Multiple fault condition recognition of gearbox with sequential hypothesis test

Abstract A novel method for the fault condition recognition in which the recognition system interrogates a propagation channel adaptively and intelligently by using the available data is proposed based on sequential hypothesis testing. The waveform of the data in the propagation channel for the fault condition recognition is designed with the kurtosis of the measured data in time domain. The sequential hypothesis testing framework is proposed when hard decisions are made with adequate confidence. The distinguished characteristic of the channel recognition is that it operates in a closed loop and makes constant optimization in response to its changing understanding of the channel. The fault condition recognition of the gearbox is to update the multiple target hypothesis/class based on the measured data, customize waveform as the class probabilities changes, and make conclusion when the sufficient understanding of the propagation channel is achieved.