A Novel Similarity-based Method for Remaining Useful Life Prediction under Multiple Fault Modes

Traditional similarity-based methods generally ignore the diversity of equipment fault modes, the difference in degradation rates, and the inconsistency among monitoring data lengths. Thus, a similarity-based multi-scale method under multiple fault modes (MS-MFM method) is proposed to enhance the accuracy of remaining useful life (RUL) prediction and depict prediction uncertainty. Firstly, the fault mode classification model is trained through feature extraction and clustering analysis. Combining with the designed time-series weighted prediction strategy, the fault mode of test equipment is recognized. In this way the test equipment is matched with the training equipment with the same fault mode to improve matching accuracy and reduce matching complexity. On this basis, a multi-scale (MS) strategy is proposed to overcome the data utilization limitation caused by single-scale matching methods. This strategy matches the similarities between test equipment and training equipment at multiple time scales, and then multi-scale prediction results are integrated to fit accurate RUL probability distribution by employing kernel density estimation. Experimental results demonstrate the superiority of the MS-MFM method in dealing with the three differences in equipment degradation.