Time-Variant Series-System Reliability of Non-Monotonic Systems Using an Extreme Safe Set  
Author Gordon J. Savage


Co-Author(s) Young Kap Son


Abstract Errors in time-variant reliability analyses arise when the safe design space is incorrectly enlarged by including previously failed events. The situation typically arises when one turns a time-variant system into a step-wise time-invariant system and uses only two contiguous events (one new failure and one past safe) at each timestep. In this paper a way to form an accurate and efficient limit-state representation of the true time-variant safe region is presented. The key to the success herein is the identification of four modes of limit-state motion: these modes may be clearly seen in a parametric polar plot over life-time comprising the magnitude of the most-likely failure point (MLFP) vector and its relative angle. The parametric plot provides the key times and the logic for the selection of a small number of limit-state surfaces to comprise an extreme safe region at any time. The results not only agree favorably to the gold-standard set by the marching-out Monte-Carlo simulation (MCS) method, but show either equivalent or significant improvements over the classical two-time event methods. A case study of series-system reliability of a supported corroding beam containing stochastic excitations and degradation conditions shows the efficacy of the method.


Keywords Time-variant reliability, Parametric polar plot, Extreme safe set, Series systems
    Article #:  RQD26-16

Proceedings of 26th ISSAT International Conference on Reliability & Quality in Design
Virtual Event

August 5-7, 2021