Multi-Layer Perceptron para Detecção de Faltas em Aerogeradores Baseados na Máquina de Indução com Rotor em Gaiola

In this study a application proposal to develop a classifier using an artificial neural network (ANN) for the problem of detecting internal faults in a wind turbine generator based on an induction machine with a caged rotor, a problem for which there is still no optimal solution. The classifier evaluates a given instance and returns whether the instance comes from a healthy machine or a machine that has a stator coil fault. Synthetic data obtained by simulation was used for training the ANN. The results obtained demonstrate that a simple network with a single intermediate layer already achieves good performance with low computational cost as the memory to store the ANN.