Workshop scheduling method and system fusing decision tree and genetic algorithm

Abstract

The invention provides a workshop scheduling method and system fusing a decision tree and a genetic algorithm. The method comprises the following steps: firstly, collecting operation process and machine information through an ERP system and extracting related features; generating a scheduling scheme by using historical orders of an ERP system or manually added orders, and constructing a training set training decision tree to accurately judge machine allocation conflicts; constructing a multi-target flexible job shop scheduling model, and constructing a target function based on a hierarchical Pareto dominance relationship; and secondly, realizing job-shop scheduling scheme coding by adopting double-layer chromosome coding, carrying out selection, intersection and mutation operations in combination with a genetic algorithm, carrying out conflict detection and repair by utilizing a decision tree, solving an objective function, continuously iterating until a convergence condition is met, and outputting a Pareto optimal solution. According to the workshop scheduling method, the scheduling efficiency and feasibility are improved by establishing a data model of a multi-target flexible job workshop scheduling problem and through dynamic conflict detection, hierarchical multi-target optimization and a closed-loop feedback mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of production scheduling, in particular to a workshop scheduling method and system fusing decision tree and genetic algorithm. BACKGROUND

[0002] Flexible job shop scheduling problem plays a crucial role in modern manufacturing industry, but it is also a typical NP-hard problem. NP-hard problem means that the problem is extremely high in computational complexity, and it is difficult to find the optimal solution in polynomial time. Solving flexible job shop scheduling problem is of great significance to improve the production efficiency of manufacturing industry, reduce cost, and enhance resource utilization efficiency.

[0003] Flexible job shop scheduling problem has the following challenges: Multi-objective conflict problem: In flexible job shop scheduling, multiple objectives need to be optimized at the same time, and these objectives often restrict each other, making it difficult to achieve the optimal solution at the same time. Order duration is an important indicator to measure production efficiency, and enterprises usually hope to shorten the duration as much as possible to quickly respond to market demand, improve customer satisfaction, and enhance market competitiveness. However, energy consumption is increasingly concerned in today's society, and reducing energy consumption not only can reduce production cost, but also conforms to the concept of sustainable development, and has a positive significance to environmental protection. At the same time, equipment utilization is also the focus of enterprises, and full utilization of equipment can reduce the production cost of unit product and improve the economic benefit of enterprises.

[0004] Resource constraints are another key challenge of flexible job shop scheduling problem. In flexible job shop, each process can usually choose multiple different machines for processing, which provides a certain flexibility for scheduling. However, this flexibility also brings new problems, that is, a single machine can only process one process at the same time. This requires that the processing order and time of each process on different machines must be reasonably arranged to avoid machine conflicts and intermittent idling.

[0005] For example, when process A and process B can both choose machine M1, if the processing time is not arranged properly, it may lead to conflict between the two processes occupying machine M1 at the same time. In order to avoid such conflicts, the processing order and time of processes need to be carefully planned. At the same time, factors such as machine maintenance time and preparation time need to be considered to ensure the normal operation and efficient use of machines. In addition, since each process can choose multiple machines, it also increases the complexity of scheduling, and needs to find the optimal solution among numerous choices.

[0006] The existing method adopts a post-collision repair strategy when dealing with the flexible job shop scheduling problem. This strategy repairs the scheduling scheme after discovering conflicts in the scheduling scheme. However, this strategy can cause low calculation efficiency. Because the backtracking method is usually used to perform the repair after discovering the conflict, it is necessary to backtrack to the previous step to make a decision again, which requires additional calculation time. According to statistics, the method can require an additional 30% of the calculation time, which is unacceptable in actual production. Because time is efficiency in actual production, the long calculation time can cause the production plan to be unable to be formulated in time, and affect the normal production.

[0007] There is an inherent conflict between the above-mentioned objectives. For example, in order to shorten the construction period, it can be necessary to increase the number of devices used or increase the running speed of the device, which often leads to an increase in energy consumption. If excessive pursuit of reducing energy consumption is pursued, a device with a slow running speed but low energy consumption can be selected, thereby prolonging the construction period. Similarly, increasing the utilization rate of the device can cause the device to be in a high-load running state for a long time, increase the wear and maintenance cost of the device, and also can affect the quality of the product, thereby adversely affecting the construction period and other indicators. Therefore, how to find a balance between these mutually contradictory objectives is a big challenge for the flexible job shop scheduling problem. SUMMARY

[0008] The present application provides a job shop scheduling method and system combining decision tree and genetic algorithm, which solves the actual scheduling problem of the multi-objective flexible job shop by establishing a data model for solving the multi-objective flexible job shop scheduling problem; and improves the scheduling efficiency and feasibility through dynamic conflict detection, hierarchical multi-objective optimization and closed-loop feedback mechanism.

[0009] According to a first aspect of the present application, a job shop scheduling method combining decision tree and genetic algorithm is provided, comprising the following steps: Collecting workpiece process information and machine information, and extracting process features and machine features; Constructing a decision tree based on the extracted process features and machine features; Constructing a multi-objective flexible job shop scheduling model, and constructing an objective function based on a hierarchical Pareto dominance relationship; Establishing a scheduling scheme combining the decision tree and the genetic algorithm, performing selection, crossover and mutation operations through the genetic algorithm, combining the decision tree to detect and repair conflicts, solving the objective function, and obtaining the optimal solution of the objective function.

[0010] On the basis of the above technical solution, the present application can also be improved as follows.

[0011] Optionally, the process of collecting workpiece process information and machine information and extracting features is as follows: Collecting workpiece process information and machine information through a manufacturing enterprise ERP system; the workpiece process information includes workpiece number, processable machine set and processing efficiency; the machine information includes machine number and initial state; the process characteristics include workpiece demand quantity and deadline, and the machine characteristics include current load and processing efficiency.

[0012] Optionally, the process of constructing a decision tree based on the extracted process and machine characteristics is as follows: In the training data generation phase, a scheduling scheme is generated using ERP system historical orders or manually added special case orders, and a training set is constructed with the extracted process and machine characteristics as input and conflict cases as output; When training the decision tree, the CART algorithm is used to train the data, the parameter optimization model is set, the model is optimized by 10-fold cross-validation, and the accuracy is ensured to be more than 92% and the AUC value is more than 0.95; The decision rules are generated in combination with the constraint conditions of the flexible job shop scheduling problem.

[0013] Optionally, the multi-objective flexible job shop scheduling model includes optimization objectives of workpiece on-time completion rate, minimum completion time and minimum number of machines; and the target function constructed based on the hierarchical Pareto dominance relationship includes: Ensure that the core objective is optimized first, and then the secondary objective is optimized step by step.

[0014] Optionally, the target function is represented as follows:

[0015] Wherein, are the on-time completion rate optimization function, the total completion time optimization function and the minimum number of machines optimization function, C is a constant, which ensures that , the parameter satisfies .

[0016] Optionally, the construction of the multi-objective flexible job shop scheduling model further includes: Determining the constraint conditions of the multi-objective flexible job shop scheduling model, the constraint conditions include: process assignment uniqueness constraint, process uniqueness constraint, process order constraint and machine capacity limit.

[0017] Optionally, the solving of the target function by selecting, crossing and mutating through the genetic algorithm combined with the decision tree for conflict detection and repair includes: Step 401: Initialize algorithm parameters, with the minimum of the target function defined in this method as the optimization objective, and terminate when the algorithm reaches the maximum number of iterations; wherein the parameters include the total number of population individuals P and the number of algorithm iterations T in the improved genetic algorithm; Step 402: complete the chromosome coding of the workpiece workshop scheduling scheme by using the double-layer chromosome coding mode; Step 403: calculate the fitness value of each chromosome in the sequence, calculate the Pareto optimal solution set of the current population by using the hierarchical Pareto dominance relationship, and save the top 5% chromosomes in the population to the historical optimal individual archive optimal solution according to the Pareto ranking of each chromosome; Step 404: judge whether the convergence criteria are met, if any of the convergence criteria is met, end the loop iteration, and output the Pareto optimal solution to the historical optimal individual archive optimal solution; otherwise, execute step 405; Step 405: calculate the fitness value of each chromosome in the sequence, and select Ps individuals by using the roulette wheel selection method; Step 406: perform the crossover operation on the selected individuals according to the random pairing principle to perform the process layer mutation and the machine layer mutation, detect and repair the decision tree conflict, and update the optimal solution; Step 407: perform the mutation operation on the process layer and the machine layer in the population, detect and repair the decision tree conflict, and update the optimal solution; Step 408: combine the generated individuals and the remaining individuals in the population with the historical optimal solution to form a new population.

[0018] Optionally, when the mutation operation is performed on the process layer and the machine layer in the population, the following hierarchical execution is performed: PS process layer mutation: the exchange mutation method is used to exchange the positions of the processes in the chromosome with a certain probability Randomly select the position of the process in the chromosome; MA machine layer mutation: the roulette wheel selection method is used to select a machine from the set of processable machines according to the production efficiency of the machine selected by the process. Optionally, the optimal solution of the target function comprises: A data model is established for a multi-objective flexible job shop scheduling problem with different job final completion times, machine total load and key machine energy efficiency, and the Pareto optimal solution is obtained by experiment and simulation on the actual workshop scheduling case through the data model.

[0019] According to the second aspect of the present application, a workshop scheduling method combining a decision tree and a genetic algorithm is provided, comprising: A data collection module is used to collect workpiece process information and machine information, and extract process features and machine features; A decision tree construction module is used to construct a decision tree based on the extracted process features and machine features; The target function construction module is configured to construct a multi-objective flexible job shop scheduling model, and construct a target function based on a hierarchical Pareto dominance relationship; The target function optimal solution output module is configured to establish a scheduling scheme combining a decision tree and a genetic algorithm, perform selection, crossover and mutation operations through the genetic algorithm, and combine the decision tree to detect and repair conflicts to solve the target function, so as to obtain an optimal solution of the target function.

[0020] Technical effects and advantages of the present application: The present application provides a kind of workshop scheduling method and system of fusion decision tree and genetic algorithm, by using hybrid genetic algorithm for the multi-objective flexible job shop scheduling problem of different job last completion time, machine total load and key machine energy efficiency establishes data model, through data model to actual workshop scheduling case is experimented and simulated, to obtain Pareto optimal solution.The present application is suitable for the resource optimization configuration of multi-species, small batch production scene in discrete manufacturing industry, improves scheduling efficiency and feasibility by dynamic conflict detection, hierarchical multi-objective optimization and closed-loop feedback mechanism. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The genetic algorithm and decision tree cooperative optimization flowchart provided by the embodiments of the present application are provided; Figure 2 The schematic diagram of workpiece bill of materials tree provided by the embodiments of the present application is provided; Figure 3 The schematic diagram of decision tree model structure provided by the embodiments of the present application is provided. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0023] It should be noted that the present application provides a kind of workshop scheduling method and system of fusion decision tree and genetic algorithm, by solving flexible job shop scheduling problem, the hybrid genetic algorithm of fusion decision tree and genetic algorithm can solve discrete flexible job shop scheduling problem after improvement, and better effect than traditional genetic algorithm is obtained.For solving multi-objective flexible job shop scheduling problem, it is suitable for resource optimization configuration in the scene of multi-species, small batch production of discrete manufacturing industry.

[0024] It can be understood that, based on the defects in the background art, the embodiments of the present application propose a kind of workshop scheduling method of fusion decision tree and genetic algorithm, specifically asFigure 1 As shown, comprising the following steps: Step 1: Collecting workpiece process information and machine information, extracting process features and machine features.

[0025] In this embodiment, the workpiece process information and machine information are collected through the manufacturing enterprise ERP system. The workpiece process information includes workpiece number, processable machine set and processing efficiency; the machine information includes machine number and initial state; then the decision tree is assisted to accurately judge the machine allocation conflict, and the process features and machine features are extracted; the process features include workpiece demand quantity and deadline, and the machine features include current load and processing efficiency. The above implementation steps are used to perfect the data processing and assist the decision.

[0026] Step 2: Constructing a decision tree based on the extracted process features and machine features; It should be noted that the embodiment of the present application is constructed around the production scheduling model of the manufacturing enterprise, from data generation to model training and rule constraint.

[0027] The process of constructing a decision tree based on the extracted process features and machine features includes: In the training data generation stage, the historical orders of the ERP system or the special situation orders added manually are used to generate a scheduling scheme, and the extracted process and machine features are used as the input and the conflict situation is used as the output to construct a training set; When training the decision tree, the CART algorithm is used to train the data, the parameter optimization model is set, the model is optimized by 10-fold cross-validation, and the accuracy is ensured to be more than 92% and the AUC value is more than 0.95; Finally, the decision rules are generated in combination with the constraint conditions of the flexible job shop scheduling problem; the constraint conditions include process sequence, processing uniqueness, order deadline and machine processing limitation.

[0028] Step 3, constructing a multi-objective flexible job shop scheduling model, and constructing a target function based on a hierarchical Pareto dominance relationship; The multi-objective flexible job shop scheduling model includes the optimization objectives of workpiece on-time completion rate, minimum completion time and minimum machine number.

[0029] The construction of the multi-objective flexible job shop scheduling model needs to determine multiple constraint conditions in the multi-objective flexible job shop scheduling model, and the decision tree is a key component (providing a conflict detection tool to ensure the feasibility of the solution) for realizing the feasibility constraint of the multi-objective flexible job shop scheduling model, and the solution process of the multi-objective flexible job shop scheduling model provides a dynamic optimization feedback mechanism for the decision tree, which jointly serves the efficient solution of the flexible job shop scheduling.

[0030] In the multi-objective flexible job shop scheduling model, it is necessary to first construct an optimization sub-model for the three objectives of on-time completion rate of workpieces, minimization of completion time, and minimum number of machines, and then construct the objective function of the multi-objective flexible job shop scheduling model based on the hierarchical Pareto dominance relationship of these three multi-objectives.

[0031] Building a multi-objective flexible job shop scheduling model includes the following steps: The construction of a multi-objective flexible workpiece shop scheduling model involves factors such as workpieces and equipment. For this purpose, symbols such as m represents the total number of machines and n represents the number of workpieces are defined. All symbols involved in the model are defined as follows: Machine Collection: , a total of m machines; Machine status: , its value meaning is 0: available, 1: waiting for maintenance; Maximum available load capacity of the machine: (unit: min); Artifact Collection: ,n artifacts; Latest completion time of workpiece: ; Required quantity of workpiece: Indicates workpiece Number of workpieces ; Workpiece process: Each workpiece Include The process is sorted by traversing the BOM tree in reverse order: , ; Process optional machine set: process The set of optional machines is ; Required quantity of each workpiece process: Indicates that the workpiece is calculated based on the BOM Corresponding to each process the number of Machining efficiency: Indicates the process Processing efficiency on optional machines (pieces / min) ); Processing start time: Indicates the process In the machine Start time on Processing end time: Indicates the process On the machine The end time of The calculation is obtained; at the same time, the constraint condition needs to be met: ; Machine allocation information: ; Workpiece on-time completion rate; since the system design generally requires the scheduling scheme to ensure that all workpieces are completed on time, the workpiece on-time completion rate is considered as an optimization index. The optimization function is defined as follows: (1) Where is a positive decimal , and the constant is a large positive number; Total time of all workpiece completion; in order to save time and improve efficiency, the minimum total time of all workpiece completion is taken as the optimization goal. The optimization function is defined as follows: (2) Number of machines used; in order to improve the service life of equipment and ensure the stable operation of equipment, only the minimum number of machine equipment is used each time, which is convenient for timely maintenance and staff rest, and the optimization function is defined as follows: (3) Where is a positive decimal .

[0032] Final objective function; since in reality enterprises have strong priority order for multiple targets in multi-objective optimization (such as "time first, efficiency second, cost third" in production optimization problem), similarly for the above three optimization objective functions , generally it has the following priority order (first meet optimization , on this basis meet optimization , and finally meet optimization ). Therefore, the hierarchical Pareto dominance relationship is selected to convert the above multi-objective optimization into a serialized single-objective optimization, which is closer to the decision logic of "first meet core demand, then optimize secondary indicators" in reality. The final objective function of the multi-objective flexible workpiece shop scheduling problem is defined as: (4) Wherein, are the optimization objective functions defined in formulas (1)-(3), C is a larger constant, and , ; through the setting of the above parameters , it is ensured that the algorithm first optimizes the core target, and then gradually optimizes the secondary target.

[0033] In constructing the multi-objective flexible job shop scheduling model, the constraint conditions of the multi-objective flexible job shop scheduling model need to be determined, and the constraint conditions are as follows: Process assignment uniqueness constraint: each process of a general workpiece must be assigned to a unique selectable machine for completion, that is, the process assignment uniqueness constraint is satisfied, that is, the following formula is satisfied: (5) Machine process uniqueness constraint: the process on each machine is unique in each period, that is, only one process can be processed at a time, so: (6) Process sequence constraint on machine: All processes on the machine must satisfy the before-and-after relationship of each process in the BOM tree of each workpiece, that is, When , it is necessary to judge whether the process and the process are on the same branch on the BOM tree of the first workpiece, if yes, return 0 (violation of constraint), otherwise return 1 (no violation of constraint).

[0034] Machine capacity limit: Since the maximum available load of each machine is limited (plus sometimes scheduled maintenance or advance occupation), the following constraint must be met in the given scheduling scheme: (7) Where the constant is the mold changing time of the machine when processing different processes.

[0035] Step 4, establish a scheduling scheme combining decision tree and genetic algorithm, perform selection, crossover and mutation operations through genetic algorithm, combine decision tree for conflict detection and repair to solve the objective function, and obtain the optimal solution of the objective function.

[0036] The establishment of the scheduling scheme combining the decision tree and the genetic algorithm, through the genetic algorithm to perform selection, crossover and mutation operations, combining the decision tree for conflict detection and repair to solve the objective function includes: Step 401: initialize algorithm parameters, take the minimum total time of all workpieces as the optimization target, and terminate the algorithm when the maximum number of iterations is reached; wherein the parameters mainly include the total number of population individuals P in the improved genetic algorithm, the number of algorithm iterations T; Step 402: adopt a double-layer chromosome coding mode to realize individual coding of the workpiece shop scheduling scheme, and perform subsequent genetic algorithm evolution on the population composed of these individuals to obtain the optimal solution; ​The individual coding scheme of all P job shop scheduling schemes in the population is as follows: The upper chromosome PS represents the processing sequence of the process, which is obtained by arranging and combining the job number and its BOM tree; wherein the upper process sequence (PS) is sorted by BOM tree inverse layer sequence, which ensures that the process sequence meets the process constraint, so as to optimize the on-time completion rate of the job (target 1).

[0037] Figure 2 The schematic diagram of the job BOM tree provided by the embodiment of the application is shown in the figure, A is the number of the process, A represents the last job code, A1 to A123 are the codes of its materials (sub jobs), in the figure, the whole BOM decision tree is located at the top of the last product, and below it are various parts in the manufacturing process.

[0038] The lower chromosome MA represents the machine allocation corresponding to each process, which is obtained by mapping the machine number; wherein the lower machine allocation (MA) is based on the production efficiency of the upper process selectable machine set in its corresponding position, selects a machine from the processable machine set by using the roulette method, reduces the total processing time (target 2), and at the same time improves the equipment centralization rate (target 3) through the load concentration strategy (selects the machine that meets the demand and has the smallest remaining capacity).

[0039] In addition, the coding design needs to be coordinated with the layered Pareto dominance relationship to ensure that the secondary target (such as energy consumption) is optimized under the premise of meeting the core target (such as meeting the time limit).

[0040] The job-machine relationship diagram is shown in the following table.

[0041]

[0042] The table gives an example, which contains 2 jobs and 8 machines, wherein the first job contains 4 processes, and the second job contains 3 processes, so the coding length of each individual upper and lower layer is 7. In the figure, J1:J2 is the job, 1:3 represents the process, and the corresponding processing efficiency of the machine in the bracket. It is assumed that the coding of the individual in the initialization solution population is [12 2 1 1 2 1] in the upper layer and [1 2 5 5 7 6 6] in the lower layer; the upper layer is the coding of the process sequence, which indicates that the first process of job 1, the first process of job 2, the second process of job 2, the second process of job 1, and so on are processed in sequence from left to right; the lower layer is the coding of the machine sequence, which indicates the machine number in the selectable machine set that can be used to process the process corresponding to the position in the upper layer.

[0043] Step 403: Calculate the fitness value of each chromosome in the sequence according to formula (4), calculate the Pareto optimal solution set of the current population using the hierarchical Pareto dominance relationship, save the top 5% chromosomes in the population to the historical optimal individual archive HOIA according to the Pareto ranking of each chromosome, and only save one individual for repeated individuals; Step 404: Determine whether the convergence criteria are met. If any of the convergence criteria are met, end the loop iteration and output the Pareto optimal solution to the historical optimal individual archive HOIA; otherwise, perform step 405. The convergence criteria are that all jobs can be completed on time and the optimal solution in the historical optimal individual archive is not updated within T2 generations or the number of iterations reaches T times.

[0044] Step 405: Calculate the fitness value of each chromosome in the sequence, and select Ps individuals into step 406 using roulette selection. Step 406: Perform crossover operations on the selected individuals according to the random pairing principle for the process layer PS and the machine layer MA, then perform decision tree conflict detection and repair, and update the optimal solution.

[0045] In this embodiment step, when performing crossover operations on individuals, the following hierarchical execution is performed: PS process layer crossover: using the position-based crossover (POX) method, randomly selecting a part of the process position of two parent individuals, copying the processes in these positions to the child individuals, and then filling the remaining processes according to the process order of the other parent individual; MA machine layer crossover: using the uniform crossover method, with a certain probability randomly exchange the machine allocation of the corresponding processes of the two parent individuals; wherein the crossover probability (Pc) is a key parameter for controlling whether the individual performs a crossover operation, and its value directly affects the exploration and development balance of the algorithm. Therefore, we choose to dynamically adjust Pc according to the evolution generation, population state, or individual fitness, as shown in formula (8): (8) where: t is the current generation, T is the total number of iterations, is the initial high probability (such as 0.9), is the late low probability (such as 0.5).

[0046] For the new population generated above, conflict detection and repair using a decision tree are needed: For each process-machine combination, extract the corresponding features (such as process processing time, current machine load, etc.) and input them into the decision tree model.

[0047] If the decision tree determines that there is a conflict (for example, the same machine is assigned to multiple processes at the same time), the machine for the conflicting process needs to be selected again, and feature extraction and conflict detection are performed again until the conflict is eliminated.

[0048] Through the above steps, the entire process of population crossover operation is completed, and it is ensured that the generated offspring chromosomes have feasibility in machine allocation.

[0049] Step 407: Perform mutation operation on the process layer PS and the machine layer MA in the population, then perform decision tree conflict detection and repair, and update HOIA; When mutating an individual, the following hierarchical execution is performed: PS process layer mutation: adopt exchange mutation method, with a certain probability Randomly select the position of the process in the mutation chromosome.

[0050] MA machine layer mutation: adopt roulette selection method, take the production efficiency of the machine available for the process as the calculation basis, and select a machine from the set of machines available for processing according to the roulette method for allocation.

[0051] Mutation probability (Pm) is a key parameter for controlling random mutation of individual genes, and its core role is to maintain population diversity, avoid premature convergence, and fine-tune the quality of local search. Therefore, we choose to adjust Pm nonlinearly according to the evolution number, as shown in equation (9): (9) In the formula: t is the current generation number, T is the iteration number, is the initial low probability, such as 0.01, which protects high-quality solutions, is the high probability in the later stage, such as 0.1, which breaks through the local optimum.

[0052] Similarly, the mutation operation (such as the neighborhood search mutation of the machine layer) will randomly change the machine allocation of a certain process, so after mutation, conflict detection needs to be performed on the process and the upstream and downstream processes that may be affected. Extract the feature vector from the mutated machine allocation and input it into the decision tree model; input the feature vector into the trained decision tree model, such as the binary tree constructed by the CART algorithm. Figure 3 The decision tree model structure diagram provided by the embodiment of the present application is shown in Figure 3 The decision tree model outputs a conflict label through the feature split node: Label = 0: machine allocation algorithm (time does not overlap and load does not exceed limit); Label = 1: conflict exists (such as total load exceeding capacity limit, etc.).

[0053] If the decision tree predicts a conflict (label = 1), trigger the machine to assign a repair mechanism, and adjust layer by layer according to the "priority selection → random selection → closed-loop feedback": Sort by priority from the optional machine set of the process: Prioritize efficient machines (highest production efficiency); If multiple efficient machines are available, select the one with the smallest remaining capacity to meet the process capacity (load concentration); If all efficient machines are in conflict, switch to the remaining machines, and also sort them by remaining capacity in descending order.

[0054] If the above selection fails (all optional machines are in conflict), randomly select one from the optional machines, and allow temporary overload (recorded as a conflict case to feed back to the decision tree); Regardless of the success or failure of the repair, the features before and after the variation, the conflict label, and the repair result are stored in the feedback dataset Dt, which is used to update the decision tree regularly (e.g., incremental training every 50 iterations) to avoid repeated conflicts; If the above repair is unsuccessful, then reselect machines for the processes that are in conflict, and perform feature extraction and conflict detection again until the conflict is eliminated.

[0055] Step 408: Combine the individual generated in step 407 with the remaining individuals in the population and the historical optimal solution (HOIA) to form a new population; including: Merge individuals: merge the individual selected in step 405 and the individual generated in step 407 with the individuals in the historical optimal individual archive (HOIA); De-duplication and sorting: remove duplicate individuals (same chromosome). Re-perform hierarchical Pareto dominance relationship on the merged individuals to generate a new ranking list; Prioritize preserving HOIA individuals: if there are Pareto frontier individuals in the HOIA that are not covered by the current population, they are prioritized for preservation; Fill the population: according to the results of the hierarchical Pareto dominance relationship, preferentially select individuals with low non-dominated levels until the population size P is reached to form a new population, which enters the next iteration. Until the number of iterations T is reached, output the final Pareto optimal solution, and complete the job shop scheduling.

[0056] The optimal solution of the target function includes: A data model is established for a multi-objective flexible job shop scheduling problem with different job final completion times, machine total load, and key machine energy efficiency. Through the data model, actual job shop scheduling cases are experimented and simulated to obtain the Pareto optimal solution.

[0057] Finally, a closed-loop feedback mechanism is introduced, which runs through the iteration process of the genetic algorithm evolution: Conflict case record: After the decision tree detects a conflict each time, the features before and after the conflict, the label, and the repair result are stored in the feedback dataset Dt.

[0058] Model incremental update: After every 50 iterations, the decision tree model is incrementally trained using Dt to improve its adaptability to new scenarios.

[0059] Parameter adaptive adjustment: The crossover probability P c and mutation probability P m of the genetic algorithm are dynamically adjusted according to the convergence state of the population (as in equations (8) and (9)), balancing global search and local development capabilities.

[0060] Through the closed-loop feedback mechanism, the algorithm is continuously optimized in the long-term running, avoiding the decline of the algorithm performance caused by the change of the production environment.

[0061] In summary, the workshop scheduling method integrating decision trees and genetic algorithms described in the embodiments of the present application solves the problems of insufficient dynamic balance and low conflict repair efficiency in multi-objective scheduling of the prior art through innovative points such as closed-loop feedback mechanism, hierarchical Pareto dominance relationship, and dynamic weight adjustment. The technical solution has significant progress in mathematical optimization logic, priority strategy, and dynamic learning ability, providing an efficient and scalable solution for flexible job shop scheduling. Through data modeling, actual workshop scheduling cases are experimented and simulated, and Pareto optimal solutions are obtained.

[0062] According to a second aspect of the present application, a workshop scheduling system integrating decision trees and genetic algorithms is provided, comprising: a data collection module for collecting workpiece process information and machine information, extracting process features and machine features; a decision tree construction module for constructing a decision tree based on the extracted process features and machine features; a target function construction module for constructing a multi-objective flexible job shop scheduling model and constructing a target function based on a hierarchical Pareto dominance relationship; a target function optimal solution output module for establishing a scheduling scheme combining decision trees and genetic algorithms, performing selection, crossover, and mutation operations through genetic algorithms, and combining decision trees for conflict detection and repair to solve the target function and obtain the optimal solution of the target function.

[0063] It can be understood that the workshop scheduling system integrating decision trees and genetic algorithms provided by the present application corresponds to the workshop scheduling method integrating decision trees and genetic algorithms provided by the aforementioned embodiments, and the related technical features of the workshop scheduling system integrating decision trees and genetic algorithms can refer to the related technical features of the workshop scheduling method integrating decision trees and genetic algorithms, which will not be repeated here.

[0064] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments by those skilled in the art once they have the benefit of the present disclosure. Therefore, the appended claims are intended to encompass within their scope all possible variations and modifications of the preferred embodiments. The present application may, however, be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The foregoing embodiments are therefore to be considered in all respects as illustrative only and not restrictive, the scope of the application being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

[0065] Finally, it should be noted that the above-mentioned merely preferred embodiments of the present application, and are not used to limit the present application, although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A workshop scheduling method integrating decision tree and genetic algorithm, characterized in that: The following steps are involved: Collect workpiece process information and machine information, and extract process features and machine features; Construct a decision tree based on the extracted process features and machine features; Construct a multi-objective flexible job shop scheduling model and build the objective function based on the hierarchical Pareto dominance relationship; A scheduling scheme combining decision tree and genetic algorithm is established. The selection, crossover and mutation operations are performed by genetic algorithm, and the decision tree is combined to perform conflict detection and repair to solve the objective function and obtain the optimal solution of the objective function.

2. The method for shop scheduling integrating decision tree and genetic algorithm according to claim 1, characterized in that: The collecting of workpiece process information and machine information and the extraction of process features and machine features include: Workpiece process information and machine information are collected through the manufacturing enterprise ERP system; the workpiece process information includes the workpiece number, the set of processable machines and the processing efficiency; the machine information includes the machine number and initial status; the process characteristics include the required number of workpieces and the deadline, and the machine characteristics include the current load and processing efficiency.

3. The workshop scheduling method integrating decision tree and genetic algorithm according to claim 1, characterized in that: The construction of a decision tree based on the extracted process features and machine features includes: In the training data generation phase, scheduling plans are generated using historical orders from the ERP system or manually added special case orders. The training set is constructed using the extracted process and machine features as input and conflict situations as output. When training the decision tree, we used the CART algorithm to train the data, set the parameters for the optimization model, and used 10-fold cross-validation to optimize the model, ensuring an accuracy rate exceeding 92% and an AUC value exceeding 0.

95. Generate decision rules based on the constraints of flexible job shop scheduling problem.

4. The method for shop scheduling integrating decision tree and genetic algorithm according to claim 1, characterized in that: The multi-objective flexible job shop scheduling model includes the optimization objectives of on-time completion rate of workpieces, minimization of completion time and minimum number of machines; the objective function constructed based on the hierarchical Pareto dominance relationship includes: Make sure to optimize your core goals first and then work your way down to your secondary goals.

5. The workshop scheduling method integrating decision tree and genetic algorithm according to claim 4, characterized in that: The objective function is expressed as follows: in, They are respectively the on-time completion rate optimization function, the total completion time optimization function and the minimum number of machines optimization function, and C is a constant, which ensures ,parameter satisfy .

6. The method for shop scheduling integrating decision tree and genetic algorithm according to claim 1, characterized in that: The constructing of the multi-objective flexible job shop scheduling model further includes: The constraints of the multi-objective flexible workpiece shop scheduling model are determined, wherein the constraints include: process allocation uniqueness constraint, process uniqueness constraint, process sequence constraint and machine capacity limitation.

7. The workshop scheduling method integrating decision tree and genetic algorithm according to claim 1, characterized in that: The objective function is solved by performing selection, crossover and mutation through a genetic algorithm combined with a decision tree for conflict detection and repair, including: Step 401: Initialize the algorithm parameters, with the optimization goal of minimizing the objective function defined by this method. The algorithm terminates when the maximum number of iterations is reached. The parameters include the total number of individuals in the population P and the number of algorithm iterations T in the improved genetic algorithm. Step 402: using a double-layer chromosome encoding method to implement individual encoding of the workpiece workshop scheduling plan, and performing subsequent genetic algorithm evolution on the population composed of these individuals to obtain the optimal solution; Step 403: Calculate the fitness value of each chromosome in the sequence, use the hierarchical Pareto dominance relationship to calculate the current population Pareto optimal solution set, and according to the Pareto ranking of each chromosome, save the chromosomes in the top 5% of the population into the optimal solution of the historical optimal individual archive; Step 404: Determine whether the convergence criteria are met. If any convergence criteria are met, terminate the loop iteration and output the Pareto optimal solution to the historical optimal individual archive optimal solution; otherwise, execute step 405. Step 405: Calculate the fitness value of each chromosome in the sequence and select P individuals using the roulette wheel selection method; Step 406: Perform crossover operations on the selected individuals according to the random pairing principle corresponding to process-level variation and machine-level variation, perform decision tree conflict detection and repair, and update the optimal solution; Step 407: Perform mutation operations on the process layer and machine layer in the population, perform decision tree conflict detection and repair, and update the optimal solution; Step 408: Combine the generated individuals and the remaining individuals in the population with the historical optimal solution to form a new population.

8. The workshop scheduling method integrating decision tree and genetic algorithm according to claim 7, characterized in that: When performing mutation operations on the process layer and machine layer in the population, the operations are performed in the following layers: PS process layer variation: using the exchange variation method, with a certain probability Randomly select the position of the process in the mutant chromosome; MA machine layer variation: The roulette wheel selection method is used to take the production efficiency of the optional machines in the process as the calculation basis, and a machine is selected from the set of processable machines for allocation according to the roulette wheel method.

9. The method for shop scheduling integrating decision tree and genetic algorithm according to claim 1, characterized in that: Obtaining the optimal solution of the objective function includes: A data model is established for the multi-objective flexible job shop scheduling problem with different job completion times, total machine load and key machine energy efficiency. Experiments and simulations are conducted on actual job shop scheduling cases using the data model to obtain the Pareto optimal solution.

10. A workshop scheduling system integrating decision tree and genetic algorithm, characterized in that: include: Data collection module, used to collect workpiece process information and machine information, and extract process features and machine features; A decision tree construction module is used to construct a decision tree based on the extracted process features and machine features; Objective function construction module, used to build a multi-objective flexible job shop scheduling model and construct the objective function based on the hierarchical Pareto dominance relationship; The objective function optimal solution output module is used to establish a scheduling scheme that combines decision trees and genetic algorithms. It performs selection, crossover and mutation operations through genetic algorithms, and combines decision trees to perform conflict detection and repair to solve the objective function and obtain the optimal solution of the objective function.

Images