Prediction of Dust Concentrations in A Cement Plant Using Artificial Neural Networks: A Process Data-Based Approach for Environmental Control

Abstract

The production of clinker in cement plants is a significant source of dust emissions, which poses a major challenge for compliance with environmental standards and worker health. This research envisions the development of an artificial neural network (ANN) model to predict dust levels in near real-time, based on easily quantifiable operational criteria. The model relies on five key variables: the rate of burned gas, the temperature of the exhaust gasses at the tower outlet, the volume of flour introduced into the oven, the differential pressure of the filter (aff2), and the percentage of excess air (O₂). For the model's learning and validation, we used an exhaustive dataset including historical records for 2023 and 2024 (more than 13,000 observations in total). The results highlight the exceptional predictive efficiency of the model, which has a solid ability to generalize across different operating conditions and accurately capture the complex nonlinear relationships that govern dust emissions. This method offers a promising tool for cement producers, promoting better environmental control and an anticipated reduction in dust emissions.