1991 Volume 77 Issue 1 Pages 79-84
As for operation of the blast furnace, it is important to recognize the distribution pattern of measured furnace data. Usually, these data are identified visually by human experts. Recently, the neural network technology is expected to be a new technology realizing the artificial pattern recognition of the data in good agreement with human judgement.
In this paper, the application of back propagation type neural network to the pattern recognition of blast furnace data is studied. This type of network can distinguish some specific pattern from others after learning typical teaching data. Such teaching data can be arranged in two ways. One way is to choose typical pattern out of the actual operation data. The other way is to create artificial teaching data, when it is difficult to find out that data from actual data. As for the recognition of the top gas temperature distribution in furnace the actual data are prepared for learning and for that of burden profile recognition, the artificial data are adopted. In both cases, the neural network succeeded in recongnition regardless of noises existing in the data.
