Abstract
Time-variant reliability sensitivity (TRS) analysis can measure the effect of input factors on the structure/mechanism failure. The traditional method for TRS analysis employs a double-loop procedure, with computational cost depend on the number of input factors. To address the above weaknesses, a single-loop method (SLM) is developed for TRS analysis. Based on Bayes theorem, the sensitivity measure is derived and expressed by the difference between the probability density function (PDF) and the failure-conditional PDF. This derivation allows for TRS analysis to be performed with just one set of samples, where the computational complexity doesn't depend on the number of inputs. Furthermore, the sensitivity index is normalized to effectively identify influential variables. Three examples involving numerical and engineering problems are employed to validate the proposed SLM, with Monte Carlo simulation introduced for comparison. The results reveal that the proposed SLM provides satisfactory sensitivity analysis for motion mechanisms while significantly saving computational resources.
