A Data-Driven Statistical Approach for Monitoring and Analysis of Large Industrial Processes

Abstract

Monitoring and fault detection of industrial processes is an important area of research in data science, helping effective management of the plant by the remote operator. In this article, a data-driven statistical model of a process is estimated using the principal component analysis (PCA) method and the associated probability density function. The aim is to use the model to monitor and detect the incurred faults in the industrial plant. The experimental data are collected by finding the suitable subsystems of a Recycle Gas in Ethylene Oxide production process, and a subset of nine variables are extracted for further statistical analysis of the system. The performance of the developed model for monitoring purpose is evaluated by using faulty and close to faulty inputs as the new test data.