Supply chain configuration optimization method and system based on multi-source driven population genetic algorithm

Abstract

The invention relates to a supply chain configuration optimization method and system based on a multi-source driven population genetic algorithm, and the method comprises the steps: dividing a production cycle formed by a plurality of orders into a plurality of scheduling units, and each scheduling unit comprises a plurality of orders; three populations are initialized, each population comprises a plurality of individuals, and each individual is obtained by randomly selecting and configuring each order of one scheduling unit; carrying out fitness evaluation on each individual of the three populations according to respective fitness functions; respectively carrying out selection operation, repairable crossover operation and mutation operation on the three populations to obtain a new main population; performing fitness evaluation on each individual in the new main population again; and according to the new fitness evaluation result, judging whether an iteration termination condition is reached or not, if so, optimizing all scheduling units, and if not, circularly executing the improved genetic algorithm to reach the iteration termination condition. According to the invention, the supply chain can be effectively configured and optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of supply chain optimization configuration, and in particular to a supply chain configuration optimization method and system based on a multi-source driven population genetic algorithm. Background Art

[0002] As globalization accelerates and market demands evolve, traditional static supply chain models are increasingly struggling to adapt to new challenges. Traditional supply chain models, based on static assumptions, struggle to effectively handle dynamic market fluctuations and unexpected disruptions, often failing to effectively respond to these changes. This not only increases production costs but also creates delays, further burdening businesses and failing to meet customer needs. In response to the demands of a complex market environment, dynamic supply chains are urgently needed.

[0003] Dynamic supply chains offer high-quality, rapid response capabilities and flexible adjustment mechanisms, which are crucial for adapting to a globalized and dynamic market. By monitoring and analyzing data from all stages of the supply chain in real time, they flexibly adjust resource allocation and configuration options, achieving dynamic and precise resource allocation, avoiding idle resources and waste, and enhancing companies' competitiveness in the global market. On the one hand, they can keenly detect dynamic changes in various processes, such as raw material procurement, manufacturing, logistics distribution, and marketing, and rapidly adjust them to ensure the supply chain meets market demand. On the other hand, they can flexibly and scientifically adjust supply chain resource allocation, taking into account factors such as cost and efficiency, and dynamically optimizing the allocation of various resources.

[0004] Some progress has been made in exploring dynamic supply chain optimization, with intelligent optimization algorithms with global search capabilities playing a significant role. Many scholars are currently conducting in-depth research on how to find the optimal solution for dynamic supply chain configurations. While existing research has validated the feasibility of intelligent optimization algorithms for dynamic supply chain optimization, a more efficient algorithmic framework has yet to be designed to address the dynamic nature of supply chain problems. For example, accelerating the search for optimal solutions in dynamic environments remains a key issue that needs to be addressed. Summary of the Invention

[0005] Therefore, the technical problem to be solved by the present invention is to overcome the problem in the prior art of finding the optimal solution for dynamic supply chain configuration and how to quickly find the optimal solution.

[0006] To solve the above technical problems, the present invention provides a supply chain configuration optimization method based on a multi-source driven population genetic algorithm, comprising:

[0007] Step S1: Divide the production cycle consisting of multiple dynamically generated orders into multiple scheduling units, each scheduling unit contains multiple orders to be optimized, and each order is obtained by selecting and configuring a number of nodes composed of different manufacturers from upstream to downstream;

[0008] Step S2: For each scheduling unit: first initialize three populations with different fitness functions, each population includes several individuals, and each individual is obtained by randomly selecting a configuration for each order of a scheduling unit;

[0009] Step S3: Evaluate the fitness of each individual in the three populations according to their respective fitness functions;

[0010] Step S4: performing selection operations on the three populations respectively according to their respective fitness evaluation results;

[0011] Step S5: performing a crossover operation on the three populations together to obtain a population after the crossover operation;

[0012] Step S6: performing a mutation operation on the population after the crossover operation, and selecting only the new main population from the three populations;

[0013] Step S7: re-evaluate the fitness of each individual in the new main population;

[0014] Step S8: Determine whether the new main population meets the termination condition based on the fitness evaluation result. If so, output the selection configuration corresponding to the optimal individual in the current scheduling unit; if not, loop through steps S3 to S7 until the termination condition is met.

[0015] Step S9: Execute steps S1 to S8 in a loop until all scheduling units have completed the optimization, and output the selected configuration corresponding to the optimal individual in each scheduling unit.

[0016] In one embodiment of the present invention, the genetic code of the individual is represented by a two-dimensional matrix, each column in the two-dimensional matrix represents a configuration option of an order, and the total configuration options of each order are represented as a row in the two-dimensional matrix, and the number of rows in the two-dimensional matrix is ​​equal to the number of orders.

[0017] In one embodiment of the present invention, the method of performing fitness evaluation on each individual in the three populations according to their respective fitness functions in step S3 includes:

[0018] Calculate the production cost and overtime cost of a single order, calculate the cost of a single scheduling unit based on the production cost and overtime cost, calculate the entire production cycle cost of the product based on the cost of a single scheduling unit, and use the entire production cycle cost of the product as the fitness function of the main population to achieve fitness evaluation of each individual;

[0019] The production cost is used as the fitness function of the first auxiliary population to evaluate the fitness of each individual;

[0020] The overtime cost is used as the fitness function of the second auxiliary population to evaluate the fitness of each individual;

[0021] Among them, the main population, the first and second auxiliary populations constitute three populations.

[0022] In one embodiment of the present invention, the formula for calculating the production cost of a single order is:

[0023]

[0024] The constraints are:

[0025]

[0026] Among them, F1 represents the production cost of a single order, nNode is the number of nodes, represents the production cost of a demand for the order in the i-th node, x i Demand represents the configuration choice of node i, i is the total demand of the i-th node, which is equal to the sum of the demand of all downstream nodes of node i; downNode i Represents the set of downstream nodes of node i, s represents the sth node in the set of downstream nodes of node i, nOption i Indicates the number of configurations for point i, Capacity j,used Indicates the production capacity that has been occupied by configuration j, Capacity j represents the production capacity of configuration j.

[0027] In one embodiment of the present invention, the formula for calculating the overtime cost of a single order is:

[0028]

[0029] The constraints are:

[0030]

[0031] Each order has a time window, which represents the latest delivery range that the order can tolerate; F2 represents the overtime cost of a single order; TW min and TW max Indicates the minimum and maximum values ​​of the order time window; LT represents the delivery time of the order, that is, the maximum value of the delivery time of all nodes. The delivery time of a node is the maximum value of the time spent by the current node plus the delivery time of all upstream nodes of the node; LTimei Indicates the delivery time of node i; maxUpLTime i is the maximum delivery time of the upstream node of node i; upNode i is the upstream node of node i; nNode represents the number of nodes; p represents the penalty coefficient for delivery time delay within the order time window, C timeout represents the penalty cost for delivery time exceeding the order time window.

[0032] In one embodiment of the present invention, the cost of a single scheduling unit is calculated based on the production cost and the overtime cost, and the formula is:

[0033]

[0034] Among them, PC m represents the cost of a single scheduling unit, nOrder represents the number of orders in a scheduling unit, F k，1 、F k,2 denote the production cost and overtime cost of the kth order in the scheduling unit respectively.

[0035] In one embodiment of the present invention, the formula for calculating the entire production cycle cost of a product based on the cost of a single scheduling unit is:

[0036]

[0037] Among them, PC represents the cost of the entire production cycle, nScheduling represents the number of scheduling units, and m represents the mth scheduling unit.

[0038] In one embodiment of the present invention, the method of performing a crossover operation on the three populations at the step S5 includes:

[0039] Determine whether the nodes of the parent individuals of the three populations are feasible:

[0040] If the nodes of the parent individuals of the three populations are feasible, the whole column crossover method is used for the feasible nodes of the parent individuals to exchange the configuration selection of all orders of the parent individuals on the feasible nodes to obtain two child individuals; if the two child individuals are still feasible on the node, the two child individuals replace the parent individuals and enter the new population;

[0041] If the nodes of the parent individuals of the three populations are infeasible, the child individuals generated by the entire column crossover are still infeasible. For the infeasible child individuals, a random value is first generated to compare with the preset repair probability. When the random value is less than or equal to the preset repair probability, the illegal configuration is not processed; when the random value is greater than the preset repair probability, the infeasible solution repair operation is performed on the illegal configuration;

[0042] Among them, individual feasibility means that the occupied production capacity of the individual configuration is less than or equal to the actual production capacity of the configuration, otherwise it means that the individual is infeasible; the illegal configuration is the configuration selection in which the occupied production capacity of the configuration is greater than the actual production capacity of the configuration.

[0043] In one embodiment of the present invention, the infeasible solution repair operation is specifically:

[0044] Repair all illegal configurations in the current illegal configuration set in sequence: For a illegal configuration, randomly replace other configuration options for the node that selected the illegal configuration until the currently occupied production capacity of the configuration is less than or equal to its production capacity. Then check whether the individual has generated a new illegal configuration during the repair process. If so, add the new illegal configuration to the current illegal configuration set; if not, reduce the number of the current illegal configuration set by one; execute the cycle until the current illegal configuration set is empty.

[0045] To solve the above technical problems, the present invention provides a supply chain configuration optimization system based on a multi-source driven population genetic algorithm, comprising:

[0046] Division module: divides the production cycle consisting of multiple dynamically generated orders into multiple scheduling units. Each scheduling unit contains multiple orders to be optimized. Each order is obtained by selecting and configuring several nodes composed of different manufacturers from upstream to downstream.

[0047] Initialization module: For each scheduling unit: First, initialize three populations with different fitness functions. Each population includes several individuals, and each individual is obtained by randomly selecting a configuration for each order of a scheduling unit;

[0048] The first fitness evaluation module: evaluates the fitness of each individual in the three populations according to their respective fitness functions;

[0049] Selection module: performs selection operations on the three populations according to their respective fitness evaluation results;

[0050] Crossover module: Perform crossover operation on the three populations together to obtain the population after crossover operation;

[0051] Mutation module: performs mutation operation on the population after the crossover operation and selects only the new main population from the three populations;

[0052] The second fitness evaluation module: re-evaluates the fitness of each individual in the new main population;

[0053] Judgment module: Determine whether the new main population meets the termination condition of iteration based on the fitness evaluation results. If so, output the selection configuration corresponding to the optimal individual in the current scheduling unit; if not, loop through the first fitness evaluation module to the second fitness evaluation module until the termination condition is met.

[0054] Optimization module: Circularly execute the partitioning module to the judgment module until all scheduling units are optimized, and output the selected configuration corresponding to the optimal individual in each scheduling unit.

[0055] The above technical solution of the present invention has the following advantages over the prior art:

[0056] This paper proposes a supply chain configuration optimization method based on a multi-source driven population genetic algorithm. It regards supply chain configuration as a dynamic and continuous process, which is closer to the dynamic characteristics of actual business scenarios. It can accurately find the optimal solution for dynamic supply chain configuration and can quickly find the optimal solution.

[0057] The present invention designs two-dimensional matrix encoding, multi-source driven population and infeasible solution repair strategy, which improves the diversity of population and enhances the convergence of the method;

[0058] The present invention utilizes a multi-source driven population genetic algorithm to continuously improve the convergence of the algorithm to minimize the cost of the production cycle. The entire method has low solution difficulty, high solution speed and solution quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to make the contents of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings.

[0060] Figure 1 is a flow chart of the method of the present invention;

[0061] Figure 2 Schematic diagram of the DSCC optimization process in an embodiment of the present invention;

[0062] Figure 3 Schematic diagram of an individual performing restorative crossover in an embodiment of the present invention. DETAILED DESCRIPTION

[0063] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0064] Example 1

[0065] Reference Figure 1 As shown, the present invention relates to a supply chain configuration optimization method based on a multi-source driven population genetic algorithm, the method comprising the following steps:

[0066] Step 1: This embodiment improves the genetic algorithm (GA) and names it the multi-source driven population genetic algorithm. First, the production cycle consisting of multiple dynamically generated orders is divided into multiple scheduling units. Each scheduling unit contains multiple orders to be optimized, and each order is processed by several nodes composed of different manufacturers from upstream to downstream.

[0067] Step 2: Initialize the population. In one embodiment, for any population, 100 individuals are initialized using real number encoding for its DNA. Each individual can be obtained by randomly selecting a configuration for each order of a scheduling unit. Each order is processed by several nodes composed of different manufacturers from upstream to downstream. Selecting a configuration means selecting different suppliers, processors, product manufacturers, etc. for each order, such as Figure 2 As shown, regarding upstream and downstream nodes, for example, suppliers are upstream nodes of processors and product manufacturers, processors are upstream nodes of product manufacturers, and processors are downstream nodes of suppliers. In this embodiment, the genetic code of an individual is represented by a two-dimensional matrix. Each column of the two-dimensional matrix represents a configuration option for an order. The total configuration options for each order are represented as a row in the two-dimensional matrix. The number of rows in the two-dimensional matrix equals the number of orders.

[0068] It should be noted that the scheduling unit needs to process the currently received but unprocessed orders, and the number of orders in each scheduling unit is inconsistent. For ease of understanding, the scheduling unit can be considered as a collection of orders received in different time periods.

[0069] Step 3: Perform fitness evaluation for each individual in the population. First, calculate the production cost and overtime cost through the individual coding gene and calculation formula, then calculate the cost of a single scheduling unit, and finally calculate the cost of the entire production cycle of the dynamic multi-stage supply chain configuration optimization (Dynamic Supply Chain Configuration, DSCC). For details, refer to Figure 2 The optimization process calculates the cost.

[0070] This embodiment includes three populations: a main population, and first and second auxiliary populations (the first and second auxiliary populations are primarily used to optimize the main population). The entire production cycle cost is used as the fitness function of the main population to implement fitness evaluation for each individual. The production cost is used as the fitness function of the first auxiliary population to implement fitness evaluation for each individual. The overtime cost is used as the fitness function of the second auxiliary population to implement fitness evaluation for each individual.

[0071] For the production cost calculation of a single order, the formula is:

[0072]

[0073] The constraints satisfied are:

[0074]

[0075] where F1 represents the production cost of a single order, nNode is the number of nodes (members of the supply chain network, such as suppliers, processors, product manufacturers), represents the production cost of the order at the i-th node for a demand; x i represents the configuration (option of the node) selected by node i, the configuration selected by node i cannot be the one whose current number of orders executed is greater than or equal to its production capacity (i.e., the maximum number of orders that can be executed simultaneously by the configuration); Demand i is the total demand of the node, whose value is equal to the sum of the demands of all downstream nodes of node i; downNode i represents the set of downstream nodes of node i, s represents the s-th node in the set of downstream nodes of node i; nOption i is the number of configurations of node i, Capacity j is the production capacity of configuration j; Capacity j,used is the production capacity of configuration j that has been occupied.

[0076] For the overtime cost calculation of a single order, the formula is:

[0077]

[0078] The constraints satisfied are:

[0079]

[0080] where in the DSCC model, each order has a time window, representing the range of the latest delivery that the order can tolerate; F2 represents the overtime cost of a single order; TW min and TW max represent the minimum and maximum values of the order time window; LT represents the delivery time of the order, which is the maximum value of the delivery times of all nodes, the delivery time of a node being the time spent by the current node plus the maximum value of the delivery times of all upstream nodes of the node; p represents the penalty coefficient of the delivery time delay within the order time window, C timeout represents the overtime penalty cost for delivery time exceeding the order time window; LTime i is the delivery time of node i, is the time selected by node i, maxUpLTime i is the maximum delivery time of the upstream nodes of node i, upNode i is the upstream node of node i.

[0081] For the cost calculation of a single dispatch unit, the formula is:

[0082]

[0083] Among them, PC m represents the cost of a single scheduling unit, nOrder represents the number of orders in a scheduling unit, then F k,1 、F k,2 They represent the production cost and overtime cost of the kth order in the scheduling unit. The formula shows that the cost of a single scheduling unit is the sum of the production cost and overtime cost of all orders.

[0084] Therefore, the cost calculation formula for the entire DSCC production cycle is:

[0085]

[0086] Where PC represents the total production cycle cost, nScheduling represents the number of scheduling units, and n represents the nth scheduling unit. The formula indicates that the total production cycle cost is the sum of the costs of all scheduling units.

[0087] Step 4: Select the main population, the first auxiliary population, and the second auxiliary population based on their fitness evaluation results. Individuals with better fitness are selected from these three populations as crossover parents. The population size ratio of the main population, the first auxiliary population, and the second auxiliary population is 3:1:1. Parent selection is performed on each population based on their respective fitness functions.

[0088] Step 5: In each iteration, perform a restorative crossover operation on the three populations together.

[0089] First, determine whether a node of the parent individual is feasible. If the nodes of the parent individuals of the three populations are feasible, then use the whole column crossover method for the feasible nodes of the parent individual, exchange the configuration selection of all orders of the parent individual on the feasible node and obtain two child individuals. If the two child individuals are still feasible on the node, then the two child individuals replace the parent individual and enter the new population. Figure 3 As shown, node 2 (i.e., the second column of the two-dimensional matrix) contains configuration 1 and configuration 2. Assuming that the production capacity of these two configurations is 2, when the whole column crossover method is used, the offspring individuals at this node do not exceed the production capacity and are both feasible.

[0090] If the nodes of the parent individuals of the three populations are infeasible, the child individuals generated by the entire column crossover are still infeasible. For these infeasible child individuals, a random value is first generated and compared with the preset repair probability. If the random value is less than or equal to the preset repair probability, no action is taken; if the random value is greater than the preset repair probability, the infeasible solution repair operation is executed. Because processing all illegal configurations would result in excessive computational complexity, this embodiment selects some illegal configurations for processing, thereby improving computational efficiency and saving computational time.

[0091] It should be noted that individual feasibility means that the occupied production capacity of the individual configuration is less than or equal to the actual production capacity of the configuration, otherwise it means that the individual is infeasible; the illegal configuration is the configuration selection in which the occupied production capacity of the configuration is greater than the actual production capacity of the configuration.

[0092] For an infeasible solution, the infeasible solution repair operation is specifically as follows: repair all illegal configurations in the current illegal configuration set in turn: for a illegal configuration, randomly replace other configuration options for the node that selects the illegal configuration until the number of orders currently executed by the configuration is less than or equal to its production capacity, and then detect whether the individual has generated a new illegal configuration during the repair process. If so, the new illegal configuration is added to the current illegal configuration set; if not, the number of the current illegal configuration set is reduced by one; execute in a loop until the current illegal configuration set is empty.

[0093] In step 6, during each iteration, mutation is performed on the three populations. First, the genes of the individuals in the new population are mutated according to the set mutation rate. The configuration of the mutated genes is changed. The mutated individuals replace the original individuals in the new population, while the unmutated individuals enter the new population directly. Finally, the mutated new population replaces the original population, and only the new master population is selected from the three populations.

[0094] Step 7: Re-evaluate the fitness of each individual in the new main population.

[0095] Step 8: Determine whether the new main population meets the termination conditions based on the fitness evaluation results. If so, output the selected configuration corresponding to the optimal individual in the current scheduling unit; if not, loop through steps 3 to 7 until the termination conditions are met.

[0096] It's worth noting that in step 8 of this embodiment, only the main population needs to be determined to have met the termination criteria, without the need for two auxiliary populations. This is because this embodiment primarily uses the two auxiliary populations to optimize the main population, allowing it to reach the optimal solution as quickly as possible. Furthermore, based on the new fitness evaluation results, the termination criteria are determined. If they are met, an elite population is selected from the final solution set.

[0097] Step 9: Determine whether all scheduling units have been optimized. If so, return the optimal solution of each scheduling unit (the selected configuration corresponding to the optimal individual) as the final optimization result of the DSCC problem; if not, loop through steps 1 to 8 until the optimization of all scheduling units is completed.

[0098] In step 9, when optimizing all scheduling units, it is determined whether it is the first scheduling unit. If not, the current scheduling unit needs to join the elite population of the previous scheduling unit.

[0099] The experimental comparison is as follows:

[0100] To illustrate the effectiveness of the method of this application, this application sets up the following examples for experimental data comparison: Algorithm 1 is the method provided by this application (MSDGA). Algorithm 2, Algorithm 3, and Algorithm 4 are the algorithms published in the 201713th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNC-FSKD) conference paper, the journal Advances in engineering software, and Cluster Computing, respectively, published by IEEE: Algorithm 2 is the ImGA algorithm in the paper "Mathematical modeling and system design of timetabling problem based on improved GA"; Algorithm 3 is the GWO algorithm in the paper "Grey wolf optimizer"; Algorithm 4 is the HGA algorithm in the paper "Exploration of intelligent computing based on improved hybrid genetic algorithm".

[0101] Three examples of different scales were used as experimental data, and algorithms 1, 2, 3, and 4 were used to solve them. These three examples contain different member sizes and configuration scales, which can well verify the performance of the solving algorithm. The specific contents of the three examples are shown in Table 1:

[0102] Table 1

[0103]

[0104] The three supply chain examples above represent small, medium and large scale examples respectively. Each example is executed for 20 seconds per scheduling unit, 40 seconds per scheduling unit and 60 seconds per scheduling unit respectively. In order to ensure the fairness of the experiment, each example is executed 25 times for each algorithm respectively.

[0105] The parameters of the algorithms one, two, three and four are set by the control variable method, and the specific parameters are shown in Table 2.

[0106] Table 2

[0107] algorithm Parameter settings MSDGA The population size is 100; the crossover rate is 0.9 and the mutation rate is 0.1 ImGA The population size is 100; the crossover rate is adaptive, and the mutation rate is adaptive. GWO The population size is 100 HGA The population size is 100; the crossover rate is 0.9, the mutation rate is 0.1, and the chaos parameter is 4

[0108] In order to compare each algorithm, the solutions are all legal solutions that meet the constraints, and the optimal solution in the average case is shown. Specifically, the smaller the cost is, the better the optimal solution is, that is, the smaller the numerical value of the optimal solution is.

[0109] The experimental results of the four algorithms on the examples are shown in Table 3.

[0110] Table 3

[0111]

[0112]

[0113] The experimental results show that in all examples, the method of the present application is better than other comparison methods. This shows that the method of the present application has better solving ability when solving the DSCC problem. At the same time, as the size of the example increases, the difficulty of solving also increases, but the advantage of the method of the present application does not diminish, which proves that the method of the present application has stronger solving ability in solving large-scale problems and higher optimization efficiency in the solution space.

[0114] Embodiment two

[0115] The embodiment provides a supply chain configuration optimization system based on a multi-source driven population genetic algorithm, comprising:

[0116] The division module is configured to divide the production cycle of all products into a plurality of scheduling units, each scheduling unit corresponding to a plurality of orders, and each order being processed through a plurality of nodes composed of different manufacturers from upstream to downstream;

[0117] The initialization module is configured to initialize three populations with different fitness functions, each population including a plurality of individuals, and each individual being obtained by randomly selecting a configuration for each order of a scheduling unit;

[0118] The first fitness evaluation module is configured to evaluate the fitness of each individual in the three populations according to the respective fitness functions;

[0119] Selection module: used to perform selection operations on the three populations according to their respective fitness evaluation results;

[0120] Crossover module: used to perform crossover operation on three populations together to obtain the population after crossover operation;

[0121] Mutation module: used to perform mutation operation on the population after the crossover operation and select only the new main population from the three populations;

[0122] The second fitness evaluation module: used to re-evaluate the fitness of each individual in the new population;

[0123] Judgment and output module: used to judge whether the termination condition of the iteration is met based on the new fitness evaluation result. If so, the order configuration of the scheduling unit corresponding to each individual in the current new population is output; if not, the first fitness evaluation module to the second fitness evaluation module is executed cyclically until the termination condition of the iteration is met.

[0124] Example 3

[0125] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the supply chain configuration optimization method based on a multi-source driven population genetic algorithm described in Example 1 are implemented.

[0126] Example 4

[0127] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the supply chain configuration optimization method based on a multi-source driven population genetic algorithm described in the first embodiment are implemented.

[0128] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiment of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal translation scripting language JavaScript, etc.

[0129] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0130] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0131] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0132] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0133] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A supply chain configuration optimization method based on a multi-source driven population genetic algorithm, characterized by: include: Step S1: Divide the production cycle consisting of multiple dynamically generated orders into multiple scheduling units. Each scheduling unit contains multiple orders to be optimized. Each order is processed by several nodes composed of different manufacturers from upstream to downstream. Step S2: For each scheduling unit: first initialize three populations with different fitness functions, each population includes several individuals, and each individual is obtained by randomly selecting a configuration for each order of a scheduling unit; Step S3: Evaluate the fitness of each individual in the three populations according to their respective fitness functions; Step S4: performing selection operations on the three populations according to their respective fitness evaluation results; Step S5: performing a crossover operation on the three populations together to obtain a population after the crossover operation; Step S6: performing a mutation operation on the population after the crossover operation, and selecting only the new main population from the three populations; Step S7: re-evaluate the fitness of each individual in the new main population; Step S8: Determine whether the new main population meets the termination condition based on the fitness evaluation result. If so, output the selection configuration corresponding to the optimal individual in the current scheduling unit; if not, loop through steps S3 to S7 until the termination condition is met. Step S9: Execute steps S1 to S8 in a loop until all scheduling units have completed the optimization, and output the selected configuration corresponding to the optimal individual in each scheduling unit.

2. The supply chain configuration optimization method based on a multi-source driven population genetic algorithm according to claim 1 is characterized by: The genetic code of the individual is represented by a two-dimensional matrix, each column in the two-dimensional matrix represents a configuration choice of an order, and the total configuration choice of each order is represented as a row in the two-dimensional matrix. The number of rows in the two-dimensional matrix is ​​equal to the number of orders.

3. The supply chain configuration optimization method based on a multi-source driven population genetic algorithm according to claim 1 is characterized by: The method of performing fitness evaluation on each individual in the three populations according to their respective fitness functions in step S3 includes: Calculate the production cost and overtime cost of a single order, calculate the cost of a single scheduling unit based on the production cost and overtime cost, calculate the entire production cycle cost of the product based on the cost of a single scheduling unit, and use the entire production cycle cost of the product as the fitness function of the main population to achieve fitness evaluation of each individual; The production cost is used as the fitness function of the first auxiliary population to evaluate the fitness of each individual; The overtime cost is used as the fitness function of the second auxiliary population to evaluate the fitness of each individual; Among them, the main population, the first and second auxiliary populations constitute three populations.

4. The supply chain configuration optimization method based on a multi-source driven population genetic algorithm according to claim 3 is characterized by: The formula for calculating the production cost of a single order is: The constraints are: Among them, F1 represents the production cost of a single order, nNode is the number of nodes, represents the production cost of a demand for the order in the i-th node, x i Demand represents the configuration choice of node i, i is the total demand of the i-th node, which is equal to the sum of the demand of all downstream nodes of node i; downNode i Represents the set of downstream nodes of node i, s represents the sth node in the set of downstream nodes of node i, nOption i Indicates the number of configurations for point i, Capacity j,used Indicates the production capacity that has been occupied by configuration j, Capacity j represents the production capacity of configuration j.

5. The supply chain configuration optimization method based on multi-source driven population genetic algorithm according to claim 3 is characterized by: The formula for calculating the overtime cost of a single order is: The constraints are: Each order has a time window, which represents the latest delivery range that the order can tolerate; F2 represents the overtime cost of a single order; TW min and TW max Indicates the minimum and maximum values ​​of the order time window; LT represents the delivery time of the order, that is, the maximum value of the delivery time of all nodes. The delivery time of a node is the maximum value of the time spent by the current node plus the delivery time of all upstream nodes of the node; LTime i Indicates the delivery time of node i; maxUpLTime i is the maximum delivery time of the upstream node of node i; upNode i is the upstream node of node i; nNode represents the number of nodes; p represents the penalty coefficient for delivery time delay within the order time window, C timeout represents the penalty cost for delivery time exceeding the order time window.

6. The supply chain configuration optimization method based on a multi-source driven population genetic algorithm according to claim 3 is characterized by: The cost of a single scheduling unit is calculated based on the production cost and overtime cost, and the formula is: Among them, PC m represents the cost of a single scheduling unit, nOrder represents the number of orders in a scheduling unit, F k,1 、F k,2 denote the production cost and overtime cost of the kth order in the scheduling unit respectively.

7. The supply chain configuration optimization method based on a multi-source driven population genetic algorithm according to claim 6 is characterized by: The formula for calculating the entire production cycle cost of a product based on the cost of a single scheduling unit is: Among them, PC represents the cost of the entire production cycle, nScheduling represents the number of scheduling units, and m represents the mth scheduling unit.

8. The supply chain configuration optimization method based on a multi-source driven population genetic algorithm according to claim 1 is characterized by: The method of performing the crossover operation on the three populations at the step S5 includes: Determine whether the nodes of the parent individuals of the three populations are feasible: If the nodes of the parent individuals of the three populations are feasible, the whole column crossover method is used for the feasible nodes of the parent individuals to exchange the configuration selection of all orders of the parent individuals on the feasible nodes to obtain two child individuals; if the two child individuals are still feasible on the node, the two child individuals replace the parent individuals and enter the new population; If the nodes of the parent individuals of the three populations are infeasible, the child individuals generated by the entire column crossover are still infeasible. For the infeasible child individuals, a random value is first generated to compare with the preset repair probability. When the random value is less than or equal to the preset repair probability, the illegal configuration is not processed; when the random value is greater than the preset repair probability, the infeasible solution repair operation is performed on the illegal configuration; Among them, individual feasibility means that the occupied production capacity of the individual configuration is less than or equal to the actual production capacity of the configuration, otherwise it means that the individual is infeasible; the illegal configuration is the configuration selection in which the occupied production capacity of the configuration is greater than the actual production capacity of the configuration.

9. The supply chain configuration optimization method based on multi-source driven population genetic algorithm according to claim 8 is characterized by: The infeasible solution repair operation is specifically as follows: Repair all illegal configurations in the current illegal configuration set in sequence: For a illegal configuration, randomly replace other configuration options for the node that selected the illegal configuration until the currently occupied production capacity of the configuration is less than or equal to its production capacity. Then check whether the individual has generated a new illegal configuration during the repair process. If so, add the new illegal configuration to the current illegal configuration set; if not, reduce the number of the current illegal configuration set by one; execute the cycle until the current illegal configuration set is empty.

10. A supply chain configuration optimization system based on a multi-source driven population genetic algorithm, characterized by: include: Division module: divides the production cycle consisting of multiple dynamically generated orders into multiple scheduling units. Each scheduling unit contains multiple orders to be optimized. Each order is obtained by selecting and configuring several nodes composed of different manufacturers from upstream to downstream. Initialization module: For each scheduling unit: First, initialize three populations with different fitness functions. Each population includes several individuals, and each individual is obtained by randomly selecting a configuration for each order of a scheduling unit; The first fitness evaluation module: evaluates the fitness of each individual in the three populations according to their respective fitness functions; Selection module: performs selection operations on the three populations according to their respective fitness evaluation results; Crossover module: Perform crossover operation on the three populations together to obtain the population after crossover operation; Mutation module: performs mutation operation on the population after the crossover operation and selects only the new main population from the three populations; The second fitness evaluation module: re-evaluates the fitness of each individual in the new main population; Judgment module: Determine whether the new main population meets the termination condition of iteration based on the fitness evaluation results. If so, output the selection configuration corresponding to the optimal individual in the current scheduling unit; if not, loop through the first fitness evaluation module to the second fitness evaluation module until the termination condition is met. Optimization module: Circularly execute the partitioning module to the judgment module until all scheduling units are optimized, and output the selected configuration corresponding to the optimal individual in each scheduling unit.

Images