Electrical conductivity is of fundamental importance in electric arc furnaces (EAF) and the interaction of this phenomenon with the process slag results in energy losses and low optimization. As mathematical modeling helps in understanding the behavior of phenomena and it was used to predict the electrical conductivity of EAF slags through artificial neural networks. The best artificial neural network had 100 neurons in the hidden layer, with 6 predictor variables and the predicted variable, electrical conductivity. Mean absolute error and standard deviation of absolute error were calculated, and sensitivity analysis was performed to correlate the effect of each predictor variable with the predicted variable.
The viscosity of lead-containing glasses is of fundamental importance for the manufacturing process, and can be predicted by algorithms such as artificial neural networks. The SciGlass database was used to provide training, validation and test data of chemical composition, temperature and viscosity for the construction of artificial neural networks with node variation in the hidden layer. The best model built with training data and validation data was compared with 7 other models from the literature, demonstrating better statistical evaluations of mean absolute error and coefficient of determination to the test data, with subsequent sensitivity analysis in agreement with the literature. Skewness and kurtosis were calculated and there is a good correlation between the values predicted by the best neural network built with the test data.
Viscosity in the metallurgical and glass industry plays a fundamental role in its production processes, also in the area of geophysics. As its experimental measurement is financially expensive, also in terms of time, several mathematical models were built to provide viscosity results as a function of several variables, such as chemical composition and temperature, in linear and nonlinear models. A database was built in order to produce a nonlinear model by artificial neural networks by variation of hyperparameters to provide reliable predictions of viscosity in relation to chemical systems and temperatures. The model produced named Viskositas demonstrated better statistical evaluations of mean absolute error, standard deviation and coefficient of determination in relation to the test database when compared to different models from literature and 1 commercial model, offering predictions with lower errors, less variability and less generation of outliers.