A method and system for multi-objective optimization dynamic scheduling of flexible workshops

By constructing a multi-objective scheduling model using the adaptive NSGA-II algorithm and combining event-driven and periodic-driven mechanisms, the problem of multi-objective optimization in the dynamic scheduling of flexible workshops is solved, achieving stable production scheduling under uncertain environments and improving production efficiency and quality.

CN117331349BActive Publication Date: 2026-11-17BEIHANG UNIV
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202210726070.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-06-24
Publication Date
2026-11-17
Estimated Expiration
2042-06-24

AI Technical Summary

Technical Problem

Existing dynamic scheduling methods for flexible workshops fail to effectively consider various uncertainties, resulting in scheduling schemes that are not applicable in actual production and cannot meet the needs of multi-objective optimization.

Method used

An adaptive NSGA-II algorithm is used to construct a multi-objective scheduling model. Combining event-driven and periodic-driven mechanisms, a new scheduling scheme is generated by minimizing the total processing time, total processing cost, and total processing quality coefficient. The disturbance impact index of the workpiece is considered to determine whether dynamic scheduling should be performed.

Benefits of technology

It achieves stable scheduling of flexible workshops in uncertain environments, reduces system oscillations caused by frequent scheduling, meets multi-objective optimization requirements, and improves production efficiency and quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117331349B_ABST
    Figure CN117331349B_ABST
Patent Text Reader

Abstract

This invention relates to a multi-objective optimization dynamic scheduling method and system for flexible workshops, belonging to the field of flexible workshop scheduling technology. It solves the problem that existing technologies, which do not consider uncertain factors or only consider a single uncertain factor, cannot meet the actual production scheduling requirements. The method includes the following steps: constructing a scheduling model with the objectives of minimizing total processing time, total processing cost, and total processing quality coefficient, and adding constraints; determining whether to perform dynamic scheduling based on the scheduling cycle and the impact of disturbances; when the scheduling cycle is reached or a disturbance occurs and the disturbance impact index is greater than a threshold, generating a new scheduling scheme based on the scheduling model and constraints using the adaptive NSGA-II algorithm.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of flexible workshop scheduling technology, and in particular to a multi-objective optimization dynamic scheduling method and system for flexible workshops. Background Technology

[0002] With the intensification of globalization and the increasingly competitive market environment, the production model of simply ensuring production capacity and having only simple management is no longer suitable for manufacturing enterprises. Advanced production management is an effective tool for enhancing competitiveness. Its core content is the formulation of production plans and the scheduling of production workshops. The specific implementation of the formulated production plans depends on the specific scheduling of the lower-level production workshops. The Job-shop Scheduling Problem (JSP), as a classic combinatorial optimization problem, is a typical NP-hard problem. Its solution calculation process becomes extremely complex with the increase of problem size, and the computational amount of finding the optimal solution increases exponentially. With the development of the manufacturing industry, the production methods of manufacturing enterprises have begun to undergo fundamental changes, from a production model mainly based on long-term regular large-batch, single-variety products to a short-term, random, and sudden order-based production model. Therefore, the classic job-shop scheduling problem is also transforming into a multi-objective, practical, and dynamic Flexible Job-shop Scheduling Problem (FJSP).

[0003] Solving the dynamic scheduling problem of flexible job shops requires addressing one or more scheduling objectives, such as maximum completion time, maximum processing cost, delivery delay, average machine load, etc. Considering the complexity of the flexible job shop problem, most current domestic and international research focuses on scheduling problems with a single objective. However, in real-world flexible job shops, production and processing need to satisfy the interests of all parties, and the scheduling problem often falls under the category of multi-objective optimization (MOP), frequently requiring the simultaneous optimization of multiple objectives in actual production. Currently, the multi-objective optimization problem of flexible job shop scheduling is generally summarized into the following two methods.

[0004] (1) Weighted Single-Objective Optimization Method. Based on actual scheduling needs, different weights are assigned to each scheduling objective, and the multi-objective optimization problem is transformed into a single-objective optimization problem using a weighted method. Then, the optimal solution to the scheduling problem is obtained using methods for solving single-objective optimization problems. The weighted single-objective optimization method is essentially still a single-objective optimization method, not a true multi-objective optimization method. The determination of the weight coefficients during the optimization process greatly affects the obtained solution, and usually, a satisfactory optimal solution cannot be obtained.

[0005] (2) Pareto-based heuristic optimization method. The Pareto-based heuristic optimization method reflects the essence of multi-objective optimization: due to the conflict between objectives, the concept of a single optimal solution no longer exists in multi-objective optimization problems. Valid solutions, optimal solution sets, and satisfactory solutions are more likely to reflect complex real-world situations.

[0006] While Pareto-based heuristic optimization methods can effectively solve multi-objective optimization problems in flexible job shop scheduling, most of their optimization methods are limited to static scheduling. That is, the various conditions for establishing the flexible job shop scheduling model are idealized (such as machines not breaking down) and do not consider various uncertainties in the actual production process (such as the absence of temporary order insertions). Therefore, these optimization methods provide overly idealized solutions that do not match the actual production process.

[0007] In reality, various uncertainties and disturbances occur during the production process. Some descriptions and conditions in current flexible workshop scheduling problems are based on ideal production environments and do not reflect actual production conditions, making them unsuitable for dynamic scheduling research in flexible workshops. For example, the assumption that once a workpiece's processing step begins, it cannot be interrupted or preempted is a simplification of the static flexible workshop scheduling problem. However, it ignores the uncertainties that arise in actual production, such as machine malfunctions, unexpected order insertions, and material shortages. These changes in workshop conditions render the original scheduling plan infeasible, necessitating a rescheduling plan for subsequent production—a process known as rescheduling or dynamic scheduling.

[0008] However, current dynamic scheduling methods in flexible workshops almost never consider uncertainties or only consider a single uncertainty. For example, simply considering equipment failure or temporary order insertions is insufficient to address different uncertainties. Actual production conditions are complex and ever-changing, requiring consideration of more uncertainties and actual production goals. Current domestic and international flexible workshop scheduling methods are no longer adequate to meet these needs. Summary of the Invention

[0009] Based on the above analysis, the embodiments of the present invention aim to provide a multi-objective optimization dynamic scheduling method and system for flexible workshops, in order to solve the problem that existing dynamic scheduling does not consider uncertain factors or only considers a single uncertain factor and cannot meet the actual production scheduling requirements.

[0010] On one hand, embodiments of the present invention provide a multi-objective optimization dynamic scheduling method for flexible workshops, comprising the following steps:

[0011] A scheduling model is constructed with the objective of minimizing total processing time, total processing cost, and total processing quality coefficient, and constraints are added.

[0012] Whether to perform dynamic scheduling is determined based on the scheduling cycle and the impact of the disturbance; when the scheduling cycle is reached or a disturbance occurs and the disturbance impact index exceeds the threshold, then...

[0013] Based on the aforementioned scheduling model and constraints, an adaptive NSGA-II algorithm is used to generate a new scheduling scheme.

[0014] Based on further improvements to the above method,

[0015] The total processing time T is calculated using the following formula: T = max(ST) ijk +MT ijk ) = max(ET ijk );

[0016] The total processing cost C is calculated using the following formula:

[0017]

[0018]

[0019] Among them, ST ijk MT represents the time when the j-th process of workpiece i begins to be processed on machine k. ijk ET represents the processing time of the j-th operation on machine k for workpiece i. ijk This represents the processing time of the j-th operation on machine k for workpiece i, where mc i wc represents the raw material cost of workpiece i. k Let Num be the processing cost per unit time for machine k, n represent the number of workpieces, m represent the total number of machines, and Num is the processing cost per unit time for machine k. i O represents the number of operations contained in workpiece i. ijk This indicates whether the j-th process of workpiece i is processed on machine k.

[0020] Furthermore, the overall processing quality coefficient Q is calculated using the following formula:

[0021]

[0022]

[0023]

[0024] in, P represents the total cost of the first j processes of workpiece i. ijk I represents the failure rate of the j-th process in which machine k processes workpiece i. ijk MT represents the cost-weighted quality uncertainty coefficient for the j-th process in which machine k processes workpiece i. ijkThis represents the processing time of the j-th operation on machine k for workpiece i, where mc h wc represents the raw material cost of workpiece h. k O is the processing cost per unit time for machine k. ijk This indicates whether the j-th process of workpiece i is processed on machine k, n represents the number of workpieces, and m represents the total number of machines.

[0025] Furthermore, according to the formula Calculate the disturbance influence index δ of workpiece i i ;

[0026] According to the formula Calculate the total disturbance impact index δ;

[0027] Among them, LT i ET represents the estimated completion time of workpiece i after the disturbance; i Indicates the normal completion time of workpiece i; φ i φ represents the tolerance coefficient of workpiece i, and φ represents the total tolerance coefficient.

[0028] Furthermore, based on the aforementioned scheduling model and constraints, an adaptive NSGA-II algorithm is used to generate a new scheduling scheme, including:

[0029] S01. Chromosome encoding is performed based on the processing steps, and an initial population is randomly generated as the parent population based on the constraints.

[0030] S02. Calculate the fitness of each individual in the population based on the objective function, perform fast non-dominated sorting of the individuals in the population according to the fitness, and calculate the crowding distance of each individual.

[0031] S03. Select individuals for crossover mutation in the parent population based on the non-dominated ordering and the crowding distance of the individuals; perform crossover mutation on the selected individuals using an adaptive crossover mutation strategy to generate offspring; merge the parent population and the offspring population into a new population.

[0032] S04. If the maximum number of iterations is reached, the iteration ends. Calculate the Pareto optimal solution set of the current population and select the scheduling scheme in the Pareto optimal solution set as the new scheduling scheme. Otherwise, calculate the fitness of each individual in the population based on the objective function, perform fast non-dominated sorting on the individuals in the population according to the fitness and calculate the crowding distance of each individual. Select an individual as the new parent population in the new population according to the non-dominated sorting and the crowding distance of the individual. Return to step S03.

[0033] Furthermore, the adaptive crossover and mutation strategy is used to perform crossover and mutation on the selected individuals, including calculating the crossover probability and mutation probability using the following formulas:

[0034]

[0035]

[0036]

[0037] Where R1, R2 ∈ [0, 1], Let represent the crossover probability of the i-th generation population. Fit represents the mutation probability of the i-th generation population. i This represents the fitness value of an individual in the population. This represents the fitness value of the individual with the larger fitness value among the two crossover individuals; This represents the average fitness value of the i-th generation population. T represents the maximum fitness value of the i-th generation population. max Indicates the number of population iterations.

[0038] Furthermore, chromosome encoding based on processing steps includes:

[0039] The m machines are encoded using binary;

[0040] The length of the chromosome segment corresponding to workpiece i is determined based on the machine code length and the number of processing steps for workpiece i.

[0041] The chromosome fragments of n workpieces are spliced ​​together to form a single chromosome.

[0042] Furthermore, the constraints include:

[0043] Each machine can only process one workpiece at a time.

[0044] Each workpiece and each process can only be processed on one machine.

[0045] The processes for each workpiece have sequential constraints (ST). i(j+1)k -ST ijk -MT ijk ≥0;

[0046] The processing of each workpiece can be interrupted by disturbances, but processing can resume from the interruption point. Completed and uncompleted processes are unaffected.

[0047] Among them, ST jsk ST represents the time when the s-th process of workpiece j begins to be processed on machine k. ilk MT represents the time when the l-th process of workpiece i begins to be processed on machine k. jskMT represents the processing time of the s-th operation on machine k for workpiece j. ilk λ represents the processing time of the l-th operation on machine k for workpiece i, where λ is a positive number and x is a negative number. ijk This indicates whether workpiece i on machine k is processed before workpiece j. When workpiece i is processed before workpiece j on machine k, x indicates whether x is processed before workpiece j. ijk =1, otherwise x ijk =0.

[0048] On the other hand, embodiments of the present invention provide a flexible workshop multi-objective optimization dynamic scheduling system, including the following modules:

[0049] The model building module is used to construct a scheduling model and add constraints with the goal of minimizing the total processing time, total processing cost, and total processing quality coefficient.

[0050] The scheduling judgment module is used to determine whether to perform dynamic scheduling based on the scheduling cycle and the impact of disturbances; when the scheduling cycle is reached or a disturbance occurs and the disturbance impact index is greater than the threshold, dynamic scheduling is determined to be performed.

[0051] The dynamic scheduling module is used to generate new scheduling schemes based on the scheduling model and constraints using the adaptive NSGA-II algorithm.

[0052] Furthermore, according to the formula Calculate the disturbance influence index δ of workpiece i i ;

[0053] According to the formula Calculate the total disturbance impact index δ;

[0054] Among them, LT i ET represents the estimated completion time of workpiece i after the disturbance; i Indicates the normal completion time of workpiece i; φ i φ represents the tolerance coefficient of workpiece i, and φ represents the total tolerance coefficient.

[0055] Compared with existing technologies, this invention constructs a multi-objective scheduling model with the goal of minimizing total processing time, total processing cost, and total processing quality coefficient, thereby meeting the needs of actual production. By combining an event-driven mechanism (disturbance) with a periodic-driven mechanism, it maintains system stability. Furthermore, when disturbances occur, the method determines whether to perform scheduling based on the degree of disturbance, thus promptly handling unexpected events and preventing frequent scheduling due to frequent disturbances, keeping the system stable. Moreover, the method does not restrict the type of disturbance event, but only determines whether to perform dynamic scheduling based on its degree of disturbance, thus better aligning with actual production needs.

[0056] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0057] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0058] Figure 1 This is a flowchart of the multi-objective optimization dynamic scheduling method for flexible workshops according to an embodiment of the present invention;

[0059] Figure 2 This is a block diagram of the flexible workshop multi-objective optimization dynamic scheduling system according to an embodiment of the present invention;

[0060] Figure 3 This is a graph showing the change in the objective function value during the iteration process of an embodiment of the present invention;

[0061] Figure 4 This is a Gantt chart of the scheduling scheme in an embodiment of the present invention. Detailed Implementation

[0062] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0063] When uncertain disturbances occur during production in a flexible workshop, the original scheduling plan often becomes inapplicable or even impossible to continue production. In such cases, dynamic scheduling is necessary. The timing of dynamic scheduling in a production system is the driving mechanism for dynamic scheduling. Currently, commonly used dynamic scheduling mechanisms include event-driven, periodic-driven, and hybrid-driven mechanisms.

[0064] Event-driven dynamic scheduling refers to the immediate dynamic scheduling operation in a production system whenever a disturbance event occurs. The event-driven mechanism can respond promptly to sudden disturbances and quickly adjust to generate new production scheduling plans. However, in production environments with frequent disturbances, a single event-driven approach can lead to frequent scheduling adjustments due to consecutive disturbances, severely reducing system stability, i.e., system oscillation. Periodic-driven dynamic scheduling refers to performing dynamic scheduling operations at preset fixed scheduling periods. The periodic-driven mechanism is easy to implement in actual production and can maintain a certain level of system stability. However, it neglects sudden emergencies that may seriously impact production, cannot make timely adjustments based on changes in the uncertain environment, and lacks an effective event response mechanism. Hybrid-driven dynamic scheduling uses appropriate driving mechanisms for different disturbances, considering the impact of sudden disturbances on the production system between fixed scheduling points. The hybrid-driven mechanism is superior to event- and periodic-driven mechanisms to some extent. However, if the production environment in the workshop is disturbed too frequently, the dynamic scheduling frequency of the production system will be higher than the disturbance frequency, which can easily cause excessive system oscillation and make the system even more unstable.

[0065] A specific embodiment of the present invention discloses a multi-objective optimization dynamic scheduling method for flexible workshops, such as... Figure 1 As shown, it includes the following steps:

[0066] S1. Construct a scheduling model with the objective of minimizing total processing time, total processing cost, and total processing quality coefficient, and add constraints.

[0067] S2. Determine whether to perform dynamic scheduling based on the scheduling period and the impact of the disturbance; when the scheduling period is reached or a disturbance occurs and the disturbance impact index is greater than the threshold, then...

[0068] S3. Based on the aforementioned scheduling model and constraints, an adaptive NSGA-II algorithm is used to generate a new scheduling scheme.

[0069] A multi-objective scheduling model is constructed to minimize total processing time, total processing cost, and total processing quality coefficient, thereby meeting actual production needs. By combining an event-driven (disturbance) mechanism with a periodic-driven mechanism, system stability is maintained. Furthermore, when disturbances occur, the degree of disturbance influences determines whether scheduling should be performed, allowing for timely handling of unexpected events and preventing frequent scheduling due to frequent disturbances, thus maintaining system stability. Moreover, the method does not restrict the type of disturbance event, but only determines whether dynamic scheduling should be performed based on its degree of disturbance, thus better aligning with actual production requirements.

[0070] During implementation, for ease of understanding, the mathematical description of the model is as follows:

[0071] Let Job = {Job1, Job2, ..., Job3} be the job to be processed. n Each workpiece contains one or more processing steps; there are m available machines Mac = {Mac1, Mac2, ..., Mac...} m};Num i Indicates the number of processes contained in workpiece i; O ij Let represent the j-th process of workpiece i; each process of workpiece i can be processed on multiple machines with different performance, and the set of available machines for the j-th process of workpiece i is . Let parameter O ijk The value is 0 or 1, if O ijk =1 indicates that the j-th process of workpiece i is processed on machine k; otherwise, it indicates that it is not processed on machine k, 1≤i≤n, 1≤j≤Num i 1≤k≤m; The processing time, cost, and quality of each step in the workpiece's operation vary depending on the machine's performance. ST ijk MT represents the time when the j-th process of workpiece i begins to be processed on machine k. ijk ET represents the processing time of the j-th operation on machine k for workpiece i. ijk This represents the processing time of the j-th operation on machine k for workpiece i, where ET ijk =ST ijk +MT ijk The workpiece is processed strictly according to the technological route, and several machines can be selected to perform each operation. Under specific constraints, all operations of the workpiece must be rationally arranged on the machines, and the scheduling scheme must achieve optimal performance indicators.

[0072] As the core of manufacturing enterprises, flexible workshops need to consider the interests of all parties and strive to meet requirements such as shortening production cycles, on-time delivery, reducing costs, and improving production efficiency. Therefore, the scheduling problem of flexible workshops has multiple scheduling optimization objectives. Currently, commonly used scheduling objectives include: minimizing the maximum completion time, minimizing production and processing costs, and minimizing the maximum machine load.

[0073] This invention, while considering common scheduling objectives, proposes a low-carbon optimization strategy that considers reducing machine energy consumption and a high-quality production strategy that considers workpiece processing quality, based on the actual production and processing environment. The multi-objective optimization scheduling of the flexible workshop in this paper uses time, cost, and quality as scheduling objectives, namely, total processing time T, total processing cost C, and total processing quality coefficient Q.

[0074] Specifically, the total processing time can be expressed as the maximum completion time. Therefore, the total processing time T is calculated using the following formula: T = max(ST) ijk +MTijk ) = max(ET ijk );

[0075] Processing costs generally include the total raw material cost MC of the workpiece and the total processing cost WC of all workpiece processes. Therefore, the total processing cost C is calculated using the following formula:

[0076]

[0077]

[0078] Among them, ST ijk MT represents the time when the j-th process of workpiece i begins to be processed on machine k. ijk ET represents the processing time of the j-th operation on machine k for workpiece i. ijk This represents the processing time of the j-th operation on machine k for workpiece i, where mc i wc represents the raw material cost of workpiece i. k Let Num be the processing cost per unit time for machine k, n represent the number of workpieces, m represent the total number of machines, and Num is the processing cost per unit time for machine k. i O represents the number of operations contained in workpiece i. ijk This indicates whether the j-th process of workpiece i is processed on machine k.

[0079] The quality of processing is often characterized by the product defect rate, while in a specific workshop processing step, it can be characterized by the machine failure rate. Different machines have different processing capabilities, and the failure rate varies depending on the processing step. A higher failure rate indicates that the quality of a particular process processed by that machine is less stable and more likely to produce defective products.

[0080] Furthermore, the workpiece typically involves multiple processing steps, and the processing cost increases as processing progresses. If processing defects occur in later stages, it can lead to significant losses; the more processing steps and the deeper the processing, the greater the impact on processing quality. Therefore, this invention proposes using a process cost-weighted quality uncertainty coefficient I to represent the workpiece's processing quality. The total processing quality coefficient Q is calculated using the following formula:

[0081]

[0082]

[0083]

[0084] in, P represents the total cost of the first j processes of workpiece i. ijk I represents the failure rate of the j-th process in which machine k processes workpiece i.ijk MT represents the cost-weighted quality uncertainty coefficient for the j-th process in which machine k processes workpiece i. ijk This represents the processing time of the j-th operation on machine k for workpiece i, where mc h wc represents the raw material cost of workpiece h. k O is the processing cost per unit time for machine k. ijk This indicates whether the j-th process of workpiece i is processed on machine k, n represents the number of workpieces, and m represents the total number of machines.

[0085] The cost-weighted quality uncertainty coefficient I increases with the number of processing steps. By calculating the processing quality coefficient based on the cost-weighted quality uncertainty coefficient, the actual quality can be expressed more accurately and effectively, thus meeting manufacturing requirements.

[0086] Specifically, the scheduling objective is to determine the processing sequence of all workpieces and the corresponding processing machines based on the following assumptions and constraints, so as to achieve the optimal scheduling objectives.

[0087] ① All machines are available immediately;

[0088] ②Each machine can only process one workpiece at a time;

[0089] ③ There are process constraints on the order of processing between the processes of each workpiece, that is, process j+1 can only be processed after process j is completed. There are no process constraints between processes of different workpieces.

[0090] ④ Each workpiece can only be processed on one machine at a time;

[0091] ⑤ The processing technology for all workpieces is determined, and all workpieces have the same priority;

[0092] ⑥ The processing time of all workpieces on different machines is known, and this time already includes the setup time of the processing operations.

[0093] ⑦ The processing of each workpiece process can be interrupted by uncertain factors (hereinafter referred to as disturbances), and processing can continue from the point of interruption. The completed and unprocessed processes of the workpiece are not affected in any way.

[0094] ⑧ Personnel must be present when the machine is running.

[0095] ⑨ Scheduling objective: F = (f1, f2, ..., f S );

[0096] Among them, constraint ② allows each machine to process only one workpiece at a time, expressed by the mathematical formula. express.

[0097] Among them, ST jsk ST represents the time when the s-th process of workpiece j begins to be processed on machine k. ilk MT represents the time when the l-th process of workpiece i begins to be processed on machine k. jsk MT represents the processing time of the s-th operation on machine k for workpiece j. ilk λ represents the processing time of the l-th operation on machine k for workpiece i, where λ is a positive number and x is a negative number. ijk This indicates whether workpiece i on machine k is processed before workpiece j. When workpiece i is processed before workpiece j on machine k, x indicates whether x is processed before workpiece j. ijk =1, otherwise x ijk =0. In practice, λ should be a large positive number, for example, a number greater than the total processing time of the process to be processed.

[0098] Constraint ④ Each workpiece and each process can only be processed on one machine at a time, using a mathematical formula. express;

[0099] Constraint ③: The sequence of operations for each workpiece is constrained using the mathematical formula ST. i(j+1)k -ST ijk -MT ijk ≥0 indicates;

[0100] It should be noted that constraint ⑦ is to respond to uncertain disturbances that may affect the processing during production and ensure the continuity of dynamic scheduling. Interrupted processing steps can continue to be processed from the point of interruption by the original machine or another machine according to the proportion of processing already completed.

[0101] For example, the processing time for the first step of workpiece 3 on machine 2 is MT. 312 At a certain moment, a machine malfunctions. Machine 2 is currently processing the first step of workpiece 3, and has completed 2 units of time (i.e., 1 / 4 of the first step of workpiece 3 has been completed). At this point, an interrupt mechanism is triggered, and dynamic scheduling is implemented. After dynamic scheduling, machine 1 resumes processing the first step of workpiece 3 from the 1 / 4 completion interruption point. If the original processing time for the first step of workpiece 3 was 8 units of time, then machine 1 only needs to process the remaining 3 / 4 (MT). 311 =6). Therefore, based on such assumptions, the scheduling model proposed in this invention introduces dynamic uncertainties that are not considered in the traditional flexible job shop scheduling model, thus providing a prerequisite for the feasibility of subsequent dynamic scheduling.

[0102] The objective function, assumptions, and constraints described above constitute a complete model of the multi-objective dynamic scheduling problem in flexible workshops, and ensure the feasibility and effectiveness of the scheduling results.

[0103] When performing multi-objective optimization and dynamic scheduling, a hybrid driving mechanism will be adopted first, and a disturbance impact index will be introduced to determine whether dynamic scheduling should be performed based on the degree of disturbance impact. This avoids frequent dynamic scheduling due to frequent disturbances in the production and processing environment, which could cause system oscillations and affect processing. Furthermore, notifications will ensure timely response to severe disturbances, preventing losses.

[0104] Specifically,

[0105] According to the formula Calculate the disturbance influence index δ of workpiece i i ;

[0106] According to the formula Calculate the total disturbance impact index δ;

[0107] Among them, LT i ET represents the estimated completion time of workpiece i after the disturbance; i Indicates the normal completion time of workpiece i; φ i φ represents the tolerance coefficient of workpiece i, and φ represents the total tolerance coefficient.

[0108] During implementation, a first threshold can be preset based on the company's production experience and business objectives. Preset corresponding thresholds according to the manufacturing requirements of each workpiece.

[0109] In actual production, delivery dates for different orders typically vary. Therefore, even if the overall completion time of the production plan is delayed within an acceptable threshold, a particular workpiece may experience an excessively long delivery delay exceeding expectations, necessitating dynamic scheduling. Therefore, this invention, in addition to... In addition to the threshold for the production plan disturbance impact index, for workpiece i, we also set the individual disturbance impact index of workpiece i as δ. i The difference between its delivery date and the original scheduled completion time of the workpiece is multiplied by the corresponding tolerance coefficient φ. i To set the threshold of the disturbance influence index for workpiece i

[0110] If the current time reaches the scheduling cycle, dynamic scheduling is performed. If a disturbance occurs, the disturbance impact index δ for each workpiece is calculated. i And the total disturbance influence index δ, if the disturbance influence index δ of workpiece i i If the threshold corresponding to workpiece i is greater than the threshold or the total disturbance influence index δ is greater than the first threshold, then proceed to step S3 to generate a new scheduling scheme based on the scheduling model and constraints using the adaptive NSGA-II algorithm.

[0111] The NSGA-II algorithm is suitable for solving multi-objective optimization problems; therefore, this application generates a new scheduling scheme based on the NSGA-II algorithm. Specifically, a new scheduling scheme is generated using an adaptive NSGA-II algorithm based on the aforementioned scheduling model and constraints, including:

[0112] S01. Chromosome encoding is performed based on the processing steps, and an initial population is randomly generated as the parent population based on the constraints.

[0113] Specifically, chromosome encoding based on processing steps includes:

[0114] S011. Encode m machines using binary;

[0115] First, encode the m machines using binary. For example, if there are 6 machines, a 3-bit binary code can be used, where 001 represents machine 1 and 110 represents machine 6.

[0116] S012. Determine the length of the chromosome segment corresponding to workpiece i based on the machine code length and the number of processing steps for workpiece i.

[0117] Because the scheduling model includes processing sequence constraints, the number and order of processing steps for a given workpiece are fixed. For example, workpiece A includes four steps: A1, A2, A3, and A4, with the step sequence being A1A2A3A4. Since all steps will fall on their corresponding machines, determining which step corresponds to which machine from beginning to end constitutes the chromosome segment encoding for workpiece A.

[0118] For example, processes A1, A2, A3, and A4 correspond to machines 1, 2, 3, and 4, respectively. Then, the chromosome segment code corresponding to workpiece A is: 001|010|011|100, or 001010011100, with a length of 12 bits.

[0119] The length of the chromosome segment corresponding to workpiece A can be obtained by multiplying the machine code length by the number of processing steps for workpiece A.

[0120] S013. Assemble the chromosome fragments of n workpieces into a single chromosome.

[0121] Once the chromosome code length of each workpiece is determined, the chromosome segments of n workpieces are spliced ​​together to form a single chromosome.

[0122] After chromosome encoding, initialization parameters need to be determined before randomly generating the initial population, including population size N, maximum number of iterations, selection probability, initial crossover probability, and initial mutation probability.

[0123] Meanwhile, constraints need to be considered when randomly generating the initial population. For example, randomly generated chromosomes may not satisfy the constraints ② Each machine can only process one workpiece at a time. Then the chromosome is removed, and chromosomes are randomly generated again until all chromosomes in the population meet the constraints and the number is N.

[0124] Specifically, the processing time for each step is fixed. Therefore, based on the machine corresponding to each step in the chromosome and the processing time of each step, the start and end times of processing can be calculated, and then the above formula can be used to determine whether the constraints are met.

[0125] S02. Calculate the fitness of each individual in the population based on the objective function, perform fast non-dominated sorting of the individuals in the population according to the fitness, and calculate the crowding distance of each individual.

[0126] Specifically, the fitness of each individual in the population is calculated based on the objective function determined in step S1, namely the total processing time T, total processing cost C, and total processing quality coefficient Q for each chromosome.

[0127] The key to multi-objective optimization problems lies in finding the Pareto optimal solution set. NSGA-II fast nondominated sorting stratifies the population N based on the non-dominated solution level of individuals to obtain F. i The purpose is to bring the solution closer to the Pareto optimal solution. This is a cyclical fitness value ranking process. First, the non-dominated solution set in the population is found, denoted as F1, and all individuals in it are assigned a non-dominated order i. rank =1 (where i rank Let i be the non-dominated ordinal value of individual i, and remove it from the entire population N. Then, continue to find the non-dominated solution set in the remaining population, denoted as F2. Individuals in F2 are assigned i. rank =2, and so on, until the entire population is stratified, F i The non-dominated order values ​​are the same in the layer.

[0128] The meaning of domination is: if individual i dominates individual j, it means that individual i is no worse than individual j in all objective functions, and is better than individual j in at least one objective.

[0129] Non-dominance means that an individual is the best in the population and is not dominated by the other individuals in the population.

[0130] On the same floor F iSelective sorting is required, based on the size of the individual's crowding distance. The closer the distance, the higher the density of the region where the individual resides, and it is assigned a smaller value; conversely, the farther the distance, the lower the density of the region where the individual resides, and it is assigned a larger value. This increases the probability of the individual being selected for the next generation, thus ensuring the diversity of the chromosome set during the genetic process. The crowding distance of each individual is calculated by summing the distance differences between its two neighboring individuals on each sub-objective function. The individual crowding distance is F... i The distance between individuals j+1 and j-1 that are adjacent to individual j is calculated as follows:

[0131] ① Initialize the distance between individuals in the same layer, let L[j] d =0 (representing the crowding distance of any individual j).

[0132] ② Individuals in the same layer are sorted in ascending order according to the s-th objective function value.

[0133] ③ Individuals on the edge of the ranking should be given a selection advantage.

[0134] ④ For the individual in the middle of the sort, calculate the crowding distance:

[0135] L[j] d =L[j] d +(L[j+1] s -L[j-1] s ) / (f s max -f s min )

[0136] Where: L[j+1] s f is the s-th objective function value for the (j+1)-th individual. s max f s min These are the maximum and minimum values ​​of the s-th objective function in the set, respectively.

[0137] ⑤ For different objective functions, repeat steps ② to ④ to obtain the crowding distance L[j] for individual j. d Prioritizing individuals with larger crowding distances can ensure that the calculation results are evenly distributed in the target space, thus maintaining the diversity of the population.

[0138] S03. Select individuals for crossover mutation in the parent population based on the non-dominated ordering and the crowding distance of the individuals; perform crossover mutation on the selected individuals using an adaptive crossover mutation strategy to generate offspring; merge the parent population and the offspring population into a new population.

[0139] Specifically, individuals with low non-dominated ranking and large crowding distances are given priority.

[0140] Existing methods also employ adaptive crossover and mutation operators, which utilize the average fitness value of the current generation population. Maximum fitness value of the population The difference relationship To determine whether the algorithm has converged, because when the population tends to converge... The value is greater than when the population is dispersed in the solution space. The value is small. At this point, the probability of the contemporary crossover operator is... and mutation operator probability They are shown below:

[0141]

[0142]

[0143] Where R1, R2 ∈ [0, 1], are used to restrict the crossover operator probabilities. and mutation operator probability The range of crossover and mutation probabilities. From formulas (2-13) and (2-14), it can be seen that the probabilities of crossover and mutation are determined solely by the maximum and average fitness values ​​of the population. Therefore, regardless of whether an individual has a high fitness value or a low fitness value, their crossover and mutation probabilities are the same. While this solves the problem of getting trapped in local optima, it may also prevent the algorithm from ever converging to the global optimum. To address this issue, the fitness of individuals needs to be considered simultaneously. Individuals with fitness higher than the average fitness should be assigned a lower crossover and mutation probability to protect potentially superior genes from being destroyed; while individuals with fitness lower than the average fitness should be assigned a higher crossover and mutation probability to eliminate them as quickly as possible. Therefore, this invention improves formulas (2-13) and (2-14) as follows:

[0144]

[0145]

[0146] Where R1, R2, R3, R4 ∈ [0, 1], This refers to the individual with the higher fitness value among two crossover individuals. i This represents the fitness value of an individual in the population.

[0147] From formulas (2-15) and (2-16), it can be observed that the closer the fitness value is to the maximum fitness value in the population, the smaller the crossover and mutation probabilities. When the fitness value equals the maximum fitness value in the current population, the crossover and mutation probabilities become 0. This improvement is more suitable for the later stages of evolution, but in the early stages of algorithm evolution, individuals in the population close to the maximum fitness value will hardly undergo crossover and mutation. At this time, the individual with the maximum fitness value is not necessarily the global optimum, thus increasing the probability of the algorithm getting trapped in a local optimum. To reduce the risk of getting trapped in a local optimum, this invention further improves formulas (2-15) and (2-16) to increase the crossover and mutation probabilities in the early stages of the algorithm. Specifically, the crossover and mutation probabilities are calculated using the following formulas:

[0148]

[0149]

[0150]

[0151] Where R1, R2 ∈ [0, 1], Let represent the crossover probability of the i-th generation population. Fit represents the mutation probability of the i-th generation population. i This represents the fitness value of an individual in the population. This represents the fitness value of the individual with the larger fitness value among the two crossover individuals; This represents the average fitness value of the i-th generation population. T represents the maximum fitness value of the i-th generation population. max Indicates the number of population iterations.

[0152] The fitness value is calculated by determining the total processing time T, total processing cost C, and total processing quality coefficient Q based on the objective function, assigning weights to each of the three objectives using a weighting method, and then summing them to obtain the fitness value. Therefore, the fitness value calculation requires individual normalization, that is, mapping the data of T, C, and Q to the range of 0 to 1, so that they are all on the same order of magnitude.

[0153] Normalization formula: x represents the calculated value of a certain objective function for a specific individual, where x is the maximum value of the objective function. max The minimum value x of a certain objective function min , where x′ represents the normalized value.

[0154] By increasing the adaptive factor and This effectively improves the early stages of evolution, where individuals nearing maximum fitness rarely crossover, thus reducing the probability of the algorithm getting trapped in local optima. Furthermore, in the later stages of the algorithm... and The value will approach 0, and the algorithm will evolve into the adaptive NSGA-II algorithm again, maintaining the advantages of the adaptive NSGA-II algorithm in the later stages.

[0155] Offspring are generated based on the crossover and mutation probabilities mentioned above; the parent and offspring populations are then merged into a new population.

[0156] S04. If the maximum number of iterations is reached, the iteration ends. Calculate the Pareto optimal solution set of the current population and select the scheduling scheme in the Pareto optimal solution set as the new scheduling scheme. Otherwise, calculate the fitness of each individual in the population based on the objective function, perform fast non-dominated sorting on the individuals in the population according to the fitness and calculate the crowding distance of each individual. Select an individual as the new parent population in the new population according to the non-dominated sorting and the crowding distance of the individual. Return to step S03.

[0157] Determine if the maximum number of iterations has been reached; if so, stop iterating. Calculate the Pareto optimal solution set for the current population. It should be noted that before calculating the optimal solution set, invalid individuals need to be removed based on whether the constraints, processing type of the operation, and the corresponding machine's processing capacity match. For example, if the processing type of operation A1 is polishing and the assigned machine's processing capacity is welding, then that chromosome should be removed.

[0158] The Pareto optimal set is the non-dominated set, meaning the individuals in this set are the best in the population and are not dominated by other individuals. The scheduling scheme corresponding to any individual with a specific chromosome from this set can be selected as a new scheduling scheme.

[0159] If the maximum number of iterations has not been reached, the fitness of each individual in the population is calculated using the method in step S02. Based on the fitness, the individuals in the population are subjected to a fast non-dominated sort and the crowding distance of each individual is calculated. Based on the non-dominated sort and the crowding distance of the individuals, individuals are selected as the new parent population in the new population, and the process returns to step S03 to continue the next iteration.

[0160] To verify the effectiveness of the adaptive NSGA-II algorithm proposed in this invention—referred to as the ADMNSGA-II algorithm—an experiment was conducted using a real machining workshop with 6 machines processing 6 workpieces. Specific data is shown in Table 1. The data in the table includes: processing time, processing quality, raw material cost, and equipment unit time processing cost for each process on different machines. "-" indicates no such data. Mac1 to Mac6 represent the 6 machines, Job1 to Job6 represent the 6 workpieces to be processed, and O... ij This represents the j-th process of the i-th workpiece; mc i This represents the raw material cost of workpiece i. The data in the table, for example, O... 11 The following 12 / 3 / 0.1 indicates that the first operation of workpiece 1 takes 12 units of time to process on Mac1, costs 3 units of money, and has a processing quality coefficient of 0.1. As described above, the multi-objective dynamic scheduling model optimizes the total processing time T, total processing cost C, and total processing quality coefficient Q. The relevant parameters are set as follows: population size 100, maximum number of generations 200, initial crossover rate 0.9, and initial mutation rate 0.1. Furthermore, the weighting coefficients for T, C, and Q are set as w1 = 0.5, w2 = 0.3, and w3 = 0.2, respectively, and the optimal solution set is calculated.

[0161] Table 1. Example data from a machining workshop

[0162]

[0163]

[0164]

[0165] The results of solving Table 1 are as follows Figure 3 As shown. From Figure 3 As can be seen, with the increase of the algorithm's algebraic number, the optimal values ​​of each objective are gradually optimized, indicating that the algorithm of this invention has good convergence. The ADMNSGA-II algorithm finally obtains a Pareto solution set with 100 individuals, providing a series of feasible optimized solutions for production scheduling. Finally, the ADMNSGA-II algorithm yields a satisfactory scheduling scheme as follows: Figure 4 As shown. From Figure 4 As can be seen, the scheduling plan generated by this invention not only meets the scheduling objective but also conforms to the actual generated schedule. This experiment verifies the effectiveness of this invention.

[0166] A specific embodiment of the present invention discloses a multi-objective optimization dynamic scheduling system for flexible workshops, such as... Figure 1 As shown, it includes the following modules:

[0167] The model building module is used to construct a scheduling model and add constraints with the goal of minimizing the total processing time, total processing cost, and total processing quality coefficient.

[0168] The scheduling judgment module is used to determine whether to perform dynamic scheduling based on the scheduling cycle and the impact of disturbances; when the scheduling cycle is reached or a disturbance occurs and the disturbance impact index is greater than the threshold, dynamic scheduling is determined to be performed.

[0169] The dynamic scheduling module is used to generate new scheduling schemes based on the scheduling model and constraints using the adaptive NSGA-II algorithm.

[0170] Preferably, according to the formula Calculate the disturbance influence index δ of workpiece i i ;

[0171] According to the formula Calculate the total disturbance impact index δ;

[0172] Among them, LT i ET represents the estimated completion time of workpiece i after the disturbance; i Indicates the normal completion time of workpiece i; φ i φ represents the tolerance coefficient of workpiece i, and φ represents the total tolerance coefficient.

[0173] The above-described method and system embodiments are based on the same principles, and their related aspects can be referenced from each other to achieve the same technical effects. For specific implementation processes, please refer to the foregoing embodiments, which will not be repeated here.

[0174] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0175] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A flexible job shop multi-objective optimization dynamic scheduling method, characterized in that, Includes the following steps: A scheduling model is constructed with the objective of minimizing the total processing time, total processing cost, and total processing quality coefficient, and constraints are added. Whether to perform dynamic scheduling is determined based on the scheduling cycle and the impact of the disturbance; when the scheduling cycle is reached or a disturbance occurs and the disturbance impact index exceeds the threshold, then... Based on the aforementioned scheduling model and constraints, an adaptive NSGA-II algorithm is used to generate a new scheduling scheme. Dynamic scheduling is performed when the scheduling cycle is reached; if a disturbance occurs, the disturbance impact index for each workpiece is calculated. Total Disturbance Impact Index If the workpiece Disturbance impact index The disturbance influence index threshold corresponding to workpiece i is greater than or total disturbance impact index If the value exceeds the first threshold, dynamic scheduling will be performed. According to the formula Calculate the workpiece Disturbance impact index ; According to the formula Calculate the total disturbance impact index ; in, Indicates workpiece Estimated completion time after disturbance; Indicates workpiece Normal completion time; Indicates workpiece Tolerance coefficient, This represents the overall tolerance coefficient; According to the workpiece The difference between the delivery date and the original scheduled completion time of the workpiece is multiplied by the corresponding tolerance factor. Setting the workpiece Disturbance impact index threshold ; The crossover and mutation probabilities of the adaptive NSGA-II algorithm are calculated using the following formulas: , , , , in, , Let represent the crossover probability of the i-th generation population. This represents the mutation probability of the i-th generation population. This represents the fitness value of an individual in the population. This represents the fitness value of the individual with the larger fitness value among the two crossover individuals; This represents the average fitness value of the i-th generation population. This represents the maximum fitness value of the i-th generation population. Indicates the number of population iterations.

2. The multi-objective optimization dynamic scheduling method for flexible workshops according to claim 1, characterized in that, The total processing time is calculated using the following formula. : ; The total processing cost is calculated using the following formula. : , , , in, Indicates workpiece The This process is done on the machine. The time when processing begins. Indicates workpiece The This process is done on the machine. Processing time on Indicates workpiece The This process is done on the machine. The above processing time, of which Indicates workpiece The cost of raw materials, For machines The processing cost per unit time, where n represents the number of workpieces and m represents the total number of machines. Indicates workpiece The number of processes included Indicates workpiece The Whether the process is carried out on machine k.

3. The multi-objective optimization dynamic scheduling method for flexible workshops according to claim 1, characterized in that, The overall processing quality coefficient is calculated using the following formula. : , in, Indicates workpiece The former All costs of each process, Indicates machine Machining workpieces The The failure rate of each process, Indicates machine Machining workpieces The Cost-weighted quality uncertainty coefficient of each process Indicates workpiece The This process is done on the machine. The processing time on the surface, of which Indicates workpiece The cost of raw materials, For machines Processing cost per unit time Indicates workpiece The Whether a process is performed on machine k, n represents the number of workpieces, and m represents the total number of machines.

4. The multi-objective optimization dynamic scheduling method for flexible workshops according to claim 1, characterized in that, Based on the aforementioned scheduling model and constraints, a new scheduling scheme is generated using the adaptive NSGA-II algorithm, including: S01. Chromosome encoding is performed based on the processing steps, and an initial population is randomly generated as the parent population based on the constraints. S02. Calculate the fitness of each individual in the population based on the objective function, perform fast non-dominated sorting of the individuals in the population according to the fitness, and calculate the crowding distance of each individual. S03. Select individuals for crossover mutation in the parent population based on the non-dominated ordering and the crowding distance of the individuals; perform crossover mutation on the selected individuals using an adaptive crossover mutation strategy to generate offspring; merge the parent population and the offspring population into a new population. S04. If the maximum number of iterations is reached, the iteration ends. Calculate the Pareto optimal solution set of the current population and select the scheduling scheme in the Pareto optimal solution set as the new scheduling scheme. Otherwise, calculate the fitness of each individual in the population based on the objective function, perform fast non-dominated sorting on the individuals in the population according to the fitness and calculate the crowding distance of each individual. Select an individual as the new parent population in the new population according to the non-dominated sorting and the crowding distance of the individual. Return to step S03.

5. The multi-objective optimization dynamic scheduling method for flexible workshops according to claim 4, characterized in that, Chromosome encoding based on processing steps includes: The m machines are encoded using binary; The length of the chromosome segment corresponding to workpiece i is determined based on the machine code length and the number of processing steps for workpiece i. The chromosome fragments of n workpieces are spliced ​​together to form a single chromosome.

6. The multi-objective optimization dynamic scheduling method for flexible workshops according to claim 1, characterized in that, The constraints include: Each machine can only process one workpiece at a time. ; Each workpiece and each process can only be processed on one machine. ; The processes for each workpiece have a sequential constraint. ; The processing of each workpiece can be interrupted by disturbances, but processing can resume from the interruption point. Completed and uncompleted processes are unaffected. in, Indicates workpiece The This process is done on the machine. The time when processing begins. Indicates workpiece The This process is done on the machine. The time when processing begins. Indicates workpiece The This process is done on the machine. Processing time on Indicates workpiece The This process is done on the machine. Processing time on It is a positive number. Indicates machine upper workpiece Is it more than the workpiece? Process first, then use the machine. upper workpiece Compared to the workpiece Preprocessing ,otherwise .

7. A flexible workshop multi-objective optimization dynamic scheduling system, characterized in that, Includes the following modules: The model building module is used to construct a scheduling model and add constraints with the goal of minimizing the total processing time, total processing cost, and total processing quality coefficient. The scheduling judgment module is used to determine whether to perform dynamic scheduling based on the scheduling cycle and the impact of disturbances; when the scheduling cycle is reached or a disturbance occurs and the disturbance impact index is greater than the threshold, dynamic scheduling is determined to be performed. The dynamic scheduling module is used to generate a new scheduling scheme based on the scheduling model and constraints using the adaptive NSGA-II algorithm. Dynamic scheduling is performed when the scheduling cycle is reached; if a disturbance occurs, the disturbance impact index for each workpiece is calculated. Total Disturbance Impact Index If the workpiece Disturbance impact index The disturbance influence index threshold corresponding to workpiece i is greater than or total disturbance impact index If the value exceeds the first threshold, dynamic scheduling will be performed. According to the formula Calculate the workpiece Disturbance impact index ; According to the formula Calculate the total disturbance impact index ; in, Indicates workpiece Estimated completion time after disturbance; Indicates workpiece Normal completion time; Indicates workpiece Tolerance coefficient, This represents the overall tolerance coefficient; According to the workpiece The difference between the delivery date and the original scheduled completion time of the workpiece is multiplied by the corresponding tolerance factor. Setting the workpiece Disturbance impact index threshold ; The crossover and mutation probabilities of the adaptive NSGA-II algorithm are calculated using the following formulas: , , , , in, , Let represent the crossover probability of the i-th generation population. This represents the mutation probability of the i-th generation population. This represents the fitness value of an individual in the population. This represents the fitness value of the individual with the larger fitness value among the two crossover individuals; This represents the average fitness value of the i-th generation population. This represents the maximum fitness value of the i-th generation population. Indicates the number of population iterations.

Citation Information

Patent Citations

  • Flexible production workshop scheduling method for snail powder production enterprise

    CN114460908A