Investigating the impact of correlation on system multimode reliability-based analysis of highway geometric design

Abstract

Reliability analysis has been used to account for uncertainties and evaluate the risk of highway geometric-designs. Despite the existence of correlations between the input design-variables, the majority of the studies applying reliability-analysis have ignored their correlations. The objective of this paper is to quantify the influence of input design-variable correlations on reliability-based highway geometric-design. Three modes of failure are considered: insufficient-sight-distance, vehicle-skidding, and vehicle-rollover, for passenger cars and heavy trucks. A series-system reliability problem of the failure modes is used to account for the joint occurrence of the failure mechanisms. Results show that ignoring the correlations between input-variables can lead to inaccurate estimation of the noncompliance probability for both the individual modes and the series-system reliability. The effect is more pronounced for the vehicle-skidding failure mode than the other modes. The input-variables' correlation significantly changes the multivariate distributions of the performance functions, leading to more extreme events in the failure domain.