Reliability-based design optimization: a state-of-the-art review of its methodologies, applications, and challenges

Abstract

Reliability-based design optimization (RBDO) integrates various uncertainties into the design optimization process, offering a more realistic and robust approach compared to traditional deterministic design optimization methods. Thus, RBDO has emerged as a highly compelling and vital research direction within the design field. However, there is currently a dearth of comprehensive reviews on RBDO methodologies presented in a clear and concise manner. This paper aims to address this gap by providing a state-of-the-art review of RBDO methodologies across four key aspects: performance function evaluation, reliability analysis, optimization strategies and algorithms, and RBDO applications in five typical engineering fields. The paper commences by presenting basic RBDO formulations and providing an overall picture of various RBDO methodologies. Subsequently, performance function evaluation methodologies are explained and then categorized into three groups: physics-based performance function evaluation, data-driven performance function evaluation, and physics-informed performance function evaluation. Following this, two types of reliability analysis methodologies are introduced: time-independent reliability analysis and time-dependent reliability analysis. The review also delves into the realm of optimization strategies, with a comprehensive examination of three types: double-loop strategy, single-loop strategy, and decoupling strategy. Moreover, two types of optimization algorithms, the gradient-based algorithm and the meta-heuristic algorithm, are extensively surveyed. Each is scrutinized in terms of their specific methods, advantages, and disadvantages. In addition to methodological exploration, the paper scrutinizes RBDO applications in five engineering fields: wind engineering, aeronautical engineering, ocean engineering, bridge engineering, and vehicle engineering. These applications are carefully surveyed regarding the various uncertainties encountered, the methods employed, and the results of specific RBDO problems. The paper concludes by summarizing key challenges and charting the future work of RBDO research. It offers valuable insights that draw from the analysis of 174 surveyed papers, enabling readers to gain a comprehensive understanding of RBDO theories and facilitating the proper selection and development of appropriate methods for different RBDO stages and problems in diverse engineering contexts.