Steel plate stack transfer sorting system and method based on hybrid optimization algorithm and storage medium

Abstract

The invention provides a hybrid optimization algorithm-based steel plate stack transfer sorting system and method and a storage medium. The method comprises the steps of firstly, obtaining a to-be-sorted steel plate set, the number of available temporary stack positions and a target sequence; generating an initial stack transfer strategy through a heuristic rule; secondly, performing iterative optimization on the initial stack transfer strategy by using a genetic algorithm to generate an optimal stack transfer strategy by taking the minimum operation time length and the number of temporary stack positions used for stack transfer sorting as targets; and finally, stack transferring equipment is controlled to execute the stack transferring step in the optimal stack transferring strategy, and moving operation of the steel plates between the stack positions is achieved till stack transferring sequencing is completed. According to the method, the heuristic rule, the GA and the DQN are fused, so that the ternary relationship among the time constraint, the equipment resource and the temporary stack position can be coordinated, and the idling rate of the equipment is reduced, so that the production process of a shipyard is optimized, the production efficiency is improved, and meanwhile, the production cost is reduced; according to the method, the limitation of a traditional single method is broken through, the algorithm robustness is higher, and the industrial application value is prominent.

Description

Technical Field

[0001] The present invention relates to the technical field of steel plate scheduling, and more specifically, to a steel plate stacking and sorting system, method, and storage medium based on a hybrid optimization algorithm. Background Art

[0002] In the process of shipbuilding, steel plates are the core raw materials, and their loading sequence has an important impact on the production efficiency and quality of the shipyard. The loading sequence of steel plates needs to be sorted according to specific rules to meet the technological requirements of the shipyard. However, the original state of steel plates in the stack is usually randomly disordered, which poses challenges to the sorting work.

[0003] Currently, shipyards face some limitations in the process of steel plate sorting. First, the number of stacks for temporarily storing steel plates is limited, which may lead to frequent movement of steel plates during sorting, thus increasing the operation time and workload. Second, the number of devices available for sorting operations is also limited, which further restricts the sorting efficiency. In addition, the total operation time available for steel plate sorting every day is limited, which requires completing the sorting task as efficiently as possible within the limited time.

[0004] Existing research on warehousing management shows that there is a strong correlation between the number of steel plate stacking operations and the stack allocation strategy. When more than 5 lifts of steel plates are stacked in a single stack, the effective operation efficiency will be lower than 30%. The traditional method adopts the first-come, first-served (FCFS) principle, but this mode is prone to a high idling rate of equipment and is difficult to adapt to the scenario of multi-device collaborative operation. Although some sorting algorithms (such as branch and bound method, dynamic programming) have been applied to the job scheduling of steel plate sorting, most algorithms only focus on optimizing a single constraint condition and fail to effectively coordinate the ternary relationship among time constraints, equipment resources, and temporary stacks, lacking a reliable solution that can take into account multi-objective optimization.

[0005] In summary, in the multi-constraint and high-dynamic scenario unique to shipbuilding, it is still necessary to establish a composite decision-making scheme that integrates equipment efficiency, space limitation, and time window. Summary of the Invention

[0006] In order to overcome the defect that the sorting algorithm in the above-mentioned prior art has a single constraint condition and is difficult to take into account the multi-dimensional constraints in the process of steel plate stacking, the present invention provides a steel plate stacking and sorting system, method, and storage medium based on a hybrid optimization algorithm, which can fully coordinate the ternary relationship among time constraints, equipment resources, and temporary stacks, reduce the idling rate of equipment, thereby optimizing the production process of the shipyard, improving production efficiency, and reducing production costs at the same time.

[0007] To solve the above technical problems, the technical solution of the present invention is as follows:

[0008] A steel plate stacking and sorting system based on a hybrid optimization algorithm, comprising:

[0009] An initial stack position for storing steel plates to be sorted;

[0010] A final stack position for storing steel plates arranged in the target order;

[0011] A number of available temporary stack positions for temporarily storing steel plates during the steel plate stacking and sorting process;

[0012] A central control module for generating an initial stacking strategy through heuristic rules, and taking minimizing the operation duration and the number of temporary stack positions used for stacking and sorting as the goal, and using a genetic algorithm to iteratively optimize the initial stacking strategy to generate an optimal stacking strategy;

[0013] A stacking device for executing the stacking steps in the optimal stacking strategy to realize the movement operation of the steel plates between the stack positions until the stacking and sorting is completed.

[0014] Preferably, the target order includes: from the bottom layer to the top layer of the final stack position, the steel plate numbers are arranged from small to large or from large to small.

[0015] Preferably, in the central control module, the heuristic rules at least include:

[0016] Docking difference rule: According to the difference between the number of the currently removed steel plate and the number of the topmost steel plate in each temporary stack position, select a temporary stack position with a larger or smaller difference for storage;

[0017] Dynamic center bisecting rule: Real-time obtain the number of steel plates in all stack positions except the final stack position, and according to the number of available temporary stack positions, evenly distribute the steel plates in all stack positions except the final stack position to each temporary stack position; after a steel plate is moved into the final stack position, remove the steel plate during the even distribution and update the temporary stack position distribution rule of the remaining steel plates.

[0018] Preferably, in the central control module, the inputs of the genetic algorithm include:

[0019] The set of steel plates to be sorted A = {a1, a2,..., a N} and its initial arrangement sequence, where a i represents the steel plate numbered i, and N is the total number of steel plates; in the initial arrangement sequence, the steel plate number on the leftmost side represents the topmost steel plate in the initial stack position, and the steel plate number on the rightmost side represents the bottommost steel plate in the initial stack position;

[0020] The target sorting direction;

[0021] The number of available temporary stack positions L, satisfying: 2 ≤ L ≤ N;

[0022] Stacking equipment information, including the comprehensive scheduling efficiency η of the stacking equipment and the duration Y required for the stacking equipment to perform one stacking step;

[0023] The preset operation time limit T within a day, and the maximum number of stacking times

[0024] The output of the genetic algorithm includes:

[0025] The stacking steps represented by the two-dimensional array [a, b], indicating moving the topmost steel plate of the a-th stack to the b-th stack, and used to indicate the moving path of the stacking equipment;

[0026] The minimum number of temporary stacks L required to complete the stacking and sorting of steel plates within the operation time limit T min , expressed as:

[0027]

[0028] where n actual is the actual number of stacking times to complete the sorting in the case of the minimum number of temporary stacks L min .

[0029] Preferably, in the central control module, the initialization process of the genetic algorithm includes:

[0030] Initializing the number of temporary stacks to 2, randomly generating an initial stacking strategy through heuristic rules, and using the initial stacking strategy as the initial solution.

[0031] Preferably, in the central control module, using the genetic algorithm to iteratively optimize the initial stacking strategy to generate an optimal stacking strategy includes the following steps:

[0032] Obtaining the number of stacking times of the initial stacking strategy, and evaluating the corresponding operation duration through the following function:

[0033]

[0034] where T total is the operation duration; n a ′ ctual is the actual number of stacking times;

[0035] Judging whether the operation duration T total is greater than the operation time limit T within a day. If it is greater, increase the number of temporary stacks by 1, generate a new stacking strategy by using selection, crossover, and mutation operations, and re-evaluate the corresponding operation duration; if it is less than or equal to, obtain the optimal stacking strategy, and output the stacking steps corresponding to the optimal stacking strategy, as well as the minimum number of temporary stacks L min .

[0036] Preferably, in the central control module, after generating the initial stacking strategy, it further includes:

[0037] Input the initial solution into the pre-trained deep reinforcement learning network to optimize the initial solution and improve the population quality;

[0038] The pre-trained deep reinforcement learning network is used to screen the effective action set in the initial solution; the effective action set includes several effective stacking steps, and the effective stacking steps are oriented to move the steel plate with the target number to the final stacking position as soon as possible.

[0039] Preferably, the stacking device is specifically an unmanned overhead crane device.

[0040] The present invention also provides a steel plate stacking sorting method based on a hybrid optimization algorithm. Based on the above-mentioned steel plate stacking sorting system based on a hybrid optimization algorithm, it includes the following steps:

[0041] S1: Obtain the set of steel plates to be sorted, the number of available temporary stacking positions, and the target order;

[0042] S2: Generate an initial stacking strategy through heuristic rules;

[0043] S3: Take minimizing the operation duration and the number of temporary stacking positions used for stacking sorting as the goal, and use the genetic algorithm to iteratively optimize the initial stacking strategy to generate the optimal stacking strategy;

[0044] S4: Control the stacking device to execute the stacking steps in the optimal stacking strategy to realize the movement operation of the steel plate between the stacking positions until the stacking sorting is completed.

[0045] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps in the above method are realized.

[0046] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0047] The present invention provides a steel plate stacking sorting system, method and storage medium based on a hybrid optimization algorithm. First, obtain the set of steel plates to be sorted, the number of available temporary stacking positions, and the target order; then generate an initial stacking strategy through heuristic rules; then take minimizing the operation duration and the number of temporary stacking positions used for stacking sorting as the goal, and use the genetic algorithm to iteratively optimize the initial stacking strategy to generate the optimal stacking strategy; finally, control the stacking device to execute the stacking steps in the optimal stacking strategy to realize the movement operation of the steel plate between the stacking positions until the stacking sorting is completed;

[0048] The present invention aims to minimize the operation duration and the number of temporary stacking positions used in the stacking sequence. Based on the Genetic Algorithm (GA), it optimizes the initial stacking strategy to generate the optimal stacking strategy, which can fully coordinate the ternary relationship among time constraints, equipment resources, and temporary stacking positions, reduce the equipment idle rate, thereby optimizing the production process of the shipyard, improving production efficiency, and reducing production costs. Secondly, the present invention uses the Deep Q-Network (DQN) algorithm to optimize the initial solution randomly generated by the heuristic rule, so as to obtain a higher-quality initial solution and effectively improve the convergence speed of the genetic algorithm. In addition, the present invention combines the heuristic rule, GA, and DQN algorithm, which not only retains the fast decision-making ability of the rule guidance but also realizes global optimization through intelligent algorithms, breaking through the limitations of traditional single methods, with stronger algorithm robustness and prominent industrial application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Schematic diagram of the initial disordered steel plates provided in Embodiment 1.

[0050] Figure 2 Schematic diagram of the gantry crane movement route provided in Embodiment 1.

[0051] Figure 3 Experimental result diagram in the case of a small example with 6 steel plates and 2 temporary stacking positions provided in Embodiment 1.

[0052] Figure 4 Experimental result diagram in the case of a large example with 20 steel plates and 2 temporary stacking positions provided in Embodiment 1.

[0053] Figure 5 Algorithm convergence curve diagram in the case of the largest-scale example with 130 steel plates and 38 temporary stacking positions provided in Embodiment 1.

[0054] Figure 6 Schematic diagram of the marginal effect of increasing temporary stacking positions provided in Embodiment 1.

[0055] Figure 7 Diagram of the relationship among three action sets provided in Embodiment 2.

[0056] Figure 8 Schematic diagram of a certain stacking state provided in Embodiment 2.

[0057] Figure 9 Schematic diagram for comparing and verifying the DQN results provided in Embodiment 2.

[0058] Figure 10 Flowchart of a steel plate stacking sequence method based on a hybrid optimization algorithm provided in Embodiment 3.

[0059] Figure 11 It is a schematic diagram of the running result of Postman provided in Example 4.

[0060] Figure 12 It is a schematic diagram of the local log provided in Example 4. Detailed implementation manners

[0061] The accompanying drawings are only for illustrative purposes and should not be construed as limitations to this application;

[0062] To better illustrate this embodiment, some components in the accompanying drawings are omitted, enlarged or reduced, and do not represent the size of the actual product;

[0063] For those skilled in the art, it is understandable that some well-known structures and their descriptions in the accompanying drawings may be omitted.

[0064] The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0065] Embodiment 1

[0066] This embodiment provides a steel plate stacking and sorting system based on a hybrid optimization algorithm, including:

[0067] An initial stack position for storing steel plates to be sorted;

[0068] A final stack position for storing steel plates arranged in the target order; the target order includes: from the bottom layer to the top layer of the final stack position, the steel plate numbers are arranged from small to large or from large to small;

[0069] A number of available temporary stack positions for temporarily storing steel plates during the steel plate stacking and sorting process;

[0070] A central control module for generating an initial stacking strategy through heuristic rules, and taking minimizing the operation duration and the number of temporary stack positions used for stacking and sorting as the goal, and iteratively optimizing the initial stacking strategy by combining the genetic algorithm and deep reinforcement learning to generate an optimal stacking strategy;

[0071] A stacking device for executing the stacking steps in the optimal stacking strategy to realize the movement operation of the steel plates between the stack positions until the stacking and sorting are completed;

[0072] In the central control module, the heuristic rules at least include:

[0073] Docking difference rule: According to the difference between the number of the currently removed steel plate and the number of the topmost steel plate in each temporary stack position, select a temporary stack position with a larger or smaller difference for storage;

[0074] Dynamic center bisection rule: Real-time obtain the number of steel plates at all stack positions except the final stack position, and evenly distribute the steel plates at all stack positions except the final stack position to each temporary stack position according to the number of available temporary stack positions; after a steel plate is moved into the final stack position, remove the steel plate during even distribution and update the temporary stack position distribution rule for the remaining steel plates.

[0075] In the central control module, the inputs of the genetic algorithm include:

[0076] The set of steel plates to be sorted A = {a1, a2,..., a N} and its initial permutation sequence, where a i represents the steel plate numbered i, and N is the total number of steel plates; in the initial permutation sequence, the steel plate number on the far left represents the topmost steel plate in the initial stack position, and the steel plate number on the far right represents the bottommost steel plate in the initial stack position.

[0077] The target sorting direction;

[0078] The number of available temporary stack positions L, satisfying: 2 ≤ L ≤ N;

[0079] Inversion equipment information, including the comprehensive scheduling efficiency η of the inversion equipment and the time duration Y required for the inversion equipment to execute one inversion step;

[0080] The preset operation time limit T within a day, and the maximum number of inversions

[0081] The outputs of the genetic algorithm include:

[0082] The inversion steps represented by the two-dimensional array [a, b], indicating moving the topmost steel plate at stack position a to stack position b, and used to indicate the movement path of the inversion equipment;

[0083] The minimum number of temporary stack positions L required to complete the steel plate inversion and sorting within the operation time limit T min , expressed as:

[0084]

[0085] where n actual is the actual number of inversions to complete the sorting in the case of the minimum number of temporary stack positions L min ;

[0086] In the central control module, the initialization process of the genetic algorithm includes:

[0087] Initialize the number of temporary stack positions to 2, randomly generate an initial inversion strategy through heuristic rules, and use the initial inversion strategy as the initial solution;

[0088] In the central control module, the genetic algorithm is used to iteratively optimize the initial stacking strategy to generate the optimal stacking strategy, including the following steps:

[0089] Obtain the number of stacking operations of the initial stacking strategy, and evaluate the corresponding operation duration through the following function:

[0090]

[0091] where, T total is the operation duration; n a ′ ctual is the actual number of stacking operations;

[0092] Judge whether the operation duration T total is greater than the operation time limit T within one day. If it is greater, increase the number of temporary stacking positions by 1, use selection, crossover, and mutation operations to generate a new stacking strategy, and re-evaluate the corresponding operation duration; if it is less than or equal to, obtain the optimal stacking strategy, and output the stacking steps corresponding to the optimal stacking strategy, as well as the minimum number of temporary stacking positions L min ;

[0093] The stacking equipment is specifically an unmanned overhead crane equipment.

[0094] In the specific implementation process, there are the following difficult-to-determine problems in the current shipyard's steel plate sorting: within the given operation time, it is necessary to determine the minimum number of temporary stacking positions for steel plates required to complete the sorting task; and given the number of temporary stacking positions and operation equipment, how long it takes to complete the sorting operation;

[0095] This embodiment is committed to solving the above problems. There is an existing unmanned overhead crane, one initial stacking position and one final stacking position, and several temporary stacking positions. A group of initially disordered steel plates are placed on the initial stacking position. It is necessary to make the initially disordered steel plates on the initial stacking position be arranged in ascending (or descending) order on the final stacking position through the stacking movement of an unmanned overhead crane for subsequent processing; the ultimate goal of this embodiment is to calculate the optimal stacking steps for different numbers of temporary stacking positions and initially disordered steel plate situations within a limited time to guide the unmanned overhead crane to complete this stacking process with fewer stacking operations;

[0096] In the central control module, first establish the mathematical model of this problem, and represent the set of steel plates to be sorted as A = {a1, a2,..., a N} and its initial arrangement sequence, where, a iIt represents the steel plate numbered i, and N is the total number of steel plates; in the initial arrangement sequence, the steel plate number on the leftmost side represents the topmost steel plate in the initial stack, and the steel plate number on the rightmost side represents the bottommost steel plate in the initial stack; the number of available temporary stacks is denoted as L, satisfying: 2 ≤ L ≤ N;

[0097] The target sorting direction includes: from the bottommost layer to the topmost layer of the final stack, the steel plate numbers are arranged in ascending or descending order, represented by a binary number, 0 means smaller on top and larger on the bottom, and 1 means larger on top and smaller on the bottom;

[0098] The comprehensive scheduling efficiency of the stack - reversing equipment is denoted as η (70% in this embodiment), and the time required for the stack - reversing equipment to execute one stack - reversing step is denoted as Y;

[0099] Assume that each slab - reversing operation step involves moving a steel plate from the current stack to the final stack or a temporary stack. For one operation, the required time is:

[0100]

[0101] Denote the operation time limit within a day as T, then the maximum number of stack - reversing times is That is, it is necessary to calculate within the number of stack - reversing times, the minimum number of temporary stacks L required to complete the sorting min , L min is expressed as:

[0102]

[0103] where n actual is the actual number of stack - reversing times to complete the sorting in the case of the minimum number of temporary stacks L min ;

[0104] After obtaining the minimum number of temporary stacks L min , the total time (operation time) used for its sorting can be calculated:

[0105]

[0106] where T total is the operation time in the case of the minimum number of temporary stacks L min ;

[0107] The goal of this embodiment is to minimize the operation time (number of stack - reversing times) and the number of temporary stacks. Priority is given to minimizing the number of stack - reversing times, that is, priority is given to minimizing the operation time. On this basis, the number of temporary stacks used is minimized to optimize resource allocation and improve operation efficiency;

[0108] For the above problem of steel plate stacking and sorting, the step of moving the steel plates from the initial stack to the final stack is essential. This step can complete the normal stacking cycle with the goal of the current state as the guidance. However, due to different sortings of the initially disordered steel plates, there will be partial temporary stacking of some steel plates. Therefore, the most core problem of stacking is how to reduce the number of times of this partial temporary stacking, so as to reduce the overall stacking times. And the calculation of the number of times of temporary stacking is attributed to which temporary stack the part of steel plates should be placed on. Therefore, this embodiment sets heuristic rules to limit the temporary stacks that the steel plates are allowed to enter, so that the specific stacking steps can be directly decoded according to the rules.

[0109] Taking six steel plates and two temporary stacks as an example, there are three fixed rules, which are respectively:

[0110] 1) Central bisecting: [1, 2, 3], [4, 5, 6];

[0111] 2) Center of gravity biased to the rear: [1, 2], [3, 4, 5, 6];

[0112] 3) Center of gravity biased to the front: [1, 2, 3, 4], [5, 6];

[0113] As the steel plate stacking process progresses, the number of remaining steel plates will become less and less. If the fixed rules of sequential allocation continue to be followed, it will cause the rear temporary stacks to be idle, resulting in waste. Therefore, this embodiment proposes three new dynamic rules, which are respectively:

[0114] 1) Larger docking difference: According to the difference between the number of the currently removed steel plate and the number of the topmost steel plate in each temporary stack, select the temporary stack with a larger difference for storage;

[0115] 2) Smaller docking difference: According to the difference between the number of the currently removed steel plate and the number of the topmost steel plate in each temporary stack, select the temporary stack with a smaller difference for storage;

[0116] 3) Dynamic central bisecting rule: Real-time obtain the number of steel plates in all stacks except the final stack, and according to the number of available temporary stacks, evenly distribute the steel plates in all stacks except the final stack to each temporary stack; after a steel plate is moved into the final stack, remove the steel plate during the even distribution and update the temporary stack allocation rule for the remaining steel plates;

[0117] Among them, the docking difference rule makes an independent judgment on each removed steel plate, and its parking position is determined according to the number of the topmost steel plate in the temporary stack, that is, subtract the number of the topmost steel plate in each temporary stack from the number of the removed steel plate, and find the stack with a larger or smaller difference for placement; the dynamic central bisecting rule follows the logic of the fixed rule 1), but after each target steel plate enters the final stack, it is removed and the fixed rule is updated to form a global dynamic rule;

[0118] After analysis, it is found that the effects of the above-mentioned fixed and dynamic rules fluctuate greatly with different initially disordered steel plates and do not dominate each other. Therefore, it can be concluded that there is a strong coupling relationship between the temporary stacking rules and the sorting of the initially disordered steel plates, and appropriate temporary stacking rules need to be given according to the initially disordered steel plates.

[0119] Therefore, in this embodiment, the genetic algorithm (GA) with heuristic initialization is used as the intelligent optimization algorithm to solve this problem. GA does not directly solve the problem, but finds the optimal stacking rule for the initially disordered steel plates, and decodes the result according to the rule to obtain the result of stack rearrangement, so as to achieve the goal of optimizing the stack rearrangement steps.

[0120] In the initialization process, the number of temporary stacking positions is initialized to 2 (starting from the minimum value and increasing sequentially), and the initial stack rearrangement strategy is randomly generated through heuristic rules (i.e., the above-mentioned dynamic rules).

[0121] The inputs of the genetic algorithm are:

[0122] 1) Randomly initially disordered steel plates, as Figure 1 shown (where L1 to L4 represent 4 available temporary stacking positions). Taking 20 pieces as an example, the list is in the form of: [10, 4, 14, 3, 12, 17, 6, 15, 9, 8, 7, 18, 16, 19, 1, 13, 5, 11, 20, 2], where the 10th number on the left side of the array is the top layer of the stacking position, the 2nd number on the right side is the bottom layer, the data in the array starts from 1 and is continuous, and 2 ≤ the length of the billet list ≤ 130.

[0123] 2) The number of available temporary stacking positions. Taking 4 temporary stacking positions as an example, both the initial stacking position and the final stacking position are 1, and the total number of stacking positions is 6; 2 ≤ the number of temporary stacking positions ≤ 38, and there are at most 40 stacking positions in total.

[0124] 3) The final target sorting direction, which is divided into two types: small on top and large on the bottom, and large on top and small on the bottom.

[0125] The outputs of the genetic algorithm are:

[0126] The minimum number of temporary stacking positions L min, and a two-dimensional array of stacking steps, in the form of [[1,3],[1,2],[1,3],[1,5],[1,5],[1,3],[1,2],[1,5],[1,2],[1,3],[1,2],[1,3],[1,5],[1,3],[1,4],[1,2],[1,4],[1,5],[1,6],[3,6],[3,6],[3,4],[3,6],[5,4],[5,6],[5,6],[3,6],[2,6],[5,6],[4,6],[3,6],[2,3],[2,6],[4,6],[3,6],[2,6],[4,6],[2,6],[5,6],[1,6],[4,6]], a total of 41 steps. The overhead crane movement route diagram during the stacking process is as Figure 2 shown (in the figure, the hollow circles are the removed stacking positions, the solid circles are the inserted stacking positions, the solid lines are the overhead crane carrying processes, and the dashed lines are the overhead crane empty-load processes; when there is no "horizontal dashed line" in the whole figure, it can be explained that the calculation result has no direct invalid slab turning process);

[0127] Then, algorithm iteration optimization is carried out. Starting from L = 2, stacking strategies are continuously generated, and at the same time, the corresponding operation duration is evaluated through the following function:

[0128]

[0129] where, T total is the operation duration; n a ′ ctual is the actual number of stacking times;

[0130] Judge whether the operation duration T total is greater than the operation time limit T within one day. If it is greater, increase the number of temporary stacking positions by 1, generate a new stacking strategy by using selection, crossover, and mutation operations, and re-evaluate the corresponding operation duration; if it is less than or equal to, obtain the optimal stacking strategy, and output the stacking steps corresponding to the optimal stacking strategy, as well as the minimum number of temporary stacking positions L min ;

[0131] In this embodiment, for 720 initial disordered steel plates in a small case (6 steel plates, 2 temporary stacking positions), the above eight schemes are selected to solve the optimal number of slab turnings. The results are as Figure 3 shown; it can be seen that among the eight methods, the exact solution mathematical model can obtain the global optimal solution, so its optimal proportion is 100%; while the optimal proportions of the six heuristic methods do not exceed 80%, indicating that the effect is average; and the result obtained by the heuristic algorithm + GA is exactly the same as the result obtained by the exact solution, indicating that the optimization effect of GA is excellent, the performance is stable, and it can find the global optimal solution in small-scale cases;

[0132] For large-scale cases (20 steel plates, 2 temporary stacking positions), 1000 kinds of initially disordered steel plates were randomly selected for experiments. Due to the limitations of precise calculation, it is impossible to completely solve medium and large-scale cases. And under the same conditions of GA, the mathematical model cannot complete effective calculation within 60s. Therefore, the results do not include the solution results of the mathematical model. The data line chart is as shown in Figure 4 ; It can be seen that compared with the optimal proportion of no more than 5% of the first six heuristic rules, the optimal proportion of GA reaches more than 99%, and the effect is more obvious, indicating that GA specifically designed according to the problem can solve this problem;

[0133] In the largest-scale case (130 steel plates, 38 temporary stacking positions), the calculation time of GA was extended to 180s, and the result convergence chart drawn is as shown in Figure 5 ; The data shows that the initial average fitness value is 532, and the optimal fitness value is 365. The process is as follows:

[0134] 1) In the first minute, both are overall optimized to 325, and the optimization amplitudes are 98.57% and 93% respectively;

[0135] 2) In the second minute, both are overall optimized to 324, and the optimization amplitudes are 0.48% and 2.3% respectively;

[0136] 3) In the third minute, both are overall optimized to 322, and the optimization amplitudes are 0.95% and 4.7% respectively;

[0137] It can be seen that GA can converge to a better value faster in the early stage of calculation, and the optimization amplitude is more than 90%. There will also be slight optimization in the middle and late stages of calculation, but the amplitude is very low on the same time scale; Therefore, this algorithm can calculate better stacking steps under the requirement of a short time and is suitable for large-scale stacking sorting tasks;

[0138] This embodiment also studied the marginal utility of the algorithm. Through the GA method, the number of temporary stacking positions increased from 3 to 8, and 1000 large cases of 20 steel plates were randomly generated for calculation, and seven groups of average data were obtained. The line chart of the optimization percentage relative to the mean of the previous group of experiments is as shown in Figure 6 ; It can be seen that when the number of temporary stacking positions increases from 2 to 3, the average number of plate inversion calculated for the same sample set decreases by 19.72%, and the decrease amplitude is obvious; However, when the number of temporary stacking positions increases from 3 to 4 (the elbow), the decrease amplitude drops sharply to 5.87%, and the subsequent increase in the number of temporary stacking positions brings little benefit; That is, when the number of steel plates remains unchanged, increasing the number of temporary stacking positions has a marginal effect. Therefore, effective suggestions can also be put forward for the setting of the number of temporary stacking positions based on this research;

[0139] In this embodiment, a stable feasible solution is first obtained through heuristic rules as the initial stacking strategy, and then a genetic algorithm is constructed to optimize the rules, so as to approach the optimal solution. It can fully coordinate the ternary relationship among time constraints, equipment resources and temporary stacking positions, reduce the equipment idle rate, optimize the production process of the shipyard, improve production efficiency, and reduce production costs at the same time.

[0140] Embodiment 2

[0141] This embodiment provides a steel plate stacking and sorting system based on a hybrid optimization algorithm, including:

[0142] An initial stacking position for storing steel plates to be sorted;

[0143] A final stacking position for storing steel plates arranged in the target order; the target order includes: from the bottom layer to the top layer of the final stacking position, the steel plate numbers are arranged from small to large or from large to small;

[0144] Several available temporary stacking positions for temporarily storing steel plates during the steel plate stacking and sorting process;

[0145] A central control module for generating an initial stacking strategy through heuristic rules, and taking minimizing the operation duration and the number of temporary stacking positions used for stacking and sorting as the goal, and iteratively optimizing the initial stacking strategy by combining a genetic algorithm and deep reinforcement learning to generate an optimal stacking strategy;

[0146] A stacking device for executing the stacking steps in the optimal stacking strategy to realize the movement operation of the steel plates between the stacking positions until the stacking and sorting are completed;

[0147] In the central control module, the heuristic rules at least include:

[0148] Docking difference rule: According to the difference between the number of the currently removed steel plate and the number of the topmost steel plate in each temporary stacking position, select a temporary stacking position with a larger or smaller difference for storage;

[0149] Dynamic center bisecting rule: Real-time obtain the number of steel plates in all stacking positions except the final stacking position, and evenly distribute the steel plates in all stacking positions except the final stacking position to each temporary stacking position according to the number of available temporary stacking positions; after a steel plate is moved into the final stacking position, remove the steel plate during even distribution and update the temporary stacking position distribution rule of the remaining steel plates;

[0150] In the central control module, the input of the genetic algorithm includes:

[0151] The set of steel plates to be sorted A = {a1, a2,..., a N} and its initial arrangement sequence, where a iIt represents a steel plate numbered i, and N is the total number of steel plates; in the initial arrangement sequence, the steel plate number on the leftmost side represents the topmost steel plate in the initial stack, and the steel plate number on the rightmost side represents the bottommost steel plate in the initial stack;

[0152] The target sorting direction;

[0153] The available number of temporary stacks L, satisfying: 2 ≤ L ≤ N;

[0154] The information of the stack-changing equipment, including the comprehensive scheduling efficiency η of the stack-changing equipment and the duration Y required for the stack-changing equipment to execute one stack-changing step;

[0155] The preset operation time limit T within a day, and the maximum number of stack changes

[0156] The output of the genetic algorithm includes:

[0157] The stack-changing steps represented by the two-dimensional array [a, b], indicating moving the topmost steel plate of stack a to stack b, which is used to indicate the moving path of the stack-changing equipment;

[0158] The minimum number of temporary stacks L required to complete the stack-changing and sorting of steel plates within the operation time limit T min , expressed as:

[0159]

[0160] where n actual is the actual number of stack changes to complete the sorting in the case of the minimum number of temporary stacks L min ;

[0161] In the central control module, the initialization process of the genetic algorithm includes:

[0162] Initializing the number of temporary stacks to 2, randomly generating an initial stack-changing strategy through heuristic rules, inputting the initial solution into a pre-trained deep reinforcement learning network to optimize the initial solution, and using the optimized initial stack-changing strategy as the initial solution to improve the population quality;

[0163] The pre-trained deep reinforcement learning network is used to screen the effective action set in the initial solution; the effective action set includes several effective stack-changing steps, and the effective stack-changing steps are oriented to move the steel plate with the target number to the final stack as soon as possible;

[0164] In the central control module, using the genetic algorithm to iteratively optimize the initial stack-changing strategy to generate the optimal stack-changing strategy, including the following steps:

[0165] Obtaining the number of stack changes of the initial stack-changing strategy, and evaluating the corresponding operation duration through the following function:

[0166]

[0167] Among them, T total is the operation duration; n a ′ ctual is the actual number of stack shifting operations;

[0168] Judge whether the operation duration T total is greater than the operation time limit T within a day. If it is greater, increase the number of temporary stack positions by 1, use selection, crossover, and mutation operations to generate a new stack shifting strategy, and re-evaluate the corresponding operation duration; if it is less than or equal to, obtain the optimal stack shifting strategy, and output the stack shifting steps corresponding to the optimal stack shifting strategy, as well as the minimum number of temporary stack positions L min ;

[0169] The stack shifting equipment is specifically an unmanned overhead crane equipment.

[0170] In the specific implementation process, the overall process of this embodiment is the same as that of Embodiment 1. The only difference is that deep reinforcement learning is introduced in the initialization process of the genetic algorithm. This embodiment focuses on explaining the deep reinforcement learning part;

[0171] Since a large number of randomly generated individuals are included in the GA initialization, the fitness values of such individuals fluctuate greatly, have poor stability, and are generally low, which will affect the optimization process of the GA; therefore, a method is needed to stably generate better initial individuals to replace some of the randomly generated individuals; here, the classic method DQN in deep reinforcement learning is selected to handle this problem. At the same time, in order to accelerate the learning and convergence of DQN, the algorithm has a unique design for actions:

[0172] The number of complete actions is fixed, which is L*2 + L + 1, that is, three types of situations: removing the initial stack position, moving the temporary stack positions relative to each other, and moving into the final stack position; specifically, for different states, an effective action function corresponding to the state is also designed. The hierarchical Venn diagram of the actions is as Figure 7 shown, where:

[0173] 1) The full action set is the set of all possible actions obtained according to the current number of temporary stack positions, and the number is L*2 + L + 1;

[0174] 2) The actionable action set is the set of all actions that can be actually moved obtained according to the current state, and the number is uncertain but is a subset of the full action set;

[0175] 3) The effective action set is the set of actions oriented to the target obtained according to the current state target value, and its purpose is to move the current target to the final stack position as soon as possible. The number is uncertain but is a subset of the actionable action set;

[0176] For a more intuitive illustration, this embodiment provides an example. As Figure 8 shown in the schematic diagram of a certain palletizing state, for Figure 8 the state in [reference number], if the target steel plate is No. 4, then:

[0177] The set of all actions is: [[1, 2], [1, 3], [1, 4], [2, 3], [2, 4], [3, 2], [3, 4]];

[0178] The set of actionable actions is: [[1, 2], [1, 3], [1, 4], [2, 3], [2, 4]];

[0179] The set of effective actions is: [[1, 2], [1, 3]];

[0180] This embodiment uses a self - learning strategy to train the DQN network. During one round of training, only one palletizing step is generated each time until the palletizing is completed, and the number of palletizing steps for this time is recorded; each case is repeated several (such as 100) times, and its palletizing effect is compared to correct the parameters of the DQN network until the model converges to complete the pre - training; since it is self - learning, at each step of the algorithm evolution, the state update needs to be strictly carried out according to constraints or objectives, etc.; when designing the state update mechanism, both whether all are completed and each step forward should be considered.

[0181] After pre - training, a complete DQN network is obtained. By setting epsilon = 1.0, three different - sized repeated experiments of 20 + 2, 50 + 5, and 130 + 13 are carried out to obtain the solution results of the DQN, calculate the mean and variance of its results, and the comprehensive results are as Figure 9 shown;

[0182] From Figure 9 it can be seen that for both the mean and variance of the results, the pure random process is closer to the GA result (the asterisk is the result of the final optimization of GA) than the results of the DQN network in its final version. The average optimization efficiency of DQN in terms of the mean reaches 30.05%, 14.54%, and 9.32% respectively, and the average optimization efficiency in terms of the variance reaches 90.41%, 87.53%, and 83.59% respectively; therefore, it can be proved that using it in the initialization process of GA can obtain a better initial solution, thereby accelerating the convergence efficiency of GA and obtaining a better result.

[0183] This embodiment uses the deep reinforcement learning algorithm to optimize the initial solution randomly generated by the heuristic rule, thereby obtaining a higher - quality initial solution and effectively improving the convergence speed of the genetic algorithm.

[0184] In this embodiment, by integrating heuristic rules, GA, and DQN algorithms, it not only retains the fast decision-making ability of rule guidance but also achieves global optimization through intelligent algorithms, breaking through the limitations of traditional single methods, with stronger algorithm robustness and prominent industrial application value.

[0185] Embodiment 3

[0186] As Figure 10 shown, this embodiment provides a method for steel plate stacking sequence based on a hybrid optimization algorithm. Based on the steel plate stacking sequence system based on the hybrid optimization algorithm described in Embodiment 1 or 2, it includes the following steps:

[0187] S1: Obtain the set of steel plates to be sequenced, the number of available temporary stacking positions, and the target order;

[0188] S2: Generate an initial stacking strategy through heuristic rules;

[0189] S3: With the goal of minimizing the operation duration and the number of temporary stacking positions used for stacking sequence, use the genetic algorithm to iteratively optimize the initial stacking strategy to generate the optimal stacking strategy;

[0190] S4: Control the stacking equipment to execute the stacking steps in the optimal stacking strategy to achieve the movement operation of the steel plates between the stacking positions until the stacking sequence is completed.

[0191] In the specific implementation process, first obtain the set of steel plates to be sequenced, the number of available temporary stacking positions, and the target order; then generate an initial stacking strategy through heuristic rules; then, with the goal of minimizing the operation duration and the number of temporary stacking positions used for stacking sequence, use the genetic algorithm to iteratively optimize the initial stacking strategy to generate the optimal stacking strategy; finally, control the stacking equipment to execute the stacking steps in the optimal stacking strategy to achieve the movement operation of the steel plates between the stacking positions until the stacking sequence is completed;

[0192] This method can fully coordinate the ternary relationship among time constraints, equipment resources, and temporary stacking positions, reduce the equipment idling rate, thereby optimizing the production process of the shipyard, improving production efficiency, and reducing production costs at the same time.

[0193] Embodiment 4

[0194] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps in the method described in Embodiment 3.

[0195] In the specific implementation process, the computer program in this embodiment can be embodied in the form of a software product. This computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method in Embodiment 3. The storage media in this embodiment include, but are not limited to: USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs and other media that can store program codes;

[0196] There are a total of 4 computer programs in this embodiment, namely: DQN initialization, GA final version, web_GA, and effect verification. Among them, only web_GA and effect verification are involved in the usage process;

[0197] For web_GA:

[0198] This algorithm integrates DQN initialization and GA final version, inputs specific parameters and calculates through webservice calls, so as to obtain a better sheet turning process for the sheet turning problem and output it;

[0199] The input parameter name of the algorithm is: data; the parameter form is: [[5,4,6,3,1,2],2,0], where the first element [5,4,6,3,1,2] is the initial disordered steel plate (sorted from bottom left to top right), the second element 2 is the number of temporary stacking positions, and the third element 0 is the final target sorting method (0 means smaller on top and larger on the bottom, 1 means larger on top and smaller on the bottom). The three are integrated into the list data in sequence and sent together for calculation;

[0200] The output of the algorithm is: [9,[[1,3],[1,2],[1,4],[3,4],[2,4],[1,4],[1,2],[1,4],[2,4]]], where the first element 9 in the list is the optimal number of sheet turning steps, and the second element [[1,3],[1,2],[1,4],[3,4],[2,4],[1,4],[1,2],[1,4],[2,4]] is the optimal sheet turning steps;

[0201] Taking Postman as an example for display:

[0202] 1) First run the algorithm: web_GA to enable the webservice function. At this time, relevant information will be displayed on the console interface of Pycharm;

[0203] 2) Call the corresponding web page in Postman, then input data = [[5,4,6,3,1,2],2,0], and wait for the running result. The final result is as followsFigure 11 as shown

[0204] such as Figure 12 As shown, when the current result is returned, local logs will be generated, which will record in detail the relevant important information of this operation, such as IP address, time, input, output, etc.; extracting the unstacking steps of the output can guide the unmanned overhead crane to perform unstacking sorting on the current state;

[0205] For effect verification:

[0206] The algorithm inputs are:

[0207] 1) list_ori0, the arrangement of the initial disordered steel plates, with the direction from upper left to lower right;

[0208] 2) L, the number of temporary stacking positions;

[0209] 3) best_action, the best plate turning action calculated in the current state;

[0210] After the input operation, a PDF file named "Result Verification File" will be generated in the same folder of this algorithm, which contains schematic diagrams of all plate turning steps drawn from left to right and top to bottom for manual verification;

[0211] Since the result verification belongs to a spot check process (not often used), this algorithm is independent of the web_GA main program and is called separately when the result needs to be checked to avoid occupying the resources of the main program.

[0212] The same or similar reference numerals correspond to the same or similar components;

[0213] The terms describing the positional relationship in the drawings are only for illustrative purposes and should not be construed as a limitation of this application;

[0214] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, rather than limitations on the implementation manners of the present invention. For those of ordinary skill 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 implementation manners here. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the claims of the present invention.

Claims

1. A steel plate stacking and sorting system based on a hybrid optimization algorithm, characterized in that, Comprising: An initial stack position for storing steel plates to be sorted; A final stack position for storing steel plates arranged in the target order; Several available temporary stack positions for temporarily storing steel plates during the process of re-stacking and sorting the steel plates; A central control module for generating an initial re-stacking strategy through heuristic rules, and aiming at minimizing the operation duration and the number of temporary stack positions used for re-stacking and sorting, iteratively optimizing the initial re-stacking strategy by using a genetic algorithm to generate an optimal re-stacking strategy; A re-stacking device for executing the re-stacking steps in the optimal re-stacking strategy to realize the movement operation of the steel plates between the stack positions until the re-stacking and sorting are completed.

2. The steel plate stacking and sorting system based on the hybrid optimization algorithm according to claim 1, wherein The target order includes: from the bottom layer to the top layer of the final stack position, the steel plate numbers are arranged from small to large or from large to small.

3. The steel plate stacking and sorting system based on the hybrid optimization algorithm according to claim 1, characterized in that In the central control module, the heuristic rules at least include: Docking difference rule: According to the difference between the number of the currently removed steel plate and the number of the topmost steel plate in each temporary stack position, select a temporary stack position with a larger or smaller difference for storage; Dynamic center bisecting rule: Real-time obtain the number of steel plates in all stack positions except the final stack position, and according to the number of available temporary stack positions, evenly distribute the steel plates in all stack positions except the final stack position to each temporary stack position; after a steel plate is moved into the final stack position, remove the steel plate during the even distribution and update the temporary stack position distribution rule of the remaining steel plates.

4. A steel plate stacking and sorting system based on a hybrid optimization algorithm according to claim 1, characterized in that, In the central control module, the inputs of the genetic algorithm include: The set of steel plates to be sorted A = {a1, a2,..., a N}, and its initial arrangement sequence, where a i represents the steel plate numbered i, and N is the total number of steel plates; in the initial arrangement sequence, the steel plate number on the leftmost side represents the topmost steel plate in the initial stack, and the steel plate number on the rightmost side represents the bottommost steel plate in the initial stack; The target sorting direction; The number L of available temporary stack positions, satisfying: 2 ≤ L ≤ N; Re-stacking device information, including the comprehensive scheduling efficiency η of the re-stacking device and the duration Y required for the re-stacking device to execute one re-stacking step; The preset operation time limit T within a day and the maximum number of stack rearrangements The outputs of the genetic algorithm include: The re-stacking steps represented by the two-dimensional array [a, b], indicating moving the topmost steel plate in the a-th stack position to the b-th stack position, for indicating the movement path of the re-stacking device; The minimum number of temporary stacks L required to complete the sorting of steel plate stacking within the operation time limit T min , expressed as: where n actual is the actual number of reshuffles to complete sorting under the condition of the minimum number of temporary stacks L min .

5. The steel plate stacking and sorting system based on the hybrid optimization algorithm according to claim 4, characterized in that, In the central control module, the initialization process of the genetic algorithm includes: Initialize the number of temporary stack positions to 2, randomly generate an initial re-stacking strategy through heuristic rules, and use the initial re-stacking strategy as the initial solution.

6. The steel plate stacking and sorting system based on a hybrid optimization algorithm according to claim 5, wherein In the central control module, using the genetic algorithm to iteratively optimize the initial re-stacking strategy to generate an optimal re-stacking strategy includes the following steps: Obtain the number of re-stacking times of the initial re-stacking strategy, and evaluate the corresponding operation duration through the following function: Among them, T total is the operation duration; n a ′ ctual is the actual number of stack rearrangements; Judge the operation duration T total Whether it is greater than the operation time limit T within one day. If it is greater, increase the number of temporary stacking positions by 1, generate a new stacking strategy using selection, crossover, and mutation operations, and re-evaluate the corresponding operation duration; if it is less than or equal, obtain the optimal stacking strategy, and output the stacking steps corresponding to the optimal stacking strategy, as well as the minimum number of temporary stacking positions L min .

7. A steel plate stacking and sorting system based on a hybrid optimization algorithm according to claim 6, characterized in that, In the central control module, after generating the initial re-stacking strategy, it further includes: Inputting the initial solution into a pre-trained deep reinforcement learning network to optimize the initial solution to improve the population quality; The pre-trained deep reinforcement learning network is used to screen the effective action set in the initial solution; the effective action set includes several effective re-stacking steps, and the effective re-stacking steps are oriented to moving the steel plate with the target number to the final stack position as soon as possible.

8. A steel plate stacking and sorting system based on a hybrid optimization algorithm according to any one of claims 1 to 7, characterized in that, The re-stacking device is specifically an unmanned overhead crane device.

9. A method for stacking and sorting steel plates based on a hybrid optimization algorithm, based on the steel plate stacking and sorting system based on the hybrid optimization algorithm described in any one of claims 1 to 8, characterized in that, Including the following steps: S1: Obtain the set of steel plates to be sorted, the number of available temporary stack positions, and the target order; S2: Generate an initial re-stacking strategy through heuristic rules; S3: Aiming at minimizing the operation duration and the number of temporary stack positions used for re-stacking and sorting, iteratively optimize the initial re-stacking strategy by using a genetic algorithm to generate an optimal re-stacking strategy; S4: Control the stack rearrangement device to execute the stack rearrangement steps in the optimal stack rearrangement strategy, so as to realize the movement operation of the steel plates between the stack positions until the stack rearrangement sorting is completed.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps in the method described in claim 9.

Images