| Beams and columns are lateral resisting components of frame structures,which are subjected to the combine action of cyclic axial forces,bending moments and shear forces under earthquake,significantly affecting the seismic performance of frame structures.Multi-axial concrete constitutive models and smeared crack theory are widely used to account the axial-shear-flexural interaction（ASFI）effect of reinforced concrete（RC）components.Although these methods present clear physical significance,they suffer from the disadvantages of large computational cost and poor convergence due to the complexity of the theoretical models,and they have difficulty in capturing the component-level hysteresis characteristics,such as cyclic degradation and pinching effect.In order to account the ASFI effect with comparable accuracy and computational efficiency,an experimental data-driven simulation approach based on the adaptive-update mechanism of constitutive parameters was proposed in this paper.The main research contents and conclusions are as follows:（1）A simulation approach incorporating ASFI based on the adaptive-update mechanism of constitutive parameters was proposed.This approach contains the bearing condition-based and degradation-based adaptive-update mechanism.The former adaptively updates the constitutive parameters according to the bearing conditions,which accounts the ASFI effect due to the change of bearing conditions and reduces the computational cost introduced by the utilization of microscopic models,smeared crack theory and multi-axial concrete constitutive models.The latter adaptively updates the constitutive parameters according to the damage measure,which avoids degradation prediction error introduced by the rule assumptions and simplified equations,and solves the problem that the existing ASFI simulation approaches fail to efficiently capture the cyclic degradation at component level.（2）A constitutive parameters updating approach based on neural network ensemble was proposed.According to the experimental results of 155 RC beams and 288 RC columns under cyclic loading,an adaptive-updated hysteretic model（AUHM）was developed and experimentally validated.It shows that AUHM can efficiently predict the hysteretic responses of RC frame components in different failure modes.In the case of flexure-critical and shear-flexure-critical specimens,the prediction errors of AUHM for bearing capacity0),deformation capacity0)and energy dissipation capacity0)were within±10,±20 and±15respectively.In the case of shear-critical specimens,0)and0)were within±10 and±20respectively,0)was within±20 for most specimens and beyond±20 for a few specimens,but was still controlled within±50 for all specimens.（3）The failure modes classification criteria and the deformation limits of the current design code were experimentally validated,and it shows that the classification criteria for RC beams lacks sufficient safety reserves.Therefore,the failure modes classification criteria for RC beams and columns at shear span ratioλ>2 were unified,resulting in an improvement of classification precision rate for flexure-critical and shear-flexure-critical RC beams from 53.8and 66.7 before unification to 79.4 and 77.8 respectively.（4）The Euler beam theory with plastic hinge lumped at both ends was proposed and assembles AUHM to develop adaptive beam column（Adaptive BC）element,which is capable of accounting ASFI by adaptive updating of constitutive parameters based on component-level experiments.The Adaptive BC was validated by experimental results of specimens under cyclic loading and varying axial loads,and performance of Adaptive BC was compared with shear-flexure-intraction fiber beam column element,shear-flexure-intraction multiple-vertical-line-element-model and layer shell.In the case of accuracy,the prediction errors of Adaptive BC for bearing capacity0)and energy dissipation capacity0)were within±10 and±25respectively,while the comparison elements significantly overestimated the energy dissipation capacity,illustrating that the Adaptive BC can reasonably predict the hysteretic behavior of flexure-critical and shear-critical specimens under varying axial loads.In the case of computational efficiency,the convergence and efficiency of Adaptive BC were superior to those of the comparison elements,which had 2.33～57.03 times more iterations and 1.98～664.67times more computational time than Adaptive BC.（5）The update thresholds of constitutive parameters in Adaptive BC were validated by experimental results of specimens under cyclic loading and varying axial loads,and the update thresholds were suggested to be taken as=0.20,=0.08 and=0.024 to achieve a balance between accuracy and efficiency.Compared to updating constitutive parameters at each analysis step,the suggested update thresholds can reduce the count of updates by 77.20.（6）A simulation approach for biaxial cyclic response,which enables the coupling calculation of biaxial adaptive hinge stiffness,was proposed to expand Adaptive BC to a 3D adaptive beam column（Adaptive BC3D）element.The Adaptive BC3D was validated by experimental results of specimens under biaxial cyclic loading,and it shows that the prediction errors of Adaptive BC3D for bearing capacity0),deformation capacity0)and energy dissipation capacity0)were within±20,±20 and±25 respectively,illustrating the Adaptive BC3D can represent the influence of diagonal loading and different bixial loading paths on the reaction paths,bearing capacity and deformation capacity.（7）The Adaptive BC3D was validated by experimental result of RC frame structure under cyclic loading.It shows that Adaptive BC3D can represent the failure modes and damage states of beams and columns in frame structure,and its accuracy of structural hysteretic response was superior to that of commonly used fiber element.More specifically,the structural prediction errors of Adaptive BC3D for bearing capacity0)and energy dissipation capacity0)were-8.69and-13.15 respectively,while0)and0)of fiber element were-6.24 and 117.65respectively,indicating that the fiber element significantly overestimated the structural energy dissipation capacity.（8）The coupling effect of shear deformation on component-level response,component damage and structural response was investigated based on a typical RC frame structure.The results show that neglecting the coupling effect of shear deformation will fail to reflect the plastic localization of components,resulting in a significant overestimation of the deformation capacity of the plastic localized beam end with a relative error of 97.5.Moreover,it will also overestimate the enegy dissipation capacity of beams.This will delay the formation of weak storey and underestimate the deformation of columns in weak storey with a relative error of-42.71,increasing the risk of underestimating the column damage.Therefore,the coupling effect of shear deformation must be considered and the elements incorporating ASFI must be utilized in structural analysis.The research achievements of this paper is capable of accounting the ASFI effect due to the change of bearing conditions through the adaptive updating of constitutive parameters,and its simulation accuracy and computational efficiency are superior to the existing ASFI element.Therefore,it can be applied in structural elastoplastic analysis to reveal the seismic mechanism of reinforced concrete structures,especially the existing structures with light transverse reinforcement.Moreover,the research achievement provides a novel approach for the combination of machine learning and elastoplastic analysis,as well as new ideas for the development of structural elastoplastic analysis technology,which contributes to improve the seismic damage evaluation system for building structures,and enhance the urban and rural seismic resilience. |
PDF Full Text Request