[0090] Represents the number of overhead cranes across F1, F2, and F3 in the storage area.
[0091] δ represents the safe distance between the sky and the vehicle;
[0092] S oe Represents the initial space-time stack set;
[0093] S of represents the set of non-empty stack positions at the initial time;
[0094] H st Represents the height of stack position s at time t;
[0095] WD is represents the width of slab i at the storage position of stack s;
[0096] T is represents the temperature of slab i at storage position s;
[0097] L is represents the length of slab i at the storage position of stack s;
[0098] V x , V y , V z Represents the moving speed of the overhead crane in three directions:
[0099] V x =61m / min, V y=120m / min, V z =10m / min
[0100] T r Represents the time required for the overhead crane to pick up / put down on the roller table once;
[0101] T s Represents the time required for the overhead crane to pick up / put down on the stack;
[0102] T c Represents the operation time of overhead crane c;
[0103] L c Represents the distance between two adjacent spans;
[0104] L bx Represents the length of the span;
[0105] T slab Represents the time interval between two slabs entering the same roller conveyor, T slab =7min;
[0106] WD permit Represents the maximum allowable width difference between the upper and lower slabs in the same stack;
[0107] L permit Represents the maximum allowable length difference between the upper and lower slabs in the same stack;
[0108] T permit Represents the maximum allowable temperature difference between the upper and lower slabs in the same stack;
[0109] (2) Model variable symbols
[0110] The variable symbol Pcitf is the allocated stack position of the slab.
[0111] P citf represents the storage location of slab i;
[0112]
[0113] According to the above description of the calculation of the amount of stacking, the total amount of stacking in the model can be expressed as follows: D i represents the number of slabs on the upper layer of slab i at the initial time;
[0114]
[0115]
[0116] The calculation and derivation of the number of pallet flipping is the key to establishing the mathematical model of warehouse scheduling. The secondary pallet flipping will affect the validity of the final calculation result of the number of pallet flipping.
[0117] However, in most cases, there are enough stacking positions in the warehouse area to store the obstructed slabs, so that secondary stacking will not occur. Therefore, the present invention only studies the situation where secondary stacking does not occur, divides the stacking phenomenon into three categories, and calculates the corresponding stacking times respectively, such as Figure 4 As shown:
[0118] The calculation of the stacking phenomenon and the corresponding stacking times is as follows:
[0119] For case (a), when That is, if there is a slab k that has been taken away earlier in the lower slab of the same stack as slab i, slab i will be moved to other stacks in advance. Since there is no secondary stacking, the number of stacking times N for slab i is i =0;
[0120] For case (b), when and That is, no slab in the lower layer of slab i is removed first, but a slab on the upper layer of slab i is removed first. When it is the turn of slab i, the number of slabs on its upper layer changes, resulting in a change in the number of stacking times. At this time, N i =D i -max(Q ik D k )-1;
[0121] For case (c), when and That is, there is no priority in the plan number of slabs in the upper and lower layers of the same stack of slab i. At this time, the number of slabs i stacked down is the initial number of slabs in the upper layer, that is, N i =D i ;
[0122] The number of stacking times N of slab i obtained in the combined cases (a)-(c) is i , the calculation formula for the number of stacking times is:
[0123]
[0124] The multi-objective optimization of the slab stacking allocation model is carried out to minimize the amount of slab stacking required by the hot rolling production plan when it is out of the warehouse.
[0125] The objective function of the multi-objective optimization model for slab stacking allocation specifically includes:
[0126] Minimize the amount of slab stacking required by the hot rolling production plan when it is out of the warehouse;
[0127] Ensure that incoming slabs are stacked as close to the incoming end as possible;
[0128] The model constraints include the calculation of the stacking amount and its prerequisites, the specification constraints of the upper and lower slabs in the same stack, the stacking height constraints, the stacking logic constraints, and the overhead crane operation instructions and motion constraints. The specific expressions include:
[0129] Ensure that each incoming slab can only be stored in one stack;
[0130] Ensure that each storage location has only one slab;
[0131] The process used to constrain the overhead crane to transport the slab i to the stacking position s is completed in a state t.
[0132] Ensure that slab i cannot be removed before the slab above it is removed;
[0133] It means that for the overhead crane, the end position of the previous stage is the starting position of the next stage;
[0134] It means that considering the stacking operation, the actual number of overhead crane operations is higher;
[0135] Indicates that there is a stacking position of the target slab, and the stacking slab is not allowed to enter before the target slab is taken away;
[0136] Used to avoid placing slabs in mid-air;
[0137] Used to constrain each slab to be repositioned at most once;
[0138]
[0139] is the calculation formula for the number of times slab i is stacked.
[0140] Ensure safe distance between overhead cranes;
[0141] The specific expressions of model constraints also include:
[0142] Ensure that one slab can only be moved by one overhead crane;
[0143] Ensure that the overhead crane can move a limited number of panels in one operation;
[0144] Limit the speed of the overhead crane;
[0145] Limit the stack height;
[0146]
[0147] Ensure that the number of overhead cranes on each span is constant;
[0148] It is the width constraint of the upper and lower slabs in the same stack;
[0149] It is the length constraint of the upper and lower slabs in the same stack;
[0150] It is for the same stack position up and down
[0151] Indicates that the operation starts when the overhead crane takes the slab away from the roller table, and the operation ends when the overhead crane returns to the roller table;
[0152] It means that plate i can only move within the span area, and the distance between two adjacent spans is LC;
[0153] Limit the moving distance of the overhead crane on the span;
[0154] Ensure that two adjacent operations of the overhead crane can be successfully connected in time;
[0155] The model is a mixed integer scale model, which mainly includes the calculation of the stacking amount and its prerequisites, the specification constraints of the upper and lower slabs in the same stack, the stacking height constraints, the stacking logic constraints, and the overhead crane operation instructions and motion constraints. The decision variable is P citf Slab distribution and P citb Operation instructions of overhead crane;
[0156] The above-mentioned stacking position decision model is a multi-objective optimization problem, and there are usually two methods for dealing with multi-objective optimization problems: one is to sum the weights of each target to simplify the multi-objective optimization into a single-objective optimization. The disadvantage of this method is that the weights of each target value are difficult to determine and need to be manually modified and determined. Different weight values will produce different optimization results, and manual intervention is relatively large. Another way is to grade the solutions by Pareto, and use the Pareto frontier of a set of optimal solutions of the objective function as the final output of the model. Since the overhead crane scheduling problem and SLAP (storage location allocation problem) are both proven to be NP-hard problems, their integrated optimization problem is also an NP-hard problem. Based on the complexity of the problem and the diversity of the targets, the present invention adopts a multi-objective intelligent optimization algorithm to solve the integrated optimization model of SLAP and CSP on the basis of considering the solution accuracy and solution efficiency.
[0157] The NSGA algorithm (Non-dominated Sorting Genetic Algorithm) is one of the most popular multi-objective evolutionary algorithms. The algorithm has a good effect in searching for the optimal solution set and maintaining the diversity of the Pareto frontier. However, some problems that restrict the algorithm have been found in practical applications:
[0158] 1) The time complexity of non-dominated sorting is large and the search is slow;
[0159] 2) Without an elite conservation strategy, the population is prone to degradation;
[0160] 3) The sharing parameter is manually specified, and this parameter has a significant impact on the diversity of the Pareto frontier;
[0161] The NSGA-II algorithm is used to improve the problems existing in the practical application of the NSGA algorithm. For the multi-objective optimization problem of minimization, the dominance relationship of the solution is redefined. The NSGA-II algorithm sorts all individuals in the population according to the fast non-dominated sorting, and continues to sort in its dominance solution until all individuals have a level.
[0162] The calculation steps of NSGA-II algorithm are as follows:
[0163] a. For each individual in the population, calculate its dominant solution set S p and the number n dominated p ;
[0164] b. The number n that will be dominated p = 0 are assigned to the F1 set with the highest pareo level;
[0165] c. For the individuals in the F1 set, do n for all the individuals in their dominant solution set. p =n p -1, for n p =0 individuals are classified into the F2 set;
[0166] d. Repeat step c to complete the classification of F3, F4, ... until all individuals in the population have completed the classification;
[0167] The main steps of the NSGA-II algorithm include problem encoding, population initialization, population neighborhood search, and problem decoding. The present invention combines problem characteristics and empirical knowledge to improve the population initialization and neighborhood search strategies in the traditional NSGA-II algorithm, so as to improve the solution quality and coverage to improve the algorithm's solution efficiency.
[0168] Chromosome encoding and decoding are important steps in genetic algorithms. They have a great impact on the accuracy of the solution, the difficulty of algorithm operation, and the subsequent population crossover and population mutation methods. Therefore, chromosome encoding has a great influence on the solution quality and efficiency of the genetic algorithm. Commonly used encoding methods for genetic algorithms include binary encoding, integer encoding, and string encoding.
[0169] See also Figure 5 and Figure 6 Binary uses a binary symbol set {0,1} to encode and decode chromosomes. The accuracy of binary code is related to the length of the code, and a good result requires a lot of calculations. Its advantages are:
[0170] (1) Fast calculation speed. Compared with numerical calculation, binary calculation is faster and takes up less memory;
[0171] (2) Occupies less memory.
[0172] Integer encoding uses a specific integer value to represent a gene value of a chromosome, and the length of the chromosome is related to the number of decision variables in the optimization model. Its advantages are:
[0173] (1) The solution has high accuracy and is suitable for solving problems with continuous variables;
[0174] (2) It is suitable for representing values with a relatively large range and for searching a large string of numbers in a large space;
[0175] (3) Reduces computational complexity and improves efficiency;
[0176] (4) Suitable for mixed use of genetic algorithms and other heuristic optimization algorithms;
[0177] (5) Suitable for setting corresponding genotypes for special problems;
[0178] (6) It is suitable for expressing complex constraints and can be used for combinatorial optimization problems.
[0179] String encoding has similar properties to floating-point encoding. Compared with floating-point encoding, string encoding can make the problem simpler and easier to understand, and facilitate special processing of some complex constraints.
[0180] In view of the particularity of CSP and SLAP problems, including the complex constraints of incoming material sequence constraints and slab specifications constraints in the same stack;
[0181] The casting slab needs to enter the slab yard, and the optimization is carried out with the goal of reducing the amount of stacking when the slab is taken away and shortening the feeding length of the slab on the roller when entering the slab yard;
[0182] The specific string encoding form for reducing the stacking amount when taking away the slab and shortening the feeding length of the slab on the roller when entering the slab yard is as follows:
[0183] Input = {slab number, length, width, temperature, hot rolling schedule number, billet picking sequence number, storage roller number, storage sequence number} N*8
[0184] Where N represents the number of slabs to be optimized.
[0185] Figure 7 This is an example of statistics of incoming slab information.
[0186] Output = {stack number 1, stack number 2, ..., stack number N} 1*N ;
[0187] The stack number is represented by the stack number-layer number; for example, F1-Y10-1-13 means that the incoming slab is stored in the 13th layer of the stack F1-Y10-1;
[0188] Figure 8 The optimal stacking position allocation is output after the scheduling model is optimized. The first line indicates the storage order of the slabs, and the second line indicates the target stacking position number and number of layers of the slabs.
[0189] However, due to the diversity of solutions brought by the random search strategy and multiple populations of the genetic algorithm, it is not sensitive to population initialization and does not need to create an initial population close to the optimal solution. Therefore, the present invention uses a method of random initialization combined with priority rules to generate the initial population. Since the present invention studies a multi-constraint problem, including some time and space constraints, executing a randomly initialized population may lead to the emergence of infeasible solutions, thereby wasting computing resources. In order to avoid the algorithm wasting most of its time on the calculation of infeasible solutions, affecting the algorithm's computing efficiency, the design standard for population initialization is to randomly initialize and generate an initial population in the feasible solution space. The stack initialization formula based on the priority rule is specifically:
[0190] P ij =a 1 WP ij +a 2 TP ij +a 3 RP ij +a 4 DP ij +a 5 HP ij
[0191] Among them, P ij represents the priority of allocating the i-th incoming slab to the j-th stack, WP ij TPij RP ij , DP ij They represent the similarity between the i-th slab and the top slab of the j-th stack in terms of size, thickness, hardness and steel grade, respectively. 1 、a 2 、a 3 、a 4 、a 5 Respectively represent WP ij TP ij RP ij , DP ij , HP ij The corresponding weight coefficient is: if mixed stacking is allowed at the jth stacking position, then HP ij =1, otherwise HP ij =0, by giving a 1 ~a 5 Assign different weight coefficients to obtain the corresponding initialization scheme, such as a 1 ~a 5 The weight coefficients can be assigned as {0.2, 0.2, 0.2, 0.2, 0.2} and {0.1, 0.1, 0.3, 0.3, 0.2};
[0192] The detailed process of population initialization includes:
[0193] Under the condition that the sequence of incoming slabs in the slab warehouse and the storage status of slabs in the warehouse area are known, firstly, according to the constraints of slab specifications, for each incoming slab, a set of available storage positions for the slab is selected from the warehouse area; then, according to the above priority rules, the stack with the highest priority is selected from the generated set of available storage positions to place the slab; then the inventory information of the slab warehouse is updated. When all incoming slabs have selected a stack position to be placed, it means that an individual has been generated. The above actions are repeated N times until a population is generated, and the population initialization is completed;
[0194] The detailed process of population initialization is as follows Fig. 9 As shown;
[0195] In order to get new solutions from the current solutions of each algorithm, some neighborhood operators are used in the algorithm. When a new solution is needed, one or a combination of these operations can be used randomly and applied to the solution. These neighborhood generation operations are included in each search algorithm. The neighborhood operators in genetic algorithms are usually crossover operators and mutation operators. Commonly used crossover and mutation methods include:
[0196] See also Figure 10-Figure 14 , exchanging single genetic information: During the exchange process, two genetic information are selected and then their positions in the chromosomes are replaced.
[0197] Swap a sequence: select two chromosome subsequences and then replace their positions in the chromosome.
[0198] Insert a single genetic information: Select and extract a genetic information from the current chromosome. Then select a second genetic information from the remaining genetic information and place the first genetic information before the genetic information.
[0199] Flip: Select a chromosome subsequence and then reverse the sequence.
[0200] Insert a genetic information sequence: select and extract a subsequence from the current chromosome. Then select one from the remaining genetic information and place the subsequence before the genetic information.
[0201] The crossover operation is achieved by exchanging the genetic information of random length at the head of two populations and regenerating the genetic information at the tail.
[0202] Crossover is essentially a process in which two parent populations exchange some genetic information and then generate offspring populations. Since the slab warehouse stacking is similar to a stack storage method, the slabs that meet the continuous casting billet plan will be stored in the lower layer of the stack and will be taken away near the tail of the queue in the hot rolling billet plan. The slabs that meet the continuous casting billet plan will be stored in the upper layer of the stack and will be taken away near the head of the queue in the hot rolling billet plan. The traditional crossover method will make the stacking order of the slabs in the offspring population inconsistent with the storage order, resulting in a large number of infeasible solutions. In order to avoid repeatedly generating a large number of infeasible populations and occupying computing resources, we implement the crossover operation by randomly exchanging the genetic information of the heads of the two populations with random lengths and regenerating the genetic information of the tails. This crossover operation meets the stack form of slab storage stacking and can ensure the legitimacy of the offspring population solution.
[0203] The specific operations are as follows Fig.15 As shown;
[0204] Crossover operation Specific crossover operation:
[0205] Step 1: Randomly select two individuals in the population and generate a random length as the length of the genetic information to be exchanged;
[0206] Step 2: Delete the remaining part of the chromosome and exchange the genetic information of the two individual heads;
[0207] Step 3: Generate the missing genetic information in an initialization manner;
[0208] Chromosome mutation changes the genetic information of the original chromosome of an individual. However, randomly changing chromosomes that are subject to time and space constraints will cause their storage order and position allocation to not conform to the form of a stack. In order to avoid this, the operation of chromosome mutation is improved based on the reference crossover operation.
[0209] Chromosome mutation operations such as Fig.16 As shown;
[0210] Specific operation of chromosome mutation:
[0211] Step 1: Generate a random length as the length of the chromosome information retained in the head;
[0212] Step 2: Delete the unretained chromosome information;
[0213] Step 3: Complete the missing genetic information according to the population initialization method;
[0214] The adaptive formula is improved to make it suitable for dual-objective operation;
[0215] The adaptive parameter operation is based on the traditional genetic algorithm, and the fitness is used to automatically change the crossover probability Pc and the mutation probability Pm. When the population just starts to search or the fitness value is small and tends to the local optimum, the crossover probability and the mutation probability are increased accordingly, thereby expanding the neighborhood search range, which is conducive to the algorithm jumping out of the local optimum. When the population search is nearing the end or the fitness value is large and close to the optimum, the crossover probability Pc and the mutation probability Pm are reduced accordingly, so that the population can quickly converge to the optimum. At the same time, the probability of crossover and mutation of individuals with strong adaptability is low, while the probability of crossover and mutation of individuals with poor adaptability is high. Different from the traditional single-objective optimization, the present invention improves the adaptive formula to make it suitable for dual-objective operation.
[0216] The specific adaptive formula is:
[0217]
[0218] f 11 , f 12 The fitness values of the first target of the two selected individuals;
[0219] f 21 , f 22 The fitness values of the second goal of the two selected individuals;
[0220] The average of the first and second target fitness values of the current population;
[0221] The maximum value of the first and second target fitness values of the current population;
[0222] The minimum value of the first and second target fitness values of the current population;
[0223] f′ 1 , f′ 2 f′ 1 =max(f 11 , f 12 ), f′ 2 =max(f 21 , f 22 );
[0224] The present invention also conducted a comparative test on the basis of controlling the parameter variables by introducing an adaptive algorithm and not introducing an adaptive algorithm. The operation results are as follows: Fig.17 As shown in the figure, the ordinate is the objective function to be minimized, and the abscissa represents the number of iterations of the algorithm. Fig.17 It can be seen that the introduction of adaptive crossover Pc and mutation probability Pm can speed up the search and convergence speed of the NSGA-II algorithm, improve the operation efficiency of the algorithm, and make the entire algorithm converge to the optimal solution in a relatively short time, which has a very significant effect on solving large-scale NP-hard problems. Therefore, the present invention adopts adaptive crossover and mutation probabilities for the underlying genetic algorithm.
[0225] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.
Claims
1. A stacking position decision method for metallurgical storage area based on multi-objective optimization, characterized in that: include: Obtain the type of steel, steel temperature, length, temperature, width of the slab, and the current material distribution in the warehouse; Simplify the abstract constraints of the warehouse area, establish constraints for the allocation of stacking positions, and establish constraints for the operation of the overhead crane; define the warehouse stacking position of the slab and the selected overhead crane operation instructions for executing the warehouse operation as decision variables, and define the parameters related to the properties of the warehoused slab, the properties of the overhead crane, and the properties of the stacking position in the warehouse as auxiliary variables; The stacking phenomenon is divided into three categories, and the corresponding stacking times are calculated respectively; The stacking phenomenon and the corresponding stacking times are calculated as follows: when That is, if there is a slab k that has been taken away earlier in the lower slab of the same stack as slab i, slab i will be moved to other stacks in advance. Since there is no secondary stacking, the number of stacking times N for slab i is i =0; when and That is, no slab in the lower layer of slab i is removed first, but a slab on the upper layer of slab i is removed first. When it is the turn of slab i, the number of slabs on its upper layer changes, resulting in a change in the number of stacking times. At this time, N i =D i -max(Q ik -D k )-1; when and That is, there is no priority in the plan number of slabs in the upper and lower layers of the same stack of slab i. At this time, the number of slabs i stacked down is the initial number of slabs in the upper layer, that is, N i =D i ; The number of times N the slab i is stacked is obtained i , the calculation formula for the number of stacking times is: Carry out multi-objective optimization of slab stacking model to minimize the amount of slab stacking required by hot rolling production plan when it is out of the warehouse; The NSGA-II algorithm is used to improve the problems existing in the practical application of the NSGA algorithm. For the multi-objective optimization problem of minimization, the dominance relationship of the solution is redefined. The NSGA-II algorithm sorts all individuals in the population according to the fast non-dominated sorting, and continues to sort in its dominance solution until all individuals have a level. The calculation steps of the NSGA-II algorithm are: a. For each individual in the population, calculate its dominant solution set S p and the number n dominated p ; b. The number n that will be dominated p = 0 are assigned to the F1 set with the highest pareo level; c. For the individuals in the F1 set, do n for all the individuals in their dominant solution set. p =n p -1, for n p =0 individuals are classified into the F2 set; d. Repeat step c to complete the classification of F3, F4, ..., and the classification ends when all individuals in the population have completed the classification; The casting slab needs to enter the slab yard, and the optimization is carried out with the goal of reducing the amount of stacking when the slab is taken away and shortening the feeding length of the slab on the roller when entering the slab yard; The specific character string encoding form of reducing the stacking amount when taking away the slab and shortening the feeding length of the slab on the roller when entering the slab yard is specifically: Input = {slab number, length, width, temperature, hot rolling schedule number, billet picking sequence number, storage roller number, storage sequence number} N*8 Where N represents the number of slabs to be optimized; The crossover operation is achieved by exchanging the genetic information of random length at the head of two populations and regenerating the genetic information at the tail. The adaptive formula is improved to make it suitable for dual-objective operation.
2. The method for stacking position decision-making in metallurgical storage area based on multi-objective optimization according to claim 1 is characterized in that: The stacking position allocation constraints specifically include: Requirements for setting up storage areas for incoming slabs: DHCR, a direct hot-charged continuous casting slab, enters the heating furnace directly through a roller table, HCR, a hot-charged continuous casting slab, is transported from the roller table to the soaking furnace by a crane, and CCR, a cold-charged continuous casting slab, is transported to the stack by a crane, and no further stacking is allowed on the stack; Stacks that need to be inspected for quality reasons will be blocked, i.e. no more stacking is allowed on these stacks; The slabs of the same specification whose length, width and thickness are within the tolerance range specified in the rolling regulations and whose materials, steel grades and hardness codes are the same shall be stacked; During storage, no slabs were withdrawn from the allocated stack and no slabs from other parts of the yard entered the allocated stack.
3. The method for stacking position decision-making in metallurgical storage area based on multi-objective optimization according to claim 1 is characterized in that: The overhead crane operation constraints specifically include: Overhead cranes in the same span cannot cross each other, and there must be a minimum safety distance between any two adjacent overhead cranes; When the slab is transferred on the roller table, the overhead crane lifts it and starts the operation; The speed of the overhead crane remains constant during the movement.
4. The method for stacking position decision-making in metallurgical storage area based on multi-objective optimization according to claim 1 is characterized in that: The auxiliary variable parameter definition specifically includes: Represents the slab collection in the warehouse; Represents time / overhead crane operation phase; (x, y, z)∈(X, Y, Z) represents the stack position (row, column, number of layers) of the slab; represents the Skytrain; b i The set of obstacle slabs for slab i; Pcitb stage t overhead crane c operation slab i starting position: m s Represents the slab at the bottom of the target slab set in stack s: N i represents the number of stacking times of slab i; D i represents the number of slabs above slab i. If slab i is at the top of the stack, then D i =0; I is represents the set of slabs in the same stack s as slab i; P ct represents the position of overhead travelling crane c at time t; A represents the maximum number of billets that can be lifted by the overhead crane in a single operation, A=1; N s represents the maximum number of slabs that can be placed in stack s, N s =13.
5. The method for stacking position decision-making in metallurgical storage area based on multi-objective optimization according to claim 1 is characterized in that: The auxiliary variable parameter definition also includes: Represents the number of overhead cranes across F1, F2, and F3 in the storage area. δ represents the safe distance between the sky and the vehicle; S oe Represents the initial space-time stack set; S of represents the set of non-empty stack positions at the initial time; H st Represents the height of stack position s at time t; WD is represents the width of slab i at the storage position of stack s; T is represents the temperature of slab i at storage position s; L is represents the length of slab i at the storage position of stack s; V x , V y , V z Represents the moving speed of the overhead crane in three directions: V x =61m / min,V y =120m / min,V z =10m / min T r Represents the time required for the overhead crane to pick up / put down on the roller table once; T s Represents the time required for the overhead crane to pick up / put down on the stack; T c Represents the operating time of overhead crane c; L c Represents the distance between two adjacent spans; Lb x Represents the length of the span; T slab Represents the time interval between two slabs entering the same roller conveyor; WD permit Represents the maximum allowable width difference between the upper and lower slabs in the same stack; L permit Represents the maximum allowable length difference between the upper and lower slabs in the same stack; T permit Represents the maximum allowable temperature difference between the upper and lower layers of slabs in the same stack.
6. The method for stacking position decision-making in metallurgical storage area based on multi-objective optimization according to claim 1 is characterized in that: The decision variables include: The decision variable is P citf Slab distribution and P citb Overhead crane operation instructions: P citf represents the storage location of slab i; 7. The method for stacking position decision-making in metallurgical storage area based on multi-objective optimization according to claim 1 is characterized in that: The decision variables are: The overall stacking volume expression formula in the model is as follows: D i represents the number of slabs on the upper layer of slab i at the initial time; 8. The method for stacking position decision-making in metallurgical storage area based on multi-objective optimization according to claim 1 is characterized in that: The objective function of the multi-objective optimization model for slab stacking allocation specifically includes: min∑ i∈I N i , so as to minimize the amount of slab stacking required by the hot rolling production plan when the slab is out of the warehouse; Ensure that the incoming slabs are stacked as close to the storage end as possible.
9. The method for stacking position decision-making in metallurgical storage area based on multi-objective optimization according to claim 1 is characterized in that: The model constraints include the calculation of the stacking amount and its prerequisites, the specification constraints of the upper and lower slabs in the same stack, the stacking height constraints, the stacking logic constraints, and the overhead crane operation instructions and motion constraints. The specific expressions include: Ensure that each incoming slab can only be stored in one stack; Ensure that each storage location has only one slab; It is used to constrain the process of the overhead crane transporting the slab i to the stacking position s to be completed in a state t; Ensure that slab i cannot be removed before the upper slab of slab i is removed; It means that for the overhead crane, the end position of the previous stage is the starting position of the next stage; Considering the stacking operation, the overhead crane actually runs more times; Indicates that there is a stacking position of the target slab, and the stacking slab is not allowed to enter before the target slab is taken away; Used to avoid placing slabs in mid-air; Used to constrain each slab to be repositioned at most once; is the calculation formula for the number of stacking times of slab i; Ensure a safe distance between overhead cranes.
10. The method for stacking position decision-making in metallurgical storage area based on multi-objective optimization according to claim 1 is characterized in that: The specific expression of the model constraint also includes: Ensure that one slab can only be moved by one overhead crane; Ensure that the overhead crane can move a limited number of panels in one operation; Limit the speed of the overhead crane; Limit the stack height; Ensure that the number of overhead cranes on each span is constant; It is the width constraint of the upper and lower slabs in the same stack; It is the length constraint of the upper and lower slabs in the same stack; It is the temperature constraint for the upper and lower slabs in the same stack; Indicates that the operation starts when the overhead crane takes the slab away from the roller table, and the operation ends when the overhead crane returns to the roller table; It means that plate i can only move within the span area, and the distance between two adjacent spans is LC; Limit the moving distance of the overhead crane on the span; Ensure that two adjacent operations of the overhead crane can be successfully connected in time.
