PREDICTION OF RESPONSE AND DAMAGE IN REINFORCED CONCRETE COMPONENTS AND STRUCTURES THROUGH NUMERICAL SIMULATIONS


Tezin Türü: Doktora

Tezin Yürütüldüğü Kurum: Yıldız Teknik Üniversitesi, İnşaat Fakültesi, İnşaat Mühendisliği Bölümü, Türkiye

Tezin Onay Tarihi: 2023

Tezin Dili: İngilizce

Öğrenci: Mehmet Ozan YILMAZ

Danışman: Serkan Bekiroğlu

Özet:

The modern approach to design of earthquake resistant reinforced concrete structures is based on predictable ductile failure of structures. The most basic principle in the operation of the ductile damage mechanism is to maintain the redistribution of plastic strains in beams under earthquake effects up to ductility levels that can meet high displacement demands. The design of columns to be stronger than beams in terms of bearing capacity has been introduced in order to ensure that plastic deformations occur in beams and has been accepted as the strongest guarantee of the aforementioned damage mechanism in many cases. In addition, it has been accepted in the analysis and design that the joints of reinforced concrete beams-columns, where large shear forces are transmitted, do not produce nonlinear deformations.

It is clear that the occurrence of nonlinear deformations in the joint regions is an obstacle to the ductile response of the frame. In recent years, some experimental studies have shown that nonlinear reactions may develop in frame members where these conditions are met, as well as in joints where the conditions set forth by modern codes are not met. This situation reveals the necessity of taking into account the nonlinear deflections occurring in the joints in the calculations performed both in the evaluation of existing structures and in the design of new structural systems.

In the literature, the models proposed for the prediction of internal forces and deformations under cyclic effects in joint regions can be categorised into two different main groups. The first one is the models in which the results of a limited numberof experimental studies, each with different experimental conditions, are used in calibration and the so-called super-elements, which are formed by one or more uniaxial springs are used. The second class of models is the plane and space models in which concrete and reinforcement are represented using more advanced finite element and constitutive relations. Both classes of models are examined and the reasons that are considered to be effective in their not finding a place in practical evaluation and design applications in structural engineering are revealed.

It has been evaluated that the ability of the super-elements presented in the literature to simulate with acceptable approximation the nonlinear strain responses of a given reinforced concrete column-beam joint region under cyclic effects is directly related to the a priori estimation of the unidirectional incremental rotation-shear force relationship of the relevant reinforced concrete column-beam region. For this reason, two different models have been proposed that accept as input the basic physical variables describing any joint region sample and perform shear strain-stress estimation. The first one is an advanced artificial neural network model that utilises the results of experimental studies reported in the literature. A second prediction model, which serves the same purpose, is constructed by nonlinear regression between shear strain-stress relationships obtained from numerical simulation models using advanced finite element techniques and physical variables related to the joint region. A prediction model using non-dominated sorting genetic algorithm is presented for the prediction of the reduction in strength and stiffness of reinforced concrete column-beam joints under cyclic effects.

Keywords: reinforced concrete joints, nonlinear finite element analysis, artificial neural networks, genetic algorithm