Application of fault monitoring and diagnostic techniques and their challenges in petrochemical industries

Abstract

Petrochemical plants are complex and highly interconnected, as a result control and optimization of the process is a challenge. Prior to control, it is important and necessary to find out the process deviation from the requisite normal operating region at an early phase to avoid the huge revenue loss. The amazing development in the soft sensor technology and information technology has been very significant in improving the online process display. They have facilitated the industries to collect and store voluminous data more frequently at all stages of the process. These available data libraries can be used to extract information and gain process knowledge. In the recent years, online process monitoring using fault monitoring and diagnostics (FMD) package has become an important component in the petrochemical plants in order to ensure safe and efficient plant operation.Presently, FMD products from different process automation giants prefer data driven models to first principle models (e.g., component balance, energy balance etc.). The main reason is that data driven models are easy to construct. FMD constitutes model construction using multivariate statistical analysis tools such as principal component analysis and its utilization to demarcate between normal and abnormal operation of the plant. Eventually, engineers use the key parameter indicators (KPIs) from the FMD tool to make decisions about the plant maintenance.The key focus of this work is on the application of data driven FMD model for online monitoring of a polymer reactor in a petrochemical plant. Moreover, this work highlights the practical challenges in developing an accurate and customized FMD models for the petrochemical plants.