Predictive modeling and optimization of paper mill using hybrid machine learning techniques

Abstract

The paper has played a vital role in the life of humans from ancient times covering a vast range of applications such as writing, packaging, and printing. The present paper is presenting a comprehensive review of various optimization and control methodologies, ranging from conventional to advanced ones, pertaining to the paper mill. The final goal of these control strategies is to upgrade the mill’s production and quality in presence of multiple technical challenges such as nonlinear and multivariable nature of the involved processes, various disturbance parameters, and time delays. In this work, the integration of machine learning with paper mill process is illustrated. For any manufacturing process, the final product quality is the key goal. There are various traditional techniques which have already been practiced for final produced paper quality in paper mills. This paper highlights the capability of support vector machine (SVM) algorithm to assess the produced paper quality, capturing the two crucial inputs viz. the pulp consistency and the headbox level. The basic goal of this research is twofold, firstly it presents an exhaustive literature survey exploring various strategies which are practiced currently in the domain of control and optimization of various paper mill processes. Secondly, it intends to develop and evaluate various SVM and SVM-RF hybrid models using MATLAB for assessment of quality of final product on basis of two parameters- pulp consistency and head box level. Finally, genetic algorithm has been employed in MATLAB for multivariate optimization.