Conventional maintenance strategies consist of corrective and preventive maintenance. In corrective maintenance, the system is maintained on "as needed" basis usually after major breakdown. In preventive maintenance, components are replaced based on a conservative schedule to "prevent" commonly occurring failures. Consequently, these conventional maintenance strategies are not adequate to fulfil the needs of expensive and high availability of industrial machineries. Condition based predictive maintenance is an alternative that uses embedded diagnostics and prognostics to determine bearing's health. Bearing is one of the important machine elements of any rotating machinery. However, predicted result and field performance differs widely as each of the operating parameters has significant contribution to the failure of bearings. in the present work, it is proposed to have two stages model- monitoring incipient failure and prediction of time elapsed from incipient failure to final failure. To detect online incipient failure, it is proposed to have wavelet transformation which is known to have susceptible to mechanical defect impact and trained using neural network for fault detection. The generalised regression neural network is using for predicting remaining life elapsed between after detecting the incipient failure using online incipient failure detection module an final failure. The developed model was validated from experiments results run under P/C ≤ 0.2 obtained from 10 station rigs and predicted model residual life is observed to comform experiments within errors range of 20%.
展开▼
