An Approach to Fault Detection Using a Unified Linear Gaussian Framework

Presented in this paper is an approach to fault diagnosis based on a unifying review of linear Gaussian models. The unifying review draws together different algorithms such as PCA, factor analysis, hidden Markov models, Gaussian mixture models and linear dynamical systems as simple variations of a single linear Gaussian model. The focus in this work is on PCA, factor analysis and linear dynamical systems in order to compare static and dynamic data models. One of the advantages of using a unified framework for these already well known models is the quantification of uncertainty, and the ability to evaluate the probability of a dataset given a model. To highlight the applicability of such an approach to a fault diagnosis application and to establish in some way its robustness to environmental variability, a set of data collected from the suspension system of a ground vehicle is used.