Titanium Dioxide Content Soft Sensor Development for Pilot‐Scale Ilmenite Electric Arc Furnace Using <scp>BiLSTM</scp> and <scp>BiGRU</scp> Recurrent Neural Networks

ABSTRACT Electric arc furnaces (EAFs) are central to various metallurgical processes for melting and upgrading ore. These furnaces use a significant amount of energy and consumables to operate, which suggests substantial potential for gains in operational efficiency. In this study, we propose a proof of concept for estimating a pilot‐scale ilmenite smelting electric arc furnace critical quality variable: the titanium dioxide content of the slag bath. This quality variable is estimated using a soft sensor based on a data‐driven machine learning (ML) model. The proposed ML model is trained using EAF sidewall temperatures, electric power, ore charge, and reducing agent charge values. To account for the nonlinear and dynamic nature of the semibatch process, models based on long short‐term memory (LSTM) and gated recurrent unit (GRU) neural network architectures are tested and evaluated. A systematic hyperparameter tuning approach allowed obtaining good estimation performance with an MSE of 0.23, an RMSE of 0.48, and an R2 of 0.78.