Fault detection and diagnosis via improved multivariate statistical process control

Multivariate Statistical Process Control (MSPC) technique has been widely used for fault detection and diagnosis (FDD). Currently, contribution plots are used as basic tools for fault diagnosis in MSPC approaches. This plot does not exactly diagnose the fault, it just provides greater insight into possible causes and thereby narrow down the search. Hence, the cause of the faults cannot be found in a straightforward manner. Therefore, this study is conducted to introduce a new approach for detecting and diagnosing fault via correlation technique. The correlation coefficient is determined using multivariate analysis techniques, namely Principal Component Analysis (PCA) and Partial Correlation Analysis (PCorrA). An industrial precut multicomponent distillation column is used as a unit operation in this research. The column model is developed using Matlab 6.1. Individual charting technique such as Shewhart, Exponential Weight Moving Average (EWMA) and Moving Average and Moving Range (MAMR) charts are used to facilitate the FDD. Based on the results obtained from this study, the efficiency of Shewhart chart in detecting faults for both quality variables (Oleic acid, xc8 and linoleic acid, xc9) are 100%, which is better than EWMA (75% for xc8 and 77.5% for xc9) and MAMR (63.8% for xc8 and 70% for xc9). The percentage of exact faults diagnoses using PCorrA technique in developing the control limits for Shewhart chart is 100% while using PCA is 87.5%. It shows that the implementation of PCorrA technique is better than PCA technique. Therefore, the usage of PCorrA technique in Shewhart chart for fault detection and diagnosis gives the best for it has the highest fault detection and diagnosis efficiency