Data-driven production line layout intelligent optimization and reconstruction method

Abstract

The invention discloses a data-driven production line layout intelligent optimization and reconstruction method, and relates to the field of intelligent manufacturing engineering. According to the method, the cuckoo search algorithm, the group collaborative search mechanism and the firefly algorithm are fused, and intelligent optimization and dynamic reconstruction of the production line layout are realized. According to the invention, the stability of layout optimization and the performability of the scheme can be improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing engineering, specifically to a data-driven method for intelligent optimization and reconfiguration of production line layout. Background Technology

[0002] Currently, production line layout optimization is an important research direction in the fields of intelligent manufacturing and industrial engineering. Traditional layout methods mostly rely on human experience or static planning models, formulating solutions by manually evaluating production data, material flow, and equipment utilization. However, in complex manufacturing environments, workstation load, resource allocation, and logistics paths fluctuate with changes in production rhythm and product structure, making it difficult for traditional methods to achieve real-time response and global coordination of multi-source data. Although existing intelligent optimization methods have introduced intelligent search mechanisms such as genetic algorithms, particle swarm optimization, or ant colony optimization, they generally suffer from problems such as being prone to getting trapped in local optima, insufficient search accuracy, and slow convergence speed, making it difficult to simultaneously consider global exploration and local refinement, resulting in unstable optimization results.

[0003] In existing research, the cuckoo search algorithm possesses strong global search capabilities due to its Levy flight mechanism, but it suffers from drawbacks such as insufficient local convergence and low utilization of population information in complex layout optimization. The firefly algorithm performs well in local search and fine-grained optimization, but it is prone to getting stuck in local optima and lacks global exploration capabilities. When applying a single algorithm to production line layout optimization, it is difficult to achieve a balance between global search and local refinement. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a data-driven intelligent optimization and reconstruction method for production line layout, which integrates the Cuckoo Search algorithm, the group collaborative search mechanism, and the Firefly algorithm to achieve intelligent optimization and dynamic reconstruction of production line layout.

[0005] The technical solution disclosed in this invention is: a data-driven intelligent optimization and reconstruction method for production line layout, comprising the following steps: 1) Collect real-time data from the production line and perform data preprocessing to obtain production status data, equipment status data, and material flow data for each workstation used in step 2); 2) Perform data analysis on the production status data, equipment status data and material flow data of each workstation obtained in step 1) to identify bottlenecks and inefficient areas in the production line and calculate the load balance index. 3) For workstations in bottleneck and inefficient areas, construct an objective function to optimize the comprehensive performance index formed by weighting the load balance of the production line, equipment utilization, and logistics smoothness. Then, perform a global search using the Cuckoo Search algorithm and combine it with the load balance index to obtain a preliminary optimized layout solution. 4) Using the Levy flight strategy in the cuckoo search algorithm, a global search is performed on the initially optimized layout solution and adjusted to obtain a further optimized layout solution; 5) Based on the further optimized layout solution obtained in step 4), through group collaborative search, information between layout solutions is shared, and workstation configuration and resource scheduling are collaboratively adjusted to obtain the initial group of optimized layout solutions; 6) Based on the initial population of the optimized layout solution obtained in step 5), and based on the brightness evaluation mechanism in the firefly algorithm, iterative optimization based on three nested layers of global search, group cooperative search and local search is performed to generate the final layout solution; 7) Based on the final layout solution, adjust the production line configuration and generate a production line reconfiguration plan. Then, perform intelligent optimization and reconfiguration of the production line layout according to the plan.

[0006] Furthermore, step 2) specifically includes: 21) The production status data, equipment status data and material flow data of each workstation obtained in step 1) are grouped and standardized to obtain a standardized dataset, and key feature variables are extracted by principal component analysis to form a dimensionality-reduced dataset. 22) Based on the dimensionality reduction dataset, a clustering algorithm is used to classify each workstation, and the Z-score method is used to detect outliers in the clustering results. After removing outlier data, a clean dataset is generated. 23) Perform similarity analysis on the cleaning dataset, use distance metric to calculate the similarity value between each workstation, arrange the similarity values ​​between each workstation in matrix form to form a workstation similarity matrix, and divide the bottleneck and inefficient areas in the production line according to the similarity matrix between each workstation. 24) Statistically analyze the task processing volume and resource utilization rate of each workstation per unit time, calculate the average load value of each workstation to obtain the average load of the entire production line, and calculate the deviation between the load of each workstation and the average load. Use the normalized result of the sum of squared load deviations as the load balance index.

[0007] Furthermore, step 3) specifically includes: 31) For workstations belonging to bottleneck and inefficient areas, construct an objective function to optimize the comprehensive performance index formed by weighting the load balance of the production line, equipment utilization rate, and material flow efficiency. The load balance of the production line, equipment utilization rate, and material flow efficiency are obtained by weighting the load deviation, equipment utilization rate, and material transfer time of each workstation, respectively. Calculate the comprehensive performance index of each solution and use the comprehensive performance index as the fitness. 32) According to the optimization objective function, a global search is performed by the cuckoo search algorithm. The cuckoo search algorithm performs a global search by selecting the initial layout solution, setting the search space, and adjusting the step size. In the initial stage, the upper limit between the coordinates of each workstation is taken as 10% to 20%. After each iteration, the step size is shortened by multiplying the convergence speed by the decay coefficient of 0.8 to 0.9. 33) During the global search process, in conjunction with the load balancing index, select the top 20% to 50% of the layout solutions with the highest fitness in the search space as the layout solutions for initial optimization. 34) Based on the preliminary optimized layout solution, reconfigure the layout of each workstation in the bottleneck and inefficient areas of the production line.

[0008] Furthermore, step 4) specifically includes: 41) Based on the reconfiguration in step 34), the initial solution of the search space is determined by analyzing the data such as the load, equipment status, and material flow of each workstation in the production line, and the initial solution is used as the initial position of the individual cuckoo to initialize the cuckoo search algorithm. 42) Initialize the position of individual cuckoos, construct a neighborhood search space around each individual with its current position as the center, and set a fixed step size interval of 5% to 10% of the variation range of layout parameters; generate candidate layout solutions in the neighborhood space using the Levy flight strategy, and input the candidate layout solutions into the optimization objective function; when the fitness of two consecutive iterations reaches the set threshold, trigger the fine local search stage, and reduce the step size to 2% to 3% of the variation range of layout parameters.

[0009] 43) During the search process, each individual cuckoo is evaluated. By calculating the fitness of individual cuckoos under the production line optimization target, the top 20% to 50% of individuals with the highest fitness are selected as the retention group for the next round of search, and the Levy flight algorithm is used to update the group. 44) During the search process, based on the information sharing mechanism between individuals, individuals are encouraged to cooperate with each other and optimize the layout solution by comprehensively considering the load and resource configuration of each workstation. 45) After finding a potential optimal solution in the global search, select the top few layout solutions in terms of fitness as the elite solution set. Construct a neighborhood space with the elite solutions as the center. Within the neighborhood, fine-tune the workstation location, resource allocation, and material path to generate candidate solutions. If the candidate solution has better fitness, replace the original candidate layout solution. Use the result optimized through the above steps as the layout solution for further optimization.

[0010] Furthermore, step 5) specifically includes: 51) Based on the further optimized layout solution obtained in step 4), perform a group collaborative search according to the layout solution and the corresponding workstation location, load and resource configuration information; 52) For workstations belonging to bottleneck and inefficient areas, perform resource reallocation and location adjustment on workstations in bottleneck areas, and allocate idle resources of workstations in inefficient areas to high-load areas, and update the information of each layout solution in the group synchronously. 53) During the group collaborative search process, the further optimized layout solution obtained in step 4) is used as a reference. Individuals are selected and updated based on load balancing, equipment utilization and production efficiency, and the group solution set is iteratively updated. 54) After each iteration, adjust the workstation connection relationship and resource scheduling according to the update results, update the layout solution, until the search results converge and form the final optimization scheme, and obtain the initial population of optimized layout solutions.

[0011] Furthermore, step 6) specifically includes: 61) In the global search layer, the fitness of each solution is calculated based on the initial population of optimized layout solutions obtained in step 5) using the cuckoo search algorithm. 62) During the global search process, the solution space is explored by the Levy flight strategy of the cuckoo search algorithm, the search range is controlled by the step size factor, and the updated population solution set is calculated as the global layout solution set, where the fitness of each solution is used as the global optimal evaluation benchmark for the current iteration. 63) After completing one iteration in the global search layer, a group collaborative search is carried out on the global layout solution set to perform information sharing and resource coordination operations between solutions. The load adjustment and position correction are performed on the solutions of workstations belonging to bottleneck and inefficient areas, and the collaborative correction solution set is output. 64) Initialize the firefly algorithm in the local refinement layer, use the cooperative correction solution set as the initial position of the firefly individuals, define the brightness value, and calculate the attraction coefficient based on the brightness difference to perform local search; 65) During the local search process, the step size coefficient and attraction coefficient are dynamically adjusted, the search direction is adjusted by comparing the rate of change of brightness, and the local optimal solution set is output; 66) When the fitness difference between the local optimal solution set and the global layout solution set is less than a preset threshold, the global layout solution set and the local optimal solution set are fused according to the fitness weighting coefficient to generate the final layout solution; otherwise, the local optimal solution set is used as a new initial group, and the optimization based on the three-layer nested structure of global search, group cooperative search and local search is restarted according to the above steps until the final layout solution is obtained.

[0012] Furthermore, the three-level nested structure in step 66) specifically includes: 661) The global search layer outputs a global layout solution set after each iteration based on the initial population, and uses the top K% of the solutions in terms of fitness as input to the population collaboration layer, where K is a preset selection coefficient; 662) The group collaboration layer executes a resource adjustment function based on the bottleneck area weight matrix. The resource adjustment function takes the load balance, equipment utilization and logistics smoothness of each workstation as input variables, calculates the resource allocation correction value, makes incremental adjustments to the resource configuration of high-load workstations, makes decremental balance to the resources of low-load workstations, and outputs the adjusted collaborative correction solution set. 663) The local refinement layer receives the collaboratively corrected solution set and uses the brightness value of each solution as the search evaluation basis. It then performs a local convergence search using the firefly algorithm to generate a locally optimal solution set. 664) When the fitness difference between the optimal solution in the local refinement layer and the optimal solution in the global search layer is less than a preset threshold, result fusion is triggered. The global layout solution set and the local optimal solution set are fused according to the fitness weighting coefficient to generate the final layout solution.

[0013] Furthermore, step 7) specifically includes: 71) Input the final layout solution into the production line control system, and adjust the equipment position, resource allocation and material path of each workstation according to the optimization scheme; 72) Based on real-time collected production operation data, dynamically calculate the workstation load changes, resource utilization rate and logistics delay as layout feedback parameters; 73) After each layout optimization iteration, compare the average fitness value of the current iteration with the rate of change of the previous iteration. When the rate of change is lower than the threshold, lock the current layout and output the layout scheme.

[0014] Furthermore, step 73) specifically also includes: 731) After the layout scheme is generated, a virtual production simulation model is constructed to verify the scheme in terms of production cycle time, material flow and equipment utilization, and to calculate the set of verification indicators. 732) Perform statistical analysis on the material transfer time, equipment switching time and workstation waiting time recorded during the simulation process, and calculate the overload rate of the bottleneck workstation; 733) When any indicator in the verification result deviates from the preset threshold, the parameter corresponding to the abnormal indicator is returned to the group collaboration layer as a correction input to locally update the solution vector of the relevant workstation; 734) Re-execute the joint search of group cooperation and local refinement, update the layout solution, and perform simulation verification again; 735) When two consecutive rounds of simulation verification meet the preset threshold of the verification index set, the final layout scheme is output.

[0015] The beneficial effects of this invention are as follows: This invention introduces an improved cuckoo search algorithm combined with the Levy flight strategy to achieve efficient global exploration of the solution space for complex production line layouts, avoiding the problem of traditional algorithms easily getting trapped in local optima, and improving the globality and convergence speed of the optimization results; by utilizing a group collaborative search mechanism, information sharing and dynamic resource balancing among layout solutions are achieved during the optimization process, enabling coordinated adjustment of load distribution and material flow paths of different workstations, significantly improving the overall coordination and operating efficiency of the production line; by introducing the brightness evaluation and attraction adjustment mechanism of the firefly algorithm, the optimization results are finely corrected, achieving local refinement and adaptive adjustment of the layout scheme, improving the stability of layout optimization and the executability of the scheme. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method in Example 1.

[0017] Figure 2 This is an optimized flowchart of the improved cuckoo search algorithm in Example 1.

[0018] Figure 3 This is a schematic diagram of the three-layer nested structure of Example 1. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments. Example

[0020] This embodiment describes a data-driven intelligent optimization and reconstruction method for production line layout, the main steps of which are as follows: Figure 1 As shown: S1. Collect real-time data from the production line and perform data preprocessing to obtain production status data, equipment status data, and material flow data for each workstation used in step S2.

[0021] S2. Perform data analysis on the production status data, equipment status data and material flow data of each workstation obtained in step S1, identify bottlenecks and inefficient areas in the production line, and calculate the load balance index.

[0022] S3. For workstations in bottleneck and inefficient areas, construct an objective function to optimize the comprehensive performance index formed by weighting the load balance of the production line, equipment utilization, and logistics smoothness. Then, perform a global search using the Cuckoo Search algorithm and combine it with the load balance index to obtain a preliminary optimized layout solution.

[0023] S4. Using the Levy flight strategy in the Cuckoo Search algorithm, perform a global search and adjustment on the initially optimized layout solution to obtain a further optimized layout solution.

[0024] S5. For the further optimized layout solution, through group collaborative search, information between layout solutions is shared, and workstation configuration and resource scheduling are coordinated and adjusted to obtain the initial group of optimized layout solutions.

[0025] S6. Based on the initial population of the optimized layout solution, and using the brightness evaluation mechanism in the firefly algorithm, perform iterative optimization based on three nested layers of global search, group cooperative search, and local search to generate the final layout solution.

[0026] Based on the final layout solution, S7 adjusts the production line configuration and generates a production line reconfiguration plan, and performs intelligent optimization and reconfiguration of the production line layout according to the plan.

[0027] In this embodiment, S2 specifically includes: S21. The production status data, equipment status data and material flow data of each workstation obtained in step 1 are grouped and standardized to obtain a standardized dataset. Key feature variables are extracted through principal component analysis to form a dimensionality-reduced dataset.

[0028] S22. Based on the dimensionality reduction dataset, a clustering algorithm is used to classify each workstation, and the Z-score method is used to detect outliers in the clustering results. After removing outlier data, a clean dataset is generated.

[0029] S23. Perform similarity analysis on the cleaning dataset, calculate the similarity value between each workstation using the distance metric method, arrange the similarity values ​​between each workstation in matrix form to form a workstation similarity matrix, and divide the bottleneck and inefficient areas in the production line according to the similarity matrix between each workstation.

[0030] S24. Statistically analyze the task processing volume and resource utilization rate of each workstation per unit time, calculate the average load value of each workstation to obtain the average load of the entire production line, and calculate the deviation between the load of each workstation and the average load. Use the normalized result of the sum of squared load deviations as the load balance index.

[0031] In this embodiment, S3 specifically includes: S31. For workstations belonging to bottleneck and inefficient areas, construct an objective function to optimize the comprehensive performance index formed by weighting the load balance of the production line, equipment utilization rate, and logistics smoothness. The load balance of the production line is obtained by weighting the load deviation of each workstation, the equipment utilization rate is obtained by weighting the equipment uptime of each workstation, and the logistics smoothness is obtained by weighting the material transfer time of each workstation. Construct the optimization objective function; calculate the comprehensive performance index of each solution, and use the comprehensive performance index as the fitness.

[0032] S32. Based on the optimization objective function, a global search is performed using the Cuckoo Search algorithm. The Cuckoo Search algorithm performs a global search by selecting the initial layout solution, setting the search space, and adjusting the step size. In the initial stage, the upper limit of the coordinates of each workstation is taken as 10% to 20%. After each iteration, the step size is shortened by multiplying the convergence speed by a decay coefficient of 0.8 to 0.9.

[0033] S33. During the global search process, in conjunction with the load balancing index, select the top 20% to 50% of the layout solutions with the highest fitness in the search space as the layout solutions for initial optimization.

[0034] S34. Based on the preliminary optimized layout solution, reconfigure the layout of each workstation in the bottleneck and inefficient areas of the production line.

[0035] like Figure 2 As shown, in this embodiment, S4 specifically includes: S41. Based on the reconfiguration, the initial solution of the search space is determined by analyzing the data such as the load, equipment status, and material flow of each workstation on the production line. The initial solution is then used as the initial position of the individual cuckoo to initialize the cuckoo search algorithm.

[0036] S42. Initialize the position of individual cuckoos, construct a neighborhood search space around each individual with its current position as the center, and set a fixed step size interval of 5% to 10% of the layout parameter variation range; generate candidate layout solutions in the neighborhood space using the Levy flight strategy, and input the candidate layout solutions into the optimization objective function; when the fitness of two consecutive iterations reaches the set threshold, trigger the fine local search stage, and reduce the step size to 2% to 3% of the layout parameter variation range.

[0037] S43. During the search process, each individual cuckoo is evaluated. By calculating the fitness of the individual cuckoo under the production line optimization objective, the individuals with the top 20% to 50% fitness are selected as the retention group for the next round of search, and the Levy flight algorithm is used to update the group.

[0038] S44. During the search process, based on the information sharing mechanism between individuals, individuals are encouraged to cooperate with each other. When optimizing the layout solution, the load and resource configuration of each workstation are comprehensively considered.

[0039] S45. After finding a potential optimal solution in the global search, select the top few layout solutions in terms of fitness as the elite solution set. Construct a neighborhood space with the elite solutions as the center. Within the neighborhood, fine-tune the workstation location, resource allocation, and material path to generate candidate solutions. If the candidate solution has better fitness, replace the original candidate layout solution. Use the result optimized through the above steps as the layout solution for further optimization.

[0040] In this embodiment, S5 specifically includes: S51. Based on the further optimized layout solution, perform a group collaborative search according to the layout solution and the corresponding workstation location, load and resource configuration information.

[0041] S52. For workstations belonging to bottleneck and inefficient areas, perform resource reallocation and location adjustment for workstations in bottleneck areas, and allocate idle resources of workstations in inefficient areas to high-load areas, and synchronously update the information of each layout solution in the group.

[0042] S53. During the group collaborative search process, the further optimized layout solution is used as a reference. Individuals are selected and updated based on load balancing, equipment utilization, and production efficiency, and the group solution set is iteratively updated.

[0043] S54. After each iteration, adjust the workstation connection relationship and resource scheduling according to the update results, update the layout solution, until the search results converge and form the final optimization scheme, and obtain the initial population of optimized layout solutions.

[0044] In this embodiment, S6 specifically includes: S61. In the global search layer, the fitness of each solution is calculated based on the initial population of optimized layout solutions using the cuckoo search algorithm.

[0045] S62. During the global search process, the solution space is explored using the Levy flight strategy of the Cuckoo Search algorithm. The step size factor is used to control the search range, and the updated population solution set is calculated as the global layout solution set. The fitness of each solution is used as the global optimal evaluation benchmark for the current iteration.

[0046] S63. After completing one iteration in the global search layer, a group collaborative search is carried out on the global layout solution set. Information sharing and resource coordination operations are performed between solutions. The load adjustment and position correction are performed on the solutions of workstations belonging to bottleneck and inefficient areas, and the collaborative correction solution set is output.

[0047] S64. Initialize the firefly algorithm in the local refinement layer, use the cooperative correction solution set as the initial position of the firefly individuals, define the brightness value, calculate the attraction coefficient based on the brightness difference, and perform local search.

[0048] S65. During the local search process, the step size coefficient and attraction coefficient are dynamically adjusted, and the search direction is adjusted by comparing the brightness change rate, and the local optimal solution set is output.

[0049] S66. When the fitness difference between the local optimal solution set and the global layout solution set is less than a preset threshold, the global layout solution set and the local optimal solution set are fused according to the fitness weighting coefficient to generate the final layout solution; otherwise, the local optimal solution set is used as a new initial group, and the optimization based on the three-layer nested structure of global search, group cooperative search and local search is restarted according to the above steps until the final layout solution is obtained.

[0050] like Figure 3 As shown, in this embodiment, the three-layer nested structure in S66 specifically includes: S661. The global search layer outputs a global layout solution set after each iteration based on the initial population, and uses the top K% of the solutions in terms of fitness as input to the population collaboration layer, where K is a preset selection coefficient.

[0051] S662. The group collaboration layer executes a resource adjustment function based on the bottleneck area weight matrix. The resource adjustment function takes the load balance, equipment utilization, and logistics smoothness of each workstation as input variables, calculates the resource allocation correction value, performs incremental adjustment on the resource configuration of high-load workstations, performs decremental balancing on the resources of low-load workstations, and outputs the adjusted collaborative correction solution set.

[0052] S663. The local refinement layer receives the collaboratively corrected solution set and uses the brightness value of each solution as the search evaluation criterion. It then performs a local convergence search using the firefly algorithm to generate a locally optimal solution set.

[0053] S664. When the fitness difference between the optimal solution in the local refinement layer and the optimal solution in the global search layer is less than a preset threshold, result fusion is triggered. The global layout solution set and the local optimal solution set are fused according to the fitness weighting coefficient to generate the final layout solution.

[0054] In this embodiment, S7 specifically includes: S71. Input the final layout solution into the production line control system, and adjust the equipment position, resource allocation and material path of each workstation according to the optimization scheme.

[0055] S72. Based on real-time collected production operation data, dynamically calculate the workstation load changes, resource utilization rate, and logistics delays as layout feedback parameters.

[0056] S73. After each layout optimization iteration, compare the average fitness value of the current iteration with the rate of change of the previous iteration. When the rate of change is lower than the threshold, lock the current layout and output the layout scheme.

[0057] In this embodiment, S73 specifically includes: S731. After the layout scheme is generated, a virtual production simulation model is constructed to verify the scheme in terms of production cycle time, material flow and equipment utilization, and to calculate the set of verification indicators.

[0058] S732. Perform statistical analysis on the material transfer time, equipment switching time and workstation waiting time recorded during the simulation process, and calculate the overload rate of the bottleneck workstation.

[0059] S733. When any indicator in the verification result deviates from the preset threshold, the parameter corresponding to the abnormal indicator is returned to the group collaboration layer as a correction input to locally update the solution vector of the relevant workstation.

[0060] S734. Re-execute the joint search of group collaboration and local refinement, update the layout solution, and perform simulation verification again.

[0061] S735. When two consecutive rounds of simulation verification meet the preset threshold of the verification index set, the final layout scheme is output.

[0062] To verify the feasibility of the method in this embodiment, it was applied to a multi-station automated assembly line. The production line contains ten workstations, and the processes include parts assembly, inspection, packaging, and logistics transfer. During long-term operation, the production line suffers from uneven load distribution, concentrated bottleneck workstations, and some workstations being idle, resulting in frequent cycle time fluctuations, lengthy logistics paths, and low equipment utilization. Traditional manual layout and single-algorithm optimization methods converge slowly under these conditions, and the optimization results lack stability, failing to achieve coordinated operation under high cycle time.

[0063] This embodiment employs a data-driven intelligent optimization and reconstruction method for production line layout, reconstructing the spatial arrangement of workstations, equipment locations, and material flow paths. A real-time data acquisition system records production status data, equipment utilization rate, material transfer time, and operation waiting time for each workstation. The data is standardized and processed using principal component analysis to extract production characteristic variables. Data analysis results show that the load rate of assembly stations 3 and 4 and inspection station 7 exceeds 85%, while the equipment utilization rate of stations 8 and 9 is below 40%, indicating a clear distribution of bottlenecks and inefficient areas.

[0064] After data cleaning, standardized data is input into the optimization model, and the Cuckoo Search algorithm is executed to form a global search layer. In the initial stage, the algorithm sets a large step size factor, with the search range covering 15% of the upper limit of the layout parameters. After 40 iterations of the global search, the step size is reduced to 3%, and the Levy flight strategy enters the fine-grained search stage, with the layout solution tending to converge. Optimization results show that the overall fitness is improved by 24.6% compared to the initial layout, and the logistics path length is shortened by 12.3%.

[0065] Based on the global solution set, a swarm collaborative search layer facilitates information sharing among multiple solutions. The algorithm adjusts resource allocation in bottleneck areas according to the load vector and equipment utilization rate, allocating idle resources from low-load workstations to high-load areas. After swarm iterative updates, the system load variance decreased from 0.087 to 0.032, improving load balance by 63%. The algorithm calculates the solution set fitness change rate in each iteration; when the change rate falls below 0.02, it enters the local refinement stage.

[0066] The local refinement layer employs a firefly algorithm based on brightness evaluation to perform spatial fine-tuning. The collaboratively corrected solution set serves as the initial position, and individual fireflies adjust their attraction coefficient and step size parameters based on brightness differences, with high-brightness individuals guiding low-brightness individuals to converge. After 20 rounds of local iteration, the solution set stabilizes, and the layout accuracy is significantly improved. Virtual production simulation verification shows that the production cycle uniformity reaches 97.2%, the overall equipment utilization rate increases from 73.5% to 89.4%, and material transfer time is reduced by 21.7%.

[0067] During the simulation verification phase, production efficiency, resource utilization, equipment changeover time, and bottleneck workstation overload rate were recorded. The optimized solution reduced the bottleneck workstation overload rate from 18.4% to 5.7%, shortened the average logistics waiting time by 2.6 seconds, and increased production capacity by 18.9%. In two consecutive verifications, all performance indicators met the preset standards. The optimized solution demonstrated stable operation and feasibility in both the virtual production model and actual operation.

[0068] This embodiment achieves dynamic reconstruction of the production line layout through global search, group collaboration, and local search. The Cuckoo Algorithm enables large-scale exploration, group collaboration achieves resource balancing and information sharing, and the Firefly Algorithm enables spatial refinement. The multi-layered coupled structure forms a unified layout optimization result with global convergence and local refinement, improving production line operating efficiency, resource utilization, and system coordination performance.

[0069] Table 1. Results of dynamic reconfiguration of production line layout

[0070] The data in Table 1 shows that the production line operating indicators changed significantly after applying this method. The data-driven multi-level optimization process eliminated the overload phenomenon at bottleneck workstations, and the production cycle time became more stable. Equipment utilization increased, and resource allocation became more balanced. Load variance decreased, task distribution among workstations became more uniform, and the coordination of production rhythm improved. Material transfer time was shortened, the material path structure was rationally adjusted, and the amount of work-in-process inventory decreased.

[0071] The algorithm achieves approximately 34% performance improvement during the iterative convergence phase, with a stable search process. The Cuckoo Search algorithm and the Firefly Algorithm form a unified optimization framework within a multi-layered nested structure, maintaining stable computational efficiency and global convergence performance. The overall performance index increases from 0.68 to 0.89, and the verification results are consistent with simulation calculations, with performance parameters meeting design standards. The optimized scheme operates stably, resource allocation is coordinated, and the economic efficiency and operational efficiency of the production line are improved. Overall operational data demonstrates that the layout optimization model is repeatable and engineering-applicable in complex production scenarios.

[0072] The above description is merely a preferred embodiment of the present invention and does not constitute a limitation on the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A data-driven production line layout intelligent optimization and reconfiguration method, characterized in that, The method comprises the following steps: 1) Collecting real-time data of the production line and performing data preprocessing to obtain production status data, equipment status data and material flow data of each workstation for use in step 2); 2) Performing data analysis on the production status data, equipment status data and material flow data of each workstation obtained in step 1) to identify bottlenecks and inefficient areas in the production line and calculate a load balancing index; 3) For the relevant workstations belonging to the bottlenecks and inefficient areas, constructing an objective function for optimizing a comprehensive performance index weighted by the load balancing degree, equipment utilization rate and logistics smoothness of the production line, and performing global search through the cuckoo search algorithm to obtain a preliminarily optimized layout solution in combination with the load balancing index; 4) Using the Levy flight strategy in the cuckoo search algorithm to perform global search and adjustment on the preliminarily optimized layout solution to obtain a further optimized layout solution; 5) For the further optimized layout solution obtained in step 4), performing collaborative adjustment on the workstation configuration and resource scheduling through group collaborative search and sharing information between layout solutions to obtain an initial group of optimized layout solutions; 6) On the basis of the initial group of optimized layout solutions obtained in step 5), performing three-layer nested iterative optimization based on global search, group collaborative search and local search based on the brightness evaluation mechanism in the firefly algorithm to generate a final layout solution; 7) Adjusting the production line configuration according to the final layout solution and generating a production line reconstruction scheme to intelligently optimize and reconstruct the production line layout according to the scheme.

2. The method of claim 1, wherein, Step 2) specifically comprises: 21) Grouping and standardizing the production status data, equipment status data and material flow data of each workstation obtained in step 1) to obtain a standardized data set, and extracting key feature variables through principal component analysis to form a reduced dimension data set; 22) Based on the reduced dimension data set, classifying each workstation using a clustering algorithm, and detecting outliers using the Z-score method to generate a cleaned data set after removing abnormal data; 23) Performing similarity analysis on the cleaned data set, calculating the similarity values between each workstation using a distance measurement method, arranging the similarity values between each workstation in matrix form to form a workstation similarity matrix, and dividing the bottlenecks and inefficient areas in the production line according to the similarity matrix between each workstation; 24) Calculating the average load of each workstation in unit time and resource utilization rate to obtain the average load of the entire production line, and calculating the deviation between the load of each workstation and the average load, taking the normalized result of the load deviation sum of squares as the load balancing index.

3. The method of claim 2, wherein, Step 3) specifically comprises: 31) For the relevant workstations belonging to the bottleneck and low-efficiency area, a target function is constructed to optimize the comprehensive performance index weighted by the load balance, equipment utilization and logistics smoothness of the production line, which are obtained by weighting the load deviation, equipment utilization and material transfer time of each workstation, respectively; the comprehensive performance index of each solution is calculated, and the comprehensive performance index is used as the fitness; 32) According to the optimization target function, the global search is performed by the cuckoo search algorithm, which performs global search by selecting the initial layout solution, setting the search space and adjusting the step size, and in the initial stage, 10% to 20% of the upper limit of the coordinates of each workstation is taken, and after each iteration, the step size is shortened according to the convergence speed multiplied by the decay coefficient 0.8-0.9; 33) During the global search process, the layout solution in the search space is searched in combination with the load balance index, and the top 20%-50% of the layout solutions with the highest fitness are selected as the preliminary optimized layout solutions; 34) The layout of each workstation in the bottleneck and low-efficiency area of the production line is reconfigured according to the preliminary optimized layout solution.

4. The method of claim 3, wherein, Step 4) specifically includes: 41) On the basis of the reconfiguration in step 34), the initial solution of the search space is determined by analyzing the load, equipment state, material flow and other data of each workstation of the production line, and the initial solution is used as the initial position of the cuckoo individual to initialize the cuckoo search algorithm; 42) The position of the cuckoo individual is initialized, and a neighborhood search space is constructed around each individual with its current position as the center, and a fixed step size interval of 5%-10% of the layout parameter variation range is set; candidate layout solutions are generated in the neighborhood space by Levy flight strategy, and the candidate layout solutions are input into the optimization target function; when the fitness of two consecutive iterations reaches the set threshold, the fine local search stage is triggered, and the step size is reduced to 2%-3% of the layout parameter variation range; 43) During the search process, each cuckoo individual is evaluated by calculating the fitness of the cuckoo individual under the optimization target of the production line, and the top 20%-50% of the individuals are selected as the reserved population for the next round of search, and the Levy flight algorithm is used for updating; 44) During the search process, based on the information sharing mechanism between individuals, the individuals are encouraged to cooperate with each other, and the load condition and resource allocation of each workstation are considered when optimizing the layout solution; 45) After the potential optimal solution is found in the global search, the top several layout solutions with the highest fitness are selected as the elite solution set, and the neighborhood space is constructed around the elite solution, and the workstation position, resource allocation and material path are fine-tuned to generate candidate solutions, and if the fitness of the candidate solution is better, the original candidate layout solution is replaced; the result optimized by the above steps is used as the further optimized layout solution.

5. The method of claim 4, wherein, Step 5) specifically includes: 51) Based on the further optimized layout solution obtained in step 4), the layout solution and the corresponding workstation position, load and resource allocation information are used for group collaborative search; 52) For the workstations belonging to the bottleneck and low-efficiency areas, resource reallocation and position adjustment are performed on the workstations in the bottleneck area, and the idle resources of the workstations in the low-efficiency area are allocated to the high-load area, and the information of each layout solution in the group is updated synchronously; 53) In the process of group collaborative search, the further optimized layout solution obtained in step 4) is used as a reference, and the individuals are screened and updated according to the load balancing degree, equipment utilization rate and production efficiency, and the group solution set is iteratively updated; 54) After each iteration, the workstation connection relationship and resource scheduling are adjusted according to the update result, the layout solution is updated, and the final optimization scheme is formed until the search result converges and the optimized layout solution initial group is obtained.

6. The method of claim 5, wherein, Step 6) specifically includes: 61) In the global search layer, the fitness of each solution is calculated based on the optimized layout solution initial group obtained in step 5) by the cuckoo search algorithm; 62) In the global search process, the solution space is explored by the Levy flight strategy of the cuckoo search algorithm, the search range is controlled by using the step factor, and the updated group solution set is calculated as the global layout solution set, wherein the fitness of each solution is used as the global optimal evaluation benchmark of the current iteration; 63) After completing one iteration in the global search layer, the group collaborative search is carried out on the global layout solution set, the information sharing and resource coordination operations are carried out between solutions, the solutions of the workstations belonging to the bottleneck and low-efficiency areas are adjusted and the positions are corrected, and the collaborative correction solution set is output; 64) In the local refinement layer, the firefly algorithm is initialized, the collaborative correction solution set is used as the initial position of the firefly individual, the brightness value is defined, and the attractive force coefficient is calculated according to the brightness difference, and the local search is carried out; 65) In the local search process, the step coefficient and the attractive force coefficient are dynamically adjusted, the search direction is adjusted by comparing the brightness change rate, and the local optimal solution set is output; 66) When the fitness difference between the local optimal solution set and the global layout solution set is less than a predetermined threshold, the global layout solution set and the local optimal solution set are fused according to the fitness weighted coefficient to generate the final layout solution; otherwise, the local optimal solution set is used as the new initial group, and the optimization based on the three-layer nested structure of global search, group collaborative search and local search is restarted according to the above steps until the final layout solution is obtained.

7. The method of claim 6, wherein, The three-layer nested structure in step 66) specifically includes: 661) The global search layer outputs the global layout solution set after each iteration of the initial group, and the solutions ranked in the top K% in terms of fitness are used as the input of the group collaborative layer, wherein K is a predetermined selection coefficient; 662) The group collaborative layer executes the resource adjustment function according to the bottleneck area weight matrix, the resource adjustment function takes the load balancing degree, equipment utilization rate and logistics smoothness of each workstation as input variables, calculates the resource allocation correction value, incrementally adjusts the resource configuration of the high-load workstation, and decrementally balances the resources of the low-load workstation, and outputs the adjusted collaborative correction solution set; 663) The local refinement layer receives the collaborative correction solution set, and uses the brightness value of each solution as the search evaluation basis, and carries out local convergent search by the firefly algorithm to generate the local optimal solution set; 664)When the fitness difference between the optimal solution in the local refinement layer and the optimal solution in the global search layer is less than the preset threshold, trigger result fusion, and perform fusion operation on the global layout solution set and the local optimal solution set according to the fitness weighting coefficient to generate the final layout solution.

8. The method of claim 6, wherein, Step 7) specifically includes: 71) input the final layout solution into the production line control system, and adjust the equipment position, resource allocation and material path of each workstation according to the optimization scheme; 72) dynamically calculate the load change, resource occupancy rate and logistics delay of the workstation according to the real-time collected production operation data as the layout feedback parameters; 73) after each layout optimization iteration, compare the change rate of the current iteration fitness average with that of the last iteration, and when the change rate is lower than the threshold, lock the current layout and output the layout scheme.

9. The method of claim 8, wherein, Step 73) specifically further includes: 731) after the generation of the layout scheme, construct a virtual production simulation model, verify the scheme in terms of production rhythm, material circulation and equipment utilization rate, and calculate a set of verification indexes; 732) statistically analyze the material transfer time, equipment switching time and workstation waiting time recorded in the simulation process, and calculate the overload rate of the bottleneck workstation; 733) when any index in the verification result deviates from the preset threshold, return the abnormal index corresponding parameter to the group collaboration layer as the correction input, and locally update the solution vector of the related workstation; 734) re-execute the group collaboration and local refinement combined search, update the layout solution, and perform simulation verification again; 735) when both the continuous two rounds of simulation verification meet the preset threshold of the set of verification indexes, output the final layout scheme.

Images