Efficient scenario analysis for optimal adaptation of bridge networks under deep uncertainties through knowledge transfer. (January 2023)

• Record Type: Journal Article
• Authors: Cheng, Minghui; Frangopol, Dan M.
• Is Part Of: Structural safety. Volume 100(2023)
• Publication Date: 2023-01
• Subjects: Scenario analysis -- Adaptation -- Bridge management -- Deep uncertainty -- Knowledge transfer -- Model-agnostic meta-learning; Structural stability -- Periodicals; Safety factor in engineering -- Periodicals; Reliability (Engineering) -- Periodicals; Constructions -- Stabilité -- Périodiques; Coefficient de sécurité en ingénierie -- Périodiques; Fiabilité -- Périodiques; 620.86
• DOI: 10.1016/j.strusafe.2022.102278
• Languages: English
• ISSNs: 0167-4730

Abstract

Highlights: Propose a novel scheme to accelerate the scenario analysis via knowledge transfer. Provide a definition of similar scenarios for adaptation of bridge networks under deep uncertainties. Utilize meta-learning-based surrogate modeling to realize knowledge transfer and perform optimization. Demonstrate the computational efficiency and flexibility of the novel scheme. Abstract: Due to deep uncertainties associated with climate change and socioeconomic growth, managing bridge networks faces the challenge to perform optimization for different scenarios. There exist a large number of scenarios when various sources of uncertainties, such as population growth and the increasing magnitude and frequency of natural hazards due to climate change, are compounded. Traditionally, scenarios are analyzed sequentially. However, when optimization for one single scenario is time-consuming, only a limited number of scenarios can be considered. To accelerate scenario analysis, this paper proposes a novel scheme through knowledge transfer between scenarios. Specifically, after finishing the optimization of a certain number of scenarios, the analyses of any new scenarios are accelerated by utilizing the knowledge obtained from optimization of previous scenarios. To implement the novel scheme, a proper definition of similar scenarios for adaptation of bridge networks under deep uncertainties is first given to stipulate the situation when knowledge transfer can occur. Then an approach based on surrogate modeling and meta-learning is used to realize the concept of knowledge transfer and perform the optimization. A bridge network is used as an illustrative example to demonstrate the computational efficiency of the proposed novel scheme.