Industrial IoT for Intelligent Steelmaking With Converter Mouth Flame Spectrum Information Processed by Deep Learning

• Published in: IEEE transactions on industrial informatics 2020-04, Vol.16 (4), p.2640-2650
• Main Authors: Han, Yang, Zhang, Cai-Jun, Wang, Lu, Zhang, Yan-Chao
• Format: Article
• Language: English
• Subjects: Algorithms; Automatic control; Bidirectional recursive; Blowing rate; Carbon; Carbon content; Chemical reactions; Convolution; Deep learning; Fires; Industrial applications; Internet of Things; Internet of Things (IoT); Mouth; multiscale convolution; Neural networks; Organic chemistry; Process control; Smelting; Steel; Steel construction; Steel converters; Steel making; steelmaking end-point
• ISSN: 1551-3203; 1941-0050
• DOI: 10.1109/TII.2019.2948100

In this article, based on the fact that the converter mouth flame is the comprehensive external appearance of the physical and chemical reactions in the converter during the steelmaking process, the continuous spectrum information of the converter mouth flame was obtained by USB4000 spectrometer. In the framework of the Internet of Things, a bidirectional recursive multiscale convolution depth neural network algorithm can take into account the characteristics of frequency domain structure and time domain dynamic sequence. It is applied to the deep-learning of the converter mouth flame spectrum information. The dynamic prediction model of carbon content and temperature value in molten steel at the later stage of steelmaking is constructed. The static control system and dynamic prediction model of automatic steelmaking are intelligently fused to realize one-key steelmaking control. The results show that the average hit rate of carbon content, temperature, and carbon temperature at the end of steelmaking is 94.78%, 98.41% and 93.43%, which makes the end-point control of steelmaking more stable and the blowing rate less than 1%.