Ship pipe fitting workshop scheduling method, system and equipment based on large model assisted evolutionary algorithm and medium

Abstract

The invention discloses a ship pipe fitting workshop scheduling method, system, equipment and medium based on a large model assisted evolutionary algorithm, and relates to the technical field of ship pipe fitting workshop production operation, and the specific steps are as follows: establishing a hybrid flow workshop scheduling model, adopting a multi-dimensional coding and decoding structure to code a workshop scheduling scheme, generating an initial population, and establishing a hybrid flow workshop scheduling model; and guiding the individuals to carry out iterative evolution by using a search operator sequence generated by the large model based on a historical evolution effect or a randomly generated search operator sequence, and outputting an optimal scheduling scheme with the shortest completion time. According to the method, modeling and solving of a multi-resource collaborative complex scheduling problem are realized, and the problem of insufficient scheduling scheme quality caused by neglect of transportation processing coupling or low algorithm search efficiency in an existing method is solved. By introducing a large-model intelligent guiding mechanism, the evolutionary process has adaptive optimization ability, and the feasibility of a scheduling scheme is ensured through a complete coding and decoding framework.

Description

A method, system, equipment, and medium for scheduling ship fitting workshops based on a large model-assisted evolutionary algorithm. Technical Field

[0001] This invention relates to the field of production operation technology in ship pipe fitting workshops, and in particular to a scheduling method, system, equipment and medium for ship pipe fitting workshops based on a large model-assisted evolutionary algorithm. Background Technology

[0002] In modern shipbuilding, pipe fittings, as key components, are widely used in hull construction, pipeline laying, and fluid transport in various systems. The scheduling efficiency of their production workshop directly affects the shipbuilding cycle and cost. Ship pipe fitting workshops typically feature multiple processes, technologies, and resources, involving multiple operations such as cutting, bending, straightening, welding, grinding, and pumping during production. Due to significant differences in pipe diameter, length, and weight, the workshop often employs various cranes for transportation, including overhead cranes and gantry cranes. In this complex production environment, the coordination between multiple cranes and their interaction with processing equipment significantly impacts overall production efficiency and workshop resource utilization.

[0003] Most existing research neglects the workpiece transportation process or addresses flow shop scheduling problems involving single-crane transportation constraints, with limited research on mixed flow shop problems involving multiple cranes cooperating in transportation. Traditional mathematical programming models can achieve good solutions on small-scale instances, but when faced with large-scale and high-dimensional complexity in real-world production, they tend to suffer from low solution efficiency, excessive computation time, or even failure to converge. In recent years, evolutionary algorithms have been widely used in scheduling optimization due to their excellent global search capabilities, but when solving complex problems involving multi-crane cooperative transportation, they may still face shortcomings such as slow convergence speed, susceptibility to local optima, and strong dependence on algorithm parameters. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is: how to optimize job sequencing, equipment allocation and transportation resource scheduling in the complex scenario of multi-crane collaborative transportation and processing equipment coupling in a ship pipe fitting workshop by establishing a scheduling model and combining it with a large model-assisted evolutionary algorithm, thereby shortening the production cycle.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a ship pipe fitting workshop scheduling method based on a large model-assisted evolutionary algorithm, comprising: establishing a hybrid flow workshop scheduling model according to the logical relationship between resource allocation and task timing during production; encoding workshop scheduling schemes using a multi-dimensional encoding and decoding structure based on the hybrid flow workshop scheduling model; generating an initial population containing several workshop scheduling schemes through random initialization; guiding individuals to iteratively evolve using a search operator sequence generated by the large model based on historical evolutionary effects or a randomly generated search operator sequence based on the initial population; determining whether the iterative evolution process meets the termination condition, and outputting the optimal scheduling scheme with the shortest completion time.

[0007] As a preferred embodiment of the ship fitting workshop scheduling method based on a large model-assisted evolutionary algorithm described in this invention, the step of establishing a hybrid flow shop scheduling model based on the logical relationship between resource allocation and task timing in the production process includes: defining several sets of decision variables based on the logical relationship between resource allocation and task timing in the production process; establishing an objective function based on the decision variables with the shortest maximum completion time as the optimization objective; setting constraints based on the limitations that should be met in the production process, and establishing the hybrid flow shop scheduling model in conjunction with the objective function.

[0008] This invention introduces decision variables related to processing machine selection, crane allocation, and predecessor relationships to couple the collaborative transportation constraints of overhead cranes and gantry cranes with the capacity constraints of processing equipment, establishing a hybrid flow shop scheduling model. This overcomes the shortcomings of traditional scheduling models that neglect multi-crane collaboration or simplify the transportation process, enabling the model to characterize the dynamic constraints between transportation and processing of pipe fittings in multiple processes such as cutting, bending, and welding.

[0009] As a preferred embodiment of the ship fitting workshop scheduling method based on a large model-assisted evolutionary algorithm described in this invention, the step of encoding the workshop scheduling scheme using a multi-dimensional encoding and decoding structure according to a hybrid flow workshop scheduling model includes: integrating the resource allocation and sequence logic of the workshop scheduling scheme according to the hybrid flow workshop scheduling model; selecting a multi-dimensional encoding and decoding structure according to the resource allocation and sequence logic of the workshop scheduling scheme; and encoding the workshop scheduling scheme according to the selected multi-dimensional encoding and decoding structure.

[0010] This invention employs a multi-dimensional encoding and decoding structure to encode workshop scheduling schemes, integrating pipe component sequencing, processing machine selection, and crane selection into a unified solution space. This allows each component to fully express an executable scheduling scheme encompassing the entire transportation and processing process. This encoding and decoding structure restricts constraints such as transportation connections between production processes, machine exclusivity, and first-come, first-served principles.

[0011] As a preferred embodiment of the ship pipe fitting workshop scheduling method based on a large model-assisted evolutionary algorithm described in this invention, the step of guiding individuals to iteratively evolve using a search operator sequence generated by the large model based on historical evolutionary effects or a randomly generated search operator sequence according to the initial population includes: setting a fitness index and establishing a fitness evaluation formula for the ship pipe fitting mixed flow workshop scheduling problem of multi-crane collaborative transportation; selecting the individual with the highest fitness as the optimal individual according to the fitness evaluation formula; dividing the population into subpopulations and guiding individuals to iteratively evolve using a search operator sequence generated by the large model based on historical evolutionary effects or a randomly generated search operator sequence.

[0012] This invention divides the population into subpopulations and introduces a large model to generate a sequence of search operators. Utilizing the pattern recognition and reasoning capabilities of the large model, it transforms fitness changes in the historical evolutionary effects of the population into targeted evolutionary guidance strategies. This method overcomes the problems of blind search direction and susceptibility to local optima in traditional evolutionary algorithms. Through intelligent operator selection, it improves the algorithm's convergence speed and global optimization capability in solving complex scheduling problems involving multi-crane collaborative transportation, ultimately obtaining the optimal scheduling scheme.

[0013] As a preferred embodiment of the ship fitting workshop scheduling method based on a large model-assisted evolutionary algorithm described in this invention, the step of dividing the population into subpopulations and using a search operator sequence generated by the large model based on historical evolutionary effects or a randomly generated search operator sequence to guide individuals in iterative evolution includes: dividing the initial population into two subpopulations of equal size; one subpopulation is guided by a search operator sequence generated by the large model based on historical evolutionary effects to guide individuals in iterative evolution; the other subpopulation is guided by a randomly generated search operator sequence to guide individuals in iterative evolution.

[0014] As a preferred embodiment of the ship pipe fitting workshop scheduling method based on a large model-assisted evolutionary algorithm described in this invention, the step of determining whether the iterative evolution process meets the termination condition and outputting the optimal scheduling scheme with the shortest completion time includes: determining whether the current individual evolution iteration count has reached the maximum iteration count; if the maximum iteration count has not been reached, then continuing the individual evolution process iteratively; if the maximum iteration count has been reached, then outputting the optimal individual; using a right-shift partition decoding strategy to decode the optimal individual to obtain the corresponding objective function value, and outputting the optimal scheduling scheme with the shortest completion time.

[0015] As a preferred embodiment of the ship fitting workshop scheduling method based on a large model-assisted evolutionary algorithm described in this invention, wherein: the step of establishing an objective function based on decision variables and with the shortest maximum completion time as the optimization objective includes: the objective function expression is: Where F represents the objective function, Let j be the set of pipe fittings, and j be the index of the pipe fitting. Where h is the number of operations, and h is the operation index. Indicates pipe fittings In the process The processing completion time.

[0016] This invention provides a scheduling system for ship pipe fitting workshops based on a large model-assisted evolutionary algorithm.

[0017] To address the aforementioned technical problems, this invention provides the following technical solution: a ship fitting workshop scheduling system based on a large-model-assisted evolutionary algorithm, comprising: a model building module, an encoding module, an initialization module, an iterative evolution module, and an output module; the model building module establishes a hybrid flow workshop scheduling model based on the logical relationship between resource allocation and task timing during production; the encoding module encodes the workshop scheduling schemes using a multi-dimensional encoding and decoding structure based on the hybrid flow workshop scheduling model; the initialization module generates an initial population containing several workshop scheduling schemes through random initialization; the iterative evolution module guides individuals to iteratively evolve based on the initial population using a search operator sequence generated by the large model based on historical evolutionary effects or a randomly generated search operator sequence; the output module determines whether the iterative evolution process meets the termination condition and outputs the optimal scheduling scheme with the shortest completion time.

[0018] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the aforementioned method for scheduling ship pipe fitting workshops based on a large model-assisted evolutionary algorithm.

[0019] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the aforementioned method for scheduling ship pipe fitting workshops based on a large model-assisted evolutionary algorithm.

[0020] The beneficial effects of this invention are as follows: By establishing a scheduling model that integrates the time-series coupling relationship between multi-crane collaborative transportation and processing, this invention makes the optimization objective closer to actual production constraints. A multi-dimensional encoding and decoding structure is used to satisfy the requirements of equipment exclusivity and process connection during the solution representation process. A large model is introduced to dynamically guide the evolutionary operator sequence, improving the algorithm's search efficiency in complex solution spaces. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 is a schematic diagram of the overall process of a ship pipe fitting workshop scheduling method based on a large model-assisted evolutionary algorithm according to an embodiment of the present invention.

[0023] Figure 2 is a schematic diagram of a multi-crane collaborative transportation mixed flow workshop for ship pipe fittings, which is a scheduling method for ship pipe fitting workshop based on a large model-assisted evolutionary algorithm according to an embodiment of the present invention.

[0024] Figure 3 is a schematic diagram of the three-layer matrix discrete encoding method of a ship pipe fitting workshop scheduling method based on a large model-assisted evolutionary algorithm according to an embodiment of the present invention.

[0025] Figure 4 shows the convergence process of three algorithms for solving the processing of 50 pipe fittings in a ship pipe fitting workshop scheduling method based on a large model-assisted evolutionary algorithm according to an embodiment of the present invention.

[0026] Figure 5 shows the convergence process of three algorithms for solving the processing of 100 pipe fittings in a ship pipe fitting workshop scheduling method based on a large model-assisted evolutionary algorithm according to an embodiment of the present invention. Detailed Implementation

[0027] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0028] Example 1, referring to Figure 1, is an embodiment of the present invention. This embodiment provides a scheduling method for a ship pipe fitting workshop based on a large model-assisted evolutionary algorithm, including: It should be noted that in a ship pipe fitting production workshop, there is a complex scenario involving multiple processes, various processing machines, and multiple types of cranes working together. Existing scheduling research often ignores the spatial and temporal coupling relationship between different types of cranes, or only considers a single transport device, making it difficult for the optimization results to directly guide actual production. At the same time, traditional evolutionary algorithms, when facing such high-dimensional, strongly constrained mixed flow shop scheduling problems, suffer from shortcomings such as blind search direction, slow convergence speed, and easy getting trapped in local optima.

[0029] Therefore, addressing the issues of existing scheduling methods failing to adequately consider the coupling relationship between multi-crane collaborative transportation and processing, resulting in low solution efficiency and poor solution quality, steps S1-S5 are implemented as follows: A hybrid flow shop scheduling model is established, and a multi-dimensional encoding / decoding structure is used to encode the shop scheduling scheme. This resolves the problem of existing methods not adequately considering the coupling relationship between multi-crane collaborative transportation and processing. An initial population is generated, and the search operator sequence generated by the large model based on historical evolutionary effects or a randomly generated search operator sequence guides the individuals in iterative evolution, outputting the optimal scheduling scheme with the shortest completion time. This resolves the problems of low solution efficiency and poor solution quality in existing methods.

[0030] S1: Establish a hybrid flow shop scheduling model based on the logical relationship between resource allocation and task timing in the production process; S2: Encode the shop scheduling scheme using a multi-dimensional encoding and decoding structure based on the hybrid flow shop scheduling model; S3: Generate an initial population containing several shop scheduling schemes through random initialization; S4: Guide individuals to iteratively evolve using a search operator sequence generated by a large model based on historical evolution effects or a randomly generated search operator sequence based on the initial population; S5: Determine whether the iterative evolution process meets the termination condition and output the optimal scheduling scheme with the shortest completion time.

[0031] Example 2, referring to Figures 2-3, is an embodiment of the present invention. Based on the previous embodiment, a method for scheduling ship pipe fitting workshop based on large model-assisted evolutionary algorithm is provided, including: in step S1, a hybrid flow workshop scheduling model is established according to the logical relationship between resource allocation and task timing in the production process, including the following steps A1-A3: A1: Define several sets of decision variables according to the logical relationship between resource allocation and task timing in the production process.

[0032] There are quantities in the production workshop. Each of the several pipe fittings to be processed requires six steps: cutting, bending, straightening, welding, grinding, and pumping. Each step involves... Each machine is selectable, and at any given time, a single machine can only complete one process for one part. The processing time for each pipe fitting on different processes and machines is known. A gantry crane is available between every two processes to transport pipe fittings. An overhead crane is also installed in the workshop to serve the entire workflow. A single crane (overhead crane or gantry crane) can only transport one pipe fitting at a time in a specific process; the transport time is obtained by dividing the path by the transport speed. At time 0, all pipe fittings arrive, and all processing machines and cranes are idle. Figure 2 shows a schematic diagram of a multi-crane collaborative transport system for ship pipe fittings in a mixed flow workshop.

[0033] Introducing 0-1 decision variables In the current marine pipe fitting processing environment, pipe fittings need to go through several processes sequentially, and a unique processing machine is selected for each process. Corresponding to pipe fittings In the Should machines be selected for each process? The expression is: in, Indicates pipe fittings The One process.

[0034] Introducing 0-1 decision variables In the current complex scheduling process, it is necessary to determine the two pipe fittings. and In the process Are they assigned to the same processing machine? Above, the expression is: Introducing 0-1 decision variables Pipe fittings For each process step, either a gantry crane or an overhead crane can be used for transportation. Corresponding to pipe fittings In the The expression for which type of crane is used for transportation in each process is: Introducing 0-1 decision variables During the scheduling process, it is also necessary to determine the two pipe fittings. and Whether they are assigned to the same crane for transport is expressed as: Introducing 0-1 decision variables For the first The same processing machine in each process In this case, several pipe fittings need to be processed sequentially. The start time of processing will be affected by its immediate preceding pipe fittings. The effect of processing end time is expressed as: Introducing 0-1 decision variables For the first For the gantry crane or overhead crane in this process, several pipe fittings need to be transported sequentially. The start time of transportation will be affected by its immediate preceding pipe fittings. The effect of the transportation end time is expressed as: Introducing 0-1 decision variables The expression is: Introducing 0-1 decision variables The expression is: Introducing 0-1 decision variables The expression is: Introducing 0-1 decision variables The expression is: A2: Based on the decision variables, establish an objective function with the goal of minimizing the maximum completion time.

[0035] The objective function is established with the goal of minimizing the maximum completion time, and its expression is: in, For pipe fittings collection, For the number of processes, Indicates pipe fittings In the process The processing completion time.

[0036] A3: Set constraints based on the limitations that should be met during the production process, and establish a hybrid flow shop scheduling model in conjunction with the objective function.

[0037] In this embodiment of the application, the specific steps of establishing the hybrid flow shop scheduling model in step A3, which combines the objective function, are as follows: for each pipe fitting In each process The completion time of the pipe fitting during transportation is later than the sum of the start time and the required transportation time. The expression is: in, Indicates pipe fittings In the The end time of transportation for each process, Indicates pipe fittings In the The start time of transportation for each process step. Indicates pipe fittings In the The transportation time required for each process.

[0038] Each pipe fitting In the process When being transported, the start time of transport shall not be earlier than its immediate predecessor. The end time of transportation is expressed as: Each pipe fitting In the process At most one preceding fitting is transported. The expression is: It is necessary to ensure that in each process Transport pipe fittings At that time, pipe fittings and its immediate preceding pipe fittings They are assigned to the same crane, as expressed in the following expression: The start time of the gantry crane in the next process shall not be earlier than the end time of the gantry crane in the previous process, as expressed in the following formula: The start time of the overhead crane in the next process shall not be earlier than the end time of the overhead crane in the previous process, expressed as: In the scheduling plan for each gantry crane, at most one pipe fitting can be designated as its first transport task, expressed as: Each gantry crane can have at most one pipe section designated as its final transport task during the scheduling process, expressed as: In the scheduling scheme of each overhead crane, at most one pipe component can be designated as its first transport task, expressed as: Each overhead crane can have at most one pipe section assigned to it as its last transport task during the scheduling process, expressed as: For each pipe fitting In each process The start time of processing is no earlier than the end time when the pipe fitting is transported, expressed as: in Indicates pipe fittings In the process The start time of processing.

[0039] For each pipe fitting In each process The completion time of the processed part is later than the sum of the start time and the required processing time of the pipe fitting, expressed as: in, Indicates pipe fittings In the process The required processing time.

[0040] Each pipe fitting In the process When being machined, the machining process should not begin earlier than its immediate predecessor fitting. The processing end time is expressed as: Each pipe fitting In the process During processing, there is at most one preceding fitting. The expression is: It is necessary to ensure that in each process Processing pipe fittings At that time, pipe fittings and its immediate preceding pipe fittings They are assigned to the same processing machine, as expressed in the following expression: A workpiece can only be processed by one machine at a time for the same operation. in, Indicate process Number of available processing machines.

[0041] In an optional implementation, step A3, which combines the objective function to establish a hybrid flow shop scheduling model, can also employ a priority-based heuristic scheduling method. By predefining resource allocation and task sequencing, and combining this with a real-time updated equipment status table, appropriate processing machines and transport cranes are assigned to the tasks to be scheduled at each decision step. All allocations are verified in real-time for equipment exclusivity and pre- and post-process constraints. Through rule coordination and status tracking, the generated scheduling sequence is ensured to meet the basic constraints in production.

[0042] In another optional implementation, the hybrid flow shop scheduling model established in step A3, based on the objective function, can also employ a scheduling method based on time window conflict detection and adjustment. An initial time window is pre-assigned to all process tasks, and the overlapping of task time windows on the same equipment is checked. If a conflict is detected, adjustments are made according to a predefined strategy (such as shifting tasks, changing order, or reselecting equipment), and the process is re-verified until all conflicts are eliminated.

[0043] It should be noted that this invention, by introducing variables related to processing machine selection, crane allocation, preceding processing relationships, and preceding transportation relationships, mathematically integrates the constraints of the collaborative transportation path between overhead cranes and gantry cranes, the temporal occupancy constraints of processing equipment, and the sequential logical constraints between processes, thus constructing a unified scheduling model. This scheduling model characterizes the coupling relationship between transportation and processing tasks in terms of time and resources during multiple processes such as pipe cutting, bending, and welding, overcoming the limitations of traditional methods that neglect multi-crane collaboration or simplify transportation links.

[0044] In step S2, the workshop scheduling scheme is encoded using a multi-dimensional encoding and decoding structure based on the hybrid flow workshop scheduling model, including the following steps B1-B3: B1: Based on the hybrid flow workshop scheduling model, the resource allocation and sequence logic of the workshop scheduling scheme are integrated.

[0045] B2: Select a multi-dimensional encoding / decoding structure based on the resource allocation and sequence logic of the workshop scheduling scheme.

[0046] B3: Encode the workshop scheduling scheme according to the selected multidimensional encoding and decoding structure.

[0047] In this embodiment of the application, the specific steps for encoding the workshop scheduling scheme in step B3 are as follows: the multidimensional encoding and decoding structure is a three-layer matrix discrete encoding.

[0048] Based on the resource allocation and sequence logic of the workshop scheduling scheme, a three-layer matrix discrete coding method is adopted to uniformly represent the pipe fitting sorting, processing machine selection and crane selection. The three-layer matrix discrete coding method is shown in Figure 3.

[0049] The first layer generates a vector representing the overall startup order of all pipe fittings to be processed, determining the order in which the fittings enter the production line. The second layer establishes a processing machine allocation matrix. Rows in the matrix correspond to each pipe fitting to be processed, and columns correspond to each production process. The value of each cell in the matrix represents the specific machine number selected for processing that pipe fitting in that process. The third layer establishes a crane scheduling matrix. This matrix has the exact same row and column dimensions as the processing machine allocation matrix. The value of each cell represents the crane type identifier assigned to the pipe fitting when it is transported to the next process after the completion of that process.

[0050] In an alternative implementation, the encoding of the workshop scheduling scheme in step B3 can also employ a composite encoding based on a process sequence linked list and a resource allocation list. The scheduling scheme is encoded as two associated lists: one list records the planned execution sequence of all processes, with each node containing a workpiece number, a process number, and a pointer to the next node; the other list records the specific processing machine and crane number assigned to each process node. By traversing the sequence linked list and querying the resource allocation list, the complete scheduling scheme can be reconstructed.

[0051] In another optional implementation, the coding of the workshop scheduling scheme in step B3 can also employ a two-dimensional table coding based on the binding relationship between time slots and resources. The scheduling time axis is divided into discrete, uniform time slots, and a two-dimensional table is constructed. The rows of the two-dimensional table correspond to each time slot, and the columns correspond to all processing machines and cranes. Each cell is filled with the specific workpiece and process number occupying that resource within that time slot.

[0052] It should be noted that this invention, by employing a three-layer matrix discrete coding structure, integrates the pipe fitting processing sequence, processing machine allocation, and crane selection into a unified and complete solution expression. This allows each coded individual to directly correspond to an executable scheduling scheme containing all resource allocation and timing logic, providing an directly operable and evaluable solution representation for evolutionary algorithms. Simultaneously, this coding structure inherently expresses the correspondence between resource selections across different processes at the data level, providing a foundation for designing evolutionary operators.

[0053] In step S3, an initial population containing several workshop scheduling schemes is generated through random initialization, including the following steps: based on individual encoding, the maximum number of iterations is set to... Population size is Generated through random initialization Each individual is used as the initial population.

[0054] In step S4, based on the initial population, the individual is guided to iteratively evolve using a sequence of search operators generated by the large model based on the historical evolution effect or a randomly generated sequence of search operators, including the following steps C1-C3: C1: For the scheduling problem of ship pipe fittings mixed flow workshop with multi-crane collaborative transportation, fitness index is set and fitness evaluation formula is established.

[0055] The core idea of ​​evolutionary algorithms is survival of the fittest. For the scheduling problem of a mixed assembly line workshop for ship pipe fittings involving multi-crane collaborative transportation, the maximum completion time is obtained by constant shifting. As a fitness indicator, among which The fitness value is a constant, and individuals with higher fitness values ​​perform better. If the fitness value of a offspring is not lower than that of its parent, the offspring will replace the parent in the next round of evolution.

[0056] C2: Based on the fitness evaluation formula, select the individual with the highest fitness as the optimal individual.

[0057] Select the individual with the highest fitness value in the population. This indicates that the processing time for all pipe fittings is the shortest. When an individual executes the search operator, it selects to interact with the optimal individual with a certain probability, so that the individual gradually evolves towards the optimal direction.

[0058] C3: Divide into subpopulations and use a sequence of search operators generated by the large model based on historical evolutionary effects or a randomly generated sequence of search operators to guide individuals in iterative evolution.

[0059] In this embodiment of the application, the specific steps in step S4 of guiding individuals to perform iterative evolution are as follows: The initial population... Divided into two subpopulations of equal size and .

[0060] One of the subpopulations is guided by a sequence of search operators generated by the large model based on historical evolutionary effects, allowing individuals to undergo iterative evolution.

[0061] For subpopulations Record the fitness value corresponding to each individual at present. Randomly generate a sequence of search operators and make Each individual executes the corresponding operator to update Subpopulation, after record update The fitness value corresponding to each individual Calculate subpopulation Changes in fitness of each individual after executing the operator .

[0062] Subpopulation The sequence of search operators executed and the changes in the fitness value of each individual after the operators are executed. The prompt words are input into the large model, along with other requirements, including the length of the generated sequence and the range of element values ​​in the generated sequence. See Table 1 for the complete input content of the large model: Table 1 Complete Input Content of the Large Model Extract the output of the large model to obtain a new sequence of search operators. Detection Whether the length, element values, etc., meet expectations for the search operator sequence. Repairs will be carried out.

[0063] Record the current subpopulation The fitness value corresponding to each individual .

[0064] Each individual executes the corresponding new search operator. If the fitness value of the new individual is not lower than that of the current individual, the new individual replaces the current individual.

[0065] Record new subpopulations The fitness value corresponding to each individual .

[0066] Record subpopulation Changes in fitness of each individual after executing the operator .

[0067] Another subpopulation guides individuals through iterative evolution based on a randomly generated sequence of search operators.

[0068] The length of the random permutation operator label generation is The sequence of search operators.

[0069] Each individual in the search algorithm executes the corresponding search operator. If the fitness value of the new individual is not lower than that of the current individual, the new individual replaces the current individual.

[0070] Update the optimal individual; find the individual with the highest fitness among all subpopulations. If the individual's fitness value is not lower than Fitness value, then use replace .

[0071] Furthermore, the search operator sequence includes 8 crossover operators and 2 mutation operators. The crossover operators further include 2 vector crossover methods and 4 matrix crossover methods, which are combined in pairs to form 8 crossover operators. When an individual executes a crossover operator, it has an equal probability of interacting with the optimal individual or with a randomly generated new individual. The individual currently executing the operator is denoted as... The optimal individual is denoted as The randomly generated new individual is denoted as The operators are as follows: Vector crossover method 1: For the pipe fitting sorting part, two positions are randomly selected. and Exchange individuals With individuals (or The sequence between two positions is used to determine whether there are duplicate parts in the entire pipe fitting sorting sequence, and the new individual is repaired.

[0072] Vector crossover method 2: For the pipe fitting sorting part, randomly select a starting position. From the individual Extract the element at this position In individuals (or Find the element Corresponding position , to individual of The search continues until the starting position is returned, forming a loop.

[0073] If the number of cycles is odd, the element at the cycle position of the new individual 1 comes from... The element at the cyclic position of the new individual 2 comes from (or If the number of iterations is even, the element at the cycle position of the new individual 1 comes from... (or The element at the cyclic position of the new individual 2 comes from .

[0074] Determine if all elements of the vector have been traversed. If not, continue searching; if so, output the new individual.

[0075] Matrix Crossing Method 1: For the machine selection section, randomly select two positions. and Exchange individuals With individuals (or The columns corresponding to the two positions are swapped, meaning the entire column is swapped.

[0076] For the crane selection section, two locations are randomly selected. and Exchange individuals With individuals (or The columns corresponding to the two positions are swapped, meaning the entire column is swapped.

[0077] Matrix Crossing Method 2: For the machine selection part, randomly select several positions. Exchange individuals With individuals (or The selected position corresponds to the column, which is the column to be swapped.

[0078] For the crane selection section, several locations are randomly selected. Exchange individuals With individuals (or The selected position corresponds to the column, which is the column to be swapped.

[0079] Matrix Crossing Method 3: For the machine selection section, in each column, i.e., each process... Choose two locations and Exchange individuals With individuals (or The corresponding column in the machine selection section. and For elements in positions between columns, note that the two positions in each column are not the same, and the swapping is performed within the corresponding column.

[0080] Regarding the crane selection section, in each column, i.e., each process... Choose two locations and Exchange individuals With individuals (or The corresponding column in the crane selection section and For elements in positions between columns, note that the two positions in each column are not the same, and the swapping is performed within the corresponding column.

[0081] Matrix Crossing Method 4: For the machine selection section, in each column, i.e., each process... Select several positions Exchange individuals With individuals (or When selecting elements in the corresponding column of the processing machine, it should be noted that the selected positions in each column are not necessarily the same, and the exchange is performed between elements within the corresponding column.

[0082] Regarding the crane selection section, in each column, i.e., each process... Select several positions Exchange individuals With individuals (or When selecting elements in the corresponding column of the crane selection section, it should be noted that the selected positions in each column are not necessarily the same, and the exchange is performed between elements within the corresponding column.

[0083] Mutation operator 1: For the pipe fitting sorting part, randomly select two elements and swap them.

[0084] Regarding the selection of processing machines, in each column, i.e., each process... Randomly select two locations and Swap the elements in these two positions. Note that the two positions selected in each column are not necessarily the same.

[0085] Regarding the crane selection section, in each column, i.e., each process... Randomly select two locations and Swap the elements in these two positions. Note that the two positions selected in each column are not necessarily the same.

[0086] Mutation operator 2: For the pipe sorting part, randomly select a pipe at a position and place it in the first position, and shift the other elements one position to the right in turn.

[0087] Regarding the selection of processing machines, in each column, i.e., each process... Randomly select a location The machine selected for the pipe fitting at that location is randomly changed to another machine available for that process.

[0088] Regarding the crane selection section, in each column, i.e., each process... Randomly select a location If the pipe fitting represented at this position is selected as a gantry crane, it will be changed to an overhead crane; if it is an overhead crane, it will be changed to a gantry crane.

[0089] In an optional implementation, step S4, which guides individuals through iterative evolution, can also employ an operator selection mechanism based on adaptive probability. An initial selection probability is preset for each evolutionary operator (such as multiple crossover or mutation methods). During algorithm iteration, the system continuously tracks the historical success rate of each operator's offspring improving population fitness and dynamically adjusts its selection probability accordingly; operators with better performance have a higher probability of being selected subsequently.

[0090] In another alternative implementation, the iterative evolution of individuals in step S4 can also employ a reinforcement learning model. The population state is considered the state, the selected evolutionary operator is considered the action, and the increase in population fitness is considered the reward. The reinforcement learning model learns which operator to choose in different population states to obtain the greatest long-term reward.

[0091] It should be noted that this invention utilizes a large model to analyze the correlation between operator sequences and fitness changes during historical evolution, dynamically generating search operator sequences adapted to the current population state, thus providing a guidance mechanism based on historical feedback for the evolutionary process. This overcomes the limitations of traditional evolutionary algorithms where operator selection relies on fixed probabilities or experience. Through data-driven intelligent decision-making, the algorithm can dynamically adjust its exploration and development strategies according to the search progress, improving the overall optimization performance of the algorithm.

[0092] In step S5, it is determined whether the iterative evolution process meets the termination condition, and the optimal scheduling scheme with the shortest completion time is output, including the following steps D1-D4: D1: Determine whether the current individual evolution iteration count has reached the maximum iteration count.

[0093] D2: If the maximum number of iterations has not been reached, continue iterating through the individual evolution process.

[0094] D3: If the maximum number of iterations is reached, output the optimal individual.

[0095] D4: Use the right-shift partition decoding strategy to decode the optimal individual to obtain the corresponding objective function value, and output the optimal scheduling scheme with the shortest completion time.

[0096] In this embodiment of the application, the specific steps of outputting the optimal scheduling scheme with the shortest completion time in step S5 are as follows: In view of the production characteristics of the hybrid flow workshop, a right-shift partition decoding strategy is adopted. For the first process, the transport queues of overhead cranes and gantry cranes are first determined, and then the transport end time of each workpiece is calculated. Subsequently, the processing workpiece queue on the parallel machine of this process is determined, and finally the processing completion time of each workpiece in this process is calculated. After the decoding of the first process is completed, the decoding of subsequent processes is performed in sequence. The first-come, first-served principle is followed when determining the crane workpiece queue (except for the first process) and the machine workpiece queue. That is, the problem studied is based on the non-replacement hybrid flow workshop.

[0097] In an optional implementation, the optimal scheduling scheme with the shortest completion time output in step S5 can also employ a decoding method based on time window recursion and conflict resolution. An initial time window based on its process sequence and baseline time is pre-allocated to each task. During decoding, all tasks are traversed according to the process sequence, and the tasks are placed in the schedule of the assigned equipment. If time overlap conflicts occur, conflict resolution is performed according to preset rules, and subsequent tasks are re-verified until all tasks are scheduled without conflict, at which point the scheduling scheme is output.

[0098] In another optional implementation, the optimal scheduling scheme with the shortest completion time in step S5 can also employ a decoding method based on greedy insertion and local optimization. Starting from an empty equipment timeline, all processes of each workpiece are scheduled as a whole according to the encoded workpiece order. Based on the equipment and crane information allocated to each process, a greedy strategy is used to insert it into the earliest available idle time slot on that equipment timeline. After the initial insertion of all tasks is completed, a local optimization check is performed on the generated schedule, and the optimized scheduling scheme is output.

[0099] It should be noted that the right-shift partitioning decoding strategy employed in this invention can convert the optimal individual's encoding structure into a complete and feasible workshop operation plan. The decoding process strictly follows the first-come, first-served principle and processes the transportation and processing queues sequentially according to the process order, satisfying core production constraints such as the exclusive constraints of processing machines and cranes, the transportation connection requirements between processes, and the requirement that the processing start time is no earlier than the transportation end time. This ensures that the output optimal scheduling scheme is a detailed set of operation instructions that can directly drive actual production.

[0100] In summary, this invention achieves a mathematical characterization of the complete production chain of a ship fitting workshop by constructing a dynamic coupled scheduling model that integrates the collaborative transportation and processing timing of multiple cranes. By designing a three-layer matrix discrete encoding and right-shift partitioning decoding mechanism, it solves the problem that traditional encoding methods cannot automatically satisfy complex workshop constraints. Furthermore, by introducing a large model to dynamically guide and generate the evolutionary operator sequence, it overcomes the shortcomings of traditional evolutionary algorithms in solving strongly coupled optimization problems, such as blind direction and premature convergence.

[0101] Example 3, referring to Figures 4 and 5, is an embodiment of the present invention. Based on the previous embodiment, a scheduling method for a ship pipe fitting workshop based on a large model-assisted evolutionary algorithm is provided, including: instance verification and analysis. Using Python programming, taking a mixed flow workshop for ship pipe fittings with multiple cranes cooperating in transportation as an example, the effectiveness of the large model-assisted evolutionary algorithm is verified. The specific experimental process is as follows: According to the actual production scenario, the basic information of the workshop is extracted as shown in Table 2. The data in the table indicate that there are six processes in the workshop, and the number of parallel machines for each process is 3, 2, 2, 3, 2, and 2, respectively, with a total of 14 machines.

[0102] Table 2 Basic Information of the Workshop

[0103] A comparative analysis of the large model-assisted evolutionary algorithm, genetic algorithm, and discrete particle swarm optimization algorithm is presented. The basic parameter settings of the three algorithms are shown in Table 3. Considering that the Kimi series large models have high inference capabilities and moderate prices, the Kimi-K2-0711-Preview large model is selected. Temperature is a parameter that controls randomness and creativity when generating text in the large model, and 0.8 is the result after sensitivity analysis. The population size, maximum number of iterations, crossover probability, and mutation probability are selected based on empirical values.

[0104] Table 3 Algorithm Basic Parameter Settings

[0105] Tables 4 and 5 show the maximum completion times obtained by the three algorithms when processing pipe fittings in batches of 50 and 100, respectively. To verify the reliability of the algorithms, each example was run 5 times, and the results are as follows: Table 4 Completion times (hours) obtained by the three algorithms when processing 50 pipe fittings

[0106] Table 5. Completion time (in hours) obtained by three algorithms when processing 100mm pipe fittings.

[0107] As can be seen from the data in Tables 4 and 5, the large model-assisted evolutionary algorithm designed in this invention achieves a completion time earlier than that obtained by the genetic algorithm and the discrete particle swarm algorithm in a multi-crane collaborative transportation mixed flow workshop for ship pipe fittings, demonstrating good practical application performance. Meanwhile, Figures 4 and 5 show the convergence process of the three algorithms when solving two examples.

[0108] Example 4 is an embodiment of the present invention, which provides a ship fitting workshop scheduling system based on a large model-assisted evolutionary algorithm, including: a model building module, an encoding module, an initialization module, an iterative evolution module, and an output module; the model building module establishes a hybrid flow workshop scheduling model based on the logical relationship between resource allocation and task timing in the production process; the encoding module encodes the workshop scheduling schemes using a multi-dimensional encoding and decoding structure based on the hybrid flow workshop scheduling model; the initialization module generates an initial population containing several workshop scheduling schemes through random initialization; the iterative evolution module guides individuals to iteratively evolve based on the initial population using a search operator sequence generated by the large model based on historical evolutionary effects or a randomly generated search operator sequence; the output module determines whether the iterative evolution process meets the termination condition and outputs the optimal scheduling scheme with the shortest completion time.

[0109] This embodiment also provides an electronic device applicable to a ship pipe fitting workshop scheduling method based on a large model-assisted evolutionary algorithm, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the ship pipe fitting workshop scheduling method based on a large model-assisted evolutionary algorithm proposed in the above embodiment.

[0110] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements a ship pipe fitting workshop scheduling method based on a large model-assisted evolutionary algorithm as proposed in the above embodiment.

[0111] The storage medium proposed in this embodiment and the method for scheduling ship pipe fitting workshop based on a large model-assisted evolutionary algorithm proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0112] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0113] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A scheduling method for ship pipe fitting workshops based on a large model-assisted evolutionary algorithm, characterized in that: This includes: establishing a hybrid flow shop scheduling model based on the logical relationship between resource allocation and task timing during the production process; encoding the shop scheduling schemes using a multi-dimensional encoding and decoding structure based on the hybrid flow shop scheduling model; generating an initial population containing several shop scheduling schemes through random initialization; guiding individuals to iteratively evolve using a search operator sequence generated by a large model based on historical evolution effects or a randomly generated search operator sequence based on the initial population; determining whether the iterative evolution process meets the termination condition and outputting the optimal scheduling scheme with the shortest completion time.

2. The ship fitting workshop scheduling method based on a large model-assisted evolutionary algorithm as described in claim 1, characterized in that: The process of establishing a hybrid flow shop scheduling model based on the logical relationship between resource allocation and task timing in the production process includes: defining several sets of decision variables based on the logical relationship between resource allocation and task timing in the production process; establishing an objective function based on the decision variables with the shortest maximum completion time as the optimization objective; setting constraints based on the limitations that should be met in the production process; and establishing the hybrid flow shop scheduling model in conjunction with the objective function.

3. The method for scheduling ship pipe fitting workshops based on a large model-assisted evolutionary algorithm as described in claim 2, characterized in that: The step of encoding the workshop scheduling scheme using a multi-dimensional encoding and decoding structure based on the hybrid flow workshop scheduling model includes: integrating the resource allocation and sequence logic of the workshop scheduling scheme according to the hybrid flow workshop scheduling model; selecting a multi-dimensional encoding and decoding structure according to the resource allocation and sequence logic of the workshop scheduling scheme; and encoding the workshop scheduling scheme according to the selected multi-dimensional encoding and decoding structure.

4. The ship fitting workshop scheduling method based on a large model-assisted evolutionary algorithm as described in claim 3, characterized in that: The step of guiding individuals to iteratively evolve based on the initial population using a sequence of search operators generated by a large model based on historical evolutionary effects or a randomly generated sequence of search operators includes: setting a fitness index and establishing a fitness evaluation formula for the scheduling problem of a mixed flow workshop for ship pipe parts in multi-crane collaborative transportation; selecting the individual with the highest fitness as the optimal individual according to the fitness evaluation formula; dividing the population into subpopulations and guiding individuals to iteratively evolve using a sequence of search operators generated by a large model based on historical evolutionary effects or a randomly generated sequence of search operators.

5. The method for scheduling ship pipe fitting workshops based on a large model-assisted evolutionary algorithm as described in claim 4, characterized in that: The process of dividing the population into subpopulations and using a large model to guide individuals in iterative evolution based on a sequence of search operators generated by the model based on historical evolutionary effects or a randomly generated sequence of search operators includes: dividing the initial population into two subpopulations of equal size; one subpopulation is guided by a sequence of search operators generated by the model based on historical evolutionary effects to guide individuals in iterative evolution; and the other subpopulation is guided by a randomly generated sequence of search operators to guide individuals in iterative evolution.

6. The method for scheduling ship pipe fitting workshops based on a large model-assisted evolutionary algorithm as described in claim 5, characterized in that: The step of determining whether the iterative evolution process meets the termination condition and outputting the optimal scheduling scheme with the shortest completion time includes: determining whether the current individual evolution iteration count has reached the maximum iteration count; if the maximum iteration count has not been reached, then continuing to iterate the individual evolution process; if the maximum iteration count has been reached, then outputting the optimal individual; using a right-shift partition decoding strategy to decode the optimal individual to obtain the corresponding objective function value, and outputting the optimal scheduling scheme with the shortest completion time.

7. The method for scheduling ship pipe fitting workshops based on a large model-assisted evolutionary algorithm as described in claim 6, characterized in that: The step of establishing an objective function based on decision variables and with the shortest maximum completion time as the optimization objective includes: the objective function expression is: Where F represents the objective function, Let j be the set of pipe fittings, and j be the index of the pipe fitting. Where h is the number of operations, and h is the operation index. Indicates pipe fittings In the process The processing completion time.

8. A ship pipe fitting workshop scheduling system based on a large model-assisted evolutionary algorithm, employing the ship pipe fitting workshop scheduling method based on a large model-assisted evolutionary algorithm as described in any one of claims 1-7, characterized in that, include: The system comprises a model building module, an encoding module, an initialization module, an iterative evolution module, and an output module. The model building module establishes a hybrid flow shop scheduling model based on the logical relationship between resource allocation and task timing during production. The encoding module encodes the flow shop scheduling schemes using a multi-dimensional encoding / decoding structure based on the hybrid flow shop scheduling model. The initialization module generates an initial population containing several flow shop scheduling schemes through random initialization. The iterative evolution module guides individuals through iterative evolution based on the initial population, using a sequence of search operators generated by a large model based on historical evolutionary effects or a randomly generated sequence of search operators. The output module determines whether the iterative evolution process meets the termination condition and outputs the optimal scheduling scheme with the shortest completion time.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the ship pipe fitting workshop scheduling method based on the large model-assisted evolutionary algorithm according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the ship pipe fitting workshop scheduling method based on the large model-assisted evolutionary algorithm as described in any one of claims 1 to 7.

Images