Reliability and Degradation Modeling with Random or Uncertain Failure Threshold

Abstract

This paper developed extensions to the existing research so that reliability assessment based on degradation modeling can address new problem domains that previously did not meet the required assumptions and modeling constraints. Degradation modeling is based on probabilistic modeling of a failure mechanism degradation path and comparison of a projected distribution to a pre-defined failure threshold. Previous approaches to this problem required that the predefined failure threshold must be considered as a fixed deterministic value, which can be problematic for several reasons. Often, the designer and producer of a part or a system have many diverse users of their products. In practice, the critical threshold value can vary appreciably among users. In this case, a probabilistic, rather than a deterministic threshold value is more appropriate. For other applications, the designer may not know with certainty what explicit level of degradation will cause a failure. In this case, specification of a range of possible threshold values is more appropriate. This also can be accommodated by considering the threshold value as a random variable with some assumed distribution to reflect the variability. New modeling approaches are presented in this paper such that this limiting assumption is no longer required. This should allow systems with more varied usage conditions and failure mechanisms to be analyzed using degradation-based reliability assessment methods.