Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Inspection and supervision by means of hierarchical fuzzy classifiers

Diagnose und Überwachung mit substrukturierten Fuzzy-Klassifikatoren
: Priber, U.; Kretzschmar, W.


Fuzzy sets and systems 85 (1997), pp.263-274 : Ill., Lit.
ISSN: 0165-0114
Journal Article
Fraunhofer IWU ()
cluster analysis; Clusteranalyse; fuzzy set; hierarchical structure; hierarchische Struktur; inspection of technical processes; membership function; Mustererkennung; pattern recognition; regelbasiertes System; rule base; Überwachung technischer Prozesse; unscharfe Menge; Zugehörigkeitsfunktion

The well-tried method of fuzzy classification is extended to hierarchical class structures to allow combinations of data-based and rule-based class representations. An appropriate design method is suggested, the relationships to system theory and fuzzy logic are shown and software features and application examples are described. The main application fields are diagnostics and supervision of technical processes. Model design is based on disjunctive aggregation of atomic classes, obtained by fuzzification of vectors in the feature space (learning samples). Membership functions of all fuzzy sets are described by an uniform multidimensional parametric concept. Rule-based components are an appropriate formal tool to represent multimodal class models and to integrate explizit expert knowledge. Class models can be seen as components of several types of systems (e.g. fuzzy controller) which transform measured data into discrete fuzzy sets. For this reason the initiation of further development of new fuzzy methods is expected. Two application cases with use of a class structure and its effect are studied.