Symbol: mbFuzzIT-Logo

Software   -   Background

[DEUTSCH? hier klicken]


My Tool mbFuzzIT [1] enables the work with Fuzzy technology under Java.

mbFuzzIT (Marcus Beyer´s Fuzzy Inference Tool) is a tool for producing and testing complex controller structures. As the first of its kind, it implements the modularization of a Fuzzy controller into fuzzifier, rule based inference machine, and defuzzifier, as been discussed in the A1-project of the Special research area 531 at the University of Dortmund. So this aproach breaks with the limitation on monolithic controller blocks. This results into a higher flexibility, so a larger class of controllers can be produced and examined. In particular, the inference machines process now "genuine" fuzzy sets, like postulated by the theory (see [4]). In order to keep the implementation expenditure small, at present mbFuzzIT concentrates on inference machines of the type Mamdani (max-min-inference).

As a special feature, mbFuzzIT implements the level reduction according to  [2] and  [3].

In order to facilitate the creation of the inference mechanisms, mbFuzzIT containts a visual editor. With it you can assemble an inference mechanism like drawing a connection diagramm. Further visual editors facilitate the manipulation of the individual inference building blocks.

Enough words. I wish you much fun with my tool :)

Marcus Beyer


[1] Marcus Beyer, Stephan Lehmke: Ein Software-Werkzeug zur Konstruktion und zum empirischen Test komplex strukturierter Fuzzy-Inferenzmechanismen. VDI Berichte 1526, VDI Verlag GmbH, Düsseldorf 2000, ISBN 3-18-091526-9

[2] Marcus Beyer: Stufenreduktion von Mamdani-Inferenzmechanismen. Tectum Verlag, Marburg 2000, ISBN 3-8288-0947-2

[3] Stephan Lehmke, Karl-Heinz Temme, Helmut Thiele: Reducing the Number of Inference Steps for Multiple-Stage FUZZY IF-THEN Rule Bases. Reihe Computational Intelligence, Sonderforschungsbereich 531, Universität Dortmund, 1998.

[4] Stephan Lehmke, Bernd Reusch, Karl-Heinz Temme, Helmut Thiele: On interpreting fuzzy if-then rule bases by concepts of functional analysis. Techn. Rep. CI-19/98, University of Dortmund, Colaborative Research Center 531, Feb. 1998.