Outshore manufacturing base-oriented outsourcing production and logistics combined scheduling method

Abstract

The invention discloses an outsourcing production and logistics combined scheduling method for an overseas manufacturing base. The method comprises the following steps: S1) acquiring data of a customer order, an outsourcing factory, an overseas manufacturing base workshop and a delivery ship; s2) establishing constraint conditions according to the factory, workshop and marine transportation data, wherein the constraint conditions comprise a processing workshop constraint, a parallel machine constraint and a task order time window constraint; s3) an outsourcing production and logistics combined scheduling model of the overseas manufacturing base is established according to the constraint conditions, and the target of the combined scheduling model is to minimize the maximum processing completion time and the total cost; and S4) designing a two-stage joint optimization strategy to combine with an improved decomposition multi-objective optimization algorithm of reinforcement learning to solve the production and logistics joint scheduling model in the distributed manufacturing environment, and obtaining an optimal scheme of production distribution scheduling. According to the constraint condition of the actual production workshop, the optimal joint scheduling scheme is solved, and the overall cost of the supply chain can be reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of special vehicle supply chain scheduling, including job shop scheduling and overseas logistics distribution scheduling, and in particular to a joint scheduling method for outsourcing production and logistics of overseas manufacturing bases. BACKGROUND

[0002] In a global manufacturing environment, manufacturing enterprises in different regions or countries face the challenge of coordination and coordination to achieve cross-regional or cross-country production shop scheduling to optimize the efficiency and performance of the entire supply chain. This becomes the core task of joint scheduling of overseas manufacturing enterprises. As mentioned earlier, due to the constraints of local industrial infrastructure, raw materials and parts required for product manufacturing often need to be introduced from outside. In this case, accurate scheduling between transportation and production becomes particularly important. Only through effective cross-regional or cross-country production shop scheduling, the optimization of the entire supply chain can be achieved to improve efficiency and performance. How to properly arrange the time between raw material transportation and shop product production, and plan the relationship between production scheduling and logistics scheduling, becomes a problem to be solved

[0003] Global manufacturing environment also brings a series of challenges to overseas production, including logistics, processing time and other issues, which need to be solved through efficient job shop scheduling. In actual production scenarios, job shop scheduling faces complex dynamic uncertainty systems, influenced by external environment and internal multiple factors. This uncertainty includes changes in customer demand, which may lead to adjustments in product parameters and delivery period; equipment failure, which may cause production delay; fluctuations in employee productivity, etc. These uncertain factors directly or indirectly affect the execution of job shop scheduling plan and have an impact on the normal production of workpieces. In order to cope with these challenges, flexible job shop scheduling strategies and timely response measures are needed to ensure the stability and efficiency of the production process.

[0004] The joint scheduling problem of outsourcing production and logistics pays more attention to the coordination between processing shop and logistics distribution than traditional job shop scheduling problem. Considering the geographical location of overseas manufacturing bases, their logistics distribution usually adopts sea transportation. At the same time, in order to reduce the overall scheduling cost of the supply chain, the selection of processing plants and distribution ships will also have a certain impact on the total cost, therefore, it is necessary to fully coordinate each stage of the supply chain to achieve the optimal overall benefit. SUMMARY

[0005] To solve the above technical problems, one technical solution adopted by the present application is as follows:

[0006] A joint scheduling method for outsourcing production and logistics of overseas manufacturing bases, comprising the following steps:

[0007] S1): Obtain data of customer orders, outsourcing factories, overseas manufacturing bases, and distribution ships;

[0008] S2): Establish constraint conditions according to factory, workshop, and sea transportation data, wherein the constraint conditions are processing workshop constraints, parallel machine constraints, and task order time window constraints;

[0009] S3): Establish an outsourcing production and logistics joint scheduling model of the overseas manufacturing base according to the constraint conditions, wherein the target of the joint scheduling model is to minimize the maximum processing completion time and minimize the total cost;

[0010] S4): Design an improved decomposition multi-objective optimization algorithm combining reinforcement learning to solve the production and logistics joint scheduling model in the distributed manufacturing environment to obtain an optimal scheme of production and distribution scheduling; specifically including the following steps:

[0011] S41): Initialize the population X, population weight vector λ, Q value table, and related parameters;

[0012] S42): The agent selects the neighborhood size T according to the current environment using an ε-greedy strategy;

[0013] S43): For each sub-problem x i , randomly select two sub-problem solutions from the neighborhood of its corresponding weight vector λ as the parents x1 and x2;

[0014] S44): Generate child sub-problem solutions x new1 , x new2 , x new3 , and x new4 using the POX and RPX operators for the parents x1 and x2;

[0015] S45): Calculate the fitness of the new individual x new1 , compare the aggregate function value g(x|λ, Z) of each neighborhood vector and the neighborhood corresponding individual x k in turn, and if x new1 is the optimal solution, replace the solution x k in the target direction of the sub-problem with x new1 . Similarly, calculate the fitness of the new individuals x new2 , x new3 , and x new4 , and decide whether to replace the neighborhood or not. After the decision is made, the evolution of the single sub-problem ends;

[0016] S46): Jump to step 43 until all sub-problems are completely searched, the population X is updated to the new population X', and the Pareto frontier of the new population X' is calculated;

[0017] S47):Calculate the diversity index DV of the PF of the new population X', update the Q value table according to the DV value size;

[0018] S48):Use the variable neighborhood search operator to locally optimize the sub-problem x i ;

[0019] S49):Update the PF of the current population, store the PF in the elite archive Elite, if the elite archive is full, update it in the form of non-dominated sorting;

[0020] S410):Check whether the termination condition is met. If the condition is met, it means that the iterative search has been completed, and the output result will be output; if the condition is not met, return to step 42.

[0021] Further, the step 1 specifically includes the number of customers, the geographic location of the customers, the demand of the customers for different workpieces, the number of workpieces to be processed, the processing time of each workpiece on different parallel machines, the time window limit of each workpiece, the unit processing cost of the workpiece, the number of sea vessels, the shipping cost of the vessel, the capacity limit of the vessel, the geographic location of the factory, and the number of outsourcing workshops.

[0022] Further, the constraint condition in the step 3 is specifically:

[0023] All workpieces to be processed can be processed only when the raw materials are sufficient for the predetermined batch processing, and all processing equipment can be used at the initial time;

[0024] There is no specific requirement for the processing sequence of different workpieces to be processed, and there is no specific processing sequence requirement for different batches of the same workpiece to be processed;

[0025] Each batch of workpieces can be processed on only one machine at the same time;

[0026] Each machine can process only one workpiece at the same time, without considering machine failure;

[0027] The transportation and adjustment time of the workpieces to be processed between machines is ignored;

[0028] For different processes of the same batch of the same workpiece, there is a priority constraint on the processing sequence;

[0029] Only one kind of raw material is selected for processing for the same workpiece;

[0030] All raw materials can be transported at the initial time, and all transportation equipment can be used at the initial time;

[0031] There is no specific transportation sequence requirement between different raw materials, and there is no specific transportation sequence requirement between different batches of the same raw material;

[0032] Each batch of raw materials can only be transported on one transport device at a time;

[0033] Each transport equipment can only transport one type of raw material at a time, without considering equipment malfunctions.

[0034] Ignoring the downtime of transport equipment and the handling time of raw materials;

[0035] Once the transport equipment returns to the port, the next batch of raw materials can be transported immediately.

[0036] The cost of purchasing outsourced raw materials is greater than the sum of the cost of purchasing and transporting raw materials from overseas.

[0037] The outsourcing factory has sufficient processing equipment. If the process to be processed is chosen to be processed in the outsourcing factory, then the process can be started at the initial moment.

[0038] The transportation time between factories for the same batch of workpieces is ignored.

[0039] Furthermore, step 41 specifically includes:

[0040] S411): Initialize the population X = (x1, x2, ..., x3) using a random initialization strategy. N ), calculate individual x i Objective function value FV i =F(x) i Based on the objective function value of the initialized target population, initialize the reference point Z = (z1, z2, ..., zn). k ),in k is the number of objective functions;

[0041] S412): Uniformly generated weight vector λ = (λ1, λ2, ... λ) k The weight vector of the i-th subproblem. Specifically, λ1=(10 -6 1-10 -6 ), λ N =(1-10) -6 10 -6 );

[0042] S413): The weight vector λ = (λ1, λ2, ..., λ3) is... k ) are sequentially assigned to all individuals X = (x1, x2, ..., x...) in the population. N Based on the initialized weight vector, calculate x for each weight vector. i The neighborhood B(i) = {i1, i2, ..., i N ).

[0043] Further, the step 42 specifically comprises:

[0044] S421) : importing the Q-value table;

[0045] S422) : according to the current state of the agent and the greed coefficient ε, greedily selecting the action corresponding to the Q-value in the Q-value table in the state to obtain the neighborhood size T;

[0046] S423) : searching for new solutions of neighborhood sub-problems in the neighborhood size T in the population to generate a new population;

[0047] S424) : calculating the diversity index DV of the new population PF, and comparing it with the diversity index DV of the old population; -1 making a difference to obtain ΔDV;

[0048] S425) : returning the reward value according to ΔDV;

[0049] S426) : updating the Q-value table and outputting, and the operator ends.

[0050] Further, the step 48 specifically comprises:

[0051] S481) : judging whether the memory matrix reaches the upper limit LP, if yes, jumping to step 482, otherwise, jumping to step 483;

[0052] S482) : summing up each column of the successful memory matrix and the failed memory matrix to obtain the success probability P(i) of each local search strategy LS i , and normalizing to make the sum of the selection probabilities of each local search strategy one;

[0053] S483) : assigning the local search strategy LS i to the sub-problem x i in a roulette manner;

[0054] S484) : generating a new solution x newi according to the assigned local search strategy, comparing the aggregate function values of the new solution x newi and the old solution x i , if x newi is the optimal solution, replacing the solution x i in the objective direction of the sub-problem with x newi ;

[0055] S485) : jumping to step 483 to complete the local search of all sub-problems.

[0056] Beneficial effects:

[0057] The application discloses an outsourcing production and logistics joint scheduling method for overseas manufacturing bases.

[0058] The above description is only a summary of the technical scheme of the application. In order to make the technical means of the application more clearly understood and implemented according to the content of the description, the preferred embodiments of the application are described in detail below with reference to the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 The flow chart of the joint scheduling model;

[0060] Figure 2 The symbol meaning in the model constructed by the application;

[0061] Figure 3 The symbol meaning of the decision variable in the model constructed by the application;

[0062] Figure 4 The process machining resource of the application;

[0063] Figure 5 The optional transport equipment resource of the application;

[0064] Figure 6 The machining equipment selectable by each process of the workpiece of the application;

[0065] Figure 7 The machining time of each example of the application. DETAILED DESCRIPTION

[0066] The preferred embodiments of the application are described in detail below with reference to the drawings, so that the advantages and features of the application can be more easily understood by those skilled in the art, and the protection scope of the application is more clearly and definitely defined. Specific embodiment 1:

[0068] The outsourcing production and logistics joint scheduling method for overseas manufacturing bases comprises the following steps:

[0069] S1): Obtain data on customer orders, outsourced factories, overseas manufacturing base workshops, and delivery ships; Step 1 specifically includes the number of customers, the geographical location of customers, the demand of customers for different workpieces, the number of workpieces to be processed, the processing time of each workpiece on different parallel machines, the time window limit of each workpiece, the processing cost per unit time of workpieces, the number of ocean-going ships, the ship delivery cost, the ship capacity limit, the geographical location of factories, and the number of outsourced factories.

[0070] S2): Establish constraints based on factory, workshop, and shipping data, including processing workshop constraints, parallel machine constraints, and task order time window constraints; in this embodiment, step 2 specifically includes:

[0071] S21): Establish the objective function equation; refer to the appendix for the meaning of the symbols in the formula. Figures 2-3 Table Formula 1: f1 = min(C max ), indicating that the objective function is to minimize the maximum processing time;

[0072] Formula 2: f2 = min(CTotal), which means minimizing the total cost is the objective function;

[0073] Formula 3: C max =max(C i,j,w ), calculate the maximum processing time;

[0074] Formula 4: The total cost of the workpiece processing is calculated as the sum of the total processing costs of all operations for all batches of workpieces.

[0075] Formula 5: Calculate the total cost of transporting raw materials, which is the sum of the transportation costs of all shipments of all raw materials.

[0076] Formula 6: CTotal = CP + CT, calculates the sum of costs for the entire transportation and processing process;

[0077] S22): Set constraints;

[0078] Formula 7: This means that the actual arrival time of the raw materials is equal to the start time of the transportation of the raw materials plus the transportation time;

[0079] Formula 8: E u，o+1 =E u，o +2*T u Constraints on transportation equipment;

[0080] Formula 9:

[0081] Formula 10: Formula 9 and Formula 10 show the relationship between the raw material transportation start time and the transportation equipment transportation start time, when a batch of raw material is transported by the transportation equipment, the raw material transportation start time should be equal to the transportation equipment transportation start time;

[0082] Formula 11: shows that the process completion time is equal to the process start time plus the process time on the processing equipment;

[0083] Formula 12: C i，j-1，w ≤ S i，j，w Constrains the processing sequence constraint between the same batch of the same workpiece, the processing start time of the next process of the same batch of the same workpiece must be after the processing completion time of the previous process;

[0084] Formula 13: Constrains that the same machine can only process on one machine at the same time, and the processing start time of the next process on the machine must be after the processing completion time of the previous process;

[0085] Formula 14: Constrains the relationship between the start processing time of the process on a batch of workpieces and the processing start time of the process on the assigned machine;

[0086] Formula 15: Constrains that the workpiece processing must have enough raw materials;

[0087] Formula 16: shows the definition of weight x i，j，k，w ;

[0088] Formula 17:

[0089] Formula 18: shows the definition of weight ;

[0090] Formula 19: Formula 17 and Formula 19 indicate that the same process on the same batch of the same workpiece can only be processed on one machine;

[0091] Formula 20:

[0092] shows the definition of weight ;

[0093] Formula 21: indicates the weight Only when the raw material is enough for the current batch of workpiece processing and not enough for the next batch of workpiece processing, it is 1, otherwise it is 0, which at the same time restricts the processing of each workpiece to only choose one kind of raw material;

[0094] Formula 22: The definition of weight x s，t，u ;

[0095] Formula 23:

[0096] Formula 24: The definition of weight ;

[0097] Formula 25: Formula 23 and formula 25 indicate that the same batch of the same raw material can only be transported on one transportation device;

[0098] Formula 26: The definition of weight z i，j，w ;

[0099] Formula 27: C i，j-1，w ≥ S i，j，w *z i，j，w Restrict the processing start time of each process of outsourcing processing to the processing completion time of the previous process.

[0100] S3) : according to the constraint condition, the outsourcing production and logistics joint scheduling model of the overseas manufacturing base is established, wherein the target of the joint scheduling model is to minimize the maximum processing completion time and minimize the total cost; The constraint condition in step 3 is specifically:

[0101] All workpieces to be processed can be processed only when the raw material is enough for the predetermined batch processing, and all processing equipment can be used at the initial time;

[0102] There is no specific requirement for the processing order of different workpieces to be processed, and there is no specific processing order requirement for different batches of the same workpiece to be processed;

[0103] Each batch of workpieces can only be processed on one machine at the same time;

[0104] Each machine can only process one workpiece at the same time, without considering machine failure;

[0105] Ignore the transportation and adjustment time of the workpiece to be processed between machines;

[0106] For different processes of the same batch of the same workpiece, there is a priority constraint on the processing order;

[0107] Only one kind of raw material is selected for the processing of the same workpiece;

[0108] All raw materials can be transported at the initial moment, and all transport equipment is available at the initial moment;

[0109] There is no specific transport order requirement between different raw materials, and there is no specific transport order requirement between different batches of the same raw material;

[0110] Each batch of raw materials can only be transported on one transport equipment at the same time;

[0111] Each transport equipment can only transport one kind of raw material at the same time, regardless of transport equipment failure;

[0112] Ignore the parking time of the transport equipment and the carrying time of the raw materials;

[0113] After the transport equipment returns to the port, the next batch of raw material transportation can be carried out immediately;

[0114] The purchase cost of outsourcing raw materials is greater than the sum of the purchase and transportation cost of overseas raw materials;

[0115] The processing equipment of the outsourcing factory is sufficient, and if the processing procedure is selected to be processed in the outsourcing factory, the processing of the procedure can be carried out at the initial moment;

[0116] The transportation time of the same batch of workpieces between factories is ignored.

[0117] S4): Design an improved decomposition multi-objective optimization algorithm based on two-stage joint optimization strategy combined with reinforcement learning to solve the production and logistics joint scheduling model in the distributed manufacturing environment, and obtain the optimal scheme of production and distribution scheduling;

[0118] Of course, in this embodiment, according to the specific characteristics of the outsourcing production and logistics joint scheduling problem facing the overseas manufacturing base, in the workshop production and workshop scheduling problem, two key aspects need to be considered: process sequencing and machine selection; And in the logistics transportation scheduling problem: then need to consider the transport batch sequencing and transport equipment selection. In order to effectively represent the individuals in the population, the present application proposes a two-layer double-level coding form:

[0119] The first level of coding is a Raw Material Sorting Vector (RV) based on the order of raw material transport batches; the second level of coding is a Transport Device Selection Vector (TV) based on the selection of raw material transport equipment; the first level of coding is an Operation Sequence Vector (0V) based on the order of operations, used to determine the sequence of operations; the second level of coding is a Machine Assignment Vector (MV) based on the machine selection of operations, used to determine which machine each operation is performed on.

[0120] Specifically, the following steps are included:

[0121] S41): Initialize the population X, the population weight vector λ, the Q-value table, and related parameters;

[0122] Specifically, step 41 includes:

[0123] S411): Initialize the population X = (x1, x2, ..., x3) using a random initialization strategy. N ), calculate individual x i Objective function value FV i =F(x) i Based on the objective function value of the initialized target population, initialize the reference point Z = (z1, z2, ..., zn). k ),in k is the number of objective functions;

[0124] S412): Uniformly generated weight vector λ = (λ1, λ2, ... λ) k The weight vector of the i-th subproblem. Specifically, λ1=(10 -6 1-10 -6 ), λ N =(1-10) -6 10 -6 );

[0125] S413): The weight vector λ = (λ1, λ2, ..., λ3) is... k ) are sequentially assigned to all individuals X = (x1, x2, ..., x...) in the population. N Based on the initialized weight vector, calculate x for each weight vector. i The neighborhood B(i) = {i1, i2, ..., i} N}

[0126] S42):Agent selects neighborhood size T according to current environment by using ε-greedy strategy;

[0127] The step 42 specifically comprises:

[0128] S421):Import Q-value table;

[0129] S422):According to the current state of the agent and the greed coefficient ε, the action corresponding to the Q value in the Q value table in the state is greedily selected to obtain the neighborhood size T;

[0130] S423):Each sub-problem in the population searches for a new solution of the neighborhood sub-problem in the neighborhood size T to generate a new population;

[0131] S424):Calculate the diversity index DV of the new population PF, and obtain ΔDV by subtracting the diversity index DV-(-1) of the old population;

[0132] S425):Return the reward value according to ΔDV;

[0133] S426):Update the Q-value table and output, and the operator ends.

[0134] S43):For each sub-problem x i , randomly select two sub-problem solutions from the neighborhood of its corresponding weight vector λ as parent x1 and x2;

[0135] S44):Generate child sub-problem solutions x new1 , x new2 , x new3 and x new4 by using POX and RPX operators on the parents x1 and x2;

[0136] S45):Calculate the fitness of the new individual x new1 , and compare the aggregate function value g(x|λ, Z) on each neighborhood vector with the neighborhood corresponding individual x k in turn, if x new1 is the optimal solution, replace the solution x k in the objective direction of the sub-problem with x new1 . Similarly, calculate the fitness of the new individuals x new2 , x new3 and x new4 , and decide whether to replace the neighborhood set or not. After the decision is made, the evolution of the single sub-problem ends;

[0137] S46):Jump to step 43 until all sub-problems are searched, the population X is updated to the new population X', and the Pareto front of the new population X' is calculated;

[0138] S47):Calculate the diversity index DV of the PF of the new population X', update the Q value table according to the DV value size;

[0139] S48):Use the variable neighborhood search operator to locally optimize the sub-problem x i ; the step 48 specifically includes;

[0140] S481):Judge whether the memory matrix reaches the upper limit LP, if yes, jump to step 482, otherwise, jump to step 483;

[0141] S482):Sum each column of the success memory matrix and the failure memory matrix to obtain the success probability P(i) of each local search strategy LS i , and normalize the sum to make the sum of the selection probability of each local search strategy equal to one;

[0142] S483):Distribute the local search strategy LS i to the sub-problem x i in the roulette wheel mode;

[0143] S484):Generate a new solution x newi according to the distributed local search strategy, compare the aggregate function value of the new solution x newi and the old solution x i , if x newi is the optimal solution, replace the solution x i in the objective direction of the sub-problem with x newi ;

[0144] S485):Jump to step 483 to complete the local search of all sub-problems.

[0145] S49):Update the PF of the current population, store the PF into the elite archive Elite, if the elite archive is full, update it in the non-dominated sorting mode;

[0146] S410):Check whether the termination condition is met. If the condition is met, it indicates that the iterative search is completed, and the output result is obtained; if the condition is not met, return to step 42.

[0147] The above only describes the embodiments of the present application, and does not limit the patent scope of the present application, any equivalent structural transformation made by using the content of the specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for joint scheduling of outsourced production and logistics for overseas manufacturing bases, characterized in that, Includes the following steps: S1): Obtain data on customer orders, outsourced factories, overseas manufacturing base workshops, and delivery vessels; S2): Establish constraints based on factory, workshop, and shipping data, including processing workshop constraints, parallel machine constraints, and task order time window constraints. S3): Establish a joint scheduling model for outsourced production and logistics of overseas manufacturing bases based on constraints. The objective of the joint scheduling model is to minimize the maximum processing completion time and minimize the total cost. S4): Design a two-stage joint optimization strategy combined with an improved decomposition multi-objective optimization algorithm based on reinforcement learning to solve the joint scheduling model of production and logistics in a distributed manufacturing environment, and obtain the optimal solution for production and distribution scheduling; specifically including the following steps: S41): Initialize the population X, the population weight vector λ, the Q-value table, and related parameters; S42): The agent selects the neighborhood size T using an ε-greedy strategy based on the current environment; S43): For each subproblem x i From its corresponding weight vector λ i Two solutions to subproblems are randomly selected from the neighborhood of x1 and x2, respectively, as the parent generation. S44): Use the POX and RPX operators to generate solutions to the subproblems of the parent x1 and x2. new1 x new2 x new3 and x new4 ; S45): Calculate the new individual x new1 The fitness is determined sequentially by the neighbor vector and the corresponding individual x in the neighborhood. k Compare the aggregation function values ​​g(x|λ, Z), if x new1 To find the optimal solution, the solution x in the objective direction of the subproblem is... k Replace with x new1 Similarly, calculate the new individual x. new2 x new3 and X new4 The fitness of the subproblem is determined by whether to replace it with a neighboring set decision. After the decision is made, the evolution of a single subproblem ends. S46): Jump to step 43 until all subproblems have been searched, update population X to the new population X', and calculate the Pareto front of the new population X'. S47): Calculate the diversity index DV of the PF of the new population X', and update the Q value table according to the DV value. S48): Using the variable neighborhood search operator to solve the subproblem x i Perform local optimization; S49): Update the PF of the current population and store the PF in the Elite archive. If the Elite archive is full, update it using a non-dominant sorting method. S410): Check if the termination condition is met. If the condition is met, the iterative search has been completed and the result will be output; if the condition is not met, return to step 42.

2. The method for joint scheduling of outsourced production and logistics for overseas manufacturing bases according to claim 1, characterized in that: Step 1 specifically includes the number of customers, the geographical location of customers, the demand of customers for different workpieces, the number of workpieces to be processed, the processing time of each workpiece on different parallel machines, the time window limit of each workpiece, the unit time processing cost of workpieces, the number of ocean-going vessels, the vessel delivery cost, the vessel capacity limit, the geographical location of the factory, and the number of outsourced factories.

3. The method for joint scheduling of outsourced production and logistics for overseas manufacturing bases according to claim 1, characterized in that: The specific constraints in step 3 are as follows: All workpieces to be processed can only be processed when there is enough raw material for the predetermined batch processing; all processing equipment can be used from the initial moment. There are no specific requirements for the processing sequence of different workpieces, and there are no specific requirements for the processing sequence of different batches of the same workpiece. Each batch of workpieces can only be processed on one machine at a time; Each machine can only process one workpiece at a time, without considering machine malfunctions; Ignore the time spent transporting and adjusting the workpiece between machines; For different processes within the same batch of the same workpiece, there are priority constraints on the processing sequence; Only one type of raw material is selected for processing the same workpiece; All raw materials can be transported from the outset, and all transport equipment is available from the outset. There are no specific transportation sequence requirements between different raw materials, and there are no specific transportation sequence requirements between different batches of the same raw material. Each batch of raw materials can only be transported on one transport device at a time; Each transport equipment can only transport one type of raw material at a time, without considering equipment malfunctions. Ignoring the downtime of transport equipment and the handling time of raw materials; Once the transport equipment returns to the port, the next batch of raw materials can be transported immediately. The cost of purchasing outsourced raw materials is greater than the sum of the cost of purchasing and transporting raw materials from overseas. The outsourcing factory has sufficient processing equipment. If the process to be processed is chosen to be processed in the outsourcing factory, then the process can be started at the initial moment. The transportation time between factories for the same batch of workpieces is ignored.

4. The method for joint scheduling of outsourced production and logistics for overseas manufacturing bases according to claim 1, characterized in that: Step 41 specifically includes: S411): Initialize the population X = (x1, x2, ..., x3) using a random initialization strategy. N ), calculate individual x i Objective function value FV i =F(x) i Based on the objective function value of the initialized target population, initialize the reference point Z = (z1, z2, ..., zn). k ),in k is the number of objective functions; S412): Uniformly generated weight vector λ = (λ1, λ2, ..., λ3) k The weight vector of the i-th subproblem. Specifically, λ1=(10 -6 1-10 -6 ), λ N =(1-10) -6 10 -6 ); S413): The weight vector λ = (λ1, λ2, ..., λ3) is... k ) are sequentially assigned to all individuals X = (x1, x2, ..., x...) in the population. N Based on the initialized weight vector, calculate x for each weight vector. i The neighborhood B{i} = {i1, i2, ..., i N } 5. The method for joint scheduling of outsourced production and logistics for overseas manufacturing bases according to claim 1, characterized in that: Step 42 specifically includes: S421): Import the Q-value table; S422): Based on the agent's current state and the greedy coefficient ε, greedily select the action corresponding to the Q value in the Q value table for that state to obtain the neighborhood size T; S423): Each subproblem in the population searches for new solutions to its neighborhood subproblems within a neighborhood size T to generate a new population; S424): Calculate the diversity index DV of the new population PF and compare it with the diversity index DV of the old population. -i By subtracting the values, we obtain ΔDV; S425): Returns the reward value based on ΔDV; S426): Update the Q-value table and output it. The operator ends.

6. The method for joint scheduling of outsourced production and logistics for overseas manufacturing bases according to claim 1, characterized in that: Step 48 specifically includes: S481): Determine whether the memory matrix has reached the upper limit LP. If so, jump to step 482; otherwise, jump to step 483. S482): Summing each column of the success memory matrix and the failure memory matrix separately yields the LS for each local search strategy. i The success probability P(i) is normalized so that the sum of the selection probabilities of each local search strategy is one; S483): Distribute the subproblem x using a roulette wheel method. i Local search strategy LS i ; S484): Generate a new solution x according to the assigned local search strategy. newi Compare the new solution x newi and the old solution x i The aggregate function value, if x newi To find the optimal solution, the solution x in the objective direction of the subproblem is... i Replace with x newi ; S485): Jump to step 483 to complete the local search for all subproblems.

Images