Application of unsupervised machine learning techniques to evaluate control performance in PID control loops
Authors
Tainara Marques
Industrial and Systems Engineering Graduate Program (PPGEPS), Pontifícia Universidade Católica do Paraná (PUCPR), Rua Imaculada Conceição 1155, Curitiba, Brazil.
Gilberto Reynoso-Meza
Industrial and Systems Engineering Graduate Program (PPGEPS), Pontifícia Universidade Católica do Paraná (PUCPR), Rua Imaculada Conceição 1155, Curitiba, Brazil; Control Systems Optimization Laboratory (LOSC), Pontifícia Universidade Católica do Paraná (PUCPR).
Jesús Carrillo-Ahumada
Ingeniería en Alimentos. Universidad del Papaloapan. Circuito Central 200, Col. Parque Industrial, 68301 Tuxtepec, Oaxaca, Mex.
Control performance assessment, unsupervised machine learning, PID control
Abstract
Seeking to ensure safe and reliable operation in industrial processes, Control Performance Assessment (CPA) has been employed in control systems to ensure they operate effectively and efficiently. Traditionally, CPA has been approached using control performance indicators. However, the integration of data science and unsupervised machine learning has emerged as a promising alternative for classification tasks related to CPA. This study is a continuation of previous research, exploring the use of unsupervised machine learning for CPA classification. We utilized a database describing 30 control performance indicators in a PID control loop. The results reveal that the effectiveness of unsupervised learning technique was not as significant as the approach in the previous study. This suggests the need to investigate alternative approaches or adjust model parameters to enhance its performance in CPA classification.
