Otimizacão do consumo de energia em moinhos de rolos do tipo Raymond através do algoritmo Particle Swarm Optimization em um modelo surrogate baseado em Random Forest

Abstract

This work proposes the use of Random Forest and Particle Swarm Optimization to optimize the energy consumption of a Raymond mill, used for grinding natural calcium carbonate at low particle sizes. Initially, the data is filtered, cleaned and organized and then goes through a preanalysis step, using statistical concepts that include structure and dispersion, distribution and finally correlation of the data. After that, Random Forest technique is applied in a supervised learning process with feature selection and tuning of execution hyperparameters, and ends with the energy consumption optimization process by applying Particle Swarm Optimization in two different implementations, for comparison. The results show a reduction potential of approximately 9 % when compared to the current consumption of the same equipment, also demonstrating the possibility of a gain in scale, when the same methodology is applied to other equipment within the same industrial plant.