Establishing Efficacy of Machine Learning Techniques for Vulnerability Information of Tubular Buildings

Muhammad Zain1

Suraparb Keawsawasvong2

Chanachai Thongchom2

Issara Sereewatthanawut3,Email

Muhammad Usman4

Lapyote Prasittisopin1,Email

1Architectural Technology Research Unit, Department of Architecture, Faculty of Architecture, Chulalongkorn University, Bangkok, 10330, Thailand.
2Department of Civil Engineering, Thammasat School of Engineering, Faculty of Engineering, Thammasat University, Pathumthani, 12120, Thailand.
3King Prajadhipok's Institute, Bangkok 10210, Thailand.
4School of Civil and Environmental Engineering, National University of Sciences and Technology (NUST), Islamabad, 44000, Pakistan.



During recent times, the emergence of artificial intelligence in structural engineering has rendered researchers to work on reducing the overall computational effort required for producing vulnerability information of infrastructural facilities. However, the supertall and tubular building analysis lacks substantial research due to their intricate structural behavior and aleatory uncertainties. This paper establishes the feasibility of using versatile Machine Learning (ML) algorithms for producing fragility relationships of high-rise tubular structures by considering a 55-story tall building, located in high seismicity area. Initially, the vibrational modes are decoupled, and Incremental Dynamic Analysis (IDA) has been conducted on each individual mode discreetly. The initial four modes were considered for analysis, constituting overall modal mass participation of more than 90%. Inference has been drawn between the efficacy of employed ML techniques to establish grounds for rapid structural vulnerability assessment of high-rise buildings considering ground motion features along with the structural characteristics. Testing datasets have suggested the overall adequacy of ML algorithms by substantiating the successful prediction of Engineering Demand Parameter (EDP), and their applicability for establishing vulnerability information of high-rise structures. 

Establishing Efficacy of Machine Learning Techniques for Vulnerability Information of Tubular Buildings