Research On Intelligent Optimization Algorithms For Several Integrated Scheduling Problems

Integrated scheduling is a scheduling method for the processing and assembly of a single complex product.At present,most of the existing integrated scheduling algorithms for complex products with tree structure are rule-based heuristic solutions.However,in the past ten years,meta-heuristic algorithms such as the genetic algorithms,the tabu search algorithms,the teaching optimization algorithms and so on have become mainstream algorithms to solve traditional production scheduling problems.Compared with the traditional production scheduling,the integrated scheduling is more complicated,and the overall constraints among operations are in a tree structure.As a result,the existing encoding methods and evolutionary operators for solving traditional production scheduling problems have become invalid,making it difficult to directly apply intelligent optimization algorithms to solve the integrated scheduling problem of complex products.Aiming at the integrated scheduling of complex products with tree structure,this paper studies several integrated scheduling solutions based on the intelligent optimization algorithms.The main research work is as follows:Aiming at the general integrated scheduling problem,from the perspective of "operations finding machines",an integrated scheduling algorithm for complex products based on the operation partial order relationship table is proposed.The algorithm first establishes a operation partial order relationship table for the tree structure product,and then uses the genetic algorithm to solve the problem based on the table.In order to ensure the feasibility of the initial solutions,a novel encoding method based on the dynamic operation partial order relationship table is proposed;two new different crossover and mutation methods are designed to ensure the legitimacy of the generated individual offspring and avoid the detection and repair of infeasible solutions;a simple greedy insertion-based decoding method is given.The comparison experiment results show that the proposed algorithm's solution quality outperform that of the other comparison algorithms.Aiming at the general integrated scheduling problem,from the perspective of "machines finding operations",an integrated scheduling algorithm based on a hybrid genetic algorithm and tabu search is proposed.The algorithm uses genetic algorithm for global search,and applies tabu search strategy to perform local search for the optimal solution generated by each iteration of genetic algorithm.Firstly,an encoding method based on a dynamic schedulable operation set is proposed to ensure that the operations in the same operation chain meet the machining constraints;Secondly,two new different crossover and mutation methods are proposed to explore the rationality of the problem solution space;Then,a local search strategy based on tabu search is shown to enhance the search capability for superior solutions.Finally,a decoding method based on machine idle time driving is presented to handle the scheduling order of operations on different machines.The algorithm is tested by the randomly generated instances,and the comparison experiment results show that the proposed algorithm's solution quality outperform that of the other comparison algorithms.Aiming at the problem of integrated scheduling of tree structure products with flexible machine selection,a flexible integrated scheduling algorithm based on remaining work probability selection coding is proposed.The algorithm is based on the framework of a genetic algorithm.Firstly,in order to ensure the diversity and goodness of the initial population,an encoding method based on remaining work probability selection is proposed.Secondly,two new different crossover and mutation methods are designed based on operation and position respectively,which ensure the rationality of the offspring's individual operation chain and machine chain;Then,in order to enhance the searchability of the algorithm for an optimal solution of the problem,a local search strategy based on the machine is proposed.Finally,a simple and effective decoding method based on the idle period is given.The algorithm is tested by the existing instance and randomly generated instances.The experimental results show that the proposed algorithm's solving speed and solution quality outperform other comparison algorithms.Aiming at the integrated scheduling problem of multi-workshops with complex tree structure products distributed in different places,a memetic algorithm-based distributed integrated scheduling algorithm is proposed.Based on the framework of the memetic algorithm,the algorithm uses a distributed estimation algorithm for global search and performs a local search strategy based on the critical operation set for the current optimal solution obtained in each evolutionary generation.A bi-chain-based individual representation method is presented and a simple greedy insertion-based decoding method is given;Two position-based probability models are built,which are used to describe the distribution of the operation priority and factory assignment respectively.Based on the designed probability models,two learning-based updating mechanisms and an improved sampling method are given,which ensures that the population evolves towards a promising region.In order to enhance the searchability for the superior solutions,9 disturbance operators based on the critical operation set are presented.The comparison experiment results show that the proposed algorithm's solution quality outperform that of the other comparison algorithms.Aiming at the integrated scheduling problem of complex tree structure products in cloud manufacturing environment,an improved teaching-learning-based optimization algorithm with differential learning is proposed.Firstly,an encoding method based on the dynamic schedulable operation set is given to handle the priority constraint problem in tree-structured complex products,which guarantees the feasibility of initial individuals.Then,the framework of teaching-learning-based optimization algorithm is used to solve the problem.In the teacher stage,two different discrete teaching operators based on position and operation are designed to ensure the legitimacy of the newly generated learners,thereby avoiding the detection and repair of infeasible individuals.In the learner stage,by dividing outstanding learner group,different learning strategies are implemented for different individuals.For individuals in outstanding learner group,the self-learning strategy based on critical operation set and the intra-group learning strategy are proposed.For the remaining learners,this paper puts forward a learning strategy based on the excellent individual information platform,so that the poor individuals can obtain knowledge from the information platform.Finally,the comparison experiment results show that the proposed algorithm's solution quality outperform that of the other comparison algorithms.