Article ID: TETSU-2018-040
This paper discussed reaction on thermal decomposition of limestone and Boudouard reaction.
Thermogravimetric analysis of mixed powder samples of limestone and carbonaceous material was carried out. The ratio of the sequential reaction, αc was 0.65 when limestone powder with large particle size and graphite powder with small particle size were used. It was found that the reactivity varies depending on the states of dispersion and mutual coating of the powder particles. Deep learning by recurrent neural network (RNN) and convolutional neural network (CNN) was applied to calibrate weight loss curve of TG analysis and predict reactivity of samples. The TG curve corrected by RNN was almost equivalent to that processed manually. CNN required more learning to evaluate the reactivity of the sample more accurately in the present conditions. We presented that the constructed models are extremely powerful tool for evaluation of metallurgical reactions.
