Mixed flow manufacturing workshop real-time scheduling method based on multi-agent D3QN
Through the multi-agent D3QN method and EfficientNet network, the Petri network model is built, which solves the problem of multi-target workshop scheduling in the existing technology, realizes efficient production scheduling and resource utilization, and improves production efficiency and economic benefits.
Patent Information
- Application Number
- CN202410934984.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-12
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-07-12
AI Technical Summary
现有技术难以同时优化多个目标的车间调度问题,尤其是在大规模制造车间中,容易陷入局部最优解且求解时间长,且现有基于D3QN的方法对多目标优化效果较差。
Using the real-time scheduling method of hybrid flow manufacturing workshop based on multi-agent D3QN, the MPN model of Petri network is constructed, and multiple agents are initialized. Each agent has different scheduling targets. The D3QN model is trained using the EfficientNet neural network to generate different scheduling targets. oriented production scheduling strategy.
Generate production scheduling strategies of mixed flow manufacturing workshops with different goal orientations during one scheduling calculation, which improves production efficiency and economic benefits, optimizes production scheduling and resource utilization, significantly reduces the amount of calculation and parameters, and improves the handling of large-scale scheduling problems performance and training speed.
Smart Images

Figure CN119067340A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of deep reinforcement learning and workshop intelligent scheduling, and more specifically, to a real-time scheduling method for a mixed-flow manufacturing workshop based on a multi-agent D3QN (Double Dueling Deep Q-Network). Background Art
[0002] As a combinatorial problem with a wide range of engineering backgrounds, the study of the theory and solution methods of the job shop scheduling problem is of great significance in guiding the production of manufacturing enterprises. In the past, the solution algorithms for this type of problem mainly include heuristic scheduling algorithms and meta-heuristic scheduling algorithms, but they have problems such as time-consuming solution process and poor applicability of algorithm scenarios.
[0003] With the introduction of deep reinforcement learning theory and its successful application in game decision-making, combinatorial optimization and other fields, new ideas and methods have been provided for solving workshop job scheduling problems. Deep reinforcement learning combined with scheduling rules can make up for the shortcomings of traditional scheduling methods in insufficient application of historical data and obtain scheduling solutions that meet actual production needs. Reinforcement learning is a process in which an intelligent agent obtains the maximum reward through continuous interaction with the environment. During the learning process, the intelligent agent will output an action based on the current state of the environment. This action will affect the environment and change the state of the environment. At the same time, the reward for the action of the intelligent agent's decision will be fed back. The intelligent agent will optimize its subsequent choice of action based on the reward obtained, and complete the strategy optimization of the intelligent agent in the process of continuous trial and error.
[0004] Although many scholars have achieved good results in workshop scheduling using deep reinforcement learning, it is concluded that in the past, solutions to workshop scheduling problems, whether using approximate computing methods or deep reinforcement learning methods, have mostly focused on a single scheduling goal, such as reducing the maximum completion time. However, since different departments of an enterprise have different production requirements, the production department's goal is to minimize product costs and complete tasks in the shortest time, the sales department hopes that the production department can deliver on time, and the company's senior management hopes to reduce resource waste, etc., workshop scheduling problems often have multiple goals that need to be optimized simultaneously.
[0005] The existing patent document discloses a real-time scheduling method for intelligent workshops with workpiece flow based on D3QN reinforcement learning. The method first collects the operation data of the workshop and processes it into Markov decision chain data, trains an online Q network and a target Q network of D3QN, and uses the trained online Q network as a knowledge network to match the real-time status of the workshop, select the best production schedule, and repeat the process until the production is completed. However, this scheme is designed to overcome the problems of traditional algorithms that are prone to falling into local optimal solutions and long solution time when facing large-scale manufacturing workshops. It still belongs to the solution of a single scheduling target, and the optimization effect for multiple scheduling targets is poor. Summary of the invention
[0006] In order to overcome the defect that the above-mentioned prior art is difficult to handle the workshop scheduling problem in which multiple objectives need to be optimized simultaneously, the present invention provides a real-time scheduling method for mixed-flow manufacturing workshops based on multi-agent D3QN, which can obtain mixed-flow manufacturing workshop production scheduling strategies oriented to different objectives in a single scheduling calculation process, thereby improving the overall production efficiency and economic benefits.
[0007] In order to solve the above technical problems, the technical solution of the present invention is as follows:
[0008] A real-time scheduling method for a mixed-flow manufacturing workshop based on a multi-agent D3QN comprises the following steps:
[0009] S1: Constructing the MPN (Manufacturing Petri Net) model based on Petri net according to the operation relationship between workpieces and processing equipment in mixed-flow manufacturing workshop;
[0010] S2: Obtain the operation data of the orders to be executed in the mixed-flow manufacturing workshop and input them into the MPN model to obtain the initialized MPN model. After the initialization, a production scenario for the mixed-flow manufacturing workshop can be constructed to facilitate the subsequent introduction of multi-agents; the operation data includes: the number and type of workpieces of the orders to be executed, and the status information of the processing equipment for each workpiece during the processing;
[0011] S3: Perform random scheduling simulation based on the initialized MPN model to obtain simulation data; construct Markov decision chain data according to the simulation data and operation data and load it into a preset sample data pool;
[0012] S4: Initialize multiple agents and build a D3QN model for each agent; each agent has a different scheduling target;
[0013] The D3QN model includes two EfficientNet neural networks with the same structure, which serve as the online evaluation network Q and the target value network Q of the D3QN model respectively. - ;
[0014] S5: Randomly extract several Markov decision chain data from the sample data pool and train the D3QN model of each agent to obtain the trained D3QN model of each agent;
[0015] S6: Obtain the real-time status of the mixed-flow manufacturing workshop and input it into the MPN model to obtain the real-time MPN model; call the D3QN model trained by each intelligent agent to perform scheduling calculations on the real-time MPN model, and obtain several workshop production scheduling strategies guided by different scheduling objectives, and schedule production according to the workshop production scheduling strategies.
[0016] Preferably, the step S1 comprises:
[0017] The multiple processing equipment running in parallel in each process in the mixed-flow manufacturing workshop is abstracted into a library, indicating that the workpiece can be transferred to the processing equipment when processing in a certain process; different library places are connected by an intermediate transition path, indicating the transfer of the workpiece between different processes, that is, a path for the workpiece to be transferred to the next processing equipment after processing on one processing equipment is completed;
[0018] The expression of the entire MPN model is as follows:
[0019] MPN= <Token,Place.Transition>
[0020] Token={Token1, Token2,…,Token i}
[0021] Place={Place1, Place2,…, Place j}
[0022] Transition={Transition1, Transition2,…,Transition h}
[0023] Among them, Token is the expression set of the artifact in the MPN model. i is the expression of the i-th workpiece; Place is the expression set of the processing equipment in the MPN model, Place j is the expression of the jth processing equipment; Transition is the expression set of the transition path in the MPN model, Transition h is the hth transition path.
[0024] Preferably, the expression of the i-th workpiece in the MPN model is as follows:
[0025]
[0026] Among them, Token i represents the i-th workpiece, n is the number of workpieces; ID1 is the number of the i-th workpiece, satisfying ID1∈N, N is a natural number set; C is the workpiece type, satisfying C∈{Color1, Color2,…, Color c}, Color1~Color c They are the 1st to cth workpiece types respectively; is the processing time of the i-th workpiece on the j-th processing equipment, m is the number of processing equipment; P is the number of the processing equipment where the i-th workpiece is currently located, satisfying P∈N; T is the remaining processing time of the i-th workpiece in the current equipment, satisfying T∈R, R is a real number set;
[0027] The expression of the jth processing equipment in the MPN model is as follows:
[0028]
[0029] Among them, Place j represents the jth processing equipment; ID2 is the number of the jth processing equipment, satisfying ID2∈N; Capacity is the number of workpieces that the processing equipment can accommodate, is the workpiece existing in the jth processing equipment;
[0030] The expression of the hth transition path in the MPN model is as follows:
[0031] Transition h =<ID3, inPlace, outPlace>
[0032] Among them, ID3 is the number of the transition, satisfying ID3∈N, inPlace is the number of the current processing equipment, satisfying inPlace∈N; outPlace is the number of the next processing equipment, satisfying outPlace∈N.
[0033] Preferably, in step S2, the workpiece quantity, workpiece number and workpiece type of each workpiece, as well as the processing equipment used by each workpiece during the processing and the remaining processing time of the workpiece are extracted from the job data, and all the extracted information is loaded into the MPN model for initialization to obtain the initialized MPN model.
[0034] Preferably, in step S3, a random scheduling strategy is used to simulate the whole process from the start to the end of all workpiece processing of the to-be-executed order, to generate batches of simulation data, and to construct a workpiece scheduling time matrix ST (Schedule Time Matrix), a workpiece position transfer identification matrix MM (Marking Matrix) and a workpiece in-equipment delay matrix MT (Machining Time Matrix) in combination with the operation data;
[0035] According to the three matrices obtained, the workshop status s of the three channels is constructed i , the workshop status s i 、s i Next selected action a k 、Action a k The corresponding reward r, and the next workshop state s i+1 Jointly construct Markov decision chain data i , a k , r, s i+1 > and load the preset sample data pool.
[0036] Preferably, the workshop status s i The expression is as follows:
[0037] s i =<ST n×m , M.M. n×m , M.T. n×m >
[0038] Among them, n×m is the size of the matrix, n is the number of workpieces in the MPN model, and m is the number of processing equipment in the MPN model;
[0039] Get the MPN model in the current workshop status i The following set of actions can be triggered Job(s) i ), where action refers to the process of transferring a specific workpiece from a certain process equipment to the next process equipment, action a k The expression is as follows:
[0040] a k =<ID1, inPlace, outPlace, Time>
[0041] Where Time is the time when the action is triggered, satisfying Time∈R;
[0042] According to the workshop status i Multiple actions that can be triggered by the workpiece constitute a set Job(s) i ), action set Job(s i ) is expressed as follows:
[0043] Job(s i ) = {a1, a2, ..., a k}
[0044] The action set Job(s i ) is used to indicate the workshop status s i The workpiece can choose the direction of transfer. When the actual production transfer is completed, one action will be selected from many actions to be executed. After the execution is completed, the workshop status is s i Updated to i+1 , action set Job(s i ) is also updated to Job(s i+1 );
[0045] According to the degree of fit between the result t of random scheduling simulation and the preset scheduling target tar, the workshop state s is assigned i Next, perform action a k Entering a new state i+1 The corresponding reward r;
[0046] The mixed-flow manufacturing workshop starts from the order placement state s0 to the completion state s n End, and form Markov decision chain data one by one according to the order of n scheduling and production scheduling i , a k , r, s i+1 >.
[0047] Preferably, in step S4, initializing multiple agents includes: setting different agents and different reward functions with the shortest completion time, the shortest average completion time and the highest equipment utilization rate as the scheduling target tar respectively;
[0048] The reward function is:
[0049]
[0050] Among them, R + and R-represent fixed positive and negative reward values, respectively, and tar is the scheduling target;
[0051] Each of the intelligent agents is respectively provided with a D3QN model;
[0052] The D3QN model includes two EfficientNet neural networks with the same structure, which serve as the online evaluation network Q and the target value network Q- of the D3QN model respectively.
[0053] Preferably, in step S5, data is read in batches from the sample data pool and input into the online evaluation network Q to obtain the estimated value, and input into the target value network Q- to obtain the target value, and the mean square error between the estimated value and the target value is used as the loss function of the D3QN model for supervised training;
[0054] The loss function of the D3QN model is specifically:
[0055]
[0056]
[0057] Among them, L e (θ e ) is the loss function value of the D3QN model; the subscript e is the training step length, represents the mean square error calculation, (s i , a k , r, s i+1 )~U(DB) represents the training sample data extracted from the sample data pool, Y e represents the target value; Q(s i , a k θ e ) indicates that the current workshop status s i and action a k The estimated value obtained by inputting into the online evaluation network Q; γ is the discount factor. The larger the discount factor, the more the network training focuses on the changes in future states; argmaxaQ(s i+1 , a; θ e ) represents the online evaluation network Q evaluating the next state s i+1 The most valuable action is a;
[0058] Use the stochastic gradient descent strategy to reversely update the weight parameters of the online evaluation network Q. After each training step, copy the last updated parameters of the online evaluation network Q to update the parameters of the target value network Q-. Repeat the training until the loss function value of the D3QN model is less than the preset threshold, and then obtain the trained D3QN model.
[0059] Preferably, in step S6, the real-time status of the mixed-flow manufacturing workshop is obtained and the MPN model is input to obtain the real-time MPN model and the production tasks to be scheduled in the real-time status;
[0060] The real-time state is input into the D3QN model trained by each agent. Each agent selects the scheduling action with the best value from the tasks to be scheduled in the real-time state according to the preset scheduling goal, and updates the next real-time state and the corresponding tasks to be scheduled. The scheduling action with the best value is repeatedly selected for execution until the production of all workpieces is completed, and several workshop scheduling strategies guided by different scheduling goals are obtained.
[0061] For the same pending orders, the corresponding workshop production scheduling strategy is directly called according to different scheduling targets to schedule production.
[0062] Preferably, the step S6 further includes:
[0063] Use the weighted comprehensive scoring method to score each workshop production scheduling strategy under different scheduling objectives, and obtain the priority of each workshop production scheduling strategy under different scheduling objectives;
[0064] According to the priority of each workshop production scheduling strategy under different scheduling objectives, a workshop production scheduling strategy whose priority under different scheduling objectives meets the preset conditions is selected as the optimal production scheduling strategy, and comprehensive production scheduling is performed according to the optimal production scheduling strategy.
[0065] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:
[0066] The present invention provides a real-time scheduling method for a mixed-flow manufacturing workshop based on a multi-agent D3QN. First, an MPN model based on a Petri net is constructed according to the operation relationship between workpieces and processing equipment in the mixed-flow manufacturing workshop; the operation data of the to-be-executed orders in the mixed-flow manufacturing workshop is obtained and input into the MPN model to obtain the initialized MPN model; then, a random scheduling simulation is performed based on the initialized MPN model to obtain simulation data; Markov decision chain data is constructed according to the simulation data and the operation data and loaded into a preset sample data pool; multiple agents are initialized and a D3QN model is constructed for each agent; the scheduling target of each agent is different; then, a number of Markov decision chain data are randomly extracted from the sample data pool and the D3QN model of each agent is trained to obtain the trained D3QN model of each agent; finally, the real-time status of the mixed-flow manufacturing workshop is obtained and input into the MPN model to obtain the real-time MPN model; after calling the trained D3QN model of each agent to perform scheduling calculation on the real-time MPN model, a number of workshop production scheduling strategies guided by different scheduling targets are obtained, and production is scheduled according to the workshop production scheduling strategies;
[0067] The present invention adopts the EfficientNet neural network as the basic network of the D3QN model, and utilizes the high efficiency and lightweight characteristics of the EfficientNet neural network to significantly reduce the amount of calculation and the number of parameters while maintaining high accuracy, thereby improving the performance and training speed of the D3QN algorithm when dealing with large-scale scheduling problems; at the same time, by introducing a multi-agent mechanism, the present invention enables the scheduling algorithm to obtain mixed-flow manufacturing workshop scheduling strategies oriented to different goals in a single scheduling calculation process, allowing production managers to consider and make decisions in multiple dimensions, thereby better optimizing production scheduling and resource utilization, and improving overall production efficiency and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS
[0068] Figure 1This is a flow chart of a real-time scheduling method for a mixed-flow manufacturing workshop based on multi-agent D3QN provided in Example 1.
[0069] Figure 2 This is a schematic diagram of the three-channel workshop state transfer provided in Example 2.
[0070] Figure 3 Schematic diagram of the multi-agent D3QN algorithm framework provided in Example 2. DETAILED DESCRIPTION
[0071] The drawings are for illustrative purposes only and should not be construed as limiting the present patent;
[0072] In order to better illustrate the present embodiment, some parts in the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product;
[0073] It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.
[0074] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.
[0075] Example 1
[0076] like Figure 1 As shown, this embodiment provides a real-time scheduling method for a mixed-flow manufacturing workshop based on a multi-agent D3QN, comprising the following steps:
[0077] S1: Constructing a MPN model based on Petri nets according to the operating relationship between workpieces and processing equipment in mixed-flow manufacturing workshops;
[0078] S2: Obtaining the operation data of the orders to be executed in the mixed-flow manufacturing workshop and inputting them into the MPN model to obtain the initialized MPN model; the operation data includes: the number and type of workpieces of the orders to be executed, and the status information of the processing equipment of each workpiece during the processing;
[0079] S3: Perform random scheduling simulation based on the initialized MPN model to obtain simulation data; construct Markov decision chain data according to the simulation data and operation data and load it into a preset sample data pool;
[0080] S4: Initialize multiple agents and build a D3QN model for each agent; each agent has a different scheduling target;
[0081] The D3QN model includes two EfficientNet neural networks with the same structure, which serve as the online evaluation network Q and the target value network Q of the D3QN model respectively. - ;
[0082] S5: Randomly extract several Markov decision chain data from the sample data pool and train the D3QN model of each agent to obtain the trained D3QN model of each agent;
[0083] S6: Obtain the real-time status of the mixed-flow manufacturing workshop and input it into the MPN model to obtain the real-time MPN model; call the D3QN model trained by each intelligent agent to perform scheduling calculations on the real-time MPN model, and obtain several workshop production scheduling strategies guided by different scheduling objectives, and schedule production according to the workshop production scheduling strategies.
[0084] In the specific implementation process, firstly, an MPN model based on Petri nets is constructed according to the operation relationship between workpieces and processing equipment in the mixed-flow manufacturing workshop; the operation data of the orders to be executed in the mixed-flow manufacturing workshop is obtained and input into the MPN model to obtain the initialized MPN model; then, a random scheduling simulation is performed based on the initialized MPN model to obtain simulation data; Markov decision chain data is constructed according to the simulation data and the operation data and loaded into the preset sample data pool; multiple agents are initialized and a D3QN model is constructed for each agent; the scheduling goals of each agent are different; then, several Markov decision chain data are randomly extracted from the sample data pool and the D3QN model of each agent is trained to obtain the trained D3QN model of each agent; finally, the real-time status of the mixed-flow manufacturing workshop is obtained and input into the MPN model to obtain the real-time MPN model; after calling the trained D3QN model of each agent to perform scheduling calculation on the real-time MPN model, several workshop production scheduling strategies guided by different scheduling goals are obtained, and production is scheduled according to the workshop production scheduling strategies;
[0085] This method uses the EfficientNet neural network as the basic network of the D3QN model, and takes advantage of the high efficiency and lightweight characteristics of the EfficientNet neural network to significantly reduce the amount of calculation and parameters while maintaining high accuracy, thereby improving the performance and training speed of the D3QN algorithm when dealing with large-scale scheduling problems; at the same time, by introducing a multi-agent mechanism, this method enables the scheduling algorithm to obtain mixed-flow manufacturing workshop scheduling strategies oriented to different goals in a single scheduling calculation process, allowing production managers to consider and make decisions in multiple dimensions, thereby better optimizing production scheduling and resource utilization, and improving overall production efficiency and economic benefits.
[0086] Example 2
[0087] This embodiment provides a real-time scheduling method for a mixed-flow manufacturing workshop based on a multi-agent D3QN, including the following steps:
[0088] S1: Constructing a MPN model based on Petri nets according to the operating relationship between workpieces and processing equipment in mixed-flow manufacturing workshops;
[0089] S2: Obtaining the operation data of the orders to be executed in the mixed-flow manufacturing workshop and inputting them into the MPN model to obtain the initialized MPN model; the operation data includes: the number and type of workpieces of the orders to be executed, and the status information of the processing equipment of each workpiece during the processing;
[0090] S3: Perform random scheduling simulation based on the initialized MPN model to obtain simulation data; construct Markov decision chain data according to the simulation data and operation data and load it into a preset sample data pool;
[0091] S4: Initialize multiple agents and build a D3QN model for each agent; each agent has a different scheduling target;
[0092] The D3QN model includes two EfficientNet neural networks with the same structure, which serve as the online evaluation network Q and the target value network Q- of the D3QN model respectively;
[0093] S5: Randomly extract several Markov decision chain data from the sample data pool and train the D3QN model of each agent to obtain the trained D3QN model of each agent;
[0094] S6: Obtain the real-time status of the mixed-flow manufacturing workshop and input it into the MPN model to obtain the real-time MPN model; call the D3QN model trained by each agent to perform scheduling calculation on the real-time MPN model, obtain several workshop production scheduling strategies guided by different scheduling goals, and schedule production according to the workshop production scheduling strategies;
[0095] The step S1 comprises:
[0096] The multiple processing equipment running in parallel in each process in the mixed-flow manufacturing workshop is abstracted into a library, indicating that the workpiece can be transferred to the processing equipment when processing in a certain process; different library places are connected by an intermediate transition path, indicating the transfer of the workpiece between different processes, that is, a path for the workpiece to be transferred to the next processing equipment after processing on one processing equipment is completed;
[0097] The expression of the entire MPN model is as follows:
[0098] MPN= <Token,Place.Transition>
[0099] Token={Token1, Token2,…,Token i}
[0100] Place={Place1, Place2,…, Place j}
[0101] Transition={Transition1, Transition2,…,Transition h}
[0102] Among them, Token is the expression set of the artifact in the MPN model. i is the expression of the i-th workpiece; Place is the expression set of the processing equipment in the MPN model, Place j is the expression of the jth processing equipment; Transition is the expression set of the transition path in the MPN model, Transition h is the hth transition path;
[0103] The expression of the i-th workpiece in the MPN model is as follows:
[0104]
[0105] Among them, Token i represents the i-th workpiece, n is the number of workpieces; ID1 is the number of the i-th workpiece, satisfying ID1∈N, N is a natural number set; C is the workpiece type, satisfying C∈{Color1, Color2,…, Color c}, Color1~Color c They are the 1st to cth workpiece types respectively; is the processing time of the i-th workpiece on the j-th processing equipment, m is the number of processing equipment; P is the number of the processing equipment where the i-th workpiece is currently located, satisfying P∈N; T is the remaining processing time of the i-th workpiece in the current equipment, satisfying T∈R, R is a real number set;
[0106] The expression of the jth processing equipment in the MPN model is as follows:
[0107]
[0108] Among them, Place j represents the jth processing equipment; ID2 is the number of the jth processing equipment, satisfying ID2∈N; Capacity is the number of workpieces that the processing equipment can accommodate, is the workpiece existing in the jth processing equipment;
[0109] The expression of the hth transition path in the MPN model is as follows:
[0110] Transition h =<ID3, inPlace, outPlace>
[0111] Among them, ID3 is the number of the transition, satisfying ID3∈N, inPlace is the number of the current processing equipment, satisfying inPlace∈N; outPlace is the number of the next processing equipment, satisfying outPlace∈N;
[0112] In the step S2, the workpiece quantity, workpiece number and workpiece type of each workpiece, as well as the processing equipment used by each workpiece during the processing and the remaining processing time of the workpiece are extracted from the job data, and all the extracted information is loaded into the MPN model for initialization to obtain the initialized MPN model;
[0113] In step S3, a random scheduling strategy is used to simulate the entire process from the start to the end of all workpiece processing of the order to be executed, to generate batches of simulation data, and to construct a workpiece scheduling time matrix ST, a workpiece position transfer identification matrix MM and a workpiece equipment delay matrix MT in combination with the operation data;
[0114] According to the three matrices obtained, the workshop status s of the three channels is constructed i , the workshop status s i 、s i Next selected action a k 、Action a k The corresponding reward r, and the next workshop state s i+1 Jointly construct Markov decision chain data i , a k , r, s i+1 > and load the preset sample data pool;
[0115] The workshop status i The expression is as follows:
[0116] s i =<ST n×m , M.M. n×m , M.T. n×m >
[0117] Among them, n×m is the size of the matrix, n is the number of workpieces in the MPN model, and m is the number of processing equipment in the MPN model;
[0118] Get the MPN model in the current workshop status i The following set of actions can be triggered Job(s) i ), where action refers to the process of transferring a specific workpiece from a certain process equipment to the next process equipment, action a k The expression is as follows:
[0119] a k =<ID1, inPlace, outPlace, Time>
[0120] Where Time is the time when the action is triggered, satisfying Time∈R;
[0121] According to the workshop status i Multiple actions that can be triggered by the workpiece constitute a set Job(s) i ), action set Job(s i ) is expressed as follows:
[0122] Job(s i ) = {a1, a2, ..., a k}
[0123] The action set Job(s i ) is used to indicate the workshop status s i The workpiece can choose the direction of transfer. When the actual production transfer is completed, one action will be selected from many actions to be executed. After the execution is completed, the workshop status is s i Updated to i+1 , action set Job(s i ) is also updated to Job(s i+1 );
[0124] According to the degree of fit between the result t of random simulation scheduling and the preset scheduling target tar, the workshop status s is assigned i Next, perform action a k Entering a new state i+1 The corresponding reward r;
[0125] The mixed-flow manufacturing workshop starts from the order placement state s0 to the completion state s n End, and form Markov decision chain data one by one according to the order of n scheduling and production scheduling i , a k , r, s i+1 >
[0126] In the step S4, initializing multiple agents includes: setting different agents and different reward functions with the shortest completion time, the shortest average completion time and the highest equipment utilization rate as the scheduling target tar respectively;
[0127] The reward function is:
[0128]
[0129] Among them, R + and R-represent fixed positive and negative reward values, respectively, and tar is the scheduling target;
[0130] Each of the intelligent agents is respectively provided with a D3QN model;
[0131] The D3QN model includes two EfficientNet neural networks with the same structure, which serve as the online evaluation network Q and the target value network Q- of the D3QN model respectively;
[0132] In step S5, data is read in batches from the sample data pool and input into the online evaluation network Q to obtain an estimated value, and input into the target value network Q- to obtain a target value, and supervised training is performed using the mean square error of the estimated value and the target value as the loss function of the D3QN model;
[0133] The loss function of the D3QN model is specifically:
[0134]
[0135]
[0136] Among them, L e (θ e ) is the loss function value of the D3QN model; the subscript e is the training step length, represents the mean square error calculation, (s i , a k , r, s i+1 )~U(DB) represents the training sample data extracted from the sample data pool, Y e represents the target value; Q(s i , a k θ e ) indicates that the current workshop status s i and action a k The estimated value obtained by inputting into the online evaluation network Q; γ is the discount factor. The larger the discount factor, the more the network training focuses on the changes in future states; argmaxaQ(s i+1 , a; θ e ) represents the online evaluation network Q evaluating the next state s i+1 The most valuable action is a;
[0137] Use the stochastic gradient descent strategy to perform reverse update training on the weight parameters of the online evaluation network Q; after each training step, copy the last updated parameters of the online evaluation network Q to update the parameters of the target value network Q-; repeatedly train until the loss function value of the D3QN model is less than the preset threshold, and then obtain the trained D3QN model;
[0138] In the step S6, the real-time status of the mixed-flow manufacturing workshop is obtained and the MPN model is input to obtain the real-time MPN model and the production tasks to be scheduled in the real-time state;
[0139] The real-time state is input into the D3QN model trained by each agent. Each agent selects the scheduling action with the best value from the tasks to be scheduled in the real-time state according to the preset scheduling goal, and updates the next real-time state and the corresponding tasks to be scheduled. The scheduling action with the best value is repeatedly selected for execution until the production of all workpieces is completed, and several workshop scheduling strategies guided by different scheduling goals are obtained.
[0140] For the same pending orders, the corresponding workshop production scheduling strategy is directly called according to different scheduling targets to schedule production;
[0141] In addition, the step S6 further comprises:
[0142] Use the weighted comprehensive scoring method to score each workshop production scheduling strategy under different scheduling objectives, and obtain the priority of each workshop production scheduling strategy under different scheduling objectives;
[0143] According to the priority of each workshop production scheduling strategy under different scheduling objectives, a workshop production scheduling strategy whose priority under different scheduling objectives meets the preset conditions is selected as the optimal production scheduling strategy, and comprehensive production scheduling is performed according to the optimal production scheduling strategy.
[0144] In the specific implementation process, we first construct a MPN model based on Petri nets according to the operating relationship between workpieces and processing equipment in the mixed-flow manufacturing workshop. The process is as follows:
[0145] The multiple processing equipment running in parallel in each process in the mixed-flow manufacturing workshop is abstracted into a library, indicating that the workpiece can be transferred to the processing equipment when processing in a certain process; different library places are connected by an intermediate transition path, indicating the transfer of the workpiece between different processes, that is, a path for the workpiece to be transferred to the next processing equipment after processing on one processing equipment is completed;
[0146] The expression of the entire MPN model is as follows:
[0147] MPN= <Token,Place.Transition>
[0148] Token={Token1, Token2,…,Token i}
[0149] Place={Place1, Place2,…, Place j}
[0150] Transition={Transition1, Transition2,…,Transition h}
[0151] Among them, Token is the expression set of the artifact in the MPN model. i is the expression of the i-th workpiece; Place is the expression set of the processing equipment in the MPN model, Place j is the expression of the jth processing equipment; Transition is the expression set of the transition path in the MPN model, Transition h is the hth transition path;
[0152] The expression of the i-th workpiece in the MPN model is as follows:
[0153]
[0154] Among them, Token i represents the i-th workpiece, n is the number of workpieces; ID1 is the number of the i-th workpiece, satisfying ID1∈N, N is a natural number set; C is the workpiece type, satisfying C∈{Color1, Color2,…, Color c}, Color1~Color c They are the 1st to cth workpiece types respectively; is the processing time of the i-th workpiece on the j-th processing equipment, m is the number of processing equipment; P is the number of the processing equipment where the i-th workpiece is currently located, satisfying P∈N; T is the remaining processing time of the i-th workpiece in the current equipment, satisfying T∈R, R is a real number set;
[0155] The expression of the jth processing equipment in the MPN model is as follows:
[0156]
[0157] Among them, Place j represents the jth processing equipment; ID2 is the number of the jth processing equipment, satisfying ID2∈N; Capacity is the number of workpieces that the processing equipment can accommodate, is the workpiece existing in the jth processing equipment;
[0158] The expression of the hth transition path in the MPN model is as follows:
[0159] Transition h =<ID3, inPlace, outPlace>
[0160] Among them, ID3 is the number of the transition, satisfying ID3∈N, inPlace is the number of the current processing equipment, satisfying inPlace∈N; outPlace is the number of the next processing equipment, satisfying outPlace∈N;
[0161] Then, the operation data of the orders to be executed in the mixed-flow manufacturing workshop is obtained and input into the MPN model to obtain the initialized MPN model;
[0162] Extracting the workpiece quantity, workpiece number and workpiece type of each workpiece, as well as the processing equipment used by each workpiece during processing and the remaining processing time of the workpiece from the job data, and loading all the extracted information into the MPN model for initialization to obtain the initialized MPN model;
[0163] Then, a random scheduling simulation is performed based on the initialized MPN model to obtain simulation data; Markov decision chain data is constructed based on the simulation data and job data and loaded into the preset sample data pool;
[0164] Use random scheduling strategy to simulate the whole process of all workpiece processing from the start to the end of the pending order, generate batch simulation data, and build workpiece production time matrix ST, workpiece position transfer identification matrix MM and workpiece equipment delay matrix MT in combination with operation data;
[0165] According to the three matrices obtained, the workshop status s of the three channels is constructed i , the workshop status s i 、s i Next selected action a k 、Action a k The corresponding reward r, and the next workshop state s i+1 Jointly construct Markov decision chain data i , a k , r, s i+1 > and load the preset sample data pool;
[0166] The workshop status i The expression is as follows:
[0167] s i =<ST n×m , M.M. n×m , M.T. n×m >
[0168] Among them, n×m is the size of the matrix, n is the number of workpieces in the MPN model, and m is the number of processing equipment in the MPN model;
[0169] Get the MPN model in the current workshop status i The following set of actions can be triggered Job(s) i ), where action refers to the process of transferring a specific workpiece from a certain process equipment to the next process equipment, action a k The expression is as follows:
[0170] a k =<ID1, inPlace, outPlace, Time>
[0171] Where Time is the time when the action is triggered, satisfying Time∈R;
[0172] According to the workshop status i Multiple actions that can be triggered by the workpiece constitute a set Job(s) i ), action set Job(s i ) is expressed as follows:
[0173] Job(s i ) = {a1, a2, ..., a k}
[0174] The action set Job(s i ) is used to indicate the workshop status s i The workpiece can choose the direction of transfer. When the actual production transfer is completed, one action will be selected from many actions to be executed. After the execution is completed, the workshop status is s i Updated to i+1 , action set Job(s i ) is also updated to Job(s i+1 );
[0175] According to the degree of fit between the result t of random scheduling simulation and the preset scheduling target tar, the workshop state s is assigned i Next, perform action a k Entering a new state i+1 The corresponding reward r;
[0176] The mixed-flow manufacturing workshop starts from the order placement state s0 to the completion state s n End, and form Markov decision chain data one by one according to the order of n scheduling and production scheduling i , a k , r, s i+1 >
[0177] like Figure 2 As shown, the workpiece scheduling time matrix ST represents the time points when different workpieces leave the current processing equipment. Figure 2 Medium i The matrix element in the second row and fourth column of the workpiece scheduling time matrix ST under the status is 79, which means that the time point when the workpiece numbered T2 is processed in the processing equipment numbered P4 is 79 unit time after the start of scheduling; -1 in the matrix has no actual meaning and is only used to fill the matrix;
[0178] The workpiece position transfer identification matrix MM indicates the processing equipment where the current workpiece is located and the processing equipment to which the next process can be transferred; Figure 2 Mediumi The elements of the second row in the workpiece position transfer identification matrix under the state, element 0 indicates that the T2 workpiece is currently in the P4 processing equipment; element 1 indicates that the T2 workpiece is in the processing equipment that can be transferred to the next process, that is, it can be transferred to the P5 and P6 processing equipment; element -1 indicates that the workpiece is not in the equipment or the processing equipment that cannot be transferred to the next process;
[0179] The workpiece delay matrix MT is a matrix that does not change with the state change, indicating the processing time of the workpiece on different equipment; Figure 2 Medium i The values of the first and last columns of the workpiece position transfer identification matrix in the state are all 0, which means that the workpiece is not processed at the initial equipment and the last equipment it arrives at, but is only stored;
[0180] like Figure 2 As shown, state s i After action a k After entering state s i+1 , where a k <T4, P1, P2, 0> means that the T4 workpiece is transferred from the P1 processing equipment to the P4 processing equipment at the 0-unit time point; Figure 2 Medium state i+1 The first column element of the fourth row of the workpiece scheduling time matrix changes from 0 to -1, and the second column element changes from -1 to 23, indicating that the T4 component leaves the P1 processing equipment and the processing completion time point of P2 is 23 unit time from the start of scheduling; Figure 2 Medium state i+1 The fourth row and first column elements of the workpiece position transfer identification matrix change from 0 to -1, the second column changes from 1 to 0, the third and fourth columns change from 1 to -1, and the fifth and sixth columns change from -1 to 1, indicating that the T4 component leaves the P1 processing equipment and transfers to the P2 processing setting, the original P3 and P4 processing equipment change from a transferable state to a non-transferable state, and the P5 and P6 processing equipment change from a non-transferable state to a transferable state;
[0181] Then, multiple agents are initialized, and a D3QN model is constructed for each agent. The scheduling objectives of each agent are different. The scheduling objectives include but are not limited to: the shortest completion time objective, the shortest average completion time objective, and the highest equipment utilization objective.
[0182] The shortest completion time, the shortest average completion time and the highest equipment utilization rate are used as the scheduling target tar, and different agents and different reward functions are set;
[0183] The reward function is:
[0184]
[0185] Among them, R + and R-represent fixed positive and negative reward values, respectively, and tar is the scheduling target;
[0186] Each of the intelligent agents is respectively provided with a D3QN model;
[0187] The D3QN model includes two EfficientNet neural networks with the same structure, which serve as the online evaluation network Q and the target value network Q- of the D3QN model respectively;
[0188] Randomly extract several Markov decision chain data from the sample data pool and train the D3QN model of each agent to obtain the trained D3QN model of each agent, including:
[0189] Reading data in batches from the sample data pool and inputting the data into the online evaluation network Q to obtain the estimated value, inputting the data into the target value network Q- to obtain the target value, and using the mean square error of the estimated value and the target value as the loss function of the D3QN model for supervised training;
[0190] The loss function of the D3QN model is specifically:
[0191]
[0192]
[0193] Among them, L e (θ e ) is the loss function value of the D3QN model; the subscript e is the training step length, represents the mean square error calculation, (s i , a k , r, s i+1 )~U(DB) represents the training sample data extracted from the sample data pool, Y e represents the target value; Q(s i , a k θ e ) indicates that the current workshop status s i and action a k The estimated value obtained by inputting into the online evaluation network Q; γ is the discount factor. The larger the discount factor, the more the network training focuses on the changes in future states; argmax a Q(s i+1 , a; θ e ) represents the online evaluation network Q evaluating the next state s i+1 The most valuable action is a;
[0194] Use the stochastic gradient descent strategy to perform reverse update training on the weight parameters of the online evaluation network Q; after each training step, copy the last updated parameters of the online evaluation network Q to update the parameters of the target value network Q-; repeatedly train until the loss function value of the D3QN model is less than the preset threshold, and then obtain the trained D3QN model;
[0195] like Figure 3 As shown, finally, the real-time status of the mixed-flow manufacturing workshop is obtained and input into the MPN model to obtain the real-time MPN model and the production tasks to be scheduled in the real-time state;
[0196] The real-time state is input into the D3QN model trained by each agent. Each agent selects the scheduling action with the best value from the tasks to be scheduled in the real-time state according to the preset scheduling goal, and updates the next real-time state and the corresponding tasks to be scheduled. The scheduling action with the best value is repeatedly selected for execution until the production of all workpieces is completed, and several workshop scheduling strategies guided by different scheduling goals are obtained.
[0197] For the same pending orders, the corresponding workshop production scheduling strategy is directly called according to different scheduling targets to schedule production;
[0198] In addition, the weighted comprehensive scoring method can be used to score each workshop production scheduling strategy under different scheduling objectives to obtain the priority of each workshop production scheduling strategy under different scheduling objectives;
[0199] According to the priority of each workshop production scheduling strategy under different scheduling objectives, a workshop production scheduling strategy whose priority under different scheduling objectives meets the preset conditions is selected as the optimal production scheduling strategy, and comprehensive production scheduling is performed according to the optimal production scheduling strategy, so as to obtain a scheduling strategy that can comprehensively consider various objectives after balancing the priorities of different strategies;
[0200] This method uses the EfficientNet neural network as the basic network of the D3QN model, and takes advantage of the high efficiency and lightweight characteristics of the EfficientNet neural network to significantly reduce the amount of calculation and parameters while maintaining high accuracy, thereby improving the performance and training speed of the D3QN algorithm when dealing with large-scale scheduling problems; at the same time, by introducing a multi-agent mechanism, this method enables the scheduling algorithm to obtain mixed-flow manufacturing workshop scheduling strategies oriented to different goals in a single scheduling calculation process, allowing production managers to consider and make decisions in multiple dimensions, thereby better optimizing production scheduling and resource utilization, and improving overall production efficiency and economic benefits.
[0201] The same or similar reference numerals correspond to the same or similar components;
[0202] The terms used in the drawings to describe positional relationships are only used for illustrative purposes and should not be construed as limiting this patent;
[0203] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.
Claims
1. A real-time scheduling method for mixed-flow manufacturing workshop based on multi-agent D3QN, characterized in that: The following steps are involved: S1: Constructing a MPN model based on Petri nets according to the operating relationship between workpieces and processing equipment in mixed-flow manufacturing workshops; S2: Obtaining the operation data of the orders to be executed in the mixed-flow manufacturing workshop and inputting them into the MPN model to obtain the initialized MPN model; the operation data includes: the number and type of workpieces of the orders to be executed, and the status information of the processing equipment of each workpiece during the processing; S3: Perform random scheduling simulation based on the initialized MPN model to obtain simulation data; construct Markov decision chain data according to the simulation data and operation data and load it into a preset sample data pool; S4: Initialize multiple agents and build a D3QN model for each agent; each agent has a different scheduling target; The D3QN model includes two EfficientNet neural networks with the same structure, which serve as the online evaluation network Q and the target value network Q- of the D3QN model respectively; S5: Randomly extract several Markov decision chain data from the sample data pool and train the D3QN model of each agent to obtain the trained D3QN model of each agent; S6: Obtain the real-time status of the mixed-flow manufacturing workshop and input it into the MPN model to obtain the real-time MPN model; call the D3QN model trained by each intelligent agent to perform scheduling calculations on the real-time MPN model, and obtain several workshop production scheduling strategies guided by different scheduling objectives, and schedule production according to the workshop production scheduling strategies.
2. According to claim 1, a real-time scheduling method for mixed-flow manufacturing workshop based on multi-agent D3QN is characterized in that: The step S1 comprises: The multiple processing equipment running in parallel in each process in the mixed-flow manufacturing workshop is abstracted into a library, indicating that the workpiece can be transferred to the processing equipment when processing in a certain process; different library places are connected by an intermediate transition path, indicating the transfer of the workpiece between different processes, that is, a path for the workpiece to be transferred to the next processing equipment after processing on one processing equipment is completed; The expression of the entire MPN model is as follows: MPN=<Token,Place.Transition> Token = {Token1, Token2, ..., Token i } Place={Place1,Place2,…,Place j } Transition={Transition1,Transition2,…,Transition h } Among them, Token is the expression set of the artifact in the MPN model. i is the expression of the i-th workpiece; Place is the expression set of the processing equipment in the MPN model, Place j is the expression of the jth processing equipment; Transition is the expression set of the transition path in the MPN model, Transition h is the hth transition path.
3. The real-time scheduling method for mixed-flow manufacturing workshop based on multi-agent D3QN according to claim 2 is characterized in that: The expression of the i-th workpiece in the MPN model is as follows: Among them, Token i represents the i-th workpiece, n is the number of workpieces; ID1 is the number of the i-th workpiece, satisfying ID1∈N, N is a natural number set; C is the workpiece type, satisfying C∈{Color1, Color2,…, Color c }, Color1~Color c They are the 1st to cth workpiece types respectively; is the processing time of the i-th workpiece on the j-th processing equipment, m is the number of processing equipment; P is the number of the processing equipment where the i-th workpiece is currently located, satisfying P∈N; T is the remaining processing time of the i-th workpiece in the current equipment, satisfying T∈R, R is a real number set; The expression of the jth processing equipment in the MPN model is as follows: Among them, Place j represents the jth processing equipment; ID2 is the number of the jth processing equipment, satisfying ID2∈N; Capacity is the number of workpieces that the processing equipment can accommodate, is the workpiece existing in the jth processing equipment; The expression of the hth transition path in the MPN model is as follows: Transition h =<ID3, inPlace, outPlace> Among them, ID3 is the number of the transition path, satisfying ID3∈N, inPlace is the number of the current processing equipment, satisfying inPlace∈N; outPlace is the number of the next processing equipment, satisfying outPlace∈N.
4. The real-time scheduling method for mixed-flow manufacturing workshop based on multi-agent D3QN according to claim 3 is characterized in that: In step S2, the workpiece quantity, workpiece number and workpiece type of each workpiece, as well as the processing equipment used by each workpiece during the processing and the remaining processing time of the workpiece are extracted from the job data, and all the extracted information is loaded into the MPN model for initialization to obtain the initialized MPN model.
5. The real-time scheduling method for mixed-flow manufacturing workshop based on multi-agent D3QN according to claim 3 is characterized in that: In step S3, a random scheduling strategy is used to simulate the entire process from the start to the end of all workpiece processing of the order to be executed, to generate batches of simulation data, and to construct a workpiece scheduling time matrix ST, a workpiece position transfer identification matrix MM and a workpiece equipment delay matrix MT in combination with the operation data; According to the three matrices obtained, the workshop status s of the three channels is constructed i , the workshop status s i 、s i Next selected action a k 、Action a k The corresponding reward r, and the next workshop state s i+1 Jointly construct Markov decision chain data i , a k , r, s i+1 > and load the preset sample data pool. 6. The real-time scheduling method for mixed-flow manufacturing workshop based on multi-agent D3QN according to claim 5 is characterized in that: The workshop status i The expression is as follows: s i =<STn ×m ,MM n×m ,MTn ×m > Among them, n×m is the size of the matrix, n is the number of workpieces in the MPN model, and m is the number of processing equipment in the MPN model; Get the MPN model in the current workshop status i The following set of actions can be triggered Job(s) i ), where action refers to the process of transferring a specific workpiece from a certain process equipment to the next process equipment, action a k The expression is as follows: a k =<ID1,inPlace,outPlace,Time> Where Time is the time when the action is triggered, satisfying Time∈R; According to the workshop status i Multiple actions that can be triggered by the workpiece constitute a set Job(s) i ), action set Job(s ii ) is expressed as follows: Job(s i )={a1,a2,...,a k } The action set Job(s i ) is used to indicate the workshop status s i The workpiece can choose the direction of transfer. When the actual production transfer is completed, one action will be selected from many actions to be executed. After the execution is completed, the workshop status is s i Updated to i+1 , action set Job(s i ) is also updated to Job(s i+1 ); According to the degree of fit between the result t of random scheduling simulation and the preset scheduling target tar, the workshop state s is assigned i Next, perform action a k Entering a new state i+1 The corresponding reward r; The mixed-flow manufacturing workshop starts from the order placement state s0 to the completion state s n End, and form Markov decision chain data one by one according to the order of n scheduling and production scheduling i , a k , r, s i+1 >. 7. The real-time scheduling method for mixed-flow manufacturing workshop based on multi-agent D3QN according to claim 6 is characterized in that: In the step S4, initializing multiple agents includes: setting different agents and different reward functions with the shortest completion time, the shortest average completion time and the highest equipment utilization rate as the scheduling target tar respectively; The reward function is: Among them, R + and R-represent fixed positive and negative reward values, respectively, and tar is the scheduling target; Each of the intelligent agents is respectively provided with a D3QN model; The D3QN model includes two EfficientNet neural networks with the same structure, which serve as the online evaluation network Q and the target value network Q- of the D3QN model respectively.
8. The real-time scheduling method for mixed-flow manufacturing workshop based on multi-agent D3QN according to claim 7 is characterized in that: In step S5, data is read in batches from the sample data pool and input into the online evaluation network Q to obtain an estimated value, and input into the target value network Q- to obtain a target value, and supervised training is performed using the mean square error of the estimated value and the target value as the loss function of the D3QN model; The loss function of the D3QN model is specifically: Among them, L e (θ e ) is the loss function value of the D3QN model; the subscript e is the training step length, represents the mean square error calculation, (s i , a k , r, s i+1 )~U(DB) represents the training sample data extracted from the sample data pool, Y e represents the target value; Q(s i , a k θ e ) indicates that the current workshop status s i and action a k The estimated value obtained by inputting into the online evaluation network Q; γ is the discount factor. The larger the discount factor, the more the network training focuses on the changes in future states; argmax a Q(s i+1 , a; θ e ) represents the online evaluation network Q evaluating the next state s i+1 The most valuable action is a; Use the stochastic gradient descent strategy to reversely update the weight parameters of the online evaluation network Q. After each training step, copy the last updated parameters of the online evaluation network Q to update the parameters of the target value network Q-. Repeat the training until the loss function value of the D3QN model is less than the preset threshold, and then obtain the trained D3QN model.
9. The real-time scheduling method for mixed-flow manufacturing workshop based on multi-agent D3QN according to claim 8 is characterized in that: In the step S6, the real-time status of the mixed-flow manufacturing workshop is obtained and the MPN model is input to obtain the real-time MPN model and the production tasks to be scheduled in the real-time state; The real-time state is input into the D3QN model trained by each agent. Each agent selects the scheduling action with the best value from the tasks to be scheduled in the real-time state according to the preset scheduling goal, and updates the next real-time state and the corresponding tasks to be scheduled. The scheduling action with the best value is repeatedly selected for execution until the production of all workpieces is completed, and several workshop scheduling strategies guided by different scheduling goals are obtained. For the same pending orders, the corresponding workshop production scheduling strategy is directly called according to different scheduling targets to schedule production.
10. A real-time scheduling method for mixed-flow manufacturing workshop based on multi-agent D3QN according to any one of claims 1 to 9, characterized in that: The step S6 further comprises: Use the weighted comprehensive scoring method to score each workshop production scheduling strategy under different scheduling objectives, and obtain the priority of each workshop production scheduling strategy under different scheduling objectives; According to the priority of each workshop production scheduling strategy under different scheduling objectives, a workshop production scheduling strategy whose priority under different scheduling objectives meets the preset conditions is selected as the optimal production scheduling strategy, and comprehensive production scheduling is performed according to the optimal production scheduling strategy.
Citation Information
Patent Citations
Hybrid flow shop scheduling method based on time sequence difference
CN112734172A
D3QN reinforcement learning-based workpiece circulation-oriented intelligent workshop real-time scheduling method
CN115526457A
Temporal difference-based hybrid flow-shop scheduling method
WO2022135066A1
Cited By
Discrete manufacturing system toughness enhancing method and system based on Petri net
CN120338457A
