Improved multivariable statistical process control (MSPSC) for chemical process fault detection and diagnosis (PFDD) - Cross-variable correlation approach

Abstract

Chemical process is inclined to be large-scale, continuous, integrated and linkable around the whole plant. It has put forward tough requirements on desired quality, high yield/production, low consumption, environmentally friendly and safe operation. Thus accurate process fault detection and diagnosis (PFDD) system at early stage of the process are very important to modern chemical plant in achieving the above vision. To ensure that the product is obtained in the acceptable limit ranges, some monitoring and control methods have to be applied. This paper focuses on the application of Multivariate Statistical Process Control (MSPC) as the fault detection and diagnosis tool. This is because MSPC method can easily monitor large volumes of highly correlated data, and the process information is clearly presented in the analysis chart. The paper explains how the cross-variable correlation coefficient and the multivariate control chart are developed. When a fault(s) is detected, the cause of the fault is diagnosed by means of contribution charts. The result, which is based on this new approach perform better compared to those based on conventional MSPC analysis.