A normal operating zone model for condition monitoring of reaction-separation-recycle processes

Abstract

Normal operation of reaction-separation-recycle processes is important for the safety of chemical plants. This study presents a normal operating zone (NOZ) model for condition monitoring of reaction-separation-recycle processes. The NOZ model is a geometric space in higher dimensions that is developed based on physical principles, historical data, and operational requirements. A main technical issue is the uncertainties of NOZ models from historical data with noise disturbance. Bayesian estimation theory is utilized to address this issue. The NOZ model represents an allowable variation region of multiple related process variables, and compensates certain limitations of existing methods. In particular, the NOZ model can account for normal conditions that are not present in historical data compared to data-based methods; the NOZ model can identify abnormal conditions that do not change physical models in contrast to model-based methods. The effectiveness of the proposed method is illustrated via Aspen-Hysys-based examples with comparison to existing methods.