Research On Models And Algorithms For Steelmaking And Continuous Casting Batch Planning Problems

The batch planning problems arising in steelmaking and continuous casting areas are difficulty problems for the iron and steel industry production management. Reasonable batch plans can help to improve the productivity and to reduce the production cost. However, the models of batch planning problems are not precise and the solving algorithms are inaccurate and need a lot of time yet. Based on the analysis of the steelmaking and continuous casting production processes, four problems are researched in this dissertation, including charge planning problem, cast planning problem, charge planning problem with width and weight flexible of slabs, and cast planning problem with charges’width flexible. By analyzing the key characters of technologic constraints and management requirements, the steelmaking and continuous casting batch planning problems are abstracted to combinatorial optimization problems. Then, the mathematical models are created, and hybrid improved cross entropy algorithms is proposed. Finally, by combining the cross entropy method and the self-orgnizing concepts of synergetics, the batch planning self-orgnazing optimum algorithm is presented. The novelty of this dissertation lies in the capacitated profitable tour problem based model of flexible charge planning problem and the generalized vehicle routing problem based flexible width cast planning problem model, the improved cross enthropy algorithm with improved initial state transition probability matrix, and the self-organizing optimization algorithm composed of competition factor, cooperation operator and order parameters. The contents of the dissertation are summarized as follows.To solve the charge planning problem involving charges and their orders for each charge with unkown charge number, a traveling salesman problem based charge planning model and the improved cross entropy algorithm are proposed. The objective of the model is to minimize the dissimilarity costs between each order and its charge center order, the open order costs, and the unselected order costs. The improved cross entropy algorithm is proposed with the improved initial state transition probability matrix, which is constructed according to the differences of steel grades and widths between orders. An actual numerical example shows that the improved initial state transition probability matrix can effectively increase the speed and accuracy of the algorithm.To solve the cast planning problem, which involvs the cast number as well as the select and sequence of charges, the traveling salesman problem model and the hybrid heuristic-cross entropy algorithm are proposed. The cast planning problem is modeled with the objective of the minimum of the sum of start cast fees, continuous-cast fees and unselected charges cost. The hybrid heuristic-cross entropy algorithm is proposed with the steel-grade group rules and the improved initial state transition probability matrix, which is constructed according to the continuous-cast fees between charges. An actual numerical example shows that the hybrid heuristic-cross entropy algorithm can achieve same or better optimum value in less time with higer stability, compare to the cross entropy method.Based on the capacitated profitable tour problem, the mathematical model is constructed, which is for charge planning problem with slabs’flexible width and weight. Then the hybrid improved cross entropy method was proposed to solve this model. Actual numerical examples are given to demonstrate the effectiveness of the proposed model and algorithm. The algorithm shows its superiority by compare with intelligent algorithms such as genetic algorithm.According to the characteristics of the flexible cast planning problem, the model based on generalized vehicle routing problem is proposed. Based on the local-global approach, the flexible width cast planning problem is decomposed to two sub-problems, charge order and charge widths, which are solved by the improved cross-entropy method and reaching algorithm, respectively. Actual numerical examples demonstrate the effectiveness of the proposed model and algorithm.Aim to reduce the randomness of cross-entropy method, a self-organizing optimization algorithm is proposed to solve the batch planning problem. The proposed self-organizing optimization algorithm, which is inspired by synergetics, is composed of competition factor, cooperation operator and order parameters. The actual example shows that the introduction of competition factor and cooperation operator can accelerate the convergence speed and improve the stability of self-organizing optimization process. Especially, the self-organizing optimization algorithm can achieve high-quality batch plans in less time for large scale batch planning problems.