Machine performance assessment based on integrated signal redundancy and bootstrap technique

Prediction of machine performance based on current states and historical data has been a challenging issue in a predictive maintenance of machine performance assessment. Traditional methods mainly focused on developing prediction algorithms, rather than paying attention to the understanding of the data. This paper presents an innovative method to quantitatively evaluate the predictability of machinery performance assessment based on information redundancy and a statistical simulation technique. The predictability of a series of simulated signals including periodicity signal, simulated periodicity signal, chaos signal and random white noise signal were simulated for testing the correctness of the proposed method. In addition, practical vibration data were analyzed and a high-precision prediction was achieved by computing the redundancies of these sample sequences. Results indicate that evaluation tool can present a clear indication of machine performance predictability and therefore can guide the development and selection of prediction algorithms.