Test frequency selection using efficient space coordinate transformation

The paper deals with a new efficient method for test frequency selection for analog fault diagnosis based on a transformed space of optimization coordinates. The original symmetrical linear space is transformed so that the mirrored optima are removed and the transformation takes into account the logarithm scale of the frequency axis. The optimum set of frequencies is found faster and the optimization algorithm has lower computational demands in comparison with the original optimization problem solved for example using Genetic Algorithms. Also, in the case of an overdetermined system of fault equations the method an be used for selection of frequencies with defined minimum and maximum spacing. The proposed procedure is demonstrated on an example of a frequency filter.