Inteligentní plánování sekvencí taveb na zařízení plynulého odlévání oceli

Abstract

This dissertation deals with the issue of intelligent planning of sequential melting on continuous steel casting equipment. The algorithmization of the support of the planning process on continuous steel casting equipment which works in the proposed methodologies and implementation of the algorithm, which at any time allows to determine the number of melts of steel marks that can be cast to maximize the technical life of the crystallizer. A secondary requirement for this support of the planning process is that the proposed algorithm allows the joining of steel grades that are chemically complex and thermally very close to a single sequence.

The proposed methodology and the implemented presented system for the support of intelligent planning is based on the methods of artificial intelligence – specifically the neuro-genetic system. The proposed methodology of intelligent support of the planning process represents an optimization task, the solution of which is based on the solution and determination of the purpose function of the solved process. In this case, the purpose of the process of supporting the planning of the continuous steel casting process will have the character of a complex function – combining the life of the crystallizer insert and the aspect of the type of castings in individual sequences and is implemented using an artificial neural network. The data for the input variables for the training set of the neural network are determined from the technological parameters of individual cast steel grades to which the clustering Fuzzy Gustafson-Kessel algorithm is applied. The optimization itself is performed by a genetic algorithm which can be replaced by an evolutionary approach.

The presented system of support for planning of melting sequences is implemented in Microsoft Excel, it offers to the user the possible and effective variants of production layout in the next period according to the specified restrictive conditions. This model is used only for partial verification and testing of the proposed models within the solution of the given problem.