Fault pattern recognition based on bispectrum entropy model

A fault pattern recognition method was developed on the basis of information entropy and bispectrum theory. The bispectrum features of vibration signal were analyzed. And a bispectrum entropy algorithm based on energy distribution was derived under the condition of subspace distribution probability. Then, the vibration signals of a gearbox under four conditions were extracted experimentally. And a BP neural network for the fault pattern recognition was established by using the bispectrum entropy feature as input. Finally, this method was verified by successfully recognizing four fault patterns of the gearbox.