Seismic performance and fragility of two-story and three-span underground structures using a random forest model and a new damage description method

• Published in: Tunnelling and underground space technology Vol. 135; p. 104980
• Main Authors: Yang, Jing, Zhuang, Haiyang, Zhang, Guangyu, Tang, Baizan, Xu, Changjie
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.05.2023
• Subjects: Random forest prediction model; Seismic damage; Seismic fragility; Seismic performance; Underground structure; Seismic damage; Random forest prediction model; Seismic fragility; Seismic performance; Underground structure
• DOI: 10.1016/j.tust.2022.104980
• ISSN: 0886-7798

•New concept was defined to evaluate the damages of the underground structure (US).•New model was firstly used to predict the seismic performance levels of US.•A quantification system for the seismic damage of shallow buried US was advanced.•IDR was suitable to describe the seismic damage of shallow-buried US.•PGV was found more suitable for the seismic intensity measure of US in soft soil. This paper proposes a new method for evaluating the seismic performance and fragility of large underground structures. A typical underground subway station consisting of a two-story and three-span structure was used as the research subject. Considering the variation of sites and the uncertainty of input ground motions, we developed a finite element model of a nonlinear dynamic soil-underground structure interaction system. Based on the incremental dynamic analysis method, the nonlinear seismic responses of the subway station structure were calculated for 900 cases. Python (software) was used to accelerate the processing of the large number of numerical simulation data. Based on the seismic damage characteristics of an underground structure, this paper proposes a new concept of average seismic damage value to evaluate the damage condition of the underground structure to overcome the inadequacies induced by an expert’s intuitive judgment. Meanwhile, the random forest prediction model was adopted to predict the seismic performance levels (SPLs) of the underground structure, and a non-parametric test was performed to verify the correlation between the degree of seismic damage and the inter-story drift ratios (IDR) of the large underground structure. This proved that the IDR could describe the seismic damage of shallow-buried underground structures; we advanced a precise quantification system of the SPL for shallow-buried subway station structures with two-story and three-span using probability statistics. The peak ground velocity (PGV) is more suitable than the peak ground acceleration (PGA) for use as the seismic intensity measure (IM) for large underground structures when the stiffness of the soil foundation becomes soft. In addition, the seismic fragility curves of the shallow-buried subway station structure were constructed and analyzed using the developed quantification system for the SPL.