A method for optimizing and reorganizing a power battery production line
By using particle swarm algorithms, multiple symmetry learning and improved neighborhood change methods in the optimization and reorganization of power battery production lines, the problem of local optimal and premature convergence in the optimization and reorganization of power battery production lines is solved, and machine failures are dealt with in dynamic production lines are dealt with, achieving efficient power battery production and production lines robustness.
Patent Information
- Application Number
- CN202210084356.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-01-19
- Publication Date
- 2025-06-03
- Estimated Expiration
- 2042-01-19
AI Technical Summary
When the existing technology solves the problem of power battery production line optimization and reorganization, the particle swarm algorithm is prone to fall into local optimization and premature convergence, and most of the research focuses on static production line optimization, and cannot effectively deal with unpredictable events in dynamic production line optimization.
By obtaining the static mathematical model of the power battery workshop, global search is performed using the particle swarm method, and multiple symmetric learning methods and improved variable neighborhood methods are used to avoid local optimality and premature convergence. At the same time, in dynamic production line optimization, local production line optimization and reorganization are carried out according to machine failure constraints.
It realizes rapid static production line optimization and reorganization in the case of failure, and obtains good dynamic production line optimization and reorganization results in the case of machine failure, improving the production efficiency of power batteries and the robustness of production lines.
Smart Images
Figure CN114529059B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method for optimizing and reorganizing a power battery production line, and belongs to the field of production line reorganization. Background Art
[0002] Power batteries are the core components of electric vehicles. With the increasing demand for power batteries year by year, the production efficiency of power batteries is particularly important. How to propose a reasonable production line optimization and reorganization strategy for power battery production workshops is a hot issue in the current research field of high-quality power battery production. The production process of power batteries usually includes front-end, middle-end and back-end processes. The back-end processes are concentrated in the formation, static and other processes, and do not require production line optimization and reorganization. Therefore, the production line optimization and reorganization of power battery workshops focus on the middle and front-end processes. At present, the middle and front-end processes of soft-pack battery production mainly include 6 processes such as batching, coating, film making, baking, stacking, and assembly. The parallel processing equipment corresponding to each process includes 6 groups of equipment such as batching machine, coating machine, film making machine, customized oven machine, automatic stacking machine and automatic assembly machine. So for How to arrange soft pack battery production orders The optimal reorganization problem of the battery production line is to determine the processing order of each order in each process and the selection of corresponding parallel machines to minimize the maximum completion time of all orders (that is, to improve the production efficiency of power batteries).
[0003] The main methods for solving the optimization and reorganization problem of power battery production lines include exact algorithms, heuristic algorithms, and swarm intelligence optimization algorithms. Among them, although the exact algorithm can obtain the theoretical optimal solution, its calculation time also increases significantly as the scale of the problem increases, so it is more suitable for solving small-scale problems; the advantage of the heuristic algorithm is that it has a fast solution speed, but its optimization results are often greatly affected by the heuristic rules. In comparison, the swarm intelligence optimization algorithm is less affected by the characteristics of the problem and has better solution results.
[0004] The particle swarm algorithm is a swarm intelligence optimization algorithm that imitates the process of birds looking for food in nature. The particle swarm algorithm has attracted widespread attention from scholars due to its advantages such as simple programming, intuitive and easy implementation. Although current research has achieved certain results, the particle swarm algorithm is very easy to fall into local optimality and premature convergence during the optimization process, which will result in the completion time of the solution given when solving the optimization and reorganization problem of power battery production lines is not the minimum solution, so this problem needs to be further optimized.
[0005] Meanwhile, most of the research focuses on the optimization and reorganization of the static production line in the power battery workshop. However, in actual industrial production, unpredictable events may occur at any time, such as emergency order insertion, raw material shortage, or machine failure, which affect normal production. At this time, it is necessary to consider the dynamic production line optimization and reorganization problem when the above situations occur, so as to minimize the impact of these events on production. Based on this, it is necessary to conduct in-depth research on the optimization and reorganization of the production line in the power battery production workshop to obtain a reorganization and optimization plan that can better solve the above problems. Summary of the Invention
[0006] In order to obtain a better optimization and reorganization plan for the power battery production line, the present invention provides a method for optimizing and reorganizing the static production line of power batteries, and the method includes:
[0007] Step1: Obtain the static mathematical model for optimizing the production line in the power battery workshop;
[0008] Step2: According to the static mathematical model determined in Step1, determine the fitness function and decoding method of the particle swarm, and use the particle swarm method for global search. During the global search process, the multiple symmetric learning method is used to find the individual optimal particle to avoid the individual particle getting trapped in the local optimum during the optimization process;
[0009] Step3: According to the individual optimal particle obtained in Step2, use the improved variable neighborhood method to avoid the group optimal particle from falling into premature convergence and obtain the optimization and reorganization result of the static production line.
[0010] Optionally, the static mathematical model for optimizing the production line in the power battery workshop in Step1 is:
[0011] Determine the processes included in the power battery production process, and the parallel processing equipment corresponding to the th process includes machines;
[0012] There are n types of battery orders to be completed. For any one order, it is processed on any machine of this process when it reaches the th process. represents the processing completion time of order for the th process;
[0013] Let the maximum completion time among all orders be . Then the objective function of the static mathematical model for optimizing the production line in the power battery workshop is:
[0014] (1)
[0015] The optimization constraints for the static production line of power batteries include:
[0016] (2)
[0017] (3)
[0018] (4)
[0019] (5)
[0020] (6)
[0021] (7)
[0022] Among them, represents the machine number of the parallel machine in each process, ; represents the order is processed on the machine in the th processing process. When the order is processed on the machine in the th processing process , otherwise ; represents the completion time of the order in the th processing process; represents the start time of processing of the order in the th processing process; represents the processing time of the order in the th processing process; represents the number of orders processed on the machine in the th processing process.
[0023] Optionally, the fitness function of the particle swarm is determined according to the static mathematical model determined in Step1 in Step2 as:
[0024] (8)
[0025] The velocity update and position update rules of the particle are as follows:
[0026] (9)
[0027] (10)
[0028] (11)
[0029] Among them, represents the number of iterations, is the inertia constant, and its value range is 0.8 to 1.2; is the self-learning factor, is the swarm learning factor; is a random number uniformly distributed on the interval [0, 1]; is the particle at the th iteration of the th dimension of the position; is the particle at the th iteration of the th dimension of the velocity; is the particle at the th iteration of the th dimension of the individual optimal position; is the global optimal position of the swarm particles at the th iteration in the th dimension, is the minimum value of the particle search range, is the maximum value of the particle search range.
[0030] Optionally, the purpose of the particle swarm method for the optimization and reorganization of the static production line of power batteries is to determine the processing sequence of each order in each processing procedure, and after the processing sequence of the first procedure in the procedures is determined, the subsequent procedures can determine the corresponding processing sequence according to the principle of first come, first served. Therefore, the vector represented by the globally optimal particle obtained by particle swarm optimization is
[0031] the processing sequence of each order in the first processing procedure;
[0032] For orders, the dimension of the corresponding particle swarm is , and the smallest value in the vector represented by each particle corresponds to order number 1, the second smallest value corresponds to order number 2, , and the largest value corresponds to order number ;
[0033] The subsequent procedures arrange their processing sequences according to the order of arrival of the orders at this procedure;
[0034] If multiple orders arrive at this process simultaneously, the workpiece with the longest remaining unprocessed time is preferentially arranged until all the order sequences of this process are arranged, and finally the order production sequence of each process is obtained;
[0035] For the problem of parallel machine selection in any process, the machine that can complete the current order first in the current process is preferentially selected.
[0036] Optionally, in the global search process of the particle swarm, a multiple symmetry learning method is adopted to find the individual optimal particle, including:
[0037] For the particle in the -dimensional space where a random integer between 1 and n is randomly generated b and c is determined the symmetry plane formed by the axis and the p axis is placed at the origin, and the particle generated by the particle passing through the symmetry plane is defined as the symmetric particle then the coordinate components of the symmetric particle
[0038] (12)
[0039] If is satisfied, and then the particle replaces the particle otherwise, the particle is continued to be used for subsequent iteration;
[0040] For the particle with stagnant update, times of multiple symmetry learning are performed to obtain symmetric particles. For a certain particle, if the fitness of a certain symmetric particle is better than its own fitness and better than the fitness of other symmetric particles of this particle, then this symmetric particle is used to replace the original particle with stagnant update;
[0041] During the iteration process, the individual particle with the best fitness is defined as the global optimal particle of the population.
[0042] Optionally, in the improved variable neighborhood method, 3 new neighborhood structures are defined:
[0043] (1) Multiple symmetry learning : For a certain particle, multiple symmetry planes are randomly generated to obtain symmetric particles;
[0044] (2) Opposite learning : Obtain the reverse solution for a certain particle;
[0045] (3) Genetic variation reverse order : For a certain particle, randomly select two position points in the particle position vector, and reverse the segment between these two points to obtain a reverse-order particle.
[0046] Optionally, the method of using the improved variable neighborhood method to avoid the population optimal particle falling into premature convergence and obtaining the static production line optimization and recombination result includes:
[0047] Utilize times of jitter operations to determine whether the population optimal particle falls into premature convergence;
[0048] If the population optimal particle falls into premature convergence, then save the optimal particle after jitter for the next times of local search to obtain the final population optimal particle;
[0049] On the contrary, directly output this population optimal particle without local search;
[0050] According to the position vector of the population optimal particle, the final static production line optimization and recombination result can be obtained according to the decoding rule.
[0051] The present invention also provides a method for optimizing and recombining a dynamic production line of a power battery. After obtaining the static production line optimization and recombination result of the power battery by using the above method, the method includes:
[0052] Step4: In the case of parallel machine failures in the power battery production workshop, obtain the machine failure constraints and the affected production processes;
[0053] Step5: According to the failure constraints in Step4, on the basis of the static production line optimization and recombination result obtained by any of the above methods, perform local production line optimization and recombination on the affected production processes to obtain the result of dynamic production line optimization and recombination.
[0054] Optionally, for the case of parallel machine failures in the power battery production workshop, obtaining the machine failure constraints and the affected production processes includes:
[0055] Suppose the machine of process fails at time , and the repair time of the machine is ;
[0056] Define , then the machine is unavailable within the time of , that is, the failure constraint is:
[0057] (13)
[0058] At this time, the process tasks of different battery production orders may be in one of the following states: completed, unprocessed, being processed, obtaining unprocessed and being processed production processes affected by faulty machines production processes affected by faulty machines
[0059] Optionally, the Step5 includes:
[0060] Adding the fault constraint condition to the static production line optimization constraint condition of the battery, and on the basis of the obtained static production line optimization and reorganization result, adjusting the affected being processed and unprocessed processes according to the first-come-first-served rule and the machine selection rule to determine the dynamic production line optimization and reorganization result of the power battery.
[0061] The beneficial effects of the present invention are as follows:
[0062] By obtaining the static mathematical model for the optimization of the production line in the power battery workshop; according to the static mathematical model, determining the fitness function and decoding method of the particle swarm, and using the particle swarm method for global search; when performing global search using the particle swarm method, designing a multiple symmetric learning method to avoid the individual optimal particle from falling into the local optimum; according to the improved individual optimal particle, designing an improved variable neighborhood method to avoid the swarm optimal particle from falling into premature convergence, and obtaining the static production line optimization and reorganization result; further, in the case of parallel machine failures in the power battery production workshop, obtaining the machine failure constraint and the affected production processes; according to the fault constraint, performing local production line optimization and reorganization on the affected production processes on the basis of the obtained static production line optimization and reorganization result to obtain the dynamic production line optimization and reorganization result. The power battery production line optimization and reorganization method provided by the present invention can not only quickly obtain the static production line optimization and reorganization result in the case of no failure, but also obtain a good dynamic production line optimization and reorganization result in the case of machine failure. Description of the Drawings
[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.
[0064] Figure 1 is a flowchart of a method for optimizing and reorganizing a static production line of a power battery disclosed in an embodiment of the present application.
[0065] Figure 2 is a schematic diagram of the problem of optimizing and reorganizing the production line in the power battery workshop disclosed in an embodiment of the present application.
[0066] Figure 3 It is a flowchart of a method for optimizing and reorganizing a dynamic production line of power batteries disclosed in an embodiment of the present application.
[0067] Figure 4 It is a Gantt chart of the static production line optimization and reorganization result of the power battery production line optimization and reorganization method of the present invention disclosed in an embodiment of the present application.
[0068] Figure 5 It is a Gantt chart of the static production line optimization and reorganization result based on the particle swarm optimization algorithm disclosed in an embodiment of the present application.
[0069] Figure 6 It is a Gantt chart of the static production line optimization and reorganization result based on the reverse particle swarm optimization algorithm disclosed in an embodiment of the present application.
[0070] Figure 7 It is a comparative analysis chart of the convergence of the optimization average values based on the algorithm of the present invention, the particle swarm optimization algorithm, and the reverse particle swarm optimization algorithm disclosed in an embodiment of the present application.
[0071] Figure 8 It is a Gantt chart of the dynamic production line optimization and reorganization result based on the local production line optimization and reorganization method disclosed in an embodiment of the present application.
[0072] Figure 9 It is a Gantt chart of the dynamic production line optimization and reorganization result based on the right shift method disclosed in an embodiment of the present application. Detailed implementation manners
[0073] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will further describe the embodiments of the present invention in detail with reference to the accompanying drawings.
[0074] Embodiment 1:
[0075] This embodiment provides a method for optimizing and reorganizing a static production line of power batteries. Refer to Figure 1 , the method includes:
[0076] Step1: Obtain a static mathematical model for optimizing the production line of the power battery workshop;
[0077] Step2: According to the static mathematical model determined in Step1, determine the fitness function and decoding method of the particle swarm, and use the particle swarm method for global search. During the global search process, a multiple symmetric learning method is adopted to find the individual optimal particle to avoid falling into the local optimum during the optimization process of individual particles;
[0078] Step3: According to the individual optimal particle obtained in Step2, adopt an improved variable neighborhood method to avoid the group optimal particle from falling into premature convergence, and obtain the static production line optimization and reorganization result.
[0079] Example 2:
[0080] This example provides a method for optimizing and reorganizing the static production line of power batteries. Refer to Figure 1 The method includes:
[0081] Step 101: Obtain the static mathematical model for optimizing the production line of the power battery workshop.
[0082] Determine the processes included in the power battery production process. The parallel processing equipment corresponding to the th process includes
[0083] The number of battery orders to be completed is n types. For any order, during the th process, it is processed on any one of the machines in this process. represents the completion time of order after completing the processing of the th process;
[0084] This invention studies the optimization and reorganization problem of the front and middle section production lines of the power battery soft pack workshop. The power battery production process includes 6 processes such as batching, coating, sheet making, baking, stacking, and assembly, that is, . The parallel processing equipment corresponding to each process includes 6 groups of equipment such as batching machines, coating machines, sheet making machines, customized oven machines, automatic stacking machines, and automatic assembly machines.
[0085] Generally, the problem of optimizing and reorganizing the power battery production line can be described as types of soft pack battery production order requirements are processed sequentially on the Figure 2 processes shown in . Each process has parallel machines. For any order, during the th process, it can be processed on any one of the parallel machines in this process, and the processing time on the parallel machines may be different.
[0086] The goal of the problem of optimizing and reorganizing the power battery production line to be solved by this invention is to design a reasonable swarm intelligence method to arrange the orders' processing sequences on each process and the selection of corresponding parallel machines, so as to minimize the maximum completion time of all orders.
[0087] First, obtain the static mathematical model for optimizing the production line of the power battery workshop. Its objective function is:
[0088] (1)
[0089] where is the maximum completion time of all orders. is the total order quantity, represents the order number, represents the total number of processes, represents the order completed the processing completion time of the process.
[0090] According to the actual production situation, the main distribution constraints for the production line optimization in the power battery production workshop are as follows:
[0091] (2)
[0092] (3)
[0093] (4)
[0094] (5)
[0095] (6)
[0096] (7)
[0097] Among them represents the processing process number, ; represents the number of parallel machines on each process; represents the machine number of the parallel machines on each process, ; represents the order in the process of the machine for processing;
[0098] represents the order at the th processing process completion time; represents the order at the th processing process start time; represents the order at the th processing process processing time;
[0099] represents the number of orders processed on the machine in the process ;
[0100] Equation (2) indicates that there is more than one machine on each process in the power battery workshop production, and there are parallel machines;
[0101] Equation (3) indicates that for each order, only one machine can be selected for processing in each process;
[0102] Equation (4) indicates that can only take two values, 0 and 1. That is, when the order is processed on the machine in the process , otherwise ; ;
[0103] Equation (5) indicates that for any order, the completion time in any process is equal to the sum of the processing start time and the processing time in that process;
[0104] Equation (6) indicates that the start processing time of the next process of any order is greater than or equal to the completion time of the previous process; Equation (7) indicates that for each process, the orders assigned to all the machines running within that process are .
[0105] Step 102: According to the static mathematical model determined in Step 101, determine the fitness function and decoding method of the particle swarm, and use the particle swarm method for global search.
[0106] The particle swarm algorithm searches continuously for a certain number of particles within a certain interval to find the particle with the optimal fitness and obtain the corresponding optimal value. Using this algorithm to globally search for the solution of the objective function in Equation (1), first, from Equation (1), the fitness function of this article can be set as:
[0107] (8)
[0108] The velocity update and position update rules of the particle are as follows:
[0109] (9)
[0110] (10)
[0111] (11)
[0112] Among them, represents the number of iterations, is the inertia constant, generally taking values from 0.8 to 1.2; is the self-learning factor, is the group learning factor; is a random number uniformly distributed in the interval [0, 1]; is the particle at the th iteration of the The position in dimension ; is the velocity of the particle in the th iteration in dimension ; is the individual optimal position of the particle in the th iteration in dimension ; is the global optimal position of the swarm of particles in the th iteration in dimension , is the minimum value of the particle search range, is the maximum value of the particle search range.
[0113] It can be seen from Equations (9) and (10) that the particle swarm algorithm solves the optimal solution of continuous problems, while the optimization and reorganization of the power battery production line is a discrete problem, which requires determining the processing order of orders in each processing procedure. Its result is composed of permutations of integers from 1 to
[0114] . Therefore, the present invention needs to design corresponding decoding rules. For orders, the dimension of the corresponding particle swarm is , and the smallest value in the vector represented by each particle corresponds to order number 1, the second smallest value corresponds to order number 2, , and the largest value corresponds to order number
[0115] ; The determination of the processing order of subsequent processes is determined by the following first-come-first-served rule:
[0116] For the first process, i.e., the batching process, the orders are completely processed in the production order obtained according to the particle vector proposed by the algorithm of the present invention and based on the decoding rule, and then the appropriate machines are preferentially selected for processing;
[0117] For the remaining processing procedures, they are arranged according to the first-come-first-served principle, that is, according to the arrival order of the orders at this process to arrange their processing order; if multiple orders arrive at this process simultaneously, the workpiece with the longest remaining unprocessed time is preferentially arranged until the processing order of all orders in this process is arranged, and finally the production order of the orders for each process is obtained.
[0118] For the problem of selecting parallel machines in any process, the machine that can complete the current order first in the current process is preferred. That is, first compare the completion time of the previous process of the current order with the earliest start time of all parallel machines in the current process to obtain the earliest processing time, then add the processing time corresponding to each parallel machine of the current order, and compare to find the machine with the minimum completion time. If there are multiple parallel machines with the same minimum completion time, select the machine with the shortest processing time for processing.
[0119] Based on the above rules, a complete optimization and reorganization plan for the power battery production line can be formed.
[0120] Step 103, according to the particle swarm method mentioned in Step 102, a multiple symmetric learning method is designed to prevent the individual optimal particle from falling into the local optimum.
[0121] When performing global search and solution, as the iteration progresses, the particle swarm algorithm often easily falls into the local optimum due to the stagnation of the update of its individual particles. To prevent the individual particles from falling into the local optimum, the present invention designs a multiple symmetric learning method to solve multiple symmetric particles of the stagnant particles.
[0122] The multiple symmetric learning method is as follows: For dimensional space particle , where , randomly generate For the random integer between 1 and n b and c determine The symmetric plane formed by the axis and axis is placed at the origin, and the particle generated by the particle p passing through the symmetric plane is defined as the symmetric particle , then The coordinate components of the symmetric particle are expressed as:
[0123] (12)
[0124] If it satisfies , and , then the particle replaces the particle , otherwise continue to use the particle for subsequent iteration.
[0125] Perform times of multiple symmetric learning on the updated stagnant particles, and symmetric particles can be obtained. If the fitness of a certain symmetric particle is better than its own fitness and better than the fitness of other symmetric particles, then use this symmetric particle to replace the original updated stagnant particle.
[0126] In step 104, based on the individual optimal particles improved in step 103, an improved variable neighborhood method is designed to avoid the group optimal particles from falling into premature convergence, and the optimized recombination result of the static production line is obtained.
[0127] Define the individual particle with the best fitness as the group optimal particle. The present invention designs an improved variable neighborhood method to avoid the group optimal particles from falling into premature convergence, and finally obtains the optimized recombination result of the static production line.
[0128] The improved variable neighborhood method of the present invention designs the following three neighborhood structures:
[0129] (1) Multiple symmetric learning : For a certain particle, randomly generate multiple symmetric planes to obtain neighborhood particles.
[0130] (2) Opposite learning : For a certain particle, find the opposite particle of this particle.
[0131] (3) Genetic mutation reverse order : For a certain particle, randomly select two position points in the particle position vector, and reverse the segment between these two points to obtain the reverse order particle.
[0132] After designing the neighborhood structure, this paper first uses jitter operations to judge whether the group optimal particles fall into premature convergence. If the group optimal particles fall into premature convergence, save the optimal particles after jitter for the next local search to obtain the final group optimal particles; otherwise, directly output the group optimal particles without local search. Thus, the pseudo-code of the improved variable neighborhood search method is shown in Table 1 below. Obtain the final group optimal particles from Table 1. According to the position vector of the group optimal particles, the optimized recombination result of the final static production line can be obtained according to the decoding rule in step 102.
[0133] Table 1: Pseudo-code of the improved variable neighborhood search method
[0134]
[0135] Example 3:
[0136] This example provides a method for optimizing and reorganizing a dynamic production line for power batteries. As Figure 3 shown, on the basis of Example 2, the method further includes:
[0137] In step 105, in the case of parallel machine failures in the power battery production workshop, obtain the machine failure constraints and the affected production processes.
[0138] When a certain parallel processing machine fails, that is, when the process The machine fails at and the repair time of the machine needs to be considered . Define , then the machine is unavailable within , that is, the failure constraint is:
[0139] (13)
[0140] At this time, the process tasks of different battery production orders may be in one of the following states: completed, unprocessed, being processed, obtain the unprocessed and being processed production processes affected by the failed machine .
[0141] Step 106: According to the failure constraint in Step 105, on the basis of the static production line optimization and reorganization result obtained in Step 104, perform local production line optimization and reorganization on the affected production processes to obtain the dynamic production line optimization and reorganization result
[0142] According to the failure constraint in Step 105, on the basis of the static production line optimization and reorganization result obtained in Step 104, adjust the affected being processed and unprocessed processes, that is, reselect the machines for the affected order processes using the decoding rule in Step 102 to obtain the dynamic production line optimization and reorganization result
[0143] To verify the effectiveness of the power battery production line optimization and reorganization method proposed in this application, this application conducts a simulation experiment. In the experiment, first obtain the actual production data of a certain battery production company. The number of parallel machines corresponding to its first 6 processes in the front and middle sections is 2, 3, 2, 3, 2, 2, a total of 14 machines. Obtain the processing times of 10 orders to be processed by this company as shown in Table 2 below
[0144] Table 2: Summary table of processing times of different order requirements of a power battery company under different processes
[0145]
[0146] Take the initial population , the number of iterations , the inertia constant , the self-learning factor , the group learning factor , the multiple symmetric learning times , the number of jitters in the improved variable neighborhood search .
[0147] Within a predetermined number of iterations, steps 101 to 104 are executed to obtain the optimized and reorganized result of the static production line in the power battery production workshop without faults.
[0148] Meanwhile, it is compared with the optimized and reorganized results of the static production line obtained by the classical particle swarm optimization algorithm and the reverse particle swarm optimization algorithm, and the comparison results are as Figures 4 - 7 shown below.
[0149] For the introduction of the classical particle swarm optimization algorithm, please refer to: Yu Meng, Liu Dehan. Improved PSO-GA algorithm for solving the hybrid flow shop scheduling problem [J]. Journal of Wuhan University of Technology (Transportation Science & Engineering), 2021, 45(03): 586-590.
[0150] For the introduction of the reverse particle swarm optimization algorithm, please refer to: Deng Zhicheng, Sun Hui, Zhao Jia, Wang Hui. Application of the aggregation-degree self-adaptive reverse learning particle swarm optimization algorithm in reservoir optimal operation [J]. Water Resources and Hydropower Engineering, 2020, 51(04): 166-174.
[0151] As can be seen from Figures 4 - 6 , the completion time of the optimized and reorganized static production line obtained by the method for optimizing and reorganizing the power battery production line of the present invention is 36.9 h, the completion time of the optimized and reorganized static production line obtained by the classical particle swarm optimization algorithm is 40.3 h, and the completion time of the optimized and reorganized static production line obtained by the reverse particle swarm optimization algorithm is 39 h. It can be seen that the method for optimizing and reorganizing the power battery production line of the present invention has a shorter maximum completion time and a more superior optimized and reorganized effect of the static production line than the particle swarm optimization algorithm and the reverse particle swarm optimization algorithm; moreover, with the expansion of the production scale, this advantage will become more obvious. As can be seen from Figure 7 , the convergence of the method for optimizing and reorganizing the power battery production line of the present invention is better than that of the other two production line optimization and reorganization algorithms.
[0152] On the basis of the optimized and reorganized result of the static production line shown in Figure 4 , when it is detected at that the No. 3 machine in the coating equipment fails, considering that the repair time of this machine is 5 h, steps 105 and 106 are executed to obtain the optimized and reorganized result of the dynamic production line, and it is compared with the classical right-shift method, and the comparison results are as Figures 8 - 9 shown below.
[0153] For the introduction of the classical right-shift method, please refer to: Zhang Guohui, Wang Yongcheng, Zhang Haijun. Multi-stage human-machine cooperation for solving the dynamic flexible job shop scheduling problem [J]. Control and Decision, 2016, 31(01): 169-172.
[0154] As can be seen from Figure 8 and 9It can be seen that the maximum completion time of the dynamic production line optimization and reorganization obtained by using the local production line optimization and reorganization method proposed in this application is 36.9h, which is only 0.2h later than that of the static production line optimization and reorganization. However, the maximum completion time of the dynamic production line optimization and reorganization obtained by using the existing right-shift method is 41.7h, which is 5h later than that of the static production line optimization and reorganization. The difference in the delay time between the two is 25 times, indicating that the local production line optimization and reorganization method proposed in this application has better robustness in minimizing the impact of the maximum completion time in case of faults than the traditional right-shift method, and the effect of dynamic production line optimization and reorganization is better.
[0155] In summary, in the case of no faults, for the classical swarm intelligence algorithm, the method of the present invention has better solution quality in solving the problem of optimizing and reorganizing the production line of the power workshop, that is, the maximum completion time for all order productions is the smallest. And in the case of faults, for the classical dynamic production line optimization and reorganization method, the dynamic production line optimization and reorganization method of the present invention has stronger robustness and better dynamic production line optimization and reorganization effect.
[0156] Some steps in the embodiments of the present invention can be implemented by software, and the corresponding software program can be stored in a readable storage medium, such as an optical disc or a hard disk, etc.
[0157] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A method for optimizing and reorganizing a static production line of power batteries, characterized in that, the method includes: Step1: Obtain a static mathematical model for optimizing the production line of the power battery workshop; Step2: According to the static mathematical model determined in Step1, determine the fitness function and decoding method of the particle swarm, and use the particle swarm method for global search. During the global search process, a multiple symmetric learning method is adopted to find the individual optimal particle to avoid falling into the local optimum during the optimization process of individual particles; Step3: According to the individual optimal particle obtained in Step2, adopt an improved variable neighborhood method to avoid the group optimal particle from falling into premature convergence, and obtain the optimized and reorganized result of the static production line; The static mathematical model for optimizing the production line of the power battery workshop in Step1 is: Determine the processes included in the production process of power batteries, and the parallel processing equipment corresponding to the th process includes Battery orders to be completed n types, any one of the orders is processed on any one of the machines at the th process, indicating the order completed processing completion time of the th process; Let the makespan of all orders be , then the objective function of the static mathematical model for the production line optimization of the power battery workshop is as follows: (1) The static production line optimization constraints of power batteries include: (2) (3) (4) (5) (6) (7) Among them, represents the machine number of the parallel machine in each process, ; represents the order in the th processing process on the machine for processing. When the order is in the th processing process on the machine for processing, otherwise ; represents the completion time of the order in the th processing process; represents the start time of processing of the order in the th processing process; represents the processing time of the order in the th processing process; represents the number of orders processed on the machine in the th processing process; The fitness function of the particle swarm determined according to the static mathematical model determined in Step1 in Step2 is: (8) Particle The velocity update and position update rules are as follows: (9) (10) (11) Among them, represents the number of iterations, is the inertia constant, and its value range is 0.8 to 1.2; is the self-learning factor, is the swarm learning factor; is a random number uniformly distributed on the interval [0, 1]; is the particle at the th iteration and the th dimension of the position; is the particle at the th iteration and the th dimension of the velocity; is the particle at the th iteration and the th dimension of the individual optimal position; is the global optimal position of the swarm particles at the th iteration and the th dimension, is the minimum value of the particle search range, is the maximum value of the particle search range; The multiple symmetric learning method is adopted to find the individual optimal particle during the global search process of the particle swarm, including: For dimensional space particles , where , randomly generate for random integers from 1 to n and b and c determine The symmetric plane formed by the axis and axis is placed at the origin, and the particles generated by the particles p passing through the symmetric plane are defined as symmetric particles , then For the symmetric particles the coordinate components are expressed as (12) If the following conditions are met and , then the particle replaces the particle , otherwise the particle is continued to be used for subsequent iterations; Perform times of multi-symmetric learning on the particles with stagnant updates to obtain symmetric particles. For a certain particle, if the fitness of a certain symmetric particle is better than its own fitness and better than the fitness of other symmetric particles of this particle, then use this symmetric particle to replace the original particle with stagnant updates; During the iteration process, define the individual particle with the best fitness as the group optimal particle; In the improved variable neighborhood method, define 3 new neighborhood structures: (1) Multiple symmetric learning : Randomly generate multiple symmetric planes for a certain particle to obtain symmetric particles; (2) Reverse learning : Find the reverse solution for a certain particle; (3) Reverse order of genetic variation : For a certain particle, randomly select two position points in the particle position vector, and reverse the segment between these two points to obtain a reversed particle; Adopt an improved variable neighborhood method to avoid the group optimal particle from falling into premature convergence, and obtain the optimized and reorganized result of the static production line, including: Utilize Determine whether the group-optimal particle falls into premature convergence by means of the secondary jitter operation; If the globally optimal particle falls into premature convergence, then the optimal particle after jittering is saved for the next sub-local search to obtain the final globally optimal particle; Conversely, directly output this group optimal particle without local search; According to the position vector of the group optimal particle, the final optimized and reorganized result of the static production line can be obtained according to the decoding rule.
2. The method according to claim 1, characterized in that, In the method, the purpose of the particle swarm method for optimizing and reorganizing the static production line of power batteries is to determine the processing sequence of each order in each processing operation. After the processing sequence of the first operation among the operations is determined, the subsequent operations can determine the corresponding processing sequence according to the principle of first come, first served. Therefore, the vector represented by the globally optimal particle obtained by particle swarm optimization is the processing sequence of each order in the first processing operation; the decoding rule of the particle swarm is: For orders, the dimension of the corresponding particle swarm is . In the vector represented by each particle, the smallest value corresponds to order number 1, the second smallest value corresponds to order number 2, , and the largest value corresponds to order number ; The subsequent processes arrange their processing sequences in the order of arrival of the orders at this process; If multiple orders arrive at this process simultaneously, give priority to arranging the workpiece with the longest remaining unprocessed time until all the order sequences of this process are arranged, and finally obtain the order production sequence of each process; For the problem of selecting parallel machines at any process, give priority to selecting the machine that can complete the current order first at the current process.
3. A method for optimizing and reorganizing a dynamic production line of power batteries, characterized in that, after the method described in any one of claims 1-2 obtains the optimized and reorganized result of the static production line of power batteries, the method includes: Step4: In the case of a parallel machine failure in the power battery production workshop, obtain the machine failure constraint and the affected production processes; Step5: According to the failure constraint in Step4, on the basis of the optimized and reorganized result of the static production line obtained by the method described in any one of claims 1-2, perform local production line optimization and reorganization on the affected production processes to obtain the optimized and reorganized result of the dynamic production line.
4. The method according to claim 3, characterized in that, in the case of a parallel machine failure in the power battery production workshop, obtaining the machine failure constraint and the affected production processes includes: Set the process The machine Fails at Time, and the repair time of the machine ; Definition , then the machine is unavailable within the time period, i.e., the fault constraint is: (13) At this time, the process tasks of different battery production orders may be in one of the following states: completed, unprocessed, in process, obtaining unprocessed and in-process production processes affected by faulty machines affected by faulty machines 5. The method according to claim 4, characterized in that, Step5 includes: Add the fault constraint conditions to the battery static production line optimization constraint conditions. Based on the obtained results of the static production line optimization and reorganization, adjust the affected in-process and unprocessed operations according to the first-come, first-served rule and the machine selection rule to determine the optimization and reorganization results of the power battery dynamic production line.