A probabilistic framework for online structural health monitoring: active learning from machining data streams

A critical issue for data-based engineering is a lack of descriptive labels for the measured data. For many engineering systems, these labels are costly and/or impractical to obtain, and as a result, conventional supervised learning is not feasible. This paper suggests a probabilistic framework for the investigation and labelling of engineering datasets; specifically, acoustic emission data streams recorded online from a turning machine. Two alternative probabilistic measures are suggested to select the most informative observations. During machining operations, these data would then be investigated and annotated by an engineer, in order to maximise the classification performance of a statistical model used to predict tool wear.