Uncertainty Assessment of Prognostics for Electronics under Vibration Loading
CALCE Team:
Jie Gu, Donald Barker and Michael Pecht
Objectives:
Develop the general uncertainty analysis approach for failure prognostics of electronics.
Demonstrate the approach by assessing the impact of identified uncertainties on remaining life predictions for electronics subject to vibration loading.
Introduction:
Prognostics is a process of predicting the future reliability of the system by assessing the extent of deviation or degradation of a product from its expected normal operating conditions [1]. Prognostics has been developed and implemented in electronics in recent years. Ramakrishnan et al. [2] and Mishra et al. [3] used a physics-of-failure (PoF) approach to perform prognostics on electronics subject to various loading conditions. Gu [4] assessed prognostics subject to random vibration, but, the effect of uncertainty and of variability in the material properties and prediction procedures were not considered at that time. As a consequence of uncertainty, prognostics methods must consider the interrelationships between accuracy, precision, and confidence [5]. In this study, accuracy is a measure of how close a point estimate of failure time is to the actual failure time. Precision is a measure of the narrowness of an interval in which the remaining life falls. Confidence is the probability of the actual remaining life falling between the defined confidence intervals.
For logistics use of prognostics, it is necessary to identify the uncertainties in the prognostic approach and assess the impact of these uncertainties on the remaining life distribution in order to make risk-informed decisions. Therefore, this study addresses the uncertainty analysis of prognostics, with a case study of an electronic circuit board assembly subjected to random vibration. An approach will be developed to assess the impact of uncertainties in measurement, parameter inputs, failure criteria, and future usage on remaining life prognostics. This approach utilizes sensitivity analysis to identify the dominant input variables that influence the model output, and uses the distribution of input parameter variables in a Monte Carlo simulation to provide a distribution of accumulated damage. From accumulated damage distributions, the remaining life is then predicted with confidence intervals.