Machine learning based dynamic cause maps for condition monitoring and life estimation

Estimating failure modes and life of electronic component in field environment is challenging because of dynamics in mission profile, component interactions, and errors and variability in field data. This paper presents a methodology to address these challenges by developing a dynamic cause map using statistical models, field data and machine learning algorithms. The cause map estimates probability of failure due to a failure mode at a given time. The Bayesian technique is used to update the cause map algorithmically based on changing usage profile and operating environment.