Earthquake damage investigation,experimental study and theoretical analysis show that the structural deformation capacity is insufficient due to unreasonable structural measures,especially the unreasonable transverse reinforcement configuration,which is one of the important reasons for the failure and even collapse of reinforced concrete(RC)structures under earthquake action.According to the current code in China,transverse reinforcement is mainly determined by shear bearing capacity design and seismic structural measures.The component ductility demand under earthquake action is not fully considered,and the transverse reinforcement design method is not accurate enough,which leads to the shortage or waste of transverse reinforcement.This paper focuses on the design method of transverse reinforcement of reinforced concrete structure based on component ductility demand,and the main research results are as follows:(1)The design method of transverse reinforcement of reinforced concrete structure based on component ductility demand is proposed,and the corresponding design flow chart is given.A method for determining component ductility demand through nonlinear response history analysis is established,which can reflect the component ductility demand under earthquake action in detail.(2)The calculation method of transverse reinforcement of reinforced concrete column based on machine learning(ML)is established,which can realize intelligent quantitative design of transverse reinforcement according to the actual ductility demand.Based on an experimental database consisting of 326 rectangular columns and 172 circular columns(498 samples in total),12 ML models are trained.In order to solve the problem of over-fitting,feature engineering aiming at dimension reduction is systematically carried out.Grid search and 10-fold cross-validation method are applied to tune the hyper-parameters.Through comprehensive performance evaluation on the testing set,an XGBoost model is selected.The reliability of the "black box" ML model is verified by the SHAP method and partial dependence plots(PDP).Compared with the empirical models,the ML model shows higher accuracy and smaller deviation.An open-source ML model to calculate transverse reinforcement ratio required for RC columns is provided worldwide,with flexibility to account for additional experimental results to improve performance.(3)An equivalent linearization method based on nonlinear response history analysis is proposed which realizes the high-precision equivalence of stiffness and damping.Fine mesh layered shell element and fiber beam element based on material nonlinear constitutive model are adopted.The equivalent stiffness of the component is determined by integrating the secant stiffness of concrete,steel and steel at the integral points.The damping energy spectrum of elastic multi-degree-of-freedom system is derived,and a more accurate method to determine the equivalent damping ratio is given.The equivalent linearization method given in this paper can well reflect the stress state of components in the nonlinear stage of the structure,and the results of equivalent linearization analysis are close to those of nonlinear analysis.(4)A refined seismic resilience assessment of buildings based on component damage states is proposed.The performance evaluation method of concrete is given by the strength degradation coefficient.Combined with the strain-based performance evaluation method of rebar and steel,the performance evaluation method of component considering the constraint effect of transverse reinforcement is given.Based on the results of component performance evaluation,the damage state of component is taken as the original sample,and the randomness of ground motion is considered through sample matrix expansion and Monte Carlo simulation.Based on a large number of simulated samples of component damage state,the seismic resilience assessment indexes,such as repair cost,repair time and casualty rate,are calculated,which expands the applicability of the seismic resilience assessment method to spatial structures and improves the accuracy of determining component damage state of structures with uneven story deformation or irregular torsion.In order to consider the influences of the number of ground motions and the dispersion of nonlinear response history analysis results on the seismic resilience assessment index,a simplified confidence interval algorithm at a given confidence level of resilience assessment index is derived,which improves the reliability of seismic resilience assessment results.(5)The applications of transverse reinforcement design method based on component ductility demand in new buildings and existing buildings are studied.The case study shows that the method in this paper can design transverse reinforcement accurately according to the actual ductility demand of components.Compared with the design method in the current code,the transverse reinforcement quantity of the components with high ductility demand is increased while the transverse reinforcement quantity of the components with low ductility demand is reduced,and the overall component performance and structural seismic resilience are better.For components that are difficult to increase transverse reinforcement in existing buildings,the original transverse reinforcement can meet the ductility demand by reducing the component ductility demand.Compared with the traditional reinforcement methods,the transverse reinforcement design method based on component ductility demand can quantify the transverse reinforcement demand of components under earthquake action,control the reinforcement position and quantity of existing buildings,and give consideration to economy and construction feasibility while improving safety. |