In this contribution a recently developed new modeling and classification approach to be used with streamed measurement data of industrial processes is applied.
This briefly repeated approach can be used for fault classification and diagnostic purposes. The approach is based on a fuzzy-like modeling using statistical features from training data. Based on the trained model classification can be realized allowing to distinguish unknown data sets to the given number of data classes each related to states.
Beside the brief introduction to the proposed approach, experimental data are used to demonstrate the approach and the complex example distinguishing different wear states of machine components during operation.