Effect of Intrinsic and Epistemic Uncertainties on the Multiobjective Optimal Design of a Hyperstatic Structure

Abstract

Although variability in material properties and loading conditions is commonly accounted for in safety and reliability studies, this is not the case for model assumptions. In addition, many other uncertainties affect the design and evaluation stage of civil engineering structures. This paper describes how a consistent modeling methodology can be developed for a multicriteria evaluation and multiobjective optimal design of constructions. A series of tools and methods [probabilistic description of variables, desirability functions, multiobjective particle swarm optimization (MOPSO), analytic hierarchy process (AHP), and decision aid method, among others] has been applied in order to be able to reach an optimal design alternative. The methodology is illustrated on the real case study of an industrial structure. The focus is given to several key points showing how uncertainties and lack of knowledge influence the optimization process. It is shown that items such as the epistemic uncertainty on correlation between material properties, consideration of soil–structure interaction, and designer preferences can affect the optimal solution. A robust optimization scheme is proposed that considers in a consistent way uncertainties of various origins.