Comparsion between Neural Networks and Fuzzy Classification for Acoustic Leak Monitoring

The capability of neuronal networks and fuzzy pattern classification is compared using measuring data that originate from experiments on leakage detection. The classification procedures are to localize simulated acoustically active leaks and to determine the leak rate. Different types of neural networks are needed to perform these two tasks. Alternatively fuzzy classifiers can be applied. Concerning the generalization capability, i.e. the recognition of positions and leak rates that could not be trained, neural networks clearly superior over fuzzy classification.