The two major advantages of bispectral analysis are: resistance to noise and the ability to detect nonlinearities, like quadratic phase coupling. The first aim was to study some of the theoretical aspects of bispectral estimation. A lot of attention was paid to the influence of noise, the number of segments, the influence of one or several harmonic deterministic components and aliasing. These aspects are typical of rotating machinery. An example of successful fault identification in DC electric motors is presented. The identification proved to be capable to identify quadratically coupled mechanical system when the power-spectra analysis failed. Further it proved to be quite resistant to noise.