Zastosowanie sieci neuronowej do badania przebiegów czasowych rozruchu maszyny indukcyjnej

The paper presents application of artificial neural network in induction machine speed determination making use of current and voltage waveforms. Also speed values in past sequences are given to the network input, to secure recurrental capability of the standard static network. Training and testing was made on real 3kW-induction machine. In the training set we have data from a few starting cycles. Final results are quite good and could be a base for further investigations. The advantage of this model is that the speed calculation based on the measurement of the real values of current and voltage. In practical, speed measurement is more difficult then current and voltage measurement, so someone may consider using this model instead of speed measurement.