A double neural network for interpretation of the frequency response in the electrical equipments

A novel identification technique for the characterization of an electrical apparatus is presented. The approach is based on two multi-valued neuron neural networks operating in a joined architecture able to extract geometrical displacements or insulation changes, not directly accessible or measurable. The inputs of the neural system are the samples of the frequency response on a given significant band, while the outputs represent the estimation of one or more geometrical parameters or other features related to the material properties. The method uses a Frequency Response Analysis (FRA) approach in order to preprocess the data to present to the net.