Semi-supervised Learning for Data-driven Soft-sensing of Biological and Chemical Processes

TL;DR

This paper explores semi-supervised learning techniques to develop soft-sensors for biological and chemical processes, aiming to improve process control when key measurements are costly or infrequent.

Contribution

It demonstrates the application of semi-supervised regression in process control, comparing its effectiveness against standard methods through two real-world case studies.

Findings

Semi-supervised regression outperforms standard regression in certain scenarios.

Enhanced soft-sensing accuracy for rarely measured process variables.

Potential for improved process control in bio-chemical industries.

Abstract

Continuously operated (bio-)chemical processes increasingly suffer from external disturbances, such as feed fluctuations or changes in market conditions. Product quality often hinges on control of rarely measured concentrations, which are expensive to measure. Semi-supervised regression is a possible building block and method from machine learning to construct soft-sensors for such infrequently measured states. Using two case studies, i.e., the Williams-Otto process and a bioethanol production process, semi-supervised regression is compared against standard regression to evaluate its merits and its possible scope of application for process control in the (bio-)chemical industry.