Strategy for Improving Data Quality for a Kraft Pulp Mill Recausticizing Plant

Steady state simulation of a kraft pulp mill recausticizing plant was completed for the purposes of investigating process changes for an area upgrade. Due to errors in the measured data as well as assumptions made due to a lack of necessary information, the initially completed simulation was not the best representation of the process at steady state. A comprehensive approach has been used to improve and validate the simulation. Firstly, a multi-scale data processing technique based on wavelet transform was used to eliminate errors caused by temporal events, including random noise and abnormalities in the raw measurements. Steady state was then detected using the wavelet-based method. Data reconciliation was used to reduce systematic errors and to calculate some of the necessary variables that were not measured. Combined with process knowledge, the simulation was validated using the reconciled data.