Identyfikacja położenia otworu kołowego w tarczy przy użyciu sieci neuronowych

The numerical tests have been carried out for rectangular plates with internal defect in a form of a circular hole. The wide variety of hole's positions and diameters were checked. Two identification procedures are presented. Neural network analysis of the calculated eigenfrequencies is the first way to obtain the hole's center coordinates and hole's diameter. The second one is based on analysis of a structural response to a harmonic excitation. The multi-layer feed forward backpropagation neural networks with one hidden layer have been applied. The two approaches of network's architectures have been applied; standard and cascade.