Využití Bayesovských sítí pro řízení procesů

Abstract

Title: Application of Bayesian networks to process control Author: Anton Švec Department: Department of Automation and Computer Application in Metallurgy Supervisor: doc. Ing. Jiří David, Ph.D. Supervisor's e-mail address: j.david@vsb.cz Abstract: The key part of this work is concerning about solution of industrial systems reliability by theory of Bayesian networks. Labor models of Bayesian network structure corresponding to the structure of an industrial object, so that the conditional reliability of individual units corresponding to relations of network. Network architecture includes information concerning the individual blocks and their relationships. Introduction is aimed at getting acquainted with the tools of artificial intelligence to solve the problem evaluation of of uncertain data, focusing on the causal network. Particular to Bayesian networks and influence diagrams. Next section explains and applies some known optimization of this theory, namely the method of progressive modifications, the transformation to decomposable model and Junction tree calculation. Further part is description of the program Genie and as a CD media supplement the Czech version reference manual. Labor shows the advantages and disadvantages of application of Bayesian networks for solving the problem compare to the solution of the problem using different methods, namely using classical identification and calculation of failure probability and reliability. Work implements the model solving the problem by using Bayesian networks used to demonstrate the calculation carried out by this method. It uses to store the structure of Bayesian network free distributed Smile library.