Prediction of residual calcium carbonate (CaCO3) in the lime kiln of a kraft pulp mill

In a kraft pulp mill, the rotary lime kiln is responsible for converting calcium carbonate (CaCO3) into calcium oxide (CaO) to be reused in the causticizing plant. An important parameter of the calcination efficiency and the product quality is the amount of CaCO3 not converted to CaO, commonly defined as residual carbonate. This parameter is usually determined through laboratory analysis, which introduces delays in process control and limits the ability to make timely operational adjustments. This work presents a predictive model that functions as a soft sensor for the residual carbonate in a lime kiln of a Brazilian kraft pulp mill. This model was able to estimate residual CaCO3 every 30 min using routinely measured online process variables, which is a considerable reduction from the 6-h average laboratory measurement. The correlation coefficient (r) between the model predictions and the mill values was 0.83, with a mean absolute error (MAE) of 0.33%. Considering the effect of the residence time throughout the lime kiln proved to be decisive in obtaining a satisfactory prediction result. After model validation, a variable importance analysis showed that the carbon monoxide (CO) concentration in the stack flue gas (in ppm), secondary air temperature (in oC), and fan speed (in %) were the most influential variables in predicting the residual carbonate content. Combined with commonly used operation expertise, this subset of variables can serve as additional information to achieve smoother lime kiln operation.