A Severity-Aware Reliability Index for Risk-Informed Structural Design

TL;DR

This paper introduces a Severity-Aware Reliability Index that incorporates failure severity into structural risk assessment, enhancing traditional probability-based measures with a focus on consequences, especially for rare but impactful events.

Contribution

A novel reliability framework based on Expected Failure Deficit (EFD) and a new severity-aware index, improving risk evaluation by accounting for failure consequences in structural design.

Findings

The new index aligns with classical measures in standard cases.

It detects hidden severity in complex, heavy-tailed distributions.

The framework supports risk classification with a five-level severity system.

Abstract

Classical measures of structural reliability, such as the probability of failure and the related reliability index, are still widely applied in practice. However, these measures are frequency-based only, and they do not give information about the severity of failure once it happens. This missing aspect can cause underestimation of risks, in particular when rare events produce very undesirable consequences. In this paper, a new reliability framework is proposed to address this issue. The framework is based on a new concept, called the Expected Failure Deficit (EFD), which is defined as the average deficiency of the system response when failure occurs. From this quantity, a new reliability index is introduced, called the Severity-Aware Reliability Index, which evaluates the consequence of failure in comparison with the Gaussian benchmark. The mathematical formulation is derived and it is shown that the inverse mapping exists in a restricted domain, which can be interpreted as an indicator of excessive tail risks. A Severity Classification System with five levels is also proposed and calibrated from analytical expressions. Numerical examples, including Gaussian, mildly nonlinear, and heavy-tailed cases, demonstrate that the proposed framework agrees with classical measures in standard situations, while being able to detect hidden severity in more complex cases. The method can therefore be used not only to quantify severity of failure, but also to classify risks in support of engineering design.