Vibration-based gearbox fault diagnosis by DWPT and PCA approaches and an adaptive neuro-fuzzy inference system

The majority of world energy is produced, consumed or transformed by rotating machines, like turbo-alternators, wind turbines, pumps, compressors... etc. Consequently, the reliability of the rotating machines, which can be the subject of breakdowns or dysfunctions in their times of use, is vital for a correct operation of the various industrial applications. In addition, for mostly configurations in all types of rotating machines, gearbox is an essential part to transfer rotating power source to other devices and provide speed and torque conversions. The goal of this paper is the proposition of a diagnosis procedure for real time gearbox fault detection and diagnosis while basing itself on the vibration signal. The analyzed faults can appear in the gear and bearing with various combinations under different speeds and loads. The proposed fault diagnosis technique is based on the application of the Discrete Wavelet Packet Transform DWPT and Principal Component Analysis PCA to extract the features of the different sub-bands frequencies in the vibration signal. The Adaptive Neural Fuzzy Inference System ANFIS is used to fault classification. Experimental results can verified that the proposed procedure can classify with precision various types of faults according to the fault location and type.