Reliability-based design optimization of systems in Voronoi tessellation-based partitioned design space

作     者：Khalid, Mohd Aman Bansal, Sahil 

作者机构：Indian Inst Technol Delhi Dept Civil Engn New Delhi 110016 India 

出 版 物：《STRUCTURES》 (Structures)

年 卷 期：2025年第71卷

核心收录：

主　　题：Reliability-based design optimization RBDO optimization under uncertainty probabilistic constraints stochastic simulation decoupling Voronoi Tessellation 

摘      要：Reliability-Based Design Optimization (RBDO) is a well-known design methodology involving reliability constraints to achieve an optimal design with a failure probability lower than the predefined threshold. Evaluating reliability constraints, or equivalently the Failure Probability Functions (FPFs), has proven challenging due to its higher computational cost within the optimization loop, lower accuracy, and limited applicability resulting from their approximations. To address these issues, this paper introduces a novel stochastic simulation-based decoupling approach for performing RBDO. The key novelty of the proposed approach is the transformation of FPFs approximating or evaluation into a density estimation problem using Voronoi tessellation while decoupling reliability constraints evaluation from the optimization loop. The proposed approach initiates with the establishment of an augmented reliability formulation. Subsequently, akin to Stochastic Subset Optimization (SSO), the original design space undergoes iterative reduction to obtain a feasible design space that satisfies reliability constraints. At each iteration, Voronoi tessellation is implemented to estimate the probability density value of design variables conditioned on the failure events. Ultimately, a deterministic optimization problem is solved within the feasible design space to minimize the cost function. The efficacy and computational efficiency of the proposed approach are demonstrated through its successful application to various RBDO benchmark examples.