Dynamic weight-based accelerated test modeling for fault degradation and lifetime analysis

Abstract

Under normal operational stress, accelerated degradation testing is employed to assess fault diagnosis, prognosis, lifetime, and maintenance decisions for highly reliable products. The effectiveness of accelerated degradation testing relies on the suitability of the model describing the product’s failure mechanism. While the traditional approach typically involves developing stochastic models to analyze degradation caused by a single failure mechanism, real-life scenarios often encompass multiple failure mechanisms that impact the degradation process. Unfortunately, using a single stochastic process fails to adequately capture these diverse multi-failure modes. This paper introduces an innovative mixed stochastic process model designed to overcome this limitation, focusing on its application to accelerated degradation testing. The mixed model combines three commonly used stochastic models, incorporating dynamic weights. The application of the Metropolis–Hastings algorithm aids in estimating the unknown parameters, while the comparison between the mixed model and single stochastic models in accelerated degradation testing relies on utilizing stress relaxation data to assess their performance. Results demonstrate that the proposed mixed model surpasses the conventional stochastic models, exhibiting superior accuracy.