Machine Learning for Soft Sensing and Process Estimation in Industrial Systems: Advances, Challenges, and Future Prospects

Abstract

Soft sensing and process estimation are critical techniques in modern industrial automation and process control systems. Conventional measurement approaches often rely on physical sensors, which can be expensive, prone to drift, or impractical for certain process variables. Machine learning (ML) has emerged as a powerful alternative for developing data-driven soft sensors that infer unmeasured or difficult-to-measure variables from readily available process data. This paper presents a comprehensive review of machine learning-based soft sensing and process estimation approaches, highlighting recent advancements, common algorithms, applications across various industries, associated challenges, and future research directions. Emphasis is placed on the advantages of ML models over traditional statistical methods, the integration of real-time monitoring with predictive control, and the potential of hybrid modeling approaches that combine first-principles knowledge with data-driven insights.

KEYWORDS: Soft sensing, Process estimation, Machine learning, Industrial automation, Predictive modeling, Data-driven control, Process monitoring.