Condition monitoring with Mean field independent components analysis

We discuss condition monitoring based on mean field independent components analysis of acoustic emission energy signals. Within this framework it is possible to formulate a generative model that explains the sources, their mixing and also the noise statistics of the observed signals. By using a novelty approach we may detect unseen faulty signals as indeed faulty with high precision, even though the model learns only from normal signals. This is done by evaluating the likelihood that the model generated the signals and adapting a simple threshold for decision. Acoustic emission energy signals from a large diesel engine is used to demonstrate this approach. The results show that mean field independent components analysis gives a better detection of fault compared to principal components analysis, while at the same time selecting a more compact model