Unbalance Localization of a Multi-Disc Rotor by Hybridizing Wavelet Transformation and Neural Network

Abstract

Rotating machinery is widely working within the industry, and fault diagnosis and prognosis of them may be a vital issue that can save money continually. Unbalance is a crucial fault in rotary systems, and it is focused by many researchers to develop methods to detect that for correcting before global failure happening within the machine. Hence, the establishment of a procedure that will estimate the unbalance location and its specifics are going to be valued and practical for correcting operations. The recent study exemplifies a model that can detect the unbalance parameters, for example, the location of unbalance mass and value of that based on the hybridizing Wavelet Transformation and artificial neural network (ANN) model. The inputs of the model are wavelet coefficients derived from the bearing acceleration signal. It includes two hidden layers constructed by six neurons within each layer. The parameters estimation accuracy was attained 95%, 97%, and 96% for the disc number, the eccentric radius, and unbalance mass values, correspondingly.