Evaluation of Features with Changing Effectiveness for Prognostics

Abstract

Feature evaluation is crucial to identify the best features and to achieve high accuracy in diagnostics and prognostics. Feature evaluation for prognostics is a developing research area with several publications in recent years. Most, if not all, of existing methods to evaluate features for prognostics base on the feature changes in the whole life of the system under observation. In other words, feature values collected throughout the failure degradation are analyzed to create a goodness value for the feature. In reality, the goodness of the features may change during the failure progression. A feature may be good representative of failure progression in the initial phase but not in the final phases, or vice versa. This paper presents dynamic nature of representation capabilities of features throughout the failure degradation and proposes a novel approach to evaluate the features considering their dynamic nature. Proposed approach involves feature segmentation based on their representation capabilities and feature fusion utilizing the segmented evaluations. The presented approach has been applied in simulated and real degradation datasets. Real degradation dataset were obtained from accelerated aging tests of Li-ion batteries in the lab environment. The results from both datasets show that dynamic feature evaluation improves SoH estimation accuracy.