ARTIFICIAL NEURAL-NETWORK-BASED FAULT DIAGNOSTICS OF ROTATING MACHINERY USING WAVELET TRANSFORMS AS A PREPROCESSOR

The purpose of condition monitoring and fault diagnostics are to detec t and distinguish faults occurring in machinery, in order to provide a significant improvement in plant economy, reduce operational and main tenance costs and improve the level of safety. The condition of a mode l drive-line, consisting of various interconnected rotating parts, inc luding an actual vehicle gearbox, two bearing housings, and an electri c motor, all connected via flexible couplings and loaded by a disc bra ke, was investigated. This model drive-line was run in its normal cond ition, and then single and multiple faults were introduced intentional ly to the gearbox, and to the one of the bearing housings. These singl e and multiple faults studied on the drive-line were typical bearing a nd gear faults which may develop during normal and continuous operatio n of this kind of rotating machinery. This paper presents the investig ation carried out in order to study both bearing and gear faults intro duced first separately as a single fault and then together as multiple faults to the drive-line. The real time domain vibration signals obta ined from the drive-line were preprocessed by wavelet transforms for t he neural network to perform fault detection and identify the exact ki nds of fault occurring in the model drive-line. It is shown that by us ing multilayer artificial neural networks on the sets of preprocessed data by wavelet transforms, single and multiple faults were successful ly detected and classified into distinct groups. (C) 1997 Academic Pre ss Limited.