Population growth, economic development, and rapid urbanization in many areas have led to increased exposure and vulnerability of structural and infrastructure systems to hazards. Thus, developing risk-based assessment and management tools is crucial for stakeholders and the general public to make informed decisions on prehazard planning and posthazard recovery. To this end, structural risk and resilience assessment has been an ongoing research topic in the past 20 years. Recently, machine learning (ML) techniques have been shown as promising tools for advancing the risk and resilience assessment of structure and infrastructure systems. To date, however, there is a lack of a holistic review on ML progress across various branches of structural engineering; an in-depth analysis of literature that can provide a timely evaluation of risk and resilience assessment methods of the built environment, where different types of structural and infrastructure facilities are interconnected. For this reason, this study conducted a comprehensive review on ML for risk and resilience assessment in four main branches of structural engineering (buildings, bridges, pipelines, and electric power systems). To cover the crucial modules in the prevailing risk and resilience assessment frameworks, existing literature is thoroughly examined and characterized in terms of six attributes of ML, including method, task type, data source, analysis scale, event type, and topic area. Moreover, limitations and challenges are identified, and future research needs are highlighted to move forward the frontiers of ML for structural risk and resilience assessment.

In performance-based earthquake engineering framework, seismic fragility functions play a fundamental role for quantifying the seismic resilience. There are three main approaches for performing the seismic fragility analysis: cloud analysis, incremental dynamic analysis (IDA), and multiple strip analysis (MSA). These approaches require different levels of computational efforts and yield different levels of accuracy on fragility curve estimates, leading to different predictions of seismic resilience. This paper presents an efficient method for converting fragility curves from the cloud to IDA and MSA and from IDA to MSA toward an efficient and high-fidelity resilience assessment. The proposed method requires two fragility points to obtain the fragility median and dispersion parameters that yield converted fragility curves agreeing with the target fragility curves. This method and associated resilience assessment are demonstrated through two case studies, one a typical two-span highway bridge in firm ground under longitudinal seismic excitations and the other an extended pile-shaft-supported bridge in liquefaction-induced laterally spreading ground under transverse seismic excitations. The results show that the converted fragility curves coincide very well with the target fragility curves. An efficient and high-fidelity resilience assessment can be achieved leveraging the proposed method. For easy implementation, codes of the proposed method are available online.