Model-driven degradation modeling approaches: Investigation and review

The second law of thermodynamics implies that any animate and inanimate systems degrade and inevitably stops functioning. It is irreversible over time that can be labeled as “the degradation arrow of time”. From perspective of products’ reliability design, it is essential to build appropriate models of describing the degradation arrow of time. The current modeling approaches mainly include the model-driven (having assumed forms based on cognitive experience of mankind) and data-driven (using data learning techniques without form hypothesis) approaches. In this paper, we just investigate and review the model-driven degradation approaches, hoping to provide suggestions of the model construction or selection for scholars or engineers. First, for the single mechanism, degradation law models and stochastic process models are classified as separately depicting the tendency and fluctuation of degradation. For the degradation law model, we propose the concept of meta-models as original types for various personal models. For the stochastic process model, two main types including the non-monotonic and monotonical types are presented. Then, four multi-mechanism degradation types are discussed, that are competitive degradation, multi-stage degradation, coexistence of degradation and impact, and coexistence of degradation and failure. Besides, for the multi-performance degradation, independent and coupling models are introduced. The forms, connotations, applicability and insufficiency of these models are described with a series of examples from the literature and our own experiences. The final explicit suggestions about the potential future work are provided for the development of new degradation models.