Researches On Distributed Scheduling Problem Of Complex Industrial Process By Using Dynamic Shuffled Frog Leaping Algorithm

Manufacturing industry is the pillar of economy.The process control of industrial production is an indispensable part of manufacturing industry and directly affects the operating efficiency of production system.As the part of complex system,the process control of industrial production is of great significance to the development of our country.With the global economic integration and market competition,the control of complex industrial processes is facing new challenges,and many distributed production scheduling problems are derived.These problems are characterized by large scale,complex structure,strong nonlinearity and uncertainty,which are difficult to be solved by traditional control methods.Intelligent control and intelligent optimization method,which do not depend on the model of the problem,are ideal methods to solve distributed scheduling problem in industrial process.Intelligent optimization is the main way to solve distributed scheduling problem,and the design of efficient intelligent optimization algorithm is the key to improving the performance of the production system and the core of increasing the production efficiency,economic benefits and social environmental benefits.Shuffled frog leaping algorithm(SFLA)is a common intelligent optimization algorithm,which has a super swarm intelligence optimization framework.The framework provides differentiated searches between multiple population,the use and selection of multiple search strategies,and the communication and cooperation between populations.It can make the population evolve rapidly while maintaining the diversity of the population,improving the computational efficiency and avoiding falling into the local optimum.Therefor,it is necessary to design a dynamic shuffled frog leaping algorithm(DSFLA)based on the characteristics of SFLA to solve the distributed scheduling problem in complex industrial processes.Distributed hybrid flow shop scheduling problem(DHFSP)is studied in this thesis.Some practical constraints,such as multiprocessor task and assembly,in DHFSP are adequately considered.The single-objective and multi-objective optimization problems in deterministic and uncertain environments are studied,and shuffled frog leaping algorithm is used to solve DHFSP.The main work of this thesis is as follows:(1)DHFSP with multiprocessor tasks is studied and a dynamic shuffled frogleaping algorithm with different search strategy(DSFLA)is proposed to minimize maximum completion time.Dynamic search process is executed in each memeplex with at least two improved solutions.Global search and dynamic multiple neighborhood search are applied,in which neighborhood structure is chosen based on its optimization effect.A new destruction-construction process is hybridized with DSFLA and population shuffling is done when shuffling condition is met.Lower bound is obtained and proved.A number of experiments are conducted on a set of instances.The computational results validate the effectiveness of the new strategies of DSFLA and the competitive performances on solving the considered DHFSP.(2)DHFSP with fabrication,transportation and assembly is considered.DSFLA with Q-learning is presented to minimize maximum completion time.A three-string representation is used.A Q-learning process is applied to dynamically select search strategy in memeplex search.It is composed of four actions based on combination of global search,neighborhood search and solution acceptance rule,six states designed by population evaluation based on its elite solution and dispersion degree,and a newly defined reward function.A number of experiments are conducted.The computational results demonstrate that DSFLA can provide promising results on the considered DHFSP with transportation and assembly.(3)DHFSP with sequence-dependent setup times is studied,in which factory assignment and machine assignment of first stage are integrated together.DSFLA with memeplex quality is designed to minimize total tardiness and maximum completion time simultaneously.Solution quality of memeplex is measured and new search process is implemented according to solution quality.Evolution quality is evaluated for each memeplex and adopted for dynamically selecting memeplexes in a novel memeplex shuffling.A number of experiments are conducted to test the new strategies and performances of DSFLA.The computational results demonstrate the effectiveness of the new strategies and the promising advantages of DSFLA.(4)Fuzzy DHFSP is considered and DSFLA with memeplexes collaboration is used to optimize fuzzy maximum completion time,total agreement index and fuzzy total energy consumption simultaneously.Iterated greedy,variable neighborhood search and global search are designed using problem-related features;memeplex evaluation based on three quality indices is presented,an effective cooperation process between the best memeplex and the worst memeplex is developed according to evaluation results and performed by exchanging search times and search ability,and an adaptive population shuffling is adopted to improve search efficiency.Extensive experiments are conducted and the computational results validate that DSFLA has promising advantages on solving fuzzy DHFSP.