DATA DRIVEN MODEL FREE SYSTEM IDENTIFICATION OF BATCH DISTILLATION COLUMN USING XGBOOST MACHINE LEARNING ALGORITHM

Abstract

The distillation process plays a crucial role in the chemical industry, enabling material separation, purification, and waste product disposal. Distillation columns, including the batch type, are widely used in industries due to their ability to produce raw materials for various applications. However, modeling and controlling batch distillation columns pose challenges due to their nonlinear and dynamic behavior. This paper presents a novel data-driven approach for system identification using XGBoost, an advanced gradient boosting algorithm, eliminating the need for explicit model equations. The proposed methodology leverages the power of XGBoost to learn the underlying system behavior directly from data. The paper provides an overview of the methodology, including data preprocessing, feature engineering, training the XGBoost model, and evaluating its performance. Techniques such as cross-validation and input feature delay tuning are also discussed to ensure robustness and optimal model performance. The effectiveness of the approach is demonstrated through various case studies and some comparisons. The results highlight the capability of the proposed model-free system identification methodology using XGBoost in accurately capturing the dynamics of batch distillation systems, offering potential for improved process control and optimization in the chemical industry.