Augmented mahalanobis distance for incipient fault detection of industrial processes

Abstract

For modern industrial processes, timely detection of incipient faults is of vital importance so as to ensure safe and optimal process operation. Though recently statistical process monitoring (SPM) has been extensively studied and widely applied in practice, conventional multivariate statistics are usually not sensitive to incipient faults. In this paper, a new multivariate statistical index called augmented Mahalanobis distance (AMD) is proposed for incipient fault detection. It can be concluded from fault detectability analysis that the AMD index is more sensitive to incipient faults than the conventional Mahalanobis distance (MD) index. Besides, the idea of augmentation utilized in the AMD index can also be applied to some other SPM models. Finally, case studies on a numerical example and the continuous stirred tank heater (CSTH) process are conducted to demonstrate the effectiveness of the proposed AMD index, in comparison with the MD index, as well as the squared prediction error (SPE) and T-square indices.