A hybrid approach for modelling multidimensional degradation processes

Abstract

Degradation processes are often multidimensional. Modeling such degradation processes needs to address two key issues: indicator fusion and degradation model selection; and they have been separately addressed in the literature. This paper proposes a hybrid approach to jointly address these two issues. The proposed approach first fuses multiple degradation indicators into a composite degradation indicator. This fusion step involves data normalization, aggregation model selection and determination of indicator weights. After the fusion step, the problem becomes one-dimensional, and the existing method to select the degradation model for a one-dimensional degradation process can be applied. The resulting model obtained from the proposed approach can be a two-phase model; and the model for the second phase has a closed-form expression. This considerably facilitates residual life prediction. A real-world example is included to illustrate the proposed approach and its appropriateness.