CN113656156A - Scheduling optimization method based on combination of tabu search algorithm and genetic algorithm - Google Patents
Scheduling optimization method based on combination of tabu search algorithm and genetic algorithm Download PDFInfo
- Publication number
- CN113656156A CN113656156A CN202110878110.8A CN202110878110A CN113656156A CN 113656156 A CN113656156 A CN 113656156A CN 202110878110 A CN202110878110 A CN 202110878110A CN 113656156 A CN113656156 A CN 113656156A
- Authority
- CN
- China
- Prior art keywords
- solution
- tabu
- algorithm
- population
- workpiece
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
- G06F9/4806—Task transfer initiation or dispatching
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
- G06F9/4881—Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
Abstract
A scheduling optimization method based on combination of a tabu search algorithm and a genetic algorithm belongs to the technical field of flexible manufacturing production scheduling. The problem of state space explosion in the bottleneck detection process in the prior art is solved. The method is suitable for the Petri net model of the flexible manufacturing system, random scheduling is adopted in the model, and genetic algorithm and tabu search algorithm are utilized to optimize the Petri net model data. The tabu search is: selecting offspring generated by a genetic algorithm as a feasible solution X; in each iteration: generating a plurality of feasible solutions through the neighborhood of the solution X to obtain candidate solutions, judging whether all the candidate solutions meet the forbidding rules, if so, replacing the solution Y meeting the forbidding rules with the solution Y, placing the taboo object corresponding to the solution Y into a taboo table, if not, selecting the optimal solution corresponding to the non-taboo object in the candidate solutions, placing the taboo object corresponding to the solution Y into the taboo table, and removing the taboo object which enters the taboo table at the earliest time. The invention is suitable for the dispatching of the flexible manufacturing system.
Description
Technical Field
The invention belongs to the technical field of flexible manufacturing production scheduling, and particularly relates to a scheduling optimization method based on combination of a tabu search algorithm and a genetic algorithm.
Background
In recent years, with the rapid development of the computer level and the increasing demand of the industry, the demand for production models has been diversified, and in modern industrial production, the analysis and processing of data has become more and more critical. Under the background, more mathematics and computational mechanism theories are applied to industrial production, and a random Petri net is one of the theories, has good mathematical properties and good simulation effect on a multi-variation and multi-dimension production model in the industrial production, so that the random Petri net has unique advantages in processing bottleneck detection and flow optimization problems in the industrial production process.
At present, stochastic Petri nets have been widely applied in various industrial production models and developed with numerous theories. The scheduling optimization problem of the manufacturing system generally belongs to the nondeterministic problem of polynomial level complexity, is usually quite complex and difficult and has no universal solution, a method needs to be selected according to actual production requirements, and common methods comprise linear programming, target programming, an engineering software simulation method, an intelligent search algorithm, a heuristic search method and the like. But the most primitive algorithms generate the problem of explosion of state space during analysis and bottleneck detection.
Disclosure of Invention
The invention aims to solve the problem of state space explosion in the process of performing bottleneck detection by using a stochastic petri network model, and provides an optimization method for an original scheduling method aiming at the problem.
The invention provides a scheduling optimization method based on combination of a tabu search algorithm and a genetic algorithm, which is suitable for a Petri net model of a flexible manufacturing system, wherein the Petri net model is randomly scheduled, and then the Petri net model data is optimized by utilizing the genetic algorithm and the tabu search algorithm.
Further defined, the composition of the Petri Net model is:
depot P1,P2,…,PnIs an initial state of n kinds of workpieces, the initial state being an unprocessed state, a library PiThe number of marks (i ═ 1,2, …, n) indicates the number of productions of the workpiece i, which is the number of workpieces that need to be processed;
for the
Presence library sequence
The above-mentioned
Showing the state of the ith workpiece in the mth step;
depot M1,M2,…,MkRepresenting the use rights of k machines, wherein the marked quantity of the libraries under the initial identification and the upper limit of the marked quantity of the libraries are both 1;
for the
Presence of transition sequences
Showing the machining process of the i-th workpiece,
is a random variable of time that is,
representing transitions
The average value of (a) is the machining time of the ith workpiece in the mth process.
Further defined, the method for optimizing the Petri net model data comprises the following steps:
constructing an initialization population, wherein 100 individuals form the initialization population, the iteration number is set to be k, and the initial value of k is set to be 0;
carrying out iterative cycle, and taking the total processing time of each individual as the fitness;
judging whether a termination condition is met, if so, selecting the optimal solution in the last generation of population as a final output result, and if not, continuing iteration;
carrying out selection operation, cross operation and variation operation on the population, and updating the population to generate a progeny population;
when k is greater than 20, optimizing each individual by adopting a tabu algorithm;
when k is greater than 200, the loop is ended, and the optimal solution in the population of the last generation is selected as the optimization result.
Further defined, the step of initializing the population is as follows:
randomly selecting one process for each workpiece as an initial process;
if the process is selected, reselecting; if not, continuing; until the initialization process selection of all workpieces is completed.
Further, the mutation operation on the population is performed with a probability of 0.05.
Further limiting, in the cross operation of the population, the cross parent selection mode is random selection; and selecting the father as the individual with the highest fitness in the father.
Further, the flow of the tabu search is defined as follows:
the initial solution yields: selecting offspring generated by a genetic algorithm as a feasible solution X, and emptying a tabu table;
judging a termination condition, if the termination condition is met, ending the iteration, otherwise, continuing the iteration;
in each iteration: generating a plurality of feasible solutions through the neighborhood of the solution X, and selecting a plurality of candidate solutions from the feasible solutions;
analyzing whether all candidate solutions meet the forbidding rules, if so, selecting the solution Y meeting the forbidding rules to replace X, putting the taboo object corresponding to Y into the taboo table, removing the taboo object which enters the taboo table at the earliest time, if not, selecting the optimal solution corresponding to the non-taboo object in the candidate solutions as the new current solution, and simultaneously putting the taboo object corresponding to the taboo object into the taboo table, and removing the taboo object which enters the taboo table at the earliest time.
Further, the tabu table is defined by taking the individual fitness as a tabu object, and the length of the tabu table is a fixed length and follows the first-in first-out principle.
Further, the structure of the neighborhood is that the neighborhood moves due to the fact that disturbance is added to the process of the workpiece on the individual, and the process is any two processes in one workpiece of the individual.
It is further defined that, when an operation movement can provide a better solution than the best solution so far, the operation movement can be regarded as a contraindication object regardless of whether the operation movement is in a contraindication table, and if all operation actions are in the contraindication table and do not accord with the forbidding rule, one solution is randomly selected from candidate solutions to continue the operation.
Has the advantages that: the invention optimizes the scheduling method of the existing flexible manufacturing system, and further solves the problem of state space explosion in the bottleneck detection process in the prior art. The processing time is shortened, and the working efficiency is improved.
The method is realized based on the existing flexible manufacturing system, in practical application, a random Petri net model of the flexible manufacturing system is established based on a physical model of the existing flexible manufacturing system, a fusion design is carried out on the coding, population generation, a fitness function, a selection operator, a crossover operator and a mutation operator of a genetic algorithm based on the model, the performance of the genetic algorithm is improved by adopting a tabu search algorithm, the genetic and tabu search algorithms are subjected to a hybrid algorithm, and finally the hybrid algorithm is applied to a specific model of the flexible manufacturing system to optimize scheduling.
Experimental results show that the convergence speed and the convergence result of the hybrid algorithm are superior to those of a standard genetic algorithm, and the hybrid algorithm is proved to have better adaptability to flexible and changeable flexible manufacturing systems. The invention is suitable for scheduling flexible manufacturing production.
Drawings
FIG. 1 is a physical model of a flexible manufacturing system;
FIG. 2 is a stochastic Petri net model of a flexible manufacturing system;
fig. 3 is a cross process of the workpiece 1;
FIG. 4 is a tabu search flow diagram;
FIG. 5 is a hybrid algorithm strategy;
FIG. 6 is a flow chart of a hybrid algorithm of the tabu search algorithm and the genetic algorithm;
FIG. 7 is a partial model of a stochastic Petri net of an intelligent manufacturing system;
FIG. 8 is a diagram showing the relationship between the population number of genetic algorithm and the convergence algebra;
FIG. 9 is a graph of genetic algorithm population number versus convergence time consumption;
FIG. 10 is a graph of the convergence of the optimal solution for the hybrid genetic and tabu search algorithm.
DETAILED DESCRIPTION OF EMBODIMENT (S) OF INVENTION
The individual is the whole production sequence, a random Petri net solves the whole production sequence, and each workpiece in the sequence mainly corresponds to the network structure; since the genetic algorithm needs to be abstracted to generate a plurality of solutions, an individual concept is presented, which corresponds to the whole production sequence, and the workpiece in the individual is the sequence of a certain workpiece in the whole production sequence.
The total processing time of an individual is the time spent in the entire production series for each individual.
Example 1.
1. Stochastic Petri net model for flexible manufacturing systems: fig. 1 shows a physical model of a flexible manufacturing system, wherein a production plan includes types and quantities of finished products to be produced and a planned production sequence, different workpieces are sequentially processed on each machine to obtain corresponding finished products, and finally the finished products are placed in a warehouse, so that a production process is completed. In order to establish the stochastic Petri net model of the flexible manufacturing system, the physical model of the flexible manufacturing system and the stochastic Petri net model of the flexible manufacturing system are related to determine a corresponding rule.
2. The workpieces in the physical model of the flexible manufacturing system correspond to libraries in the stochastic Petri net model, the number of the labels in the libraries represents the number of the workpieces of the type needing to be processed, the engineering of the machine processing corresponds to the transition in the stochastic Petri net model, wherein the average time t of the ith machine processing the jth workpieceijThe average time delay corresponding to the transition in stochastic Petri is determined byThe time for transport etc. is very short in relation to the processing time and therefore negligible, while the usage rights of the machine can be represented by a library with a unique mark. Through the analysis, a stochastic Petri net model corresponding to the physical model can be obtained, as shown in FIG. 2.
A stochastic Petri Net model with 2 workpieces corresponding to a Flexible manufacturing System of 4 production machines is depicted in FIG. 2, in which the library is
Representing the respective state, station, of the first workpiece during the machining process
Indicating the respective state, station, of the second workpiece during the machining process
Showing the use of 4 production machines, functioning as a transition catalyst in the stochastic Petri network
Showing the respective processes of the first type of workpiece,
each step of the second workpiece is shown. The random Petri net model can obtain that the first type of processing machine, the third type of processing machine and the fourth type of processing machine are needed in the processing process of the first type of workpiece, and the first type of processing machine, the second type of processing machine and the fourth type of processing machine are needed in the processing process of the second type of workpiece, so that the scheduling optimization problem is generated. For the more general case, the composition of the stochastic Petri net defining the flexible manufacturing system from which the stochastic Petri net model of the flexible manufacturing system can be derived is as follows:
(1) depot P1,P2,…,PnIs an initial state of n kinds of workpieces, the initial state being an unprocessed state, a library PiThe number of marks (i ═ 1,2, …, n) indicates the number of productions of the workpiece i, which is the number of workpieces that need to be processed;
(2) for the
Presence library sequence
The above-mentioned
Showing the state of the ith workpiece in the mth step;
(3) depot M1,M2,…,MkRepresenting the use rights of k machines, wherein the marked quantity of the libraries under the initial identification and the upper limit of the marked quantity of the libraries are both 1;
(4) for the
Presence of transition sequences
Showing the machining process of the i-th workpiece,
is a random variable of time that is,
representing transitions
The average value of (a) is the machining time of the ith workpiece in the mth process.
To relatively simplify the scheduling problem for flexible manufacturing systems, the following constraints are specified:
(1) each machine only processes one workpiece at the same time;
(2) one process of a workpiece can be started only after the previous process is finished;
(3) the machine preheating process is ignored;
(4) the machine cannot stop running due to unexpected factors such as faults and the like;
(5) the working procedures of the workpieces have processing sequences and cannot influence the working procedures of other workpieces;
(6) the priority of all workpieces is the same.
2. The core concept of Genetic Algorithm (GA) is to simulate Genetic selection and natural elimination occurring in the biological evolution process, and is an evolutionary Algorithm for exploring an optimal solution based on a simulated natural evolution process.
The genetic algorithm mainly comprises key parts such as coding, initial population formation, decoding, fitness analysis, selection, crossing, variation and the like. The flow of the genetic algorithm is as follows:
the genetic algorithm is an iterative algorithm that continually performs hybridization, mutation, and selection. Each genetic mechanism is only related to the current population state, but not to the previous population state, so the genetic algorithm conforms to the definition of markov chain.
Due to the characteristics of complexity and flexibility of a flexible manufacturing system, the algorithm and most genetic algorithms adopt matrix coding, the row number of the matrix represents the processing step, the column number of the matrix represents the type of a workpiece, and the element value in the matrix represents the number of a processed machine.
Wherein g isijThe ith process for indicating the jth workpiece is performed in the machine gijAnd (6) processing.
Step 1: the method for randomly generating the initial solution is adopted, the method is used for improving the randomness of the solution to a certain extent, and the flow is as follows:
(1) randomly selecting one process for each workpiece as an initial process;
(2) randomly selecting a process for each workpiece, and if the process is selected, reselecting; if not, continue.
(3) Until all process selections for all workpieces are completed.
Step 2: fitness function
The fitness is used as a parameter for describing the adaptation degree of the individual to the environment, the convergence tendency of the genetic algorithm is directly determined, and a reasonable fitness function is the core of the algorithm which can converge to an optimal solution.
And step 3: selection, crossing, mutation
The genetic process comprises selection, crossing and variation, and the genetic algorithm simulates the natural evolution process of the population through a genetic mechanism, continuously generates a new generation of population and enables the new generation of population to have better properties. The selection operator selects the individuals with better fitness from the parent generation and enables the individuals with better fitness to be retained in the child generation with higher probability. It can be seen from the selection rule of the selection operator that individuals with high fitness are easier to select and individuals with low fitness are easier to eliminate, and the individuals with high fitness repeatedly appear through continuous selection, but the selection operator has the problem that the selection operator can only be in a fixed range, the selected result is always in the population range, and better individuals outside the population cannot be selected, so that the selected result completely depends on the initial population, and at this time, the crossing and variation are needed to obtain the results outside the population.
Crossover is a genetic operator used for changing the operation of local information in the process of iteration from one generation to the next generation, good offspring are generated through the crossover operator, the selection method of crossover depends on the coding mode, and the logic of corresponding procedures and the correspondence of machines must be checked in the crossover process. In order to ensure randomness, the crossed parent selection mode is random selection.
Taking the process of the workpiece 1 as an example, let P1，P2Representing the column vector of parent matrix 1, C1，C2Representing the column vector of the 1 st column of the offspring matrix, and randomly dividing the work piece 1 procedure into two sets K1，K2A 1 is to P1Is contained in K1The gene of (5) is replicated to C1A 1 is to P2Is contained in K2The gene of (5) is replicated to C2And keep their positions unchanged. Similarly, P2Is contained in K2The gene of (5) is replicated to C1A 1 is to P1Is contained in K1The gene of (1)Copy to C2And keep their positions unchanged. As shown in fig. 3, two completely new individuals can be obtained by exchanging procedures for 12342314 and 12341234. In order to ensure the randomness of the generated result and the preservation of the individual with the highest fitness, the individual with the highest fitness in the parent retention is selected in the crossing process, the individual with the higher fitness is selected for 80% of the probability in the filial generation, and the individual with the lower fitness is selected in other cases.
The mutation operation is to replace the value at the random position in a certain individual code with other reasonable values, and then the randomness of the population is increased after the selection operation, and the purpose of the mutation operation is to improve the local search capability by improving the randomness. In order to ensure the feasibility of the variant individuals, the sequence variant mode is adopted, namely the sequence of two processes is arbitrarily exchanged in the process of random workpieces, and p is takenm＝0.05。
3. Tabu Search (TS) is a sub-heuristic random Search algorithm that implements a Search algorithm that allows the maximum variation of a specific objective function value by selecting a series of feature Search directions as tests from an initial feasible solution. In order to avoid trapping in a local optimal solution, a flexible memory method is used in the search to record and select an optimization process so as to guide the next search direction. On the basis, the generation of cyclic search is avoided through the taboo standard, and good states are released through the forbidding standard, so that various and effective searches are guaranteed. A tabu search flow diagram is shown in fig. 4.
The general flow for tabu search is as follows:
(1) generation of the initial solution: selecting an initial solution as a descendant generated by each genetic algorithm, setting detailed algorithm parameters, taking a feasible solution X, and emptying a tabu table; taking individual fitness as a tabu object, wherein the length of the tabu table is fixed, and the first-in first-out principle is followed
(2) Judging a termination condition, if the termination condition is met, ending the iteration, otherwise, continuing the iteration;
(3) generating a plurality of feasible solutions through the neighborhood of the solution X, and selecting a plurality of candidate solutions from the feasible solutions; adding disturbance to the process of the workpiece on the individual to generate the movement of the neighborhood, namely exchanging any two processes in a certain workpiece of the individual;
(4) analyzing whether all candidate solutions meet a forbidding rule, if so, selecting a solution Y meeting the forbidding rule to replace X, putting a tabu object corresponding to Y into a tabu table, removing the tabu object which enters the tabu table at the earliest time, if not, selecting the best solution corresponding to a non-tabu object in the candidate solutions as a new current solution, and simultaneously putting the tabu object corresponding to the tabu object into the tabu table, and removing the tabu object which enters the tabu table at the earliest time; the rule of breaking is when an operation move can provide a better solution than the best solution so far, whether the operation move is acceptable in the tabu table or not, if all operation actions are in the tabu table and do not conform to the rule of breaking, then one of the candidate solutions is randomly selected.
(5) Go to step (2).
In the implementation process of the tabu search algorithm, the local search capability is superior to that of the genetic algorithm, but the characteristics of the tabu search algorithm determine that the algorithm has high requirements on the initially set feasible solution, if the initial solution is close to the optimal solution, the efficiency in the algorithm iteration process is very high, a better solution can be searched in a solution space without a large amount of iteration, and if the initial solution is far away from the optimal solution or close to the local optimal solution, the initial solution is easily trapped in the local optimal solution or the iteration efficiency is very low, so that the convergence speed of the solution is seriously influenced.
The tabu search has a disadvantage over the genetic algorithm in that only one initial solution, i.e., point-to-point operation, can be used in each iteration, the serial iterative search process is different from the parallel search of the genetic algorithm, and it operates multiple solutions, i.e., group-wise operation, simultaneously in each generation, so the tabu search algorithm is not suitable for solving large-scale problems.
4. And (3) hybrid algorithm: iteration is carried out through a designed genetic algorithm, the solution space can be covered as much as possible by all individuals in the iteration process, premature convergence to a local optimal solution is guaranteed through the method, and then a local search algorithm is used for each individual in a group, so that the fitness of the individual solution is rapidly improved. Meanwhile, in order to reduce the calculation amount and improve the efficiency, after the genetic algorithm is iterated to a certain number of times, the individual position with high fitness is relatively fixed, and then the local search is performed by using tabu search, wherein a specific implementation strategy is shown in fig. 5.
The specific process of the genetic and tabu search hybrid algorithm can be obtained by the above strategy:
(1) initializing algorithm parameters, randomly generating an initial population, and setting the iteration number k to be 0;
(2) calculating the fitness of each individual in the population;
(3) judging whether a termination condition is met, if so, selecting the optimal solution in the population of the last generation, and if not, continuing iteration;
(4) executing selection, crossover and mutation operators designed in the genetic algorithm to generate an offspring population;
(5) if the iteration times k are larger than a set value, carrying out tabu search on each individual in the current population to obtain an optimized population;
(6) go to step (2).
And (4) analyzing results: convergence to a locally optimal solution: the genetic algorithm converges to the locally optimal solution for two reasons: the first is that the fitness of a small number of individuals greatly exceeds that of other individuals in the population iteration process, a large number of iterations are not needed, and the individuals with high fitness in the iteration processes quickly cover the whole population to cause convergence to a local optimal solution; the second is that in the genetic algorithm optimization process, competition among individuals only exists in the filial generation, but competition between the filial generation and the parent generation does not exist, which leads to the loss of the individuals with high fitness of the parent generation.
The efficiency is reduced after the iteration times reach a certain number: when the genetic algorithm is iterated to be close to the optimal solution or the iteration times reach a certain number, the problem that the optimal solution cannot be accurately positioned easily occurs due to the randomness in the iteration process, so that the searching efficiency after the iteration is performed for a certain number of times is reduced. Moreover, the crossover and mutation cause a problem that it jumps out of the searched optimal solution, reducing efficiency.
Aiming at the problems that the genetic algorithm is insufficient in utilization of feedback information of the system, poor in local searching capability and easy to converge too early, the problem is improved by utilizing tabu search in the local searching algorithm, potential parallel optimization capability and global searching capability of the genetic algorithm are reserved, and meanwhile the local searching capability of the tabu search can be combined to solve the global optimal solution more effectively.
Fig. 6 is a flow chart of a hybrid algorithm of a genetic algorithm and a tabu search algorithm of an optimization method for optimizing a flexibly manufactured peri network model:
constructing an initialization population, wherein 100 individuals form the initialization population, the iteration number is set to be k, and the initial value of k is set to be 0;
carrying out iterative cycle, and taking the total processing time of each individual as the fitness;
judging whether a termination condition is met, if so, selecting the optimal solution in the last generation of population as a final output result, and if not, continuing iteration;
carrying out selection operation, cross operation and variation operation on the population, and updating the population to generate a progeny population;
when k is greater than 20, optimizing each individual by adopting a tabu algorithm;
when k is greater than 200, the loop is ended, and the optimal solution in the population of the last generation is selected as the optimization result.
Example 2.
Carrying out experimental simulation on a genetic and tabu search hybrid algorithm, setting algorithm parameters through examples, and configuring the experimental scheme as follows: intel core i 79750H, 2.6GHz, memory 16GB, operating system MacOS, algorithm based on Java programming.
1. Stochastic Petri network model and algorithm parameters
The automatic processing unit of the production line of the manufacturing system mainly comprises 5 intelligent processing machines with the numbers of 1,2, 3, 4 and 5, wherein different workpieces need to be processed on different machines, the processing time of different processes of different workpieces on the machines is different, and the machines are centralized, so that the production of the machines can generate conflicts.
Production schedules are arranged reasonably to minimize overall processing time and thereby maximize production benefits. After abstraction, the following calculation example is obtained, taking 16 workpieces as an example, each workpiece requires 4 machining processes, and the machining requirements and the machining time are shown in table 1 and table 2.
TABLE 1 machine numbering for workpiece processing
TABLE 2 machining time for different processes of each workpiece
The stochastic Petri net model of the intelligent manufacturing system can be obtained based on the information, a part of the model of the stochastic Petri net is shown in FIG. 7, the stochastic Petri net is subjected to path optimization by using a genetic and tabu search hybrid algorithm, and because parameters of the algorithm in the genetic and tabu search hybrid algorithm have great influence on the performance of the algorithm, the reasonable selection of the parameters is very important.
For the taboo search algorithm, because the algorithm is optimized based on the genetic algorithm, is different from the point-to-point optimization of the original taboo search algorithm, and is optimized for the population obtained by the genetic algorithm, large-scale search is not needed, and only small-range search needs to be carried out on each individual of the population obtained by the genetic algorithm, so that the length of the taboo table is set to be 10, the maximum iteration frequency is 20, the number of search neighborhoods of each individual is 50, and the taboo search is carried out after the iteration frequency of the genetic algorithm is 20 because the quality of the solution obtained by early-stage search of the genetic algorithm is low.
The taboo search parameters are used for carrying out simulation test on the genetic algorithm, the population number is respectively 100, 200, 300, 400 and 500, and the test result is shown in fig. 8. It can be seen that as the population scale increases, the number of individual solutions increases, the coverage of the whole population for the solution space increases, and the possibility of obtaining a solution with high fitness also increases, so that the search degree for the solution space becomes high, and the number of iterations required for convergence also becomes small, but the increase in the population scale directly results in the increase of the calculated amount and the time consumption in each iteration process, as shown in fig. 9, the population scale parameter of 100 is selected and used in this document in consideration of the time cost comprehensively, and since the number of iterations until convergence does not exceed 200 under the condition that the population scale is 100 during the test, the termination condition of the genetic algorithm is set such that the number of iterations reaches 200, and the iteration is terminated.
Comparative example 1.
And comparing the performance of the genetic and tabu search hybrid algorithm with the performance of the standard genetic algorithm in solving the model scheduling problem of the flexible manufacturing system. In order to reduce the randomness of the experiment, the same feasible solution is used as the initial solution, under the condition that the rest experiment parameters are the same, the results of the optimization schemes of the genetic and tabu search hybrid algorithm and the standard genetic algorithm are compared, and the results are respectively tested by 10 times of experiments to obtain the results shown in table 3
TABLE 3 genetic and tabu search hybrid algorithm and Standard genetic Algorithm Experimental results
The experimental result shows that under the model condition, the optimal values obtained by the mixed genetic and tabu search algorithm and the standard genetic algorithm are 503 and 525 respectively, the mixed genetic and tabu search algorithm is more excellent in average value, the average processing time is only 518.1, and the unit time is reduced by 27.4 compared with the standard genetic algorithm, so that the mixed genetic and tabu search algorithm is improved for the problem that the standard genetic algorithm is easier to converge on the local optimal solution from the experimental data of the random Petri net, and the feasibility of the mixed genetic and tabu search algorithm is proved.
Fig. 10 compares the optimal solution convergence conditions of the standard Genetic Algorithm (GA) and the genetic and tabu search hybrid algorithm (GA & TS), and it can be seen from the curve in fig. 10 that the performance of the standard genetic algorithm at the initial stage of convergence does not differ much from the performance of the genetic and tabu search hybrid algorithm, and as the number of iterations increases, the convergence rate of the standard genetic algorithm gradually slows down, gradually lags behind the genetic and tabu search hybrid algorithm, and finally the convergence solution is also lower than the hybrid algorithm. Theoretically, the optimal solution of the standard genetic algorithm is unsatisfactory compared with the hybrid genetic and tabu search algorithm, mainly because the general genetic algorithm is trapped in the phenomenon of "precocity", which results in that the genetic algorithm does not converge to a more excellent solution but converges to a certain locally optimal solution. The local searching capability of the hybrid algorithm is enhanced by using the tabu search, and the hybrid algorithm has higher probability in the iteration of the genetic algorithm and is not limited to the local optimal solution, so that the experimental result of the hybrid algorithm of the genetic and tabu search is better than that of the general genetic algorithm in terms of the optimal result and the average result. The optimal solution obtained by the genetic and tabu search hybrid algorithm is
The method meets the processing logic requirements of various workpieces, and therefore, the practical feasibility of the optimal solution obtained through a genetic and tabu search hybrid algorithm is further proved.
Aiming at the defects of the genetic algorithm, the taboo search algorithm is utilized to improve the performance of the genetic algorithm, a genetic and taboo search mixed algorithm is provided, and finally the mixed algorithm is applied to a specific model of the flexible manufacturing system.
Claims (10)
1. A scheduling optimization method based on a taboo search algorithm and a genetic algorithm is suitable for a Petri net model of a flexible manufacturing system, and is characterized in that the Petri net model is randomly scheduled, and then the Petri net model data is optimized by the genetic algorithm and the taboo search algorithm.
2. The schedule optimization method according to claim 1, characterized in that the Petri Net model is composed of:
depot P1,P2,…,PnIs an initial state of n kinds of workpieces, the initial state being an unprocessed state, a library PiThe number of marks (i ═ 1,2, …, n) indicates the number of productions of the workpiece i, which is the number of workpieces that need to be processed;
for the
Presence library sequence
The above-mentioned
Showing the state of the ith workpiece in the mth step;
depot M1,M2,…,MkRepresenting the use rights of k machines, wherein the marked quantity of the libraries under the initial identification and the upper limit of the marked quantity of the libraries are both 1;
for the
Presence of transition sequences
Showing the machining process of the i-th workpiece,
is a random variable of time that is,
representing transitions
The average value of (a) is the machining time of the ith workpiece in the mth process.
3. The scheduling optimization method of claim 1, wherein the method for optimizing Petri Net model data comprises:
constructing an initialization population, wherein 100 individuals form the initialization population, the iteration number is set to be k, and the initial value of k is set to be 0;
carrying out iterative cycle, and taking the total processing time of each individual as the fitness;
judging whether a termination condition is met, if so, selecting the optimal solution in the last generation of population as a final output result, and if not, continuing iteration;
carrying out selection operation, cross operation and variation operation on the population, and updating the population to generate a progeny population;
when k is greater than 20, optimizing each individual by adopting a tabu algorithm;
when k is greater than 200, the loop is ended, and the optimal solution in the population of the last generation is selected as the optimization result.
4. The schedule optimization method according to claim 3, wherein the step of initializing a population is as follows:
randomly selecting one process for each workpiece as an initial process;
if the process is selected, reselecting; if not, continuing; until the initialization process selection of all workpieces is completed.
5. The method of claim 3, wherein the mutation is performed with a probability of 0.05.
6. The scheduling optimization method according to claim 3, wherein in the cross operation of the population, the parent selection mode of the cross is random selection; and selecting the father as the individual with the highest fitness in the father.
7. The scheduling optimization method of claim 1, wherein the tabu search flow is as follows:
the initial solution yields: selecting offspring generated by a genetic algorithm as a feasible solution X, and emptying a tabu table;
judging a termination condition, if the termination condition is met, ending the iteration, otherwise, continuing the iteration;
in each iteration:
generating a plurality of feasible solutions through the neighborhood of the solution X, and selecting a plurality of candidate solutions from the feasible solutions;
analyzing whether all candidate solutions meet the forbidding rules, if so, selecting the solution Y meeting the forbidding rules to replace X, putting a tabu object corresponding to Y into a tabu table, and removing the tabu object which enters the tabu table at the earliest time; if not, selecting the best solution corresponding to the non-tabu object in the candidate solutions as the new current solution, and simultaneously placing the corresponding tabu object into the tabu table to remove the tabu object which enters the tabu table at the earliest time.
8. The scheduling optimization method of claim 7, wherein the tabu table is a tabu object with an individual fitness, and the tabu table has a fixed length and follows a first-in-first-out principle.
9. The scheduling optimization method of claim 7 wherein the neighborhood structure is a neighborhood movement generated by adding perturbation to the individual workpiece processes, and is any two process steps in exchanging an individual workpiece.
10. The method of claim 7, wherein the rule is that when an operation move can provide a better solution than the best solution so far, the operation move is used as a tabu object, and if all operation actions are in a tabu table and do not conform to the rule, a solution is randomly selected from the candidate solutions to continue.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110878110.8A CN113656156A (en) | 2021-07-30 | 2021-07-30 | Scheduling optimization method based on combination of tabu search algorithm and genetic algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110878110.8A CN113656156A (en) | 2021-07-30 | 2021-07-30 | Scheduling optimization method based on combination of tabu search algorithm and genetic algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113656156A true CN113656156A (en) | 2021-11-16 |
Family
ID=78490201
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110878110.8A Pending CN113656156A (en) | 2021-07-30 | 2021-07-30 | Scheduling optimization method based on combination of tabu search algorithm and genetic algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113656156A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115759407A (en) * | 2022-11-18 | 2023-03-07 | 宝开(上海)智能物流科技有限公司 | Sorting system-based multidimensional scheduling scheme optimization method, system and storage medium |
| CN117035255A (en) * | 2023-05-31 | 2023-11-10 | 南通大学 | Robust scheduling method for manufacturing system containing unreliable resources |
| CN117707745A (en) * | 2024-02-05 | 2024-03-15 | 国网湖北省电力有限公司信息通信公司 | Metering task synchronous scheduling method based on self-adaptive tabu search algorithm |
-
2021
- 2021-07-30 CN CN202110878110.8A patent/CN113656156A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115759407A (en) * | 2022-11-18 | 2023-03-07 | 宝开(上海)智能物流科技有限公司 | Sorting system-based multidimensional scheduling scheme optimization method, system and storage medium |
| CN117035255A (en) * | 2023-05-31 | 2023-11-10 | 南通大学 | Robust scheduling method for manufacturing system containing unreliable resources |
| CN117035255B (en) * | 2023-05-31 | 2024-02-06 | 南通大学 | Robust scheduling method for manufacturing system containing unreliable resources |
| CN117707745A (en) * | 2024-02-05 | 2024-03-15 | 国网湖北省电力有限公司信息通信公司 | Metering task synchronous scheduling method based on self-adaptive tabu search algorithm |
| CN117707745B (en) * | 2024-02-05 | 2024-04-30 | 国网湖北省电力有限公司信息通信公司 | Metering task synchronous scheduling method based on self-adaptive tabu search algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113656156A (en) | Scheduling optimization method based on combination of tabu search algorithm and genetic algorithm | |
| Guo et al. | Stochastic hybrid discrete grey wolf optimizer for multi-objective disassembly sequencing and line balancing planning in disassembling multiple products | |
| Abbasi et al. | Production planning and performance optimization of reconfigurable manufacturing systems using genetic algorithm | |
| CN101901425A (en) | Flexible job shop scheduling method based on multi-species coevolution | |
| CN112561225B (en) | Flexible job shop scheduling method based on marker post co-evolution algorithm | |
| Mahmudy et al. | Modeling and optimization of part type selection and loading problem in flexible manufacturing system using real coded genetic algorithms | |
| Rezaeipanah et al. | Modeling the scheduling problem in cellular manufacturing systems using genetic algorithm as an efficient meta-heuristic approach | |
| CN107831740A (en) | A kind of Optimization Scheduling during the distributed manufacturing applied to notebook part | |
| Khoo et al. | A prototype genetic algorithm-enhanced multi-objective scheduler for manufacturing systems | |
| CN114841581A (en) | Feature selection method in dynamic job shop scheduling rule based on GEP-VNS evolution | |
| Rifai et al. | Reentrant FMS scheduling in loop layout with consideration of multi loading-unloading stations and shortcuts | |
| Zhou et al. | Multi-mechanism-based modified bi-objective Harris Hawks optimization for sustainable robotic disassembly line balancing problems | |
| CN114282370A (en) | Disassembly line setting method considering physical and mental loads of operator | |
| Fattahi et al. | Some heuristics for the hybrid flow shop scheduling problem with setup and assembly operations | |
| CN113050644A (en) | AGV (automatic guided vehicle) scheduling method based on iterative greedy evolution | |
| Bensmaine et al. | A simulation-based genetic algorithm approach for process plans selection in uncertain reconfigurable environment | |
| Kucukoglu et al. | Application of precedence constrained travelling salesman problem model for tool path optimization in CNC milling machines | |
| Nearchou | Flow-shop sequencing using hybrid simulated annealing | |
| Dalavi et al. | Optimal drilling sequences for rectangular hole matrices using modified shuffled frog leaping algorithm | |
| Chen et al. | Integrating genetic algorithm with time control for just-in-time scheduling problems | |
| CN116663866B (en) | Method for optimizing initial position of machining unit cutter | |
| Zhang | A scheduling model of intelligent manufacturing system based on GA optimization | |
| Fichera et al. | Genetic algorithms efficiency in flow shop scheduling | |
| Skakalina | Hybridization of the genetic algorithm with the apparatus of fuzzy sets | |
| CN113505910B (en) | Mixed workshop production scheduling method containing multi-path limited continuous output inventory |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |