Abstract
The Fuzzy-Pattern-Classification is a useful approach for modelling of complex nonlinear systems. A multi-dimensional classificator is trained in a learning phase. New unknown objects may be classified using the trained modell in the working phase. A algorithm for fuzzy-logical combinations may be added to be classification. This improves the classification results of many applications. Modules for evaluation, statistics and fuzzy-control applications are provided. A modular hardware-based accelerator was developed for time-critical on-line applications. The article describes the use of the accelerator for surface checking of metal remodelling processes.