Non-probabilistic system reliability analysis of structures with interval uncertainty considering correlated failure modes

Large and complex engineering structures consist of multiple components with various failure paths. Thus, it is more reasonable to assess reliability by treating the structure as a structural system with multiple components rather than as a whole. Aiming at the problem that traditional methods are difficult to assess reliability for structural systems based on the correlation and logical relationship between failure mode with limited samples, this paper proposes a novel method based on the non-probabilistic interval model of principal component. This method utilizes the surrogate model and principal component analysis to obtain independent principal components, followed by optimization to construct the non-probabilistic interval model of principal components. The reliability of the structural system can then be assessed using the non-probabilistic volumetric method. Two numerical examples show that results obtained using the proposed method at 95 % and 98 % confidence levels have errors of 1.4286 %, 1.1097 %, and 1.4213 %, 0.8772 %, respectively, compared to those from Monte Carlo Simulation. This demonstrates that the proposed method efficiently addresses the challenge of reliability assessment for structural systems with limited samples, providing more realistic and accurate assessment for large and complex structures.