Multi-objective scheduling method and application of hybrid flow shop based on improved genetic algorithm

Abstract

The application discloses a mixed flow shop multi-objective scheduling method based on an improved genetic algorithm and application. The mixed flow shop multi-objective scheduling method comprises the following steps: a multi-objective optimization model with multiple optimization objectives is established; an initial population is generated by combining random initialization and greedy initialization based on double-layer encoding / decoding rules; a selection operation is performed to generate a next generation population; a crossover and mutation operation is performed; and iteration updating is repeatedly performed. The multi-objective scheduling method based on the improved genetic algorithm can realize multi-objective optimization scheduling by adopting a double-layer encoding rule combining a first process processing sequence code and a device allocation code. The improved genetic algorithm adopts the combination of random initialization and greedy initialization in initialization, and an elite reservation strategy is introduced in the selection operation, so that the solution quality is improved, and the convergence speed of the algorithm is also accelerated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of scheduling optimization and intelligent decision-making, and particularly relates to a hybrid flow shop multi-objective scheduling method based on an improved genetic algorithm and application. BACKGROUND

[0002] With the advent of the fourth industrial revolution, in order to improve the competitiveness of enterprises and obtain greater production efficiency, intelligent factories have gradually become an inevitable trend for sustainable development of manufacturing enterprises. As an important part of intelligent factories, the hybrid flow shop scheduling problem (HFSP) has been proven to be an NP-hard problem. Most current production scheduling methods are mainly aimed at optimizing a single objective of minimizing the completion time, i.e., production efficiency. However, in actual production, optimizing only around this single objective cannot meet the current production requirements. With the emergence of the concept of green manufacturing, the future direction of production and manufacturing is towards green energy and low-carbon economy, and more and more enterprises are paying attention to the hybrid flow shop scheduling problem that considers how to reduce energy consumption.

[0003] The hybrid flow shop scheduling problem (HFSP) is derived from the traditional flow shop scheduling problem in the manufacturing industry, and has higher complexity, which can be divided into two sub-problems: arranging the processing sequence of workpieces and allocating processing equipment. Traditional production scheduling methods are mostly based on historical experience for manual production scheduling, which has high production cost and low efficiency.

[0004] The commonly used method to solve this problem is meta-heuristic algorithm, among which the genetic algorithm has strong parallelism and global search ability, and is very suitable for large-scale scheduling problems. However, the traditional genetic algorithm has problems such as not considering equipment allocation, slow convergence speed, and poor solution stability when solving HFSP.

[0005] For example, Chinese invention patent CN111667071A introduces a traditional job shop scheduling method based on an improved genetic algorithm, the implementation process of which is as shown in Figure 1 The method specifically includes the following steps:

[0006] Step 1: set the basic parameters of the genetic algorithm;

[0007] Step 2: initialize the population according to the population size;

[0008] Step 3: calculate the fitness value of each individual in the initial population according to the fitness function;

[0009] Step 4: select the next generation of individuals by roulette wheel selection according to the fitness value of each individual, and screen out the next generation of individuals;

[0010] Step 5: cross operation is performed on the screened chromosomes to increase population diversity;

[0011] Step 6: variation operation is completed on the chromosomes generated in step 5 by using the exchange method;

[0012] Step 7: the optimal chromosome and its fitness value are recorded;

[0013] Step 8: it is judged whether the maximum iteration number is reached, if yes, the iteration process diagram and the scheduling Gantt chart are output, otherwise, step 3 is performed.

[0014] The prior art above is used to improve the genetic algorithm to solve the traditional job shop scheduling problem, in which different processing equipment may exist in the same process, and the processing time of the workpiece to be processed on different equipment is also different. Since the single-layer coding based on the process is adopted, the allocation of the equipment is not considered, and therefore only the single target of the job completion time can be optimized, and the multi-target optimization cannot be realized. SUMMARY

[0015] In view of the deficiencies of the prior art, the purpose of the present application is to provide a hybrid flow shop multi-target scheduling method based on an improved genetic algorithm and application.

[0016] To achieve the above-mentioned purpose, the technical scheme adopted by the present application comprises:

[0017] In the first aspect, the present application provides a hybrid flow shop multi-target scheduling method based on an improved genetic algorithm, comprising:

[0018] establishing a multi-target optimization model with multiple optimization targets;

[0019] generating an initial population by combining random initialization and greedy initialization based on a double-layer coding / decoding rule, wherein the chromosome coding of the individual generated by the double-layer coding / decoding rule comprises a workpiece code and a device code, the workpiece code represents the processing sequence of each workpiece in the first process, and the device code represents the selection of the processing machine in each process;

[0020] selecting the individual in the initial population to generate a next generation population;

[0021] performing cross and variation operations on the individual, wherein the cross operation and / or the variation operation adopts different operation modes for the workpiece code and the device code of the individual;

[0022] repeating the iteration update with the next generation population as the new initial population until the end condition is met.

[0023] In the second aspect, the present application further provides a hybrid flow shop multi-target scheduling system based on an improved genetic algorithm, comprising:

[0024] a model establishing module configured to establish a multi-objective optimization model with multiple optimization objectives;

[0025] an initial population module configured to generate an initial population by combining random initialization and greedy initialization based on a double-layer encoding / decoding rule, wherein the double-layer encoding / decoding rule generates a chromosome code of an individual, the chromosome code including a workpiece code and a device code, the workpiece code representing a processing sequence of each workpiece in a first process, and the device code representing a selection of a processing machine in each process;

[0026] an individual selection module configured to perform a selection operation on the individuals in the initial population to generate a next generation population;

[0027] a crossover and mutation module configured to perform a crossover operation and a mutation operation on the individuals, the crossover operation and / or the mutation operation adopting different operation modes for the workpiece code and the device code of the individuals;

[0028] an iterative updating module configured to repeat the iterative updating by taking the next generation population as a new initial population until an end condition is met.

[0029] In a third aspect, the present application also provides a readable storage medium, wherein the readable storage medium stores a computer program, and the computer program is executed to perform the steps of the hybrid flow shop scheduling method.

[0030] Compared with the prior art, the technical scheme has at least the following beneficial effects:

[0031] 1. Compared with the prior art, the multi-objective scheduling method based on the improved genetic algorithm provided by the present application adopts a double-layer encoding rule combining a first process processing sequence code and a device allocation code, and can realize multi-objective optimization scheduling.

[0032] 2. The improved genetic algorithm provided by the present application combines random initialization and greedy initialization in initialization, and introduces an elite reservation strategy in the selection operation, which not only improves the quality of the solution, but also speeds up the convergence speed of the algorithm.

[0033] The above description is only a summary of the technical scheme of the present application, in order to enable those skilled in the art to more clearly understand the technical means of the present application, and to implement the content of the description, the following describes the preferred embodiments of the present application with reference to the detailed description and the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 is a flowchart of the hybrid flow shop scheduling method provided by the background art of the present application;

[0035] Figure 2 is a flowchart of a hybrid flow shop multi-objective scheduling method provided by a typical embodiment of the present application;

[0036] Figure 3 is a structural diagram of a hybrid flow shop scheduling model provided by a typical embodiment of the present application;

[0037] Figure 4 is a schematic diagram of chromosome coding of an individual in a hybrid flow shop multi-objective scheduling method provided by a typical embodiment of the present application;

[0038] Figure 5 is a schematic diagram of a crossover process of a partial match crossover operator in a hybrid flow shop multi-objective scheduling method provided by a typical embodiment of the present application;

[0039] Figure 6 is a schematic diagram of a crossover process of a two-point crossover operator in a hybrid flow shop multi-objective scheduling method provided by a typical embodiment of the present application;

[0040] Figure 7 is a schematic diagram of a mutation process of a swap mutation operator in a hybrid flow shop multi-objective scheduling method provided by a typical embodiment of the present application;

[0041] Figure 8 is a schematic diagram of a mutation process of a single-point mutation operator in a hybrid flow shop multi-objective scheduling method provided by a typical embodiment of the present application;

[0042] Figure 9 is a comparison diagram of algorithm iteration curves provided by a typical embodiment and a comparative example of the present application;

[0043] Figure 10 is a scheduling Gantt example diagram obtained by optimization for a specific situation provided by a typical embodiment of the present application;

[0044] Figure 11 is a scheduling Gantt example diagram obtained by optimization for another specific situation provided by a typical embodiment of the present application;

[0045] Figure 12 is a scheduling Gantt example diagram obtained by optimization for another specific situation provided by a typical embodiment of the present application. DETAILED DESCRIPTION

[0046] In view of the deficiencies in the prior art, the present inventors have obtained the technical solution of the present application through long-term research and a large number of practices. The technical solution, its implementation process and principles will be further explained as follows.

[0047] Specifically, the application aims at the HFSP problem, establishes a mixed flow shop multi-objective scheduling model with the optimization objectives of minimizing the makespan and minimizing the energy consumption, and proposes a multi-objective scheduling method based on an improved genetic algorithm. Compared with the single-layer workpiece processing sequence code which cannot realize multi-objective optimization, the patent adopts a double-layer encoding rule combining the first work sequence processing sequence code and the equipment allocation code. Meanwhile, in order to improve the quality of the individual solutions in the initial population, the random initialization and the greedy initialization are adopted to generate the initial population. And the elite reservation strategy is introduced in the selection operation, so as to accelerate the convergence speed of the algorithm.

[0048] In the following description, a lot of specific details are set forth in order to provide a thorough understanding of the application, however, the application can be practiced without resorting to the details that are different from those described herein, therefore, the scope of protection of the application is not limited by the specific embodiments disclosed below.

[0049] Referring to Figures 2-4 The multi-objective scheduling method based on the improved genetic algorithm provided by the embodiments of the application includes the following steps:

[0050] establishing a multi-objective optimization model with multiple optimization objectives;

[0051] generating an initial population by combining the random initialization and the greedy initialization based on the double-layer encoding / decoding rule, wherein the chromosome encoding of the individual generated by the double-layer encoding / decoding rule includes a workpiece code and an equipment code, the workpiece code represents the processing sequence of each workpiece in the first work sequence, and the equipment code represents the selection of the processing machine in each work sequence;

[0052] generating a next generation population by performing a selection operation on the individuals in the initial population;

[0053] performing a crossover operation and a mutation operation on the individuals, and the crossover operation and / or the mutation operation adopts different operation modes for the workpiece code and the equipment code of the individuals;

[0054] repeating the iterative update by taking the next generation population as a new initial population until a termination condition is met.

[0055] It should be noted that the order of the selection operation, the crossover operation and the mutation operation on the individuals in the application is not limited to the order disclosed above. In the target genetic algorithm, the selection operation can be performed first, and then the crossover operation and the mutation operation can be performed. Alternatively, the crossover operation and the mutation operation can be performed first, and then the selection operation can be performed. Since the number of iterations is usually not a small number of times, the above order does not have a significant impact on the final iteration result, and the person skilled in the art can make appropriate selection.

[0056] As some typical examples of the above technical solutions, the main flow of the scheduling method used is as shown in Figure 2 The specific steps include steps 1-8.

[0057] Step 1: Establish a mixed flow shop multi-objective optimization scheduling model, as shown below.

[0058] First, establish the constraint conditions. As shown in Figure 3 The simplified mixed flow shop scheduling model is shown in the figure. The HFSP can be described as: N workpieces are processed on M devices, each workpiece needs to go through S processing procedures, at least one processing device exists for each processing procedure, and at least two processing devices exist for at least one processing procedure, and the workpiece can be processed on any device in the procedure. The constraint conditions include:

[0059] (1) Each workpiece can only be processed on one device at the same time;

[0060] (2) A device can only process one workpiece at the same time;

[0061] (3) The next procedure can only be processed after the current procedure is completed;

[0062] (4) Once a workpiece starts processing, it cannot be interrupted;

[0063] (5) The completion time of the workpiece in the same procedure is equal to the sum of the start time and the processing time;

[0064] (6) Each device can start at 0 time, and the device start time is the start time of the first workpiece processing.

[0065] After defining the basic conditions and constraint conditions of the model, in some embodiments, the optimization objectives can include minimizing the completion time and minimizing the energy consumption;

[0066] In some embodiments, the calculation method of minimizing the completion time can be:

[0067]

[0068] Wherein, C max represents the completion time; is the processing end time of the last workpiece on device k;

[0069] In some embodiments, the calculation method of minimizing the energy consumption can be:

[0070] minE total = min(E p +E w )

[0071] wherein E total represents energy consumption; E p represents processing energy consumption; E w represents standby energy consumption;

[0072] In some embodiments, the completion time and energy consumption can be normalized and linearly weighted and summed to obtain a comprehensive objective function, which can be used to indicate the selection operation of the individual.

[0073] As a typical example of the above scheme, a mathematical model is established based on the above conditions, with the optimization objectives being to minimize the completion time and to minimize the energy consumption. The completion time C max is the latest completion time of the equipment. wherein is the end time of the last workpiece on the equipment k. The energy consumption E total includes the processing energy consumption E p and the standby energy consumption E w : min E total = min (E p +E w ). Since the two optimization objectives have different dimensions, they need to be normalized: wherein f1 * , f1, f 1，max respectively represent the normalized objective function value, the current objective value, and the maximum value of the objective value in the population. Then, a linear weighted sum method is used to convert the multi-objective problem into a single-objective problem, and a comprehensive objective function is obtained: f = w1f1 * +w2f2 * wherein w1 and w2 represent the weight coefficients of the two objective values, satisfying 0≤w1≤1, 0≤w2≤1, and w1+w2=1.

[0074] In practice, after the model is established, step 2: defining the encoding / decoding rule is needed, but this rule does not need to be redefined every time it is operated. After one optimization scheduling, the encoding / decoding rule often does not need to be redefined, and only needs to be repeatedly used.

[0075] In a specific example, the present application adopts a double-layer encoding rule combining the first process processing sequence code and the equipment allocation code. The first layer of the chromosome code is composed of N gene codes of the processing sequence code, indicating the processing sequence of each workpiece in the first process; the second layer of N×S gene codes is the equipment allocation code, indicating the equipment allocation in each process.

[0076] For example, as Figure 4As shown, it is an HFSP example of 5 workpieces, 2 processes, and 2 devices per process. Its chromosome is [4, 1, 5, 3, 2; 1, 2, 2, 1, 1; 4, 4, 3, 3, 4], indicating that the processing order in the first process is 4-1-5-3-2, wherein workpiece 4, workpiece 3 and workpiece 2 are processed on device 1 in the first process, workpiece 1 and workpiece 5 are processed on device 2; according to the end time of each workpiece in the first process, the first two and the last workpiece are processed in turn on device 4 in the second process according to the end time order, the third and fourth workpiece are processed on device 3; in this way, the processing order and device allocation of each process can be determined.

[0077] In the subsequent decoding process, for the workpiece code, considering that the processing start order in the subsequent processes mainly depends on the completion order of the previous process, the workpieces are processed in turn in the first process according to the order of the workpiece code from left to right, and the processing order of the other processes is sorted according to the first-come-first-served method according to the completion order of the previous process; for the device code, the device allocation of each workpiece in each process is determined in turn according to the device code in the chromosome.

[0078] A key means of the present application is also to combine two ways of initialization on the double-layer encoding strategy. Specifically, in some embodiments, the mixed flow shop multi-objective scheduling method can include the following steps:

[0079] The workpiece code is generated by a random initialization method, and the device code is generated by a combination of random initialization and greedy initialization.

[0080] In some embodiments, the greedy initialization can include time greedy and energy greedy;

[0081] In some embodiments, when generating the device code, the random initialization can be provided with a first preset proportion, the time greedy can be provided with a second preset proportion, and the energy greedy can be provided with a third preset proportion, and the first preset proportion is greater than the second preset proportion and / or the third preset proportion.

[0082] As a typical example of the above scheme, corresponding to step 3: population initialization in the overall process provided by the present application, the specific steps are as follows:

[0083] The application adopts two methods of random initialization and greedy initialization to generate the initial population. Among them, the first process workpiece code of the individual in the initial population is generated by random initialization method; for the device assignment code, 50% is generated by random initialization to ensure the diversity of the initial population, and the remaining 50% is generated by greedy initialization method, among which 25% is generated by time greedy initialization, that is, the workpiece is first assigned to the device with the shortest current completion time, and if the completion times of multiple devices are the same, the device with the smallest processing energy is preferentially assigned, and the remaining 25% is generated by energy greedy initialization, that is, the workpiece is first assigned to the device with the smallest energy, and if the processing energies of multiple devices are the same, the device with the shortest current completion time is preferentially assigned. The preset proportions of the above 50%, 25% and 25% can be adjusted appropriately, but it is best to still meet the above priority order.

[0084] In some embodiments, the selection operation can be based on the calculation of the fitness value of the individual;

[0085] In some embodiments, the fitness value can be the inverse of the objective function used to calculate the optimization target.

[0086] As an example of the above technical solution, corresponding to step 4 in the overall process provided by the application: calculating the fitness value of each individual in the population. In the application, the fitness function can be set as the inverse of the comprehensive objective function, for example: Wherein fit is the fitness value.

[0087] In some embodiments, the selection operation can be carried out in a roulette manner, and the probability of each individual being inherited into the next generation population is proportional to the size of its fitness value;

[0088] In some embodiments, the selection operation can also use an elite reservation strategy to reserve the individual with the highest fitness value in the initial population into the next generation population. That is, the individual with the highest fitness value does not participate in the selection in the above probability gambling manner.

[0089] As an example of the above technical solution, corresponding to step 5 in the overall process provided by the application: selection operation. The patent adopts a roulette selection method, in which the probability of each individual in the population being inherited into the next generation is proportional to the size of its fitness value. At the same time, in order to obtain high-quality solutions and speed up the convergence speed of the algorithm, an elite reservation strategy is introduced, that is, the individual with the highest fitness value in the population is reserved.

[0090] In some embodiments, for the workpiece code, the crossover operation can adopt a partial matching crossover operator;

[0091] In some embodiments, for the device code, the crossover operation can employ a two-point crossover operator;

[0092] In some embodiments, for the workpiece code, the mutation operation can employ an exchange mutation operator;

[0093] In some embodiments, for the device code, the mutation operation can employ a single-point mutation operator.

[0094] As an example of the above technical solution, corresponding to step 6: crossover operation in the overall process provided by the present application. Since the present patent adopts the double-layer encoding rule combining the first process processing sequence code and the device allocation code, different crossover methods are employed for the two parts of the code.

[0095] For the workpiece code part, a partially-matched crossover (PMX) operator is employed, that is, two crossover points are randomly selected in the parent chromosomes, the gene codes between the two crossover points in the chromosomes are exchanged, and conflict detection is performed on the repeated gene codes in the exchanged chromosomes. The specific example operation is as shown in the following: Figure 5 and as shown in the following:

[0096] (1) Two parent chromosomes are randomly selected, and two crossover points are randomly selected.

[0097] (2) The gene codes between the two crossover points in the two parent chromosomes are exchanged.

[0098] (3) Conflict detection. The exchanged chromosome segment is kept unchanged, repeated gene codes are searched for in the segment without exchange, and the gene codes in the corresponding positions in the exchanged segment are replaced.

[0099] (4) If there are still repeated gene codes, step (3) is repeated.

[0100] In addition, for the device code part, a two-point crossover operator is employed, and the operation process is similar to the PMX operator, but conflict detection is not required. That is, two crossover points are randomly selected in the parent chromosomes, the gene codes between the two crossover points in the chromosomes are exchanged, and the specific example is as shown in the following: Figure 6 .

[0101] In addition, corresponding to step 7: mutation operation in the overall process. Like the crossover operator, different mutation methods are employed for different segments of gene codes.

[0102] For the workpiece code part, an exchange mutation operator is employed, that is, two mutation points are randomly selected on the parent chromosomes, and the gene codes at the two mutation points are exchanged. The specific example is as shown in the following: Figure 7 .

[0103] And correspondingly, a single-point mutation operator is used for the equipment code part, that is, a mutation point is randomly selected on the parent chromosome, and the gene code at the mutation point position is randomly changed into other gene codes. Figure 8 As shown in the specific example.

[0104] In some embodiments, the end condition can include reaching a maximum number of iterations. This technical solution corresponds to step 8 in the overall process: determining whether the maximum number of iterations is reached, and if so, outputting the optimal scheduling scheme, otherwise executing step 4 in the overall process, and continuously iterating.

[0105] This way is the most convenient ending way, but it is not limited thereto, for example, when the value of the objective function is above the preset value, the iteration can also be ended. This way can generally improve the accuracy of the optimization result, but it may also increase the risk of significantly increasing the number of iterations. More preferably, either of the above two ways can be used to end the iteration first.

[0106] In the above technical solution, the key technical means in the application is:

[0107] 1. For the multi-objective scheduling problem in a hybrid flow shop, a multi-objective scheduling method based on an improved genetic algorithm is proposed, which has better optimization results and faster convergence speed than the traditional genetic algorithm.

[0108] 2. Existing technologies are mostly based on historical experience for manual production scheduling, or use other intelligent optimization algorithms for solution, but generally optimize for the single objective of completion time, without considering the allocation problem. The present patent adopts a double-layer encoding rule combining the first process processing sequence code and the equipment allocation code, which not only solves the workpiece processing sequence problem, but also considers the equipment allocation problem.

[0109] 3. The initial population of the traditional genetic algorithm is generally generated by random initialization method, which has poor solution quality stability and slow convergence. The improved genetic algorithm proposed in the present patent uses population initialization and greedy initialization methods, which not only improves the solution quality, but also accelerates the convergence speed of the algorithm to some extent. At the same time, the elite reservation strategy is introduced in the selection operation, which further accelerates the convergence speed of the algorithm.

[0110] Correspondingly, the embodiment of the present application also provides a hybrid flow shop multi-objective scheduling system based on an improved genetic algorithm, which comprises:

[0111] A model establishment module for establishing a multi-objective optimization model with multiple optimization objectives;

[0112] An initial population module is configured to generate an initial population by combining random initialization and greedy initialization based on a double-layer encoding / decoding rule, wherein the chromosome encoding of an individual generated by the double-layer encoding / decoding rule comprises a workpiece code and a device code, the workpiece code represents the processing sequence of each workpiece in the first process, and the device code represents the selection of processing machines in each process;

[0113] An individual selection module is configured to perform a selection operation on the individuals in the initial population to generate a next-generation population;

[0114] A crossover and mutation module is configured to perform a crossover operation and a mutation operation on the individuals, and the crossover operation and / or the mutation operation adopts different operation modes for the workpiece code and the device code of the individuals;

[0115] An iterative update module is configured to repeat the iterative update by taking the next-generation population as a new initial population until a termination condition is met.

[0116] In addition, the embodiment of the present application further provides a readable storage medium, wherein the readable storage medium stores a computer program, and the computer program performs the steps of the mixed flow shop scheduling method in any of the above embodiments when being executed.

[0117] Based on the above embodiment, it can be clearly seen that, compared with the prior art, the multi-objective scheduling method based on the improved genetic algorithm provided by the embodiment of the present application adopts a double-layer encoding rule combining a first process processing sequence code and a device allocation code, and can realize multi-objective optimization scheduling.

[0118] The improved genetic algorithm provided by the embodiment of the present application adopts random initialization and greedy initialization in initialization, and introduces an elite reservation strategy in the selection operation, which not only improves the quality of the solution, but also accelerates the convergence speed of the algorithm.

[0119] In more specific practical applications, the present inventors have found that, by using the prior art in the above background technology, not only can multi-objective optimization scheduling be realized, but also the quality of the solution is poor, and usually only one of several solutions can be practically applied, in addition, the convergence speed of the model is slow, and usually a large number of iterations are required to converge; in contrast, the overall process of the above specific example can not only realize multi-objective optimization, but also has very strong reference value for the solution, and can usually be directly used to guide the shop scheduling, and only a small number of iterations are required to realize the convergence of the model.

[0120] More specifically, in order to verify the effectiveness of the algorithm in the application, the actual production data in the existing literature is used, as shown in Table 1. By comparing the algorithm with the traditional genetic algorithm through the example, the algorithm parameters are set to be the same, the maximum stagnation algebra is 200, the population size is 80, the crossover rate is 0.8, and the mutation rate is 0.2. Each group is run for 20 times, and the average value is taken. The comparison results are shown in Table 2, and the results show that whether it is single-objective optimization or multi-objective optimization, the algorithm is not only superior to the traditional genetic algorithm in the optimization result, but also faster than the traditional genetic algorithm in the convergence time, and the convergence speed is increased by about 50%.

[0121] This is because the algorithm introduces the greedy initialization rule in the initialization process, so that part of the individuals in the initial population have a guiding effect on the target component, thereby obtaining a better solution, and to some extent, the convergence speed of the algorithm is accelerated. Because the algorithm introduces the elite reservation strategy in the selection operation, the convergence speed of the algorithm is further accelerated.

[0122] Through optimization test, the iteration curve comparison chart of the above two algorithms is as shown in Figure 9 , and as a better optimization result example, the scheduling Gantt chart after optimization of the algorithm is as shown in Figures 10-12 . Through the test, it is verified that the technical scheme provided by the application is feasible and effective in solving the HFSP problem, and has good stability and optimization ability.

[0123] Table 1: Existing workpiece production data

[0124]

[0125]

[0126] Table 2: Comparison results of algorithms

[0127]

[0128] Finally, it should be noted that the above embodiments are only for illustrating the technical concept and characteristics of the application, and the purpose is to enable those skilled in the art to understand the content of the application and implement it, and cannot limit the protection scope of the application. Any equivalent changes or modifications made in accordance with the spirit and essence of the application should be covered within the protection scope of the application.

Claims

1. A multi-objective scheduling method for hybrid flow shop based on an improved genetic algorithm, characterized in that, include: Establish a multi-objective optimization model with multiple optimization objectives; Based on a two-layer encoding / decoding rule, an initial population is generated using a combination of random initialization and greedy initialization methods. The chromosome code of each individual generated by the two-layer encoding / decoding rule includes a workpiece code and a device code. The workpiece code represents the processing order of each workpiece in the first process, and the device code represents the selection of the processing machine in each process. During processing, based on the end time of each workpiece in the previous process, processing is performed sequentially according to the device indicated by the device code, following a first-come, first-served principle. The specific process of generating the initial population includes: The workpiece code is generated using a random initialization method, and the device code is generated using a combination of random initialization and greedy initialization. The greedy initialization includes two modes: time-greedy and energy-greedy. When generating the device code, the random initialization is set with a first preset percentage, the time-greedy initialization is set with a second preset percentage, and the energy-greedy initialization is set with a third preset percentage. The first preset percentage is greater than the second and third preset percentages. The time-greedy initialization means that the workpiece is first assigned to the device with the shortest current completion time. If multiple devices have the same completion time, the device with the lowest processing energy consumption is assigned first. The energy-greedy initialization means that the workpiece is first assigned to the device with the lowest energy consumption. If multiple devices have the same processing energy consumption, the device with the shortest current completion time is assigned first. Based on the optimization objectives of minimizing completion time and minimizing energy consumption, individuals in the initial population are selected to generate the next generation population; The individual is subjected to crossover and mutation operations, and the crossover and / or mutation operations employ different operation methods for the individual's workpiece code and device code; Using the next generation population as the new initial population, the iterative update is repeated until the termination condition is met.

2. The multi-objective scheduling method for a hybrid flow workshop according to claim 1, characterized in that, The method for calculating the minimum completion time is as follows: ； in, Indicates the completion time; For equipment The processing end time of the last workpiece; The method for calculating the minimum energy consumption is as follows: ； in, Represents energy consumption; Represents processing energy consumption; Represents standby power consumption; The completion time and energy consumption are normalized and then linearly weighted and summed to obtain a comprehensive objective function, which is used to indicate the selection operation for the individual.

3. The multi-objective scheduling method for a hybrid flow workshop according to claim 2, characterized in that, The selection operation is based on the calculation of the fitness value of the individual; The fitness value is the reciprocal of the comprehensive objective function used to calculate the optimization objective.

4. The multi-objective scheduling method for a hybrid flow workshop according to claim 3, characterized in that, The selection operation is performed using a roulette wheel method, and the probability of each individual being inherited into the next generation population is proportional to its fitness value. The selection operation employs an elite retention strategy, retaining the individual with the highest fitness value in the initial population to enter the next generation population.

5. The multi-objective scheduling method for a hybrid flow workshop according to claim 1, characterized in that, For the workpiece code, the crossover operation uses a partial matching crossover operator; for the device code, the crossover operation uses a two-point crossover operator. For the workpiece code, the mutation operation uses a swap mutation operator; for the device code, the mutation operation uses a single-point mutation operator.

6. The multi-objective scheduling method for a hybrid flow workshop according to claim 1, characterized in that, The termination condition includes reaching the maximum number of iterations.

7. A multi-objective scheduling system for hybrid flow shop based on an improved genetic algorithm, used to implement the multi-objective scheduling method for hybrid flow shop as described in any one of claims 1-6, characterized in that, include: The model building module is used to build multi-objective optimization models with multiple optimization objectives. The initial population module is used to generate an initial population based on a combination of random initialization and greedy initialization methods using a two-layer encoding / decoding rule. The chromosome code of the individuals generated by the two-layer encoding / decoding rule includes a workpiece code and a device code. The workpiece code represents the processing order of each workpiece in the first process, and the device code represents the selection of processing machines in each process. The individual selection module is used to select individuals from the initial population to generate the next generation population; The crossover and mutation module is used to perform crossover and mutation operations on the individual, wherein the crossover and / or mutation operations adopt different operation methods for the workpiece code and device code of the individual; The iterative update module is used to repeatedly perform iterative updates using the next generation population as the new initial population until the termination condition is met.

8. A readable storage medium, characterized in that, The readable storage medium stores a computer program, which, when run, executes the steps of the hybrid flow shop multi-objective scheduling method according to any one of claims 1-6.

Images