Estudo de variáveis que afetam a vibração do corpo de um moinho vertical de rolos

Abstract

Cement production is one of the most energy-intensive manufacturing industries, with the factories’ cement grinding circuit consuming around 4% of global annual electrical energy production. Based on this, there are several types of mills used in cement grinding, the main ones being the ball mill and the vertical roller mill. Vertical mills are more advantageous because they use 30% to 50% less energy than ball mill systems and because they have a simple layout, thus reducing the number of machines in the circuit, guaranteeing a high coefficient of operation and low maintenance costs. Controversially, the vertical roller mill is very sensitive to sudden variations in the process, thus causing high rates of vibration in the mill body, generating cracks in the welds, breakage of the roller covers, severe wear on the grinding table, and other problems. That said, the central objective of the work is to analyze the variables that most contribute to the vibration of the body of a vertical roller mill intended for grinding cement in a cement plant. Thus, data collection was carried out for 10 months of operation of a vertical mill and then the scattered data was integrated, excluding outliers, empty values and variables that are not related to the target. The methods used to perform data analysis were multiple linear regression, moving average and Lag or delay. Therefore, the results showed that the relationship between the variables are not linear, requiring a different model that can capture such relationships. It is also noted in the analysis some spurious correlations due to the interference of the command center operator in the increase and reduction of productivity, thus affecting the behavior of the response variable. Therefore, a new analysis was carried out maintaining an almost constant total power supply range, to avoid interference from the control panel operator, thus revealing a better adjusted model. However, there are improvements to be made with the use of a model that can assimilate the non-linear relationships between the explanatory variables and the response. These analyzes were not carried out in this study, due to lack of time, and will be left here as future work.