Data-based monitoring of model-based PLT safety functions

Anomaly detection in reproducible, chemical batch processes

Abstract

In complex and model-based PCT safety functions,  uncertain process measurement values are used to  calculate shutdown values that are not directly observable,  leading to high safety margins when using manufacturers’  safety measurement accuracy specifications and common  error propagation methods. This paper outlines an  approach to data-based monitoring and anomaly detection for reproducible chemical batch processes. which would  justify the use of lower safety margins based on measurement repeatability, and thus have a highly positive  impact on the operability of complex and model-based PLT  safety functions. The potential of this method to increase  the degree of diagnosis and detection of unknown  anomalies in the production process is demonstrated and  discussed using historical process data.