Failure time prediction for small sample devices based on degradation data and calculation

Failure time prediction for mechanical devices is an important part in prognostics and health management which provides critical information for system health estimation. Modern mechanical devices usually have high reliability and insufficiency of failure history, which leads to small failure sample and is difficult to apply classical method to failure time prediction. Therefore, to solve the failure time prediction problem on small sample mechanical devices, a simple degradation-based prediction approach to estimate the failure time from observed degradation data is proposed in this paper, which combines the exponential fitting method with Bayesian updating technology. In this approach, exponential fitting method is used to integrate real time degradation data, simulate the degradation paths of devices, and the failure time observations can be computed through inverse operation at each new monitored epoch. Inverse gamma distribution is chosen as the prior distribution of failure time, and the posterior distribution is estimated by Bayesian updating technology. The failure time distribution can be predicted with the estimated prior distribution and failure time observations. Finally, the effectiveness of the proposed approach is illustrated through a simulation case.