Multivariate statistical based process monitoring using principal component analysis: An application to chemical reactor

The monitoring of industrial chemical plants and diagnosing the abnormalities in those set ups are crucial in process system domain as they are the deciding factors for the betterment of overall production quality in the process. Various statistical based malfunction detection methods have been included in the literature, namely, univariate and multivariate techniques. The univariate techniques are limited for monitoring only a single variable at a time whereas multivariate techniques can handle multiple correlated variables. Principal component analysis (PCA), a multi-variate technique, has been successfully used in the domain of process monitoring. PCA is used along with its two fault detection indices, T2 and Q statistics for detecting faults in any process. In the present study, a benchmark Continuous stirred tank reactor (CSTR) model is used to test the performance of the proposed PCA method. The simulated results show the effectiveness of the proposed method in handling different sensor faults in a CSTR process.