Development and assessment of response and strength models for bolted steel connections using refined nonlinear 3D finite element analysis

3D detail models;Bolted steel connection;Finite element;Failure criteria;Neural network;Micromechanics;Prying;Nonlinear;Strength models;Connection response

Citipitioglu, Ahmet Muhtar ; Civil and Environmental Engineering

University:Georgia Institute of Technology

Department:Civil and Environmental Engineering

关键词: 3D detail models;    Bolted steel connection;    Finite element;    Failure criteria;    Neural network;    Micromechanics;    Prying;    Nonlinear;    Strength models;    Connection response;   

Others  :  https://smartech.gatech.edu/bitstream/1853/31691/1/citipitioglu_ahmet_m_200912_phd.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

The difficulty in developing bolted connection designs lies in the limitations in existing methods to characterize their strength and typically nonlinear response due to the complex interaction of the bolts and structural components.Yet it is necessary for the engineer to be able to determine the three main connection response characteristics: stiffness, strength, and ductility to account for their influence on the overall structural response behavior.The need for better connection response characterization becomes even more crucial in a performance based design approach or when designing partially-restrained moment frames.Several welded moment resisting frame connections were found to have serious failures following the 1994 Northridge earthquake leading to more interest and research on bolted connections as an alternative.In this study a refined three dimensional nonlinear finite element modeling approach to accurately simulate the response of bolted connections is presented.Sensitivity studies of modeling parameters are also performed.A nonlinear response dataset of over 400 connection cases is generated using this approach with a parametric bolted angle connection model.The use of a parametric Richard-Abbott type function and a neural network, calibrated using the response dataset, as practical tool to model the nonlinear stiffness response of bolted connections under monotonic loading is demonstrated and assessed.Failure criteria that can be practically used in conjunction with the refined three dimensional finite element models without any additional modeling requirements are developed.The stress modified critical strain (SMCS) criterion based on the void growth and coalescence mechanism initiating ductile fracture in steel is used for determining failure in the connection member.The bolt failure criterion developed is a mechanics based model using the elliptical interaction of the tensile and shear capacity envelope.The failure criteria and bolted angle response dataset is combined to assess in detail the impact of geometry and topography of the bolted angle connections on the following response characteristics: strength, initial stiffness, plastic stiffness, and absolute ductility or the displacement capacity.Finally, using the dataset of bolted angle connection response, along with their capacities and failure modes determined using the developed failure criteria, the prying strength models in the AISC LRFD Specifications, Eurocode, and a hybrid model are assessed and found to be very conservative for some cases.Based on these results a modified Eurocode and hybrid prying strength model is proposed which greatly improves the prying strength prediction.These prying models are assessed and verified using experimental data found in literature.