Research And Application Of Parallel Line-up Competition Algorithm For Batch Scheduling In Process Industry

Process industry is a major basic industry of national economy. In recent years, with the development of economic globalization, the traditional process industry is facing great challenges. With increasingly keen market competition and more complex business environment, each enterprise in process industry is searching for good solutions for production and operation to improve the efficiency of production operation, thereby enhance its own core competitive advantage. Production scheduling is the core content of enterprise management, and the reasonable production schedule not only can improve the service level of enterprises, but can bring significant economic benefits for enterprises. At present, the bottleneck problems for solving the production scheduling problem in process industry are modeling and optimization algorithms. The existing methods often used a large number of integer variables and nonlinear inequalities to represent various constraints in actual production, leading to large model size and computational intractable. Therefore, it is very difficult for the existing methods to obtain acceptable solutions to large-scale problems within reasonable time.This thesis focuses on the batch production scheduling problems in process industry. New ways of modeling and algorithm are presented for solving the large-scale complex scheduling problems in batch plants effectively. The key problems addressed in the thesis are as follows.Based on the combination of conventional line-up competition algorithm (CLCA) and parallel computing, a parallel line-up competition algorithm (PLCA) for solving large-scale complex optimization problems is proposed. An improved coarse-grained parallel model with virtual master node has been presented to implement the parallelization. A new dynamic migration topology, dynamic offspring reproduction scheme and multilevel competition model are employed to balance well global search and local search, leading to the rapid convergence of the proposed algorithm. Comparative study on a group of benchmark functions show that PLCA is superior to CLCA for solving large-scale problems. Computational results on several high-constrained real-life engineering design examples further demonstrate PLCA has good performance as well as good robustness compared with other evolutionary algorithms in literature.A new model for single-stage multi-product scheduling problem (SMSP) in batch plants with parallel units is proposed. In this model the complex scheduling problem is decomposed into order assignment and order sequencing subproblems. The assignment subproblem is solved using a group of heuristic order assignment rule and flexible constraint handling strategies, while the sequencing subproblem is solved using parallel line-up competition algorithm that mutates the sequence of orders and the order assignment rule simultaneously to search the better solutions. Computational results on examples in literature show that the proposed approach can quickly obtain the same optimal solutions as obtained by the approaches in the literature for solving medium-scale and small-scale problems. However, for all 50-order problems with different complex constraints, the proposed approach has obtained better solutions than those of literature.The multi-stage multi-product scheduling problem (MMSP) in batch plants with parallel units is studied. The decomposition scheduling method and forward-backward assignment strategy are proposed to solve the MMSP. In this basis, a new model for MMSP in batch plants with parallel units is presented. Based on a mixed swap-reverse mutation strategy for the order sequence and a proportion-based search space allocation strategy for the order assignment rules, PLC A is used to solve the high-constrained examples with different scheduling objectives in literature. Computational results indicate that the proposed approach has obtained the better solutions than those of the literature for all scheduling problems with more than 10 orders. Moreover, with the problem size increasing, the solutions obtained by the proposed approach are improved remarkably.A new model for multi-purpose scheduling problem (MSP) in batch plants with parallel units is proposed. The objective function of this new model is the weighted production completion time, and the variables are sequence of order process steps and order assignment rules in each stage. To solving this model, more effective mixed mutation strategy and migration strategy are proposed to improve the performance of PLCA. Computational results on examples in literature show that the proposed approach is suitable for both sequential MSP and MSP with parallel units. Especially, for the complex MSP with parallel units, the proposed approach has obtained better results than those of the literature. On the other hand, the computational results on virtual examples show that a reasonable scheduling scheme that can meet the customer needs and improve production efficiency simultaneously can be obtained.The scheduling problem in paint production process is studied. According to the actual situation in paint enterprise, a new weighted scheduling objective consisting of total completion time of orders and total tardiness is proposed. The approach for solving MMSP with parallel units is applied to the actual paint plant. And then scheduling schemes that can maximize customer satisfaction and minimize the total completion time of orders simultaneously can be obtained.