Research On Intelligent Scheduling Algorithm Of Fuzzy Flexible Job Shop

Production scheduling is a key link in management decision-making and resource allocation in intelligent manufacturing systems,using an efficient scheduling method can effectively improve the production efficiency of enterprises and achieve energy conservation and emission reduction.Flexible job shop scheduling problem is a typical type of production scheduling problem,which exists widely in various flexible manufacturing systems.At present,certain researchers have conducted some research on the deterministic model of this problem.However,there are often many uncertain factors in the actual processing process.Therefore,the fuzzy numbers in fuzzy theory are used to represent uncertain processing data,and then a fuzzy flexible job shop scheduling problem model is established,which can more objectively describe the actual production process.The study of fuzzy flexible job shop scheduling problems and corresponding solving algorithms has high Theoretical and practical value.This paper studies the fuzzy flexible job shop scheduling problem and the intelligent scheduling algorithm to solve the problem.The main contents are as follows:(1)A hyper-heuristic genetic algorithm(HGA)is proposed to minimize the maximum fuzzy completion time(i.e,makespan)for the fuzzy flexible job shop scheduling problem(FFJSP),in which the job’s processing time is represented by using triangular fuzzy number.Firstly,after analyzing the properties of the existing sorting rules on triangular fuzzy number in detail,and fully considering the approximate error and the ambiguity of the operation of taking the bigger,a more accurate triangular fuzzy number sorting rule is designed,which can reasonably calculate the objective function values of the solutions for FFJSP and other various scheduling problems.Secondly,to realize the effective search in different regions of FFJSP’s solution space,HGA divides the solving process into two layers.The upper layer uses the genetic algorithm with adaptive mutation operator to optimize the permutation of six special operations,i.e.,six effective neighbor operations.The lower layer uses each permutation obtained from the upper layer as a heuristic to perform operations on the corresponding individual of the lower layer for executing a compact variable neighborhood local search and generating new individual,and meanwhile adds the simulated annealing mechanism to overcome the local-optimality trap.Finally,simulation experiments and algorithm comparisons verify the effectiveness of the proposed sorting rules and HGA.(2)Based on the problem of(1),further considering the uncertainty of the due date in the actual production process,A novel Discrete Teaching-Learning-Based Optimization(NDTLBO)is proposed to maximize the average agreement index for the fuzzy flexible job shop scheduling problem(FFJSP),in which the job’s processing time and the due date are represented by using fuzzy number.Firstly,considering the membership of fuzzy completion time and due date,a more accurate agreement index calculation method is designed.Secondly,in the teaching and learning phases of the algorithm,two effective cross-operations are used to search for new solutions.Meanwhile,the learning history matrixs(LHMs)are designed to learn and retain the information of positions and values of elements in each new solution,and in the tutoring phase,heuristic operations constructed via the LHMs are used to efficiently search different high-quality regions in the solution space of FFJSP.Finally,simulation experiments and algorithm comparisons verify the effectiveness of the proposed agreement index calculation method and NDTLBO.(3)further consider green indicators based on the model of(2),establishes a permutation model for the low-carbon scheduling of fuzzy flexible job shop problem(FFJSP＿LC),whose criteria include the maximization of average agreement index and minimum agreement index as well as the minimization of fuzzy total carbon emissions.Furthermore,combining the advantages of the two types of algorithms in(1)and(2),a hybrid multi-objective hyper-heuristic algorithm(HMOHA)is proposed to solve FFJSP＿LC.Firstly,for the upper layer of the algorithm,a long-term evaluation method is proposed to decode the upper layer individuals in the algorithm,which can more objectively evaluate the search performance of each individual,and use the genetic algorithm as the upper layer strategy to use the lower layer operation order And times to optimize.Secondly,for the lower layer of the algorithm,an efficient two-phase encoding method and an improved greedy active decoding strategy are designed according to the nature of the problem,which can make full use of the idle time of machine processing and select a better processing machine,thereby improving The quality of the solution,while using the efficient update operation of the NDTLBO algorithm in(2)as a lower layer strategy set,to search for different high-quality regions in the solution space.In addition,based on the learning history matrix,a teacher self-learning operation is designed to conduct a more detailed local search of the area near the non-inferior solution,further improving the quality of the solution.Finally,simulation experiments and algorithm comparisons verify the effectiveness of the proposed algorithm.