Análise de regressão: uma aplicação à previsão da potência do motor de um moinho de cimento

Abstract

About 2% of global energy is consumed by cement production. Most of this energy is used in the clinkerization and raw material grinding processes. With a constant yearly increase in the total amount of cement produced, cement industries are increasingly opting for vertical mills. In addition to being compact and more easily adaptable to factory layouts, vertical mills combine several operations into a single unit. Thus, this work aims to optimize energy efficiency and promote greater sustainability in the clinker grinding circuit of a vertical cement mill. For this purpose, linear regression analyses were conducted to understand the behavior of the process variables and, subsequently, a machine learning model was implemented to accurately predict the mill motor’s power. Since this is a variable that affects the quality of the cement produced, due to its direct influence on the mill’s operation, preprocessing is necessary before conducting analyses and applying the model. This preprocessing involves using statistical models to recognize patterns in the dataset, as well as identifying and removing outliers. After constructing the dataset, different predictive models are evaluated and tested, and the one that best fits the power flow will be implemented. The expected outcome is that the predictions provided by the selected predictive model will enhance the production process by promoting greater energy efficiency, reducing water consumption, minimizing wear on the mill’s components, and decreasing vibration.