Abstract
An on-line fault detection and isolation technique is proposed for the diagnosis of rotor vibration. The architecture of the systems mainly consists of feature generation and fault inference. A signal-based method is used for generating the features required by the subsequent neural fuzzy inference. In the signal-based approach, both lateral and axial vibration data are used for calculating signal features such as the average, the standard deviation, the maximum, and the harmonic multiples. A neural fuzzy network is exploited for intelligent inference of faults based on the extracted features. The proposed systems are implemented on the platform of a digital signal processor. Experiments carried out for a rotor kit, a centrifugal fan, and a centrifugal pump indicate the potential of the proposed techniques in predictive maintenance.