Intelligent decision support system based on video recognition of the blast furnace tuyeres

Abstract

An approach to creation of an intelligent system for predicting the state of a technological process in real time is presented. The approach is based on the analysis of a video sequence of images obtained as a result of streaming video by cameras installed on the tuyeres of a blast furnace. Algorithms for recognizing video images of tuyere foci, as well as scenario forecasting of the evolution of technological situations are proposed. A historical background regarding the development of methods for automatic control of the blast-furnace process, in particular, the use of artificial intelligence, is presented. The study is aimed at the ability of rapid analysis of the production situation (PS) and prediction of the PS evolution in the course of functioning of the blast furnace process, which will provide the possibility of timely decisions on adjusting control in an automatic or automated mode. Using the developed algorithm for analysis and prediction of the process dynamics and proceeding from the revealed regularities of the change in video data, a method for early detection of a tendency to the occurrence of certain events on tuyeres, including those leading to the destabilization of the blast furnace process, is proposed. The novelty of the presented approach lies in the fact that not only the state of the process at the next moment of time, but also the most probable chain of several subsequent states is predicted. Real-time forecasting algorithms are based on the construction and replenishment of the base of inductive knowledge — regularities revealed through the intellectual analysis of the revealed information — in the course of real functioning. Methods of studying Markov chains, machine learning and wavelet analysis are used for the associative search for patterns. The algorithms developed by the authors can be used in decision support systems for blast-furnace control. The results of practical research, confirming the effectiveness and viability of the proposed approach, are presented.