Real-time bispectral analysis of gear noise and its application to contactless diagnosis

Real-time bispectral analyses of gear noises are carried out, intending to distinguish abnormal states from the normal one without stopping the machine. Firstly, a real-time bispectral analyzer which consists of bandpass filters, multipliers, and integrators is constructed. Then, the noises of gears are analyzed. The results show that when scorings have been grown on the gear surfaces the moduli of bispectra of the nosies are reduced markedly comparing to those of normal state, while conventional spectral analysis fails to distinguish the difference. And it is shown that this fact can be used to diagnose the conditions of gear surfaces. Finally, a proper stochastic model of gear noises based on physical considerations as well as the spectral and bispectral characteristics obtained by our study, is proposed.