A Framework of Multi-Index Modeling for Similarity-Based Remaining Useful Life Estimation

As an emerging technique, the similarity-based residual life prediction (SbRLP) method is a significant method for residual useful life (RUL) prediction. However, related researches on the SbRLP method with multiple degradation variables are rare. Hence, a framework of the SbRLP method with multiple degradation indicators is advanced. Within the framework, two solutions (e.g. solution A and B) are presented to estimate RUL of the operating system. Then, the case study of RUL prediction of dynamically tuned gyroscope (DTG) proves their effectiveness and superiority. Meanwhile implementation results demonstrate that by carefully setting the weight adjustment coefficient of the recommended framework, more accurate RUL prediction can be achieved statistically. In addition, findings in case study recommend that the selection of the suggested framework using solution A or B should be referenced to specific application requirements like cost, fault tolerance and prediction accuracy.