Process analytical technology (PAT) and real-time control in chemical manufacturing

Process analytical technology (PAT) has expanded the ability to obtain timely information about manufacturing processes and product-related attributes. However, much of the existing literature focuses on analytical technologies, monitoring applications, or sector-specific implementation, while the connection between PAT measurements, dynamic process models, state estimation, and closed-loop control has received less integrated treatment. This review addresses this gap by examining PAT-enabled real-time manufacturing from a control-oriented perspective across pharmaceutical, chemical, polymer, food, and bioprocess applications. Analytical technologies and their limitations are evaluated together with feedback, feedforward, model predictive control (MPC), nonlinear model predictive control (NMPC), state-estimation methods, and hybrid modelling approaches. A representative dynamic process and MPC framework are also considered to show how measurements, process states, manipulated variables, disturbances, constraints, and quality objectives can be linked within a control system. The reviewed evidence indicates that PAT should not be treated as a control strategy by itself. Its value depends on the reliability of analytical measurements and their integration with suitable models, estimators, and validated control approaches. Implementation maturity varies considerably among industries, with stronger evidence available in pharmaceutical continuous manufacturing and selected crystallisation and process-control applications. Important limitations include calibration transfer, sensor drift, sampling representativeness, plant-model mismatch, model maintenance, validation, and data integration. Emerging AI and digital-twin approaches may extend prediction, estimation, and decision support, but their industrial use remains dependent on reliable data, validated models, and appropriate lifecycle management.