钢筋混凝土柱基于能量等效的损伤状态量化方法

Quantification Method of Damage States for Reinforced Concrete Columns Based on Energy Equivalence

摘要:

钢筋混凝土柱在地震作用下具有不同的失效模式。不同失效模式钢筋混凝土柱的损伤状态量化方法目前存在定义方式不一且预测精度较差的问题。基于246根钢筋混凝土矩形截面柱的拟静力往复加载试验数据，提出了一种基于能量等效原则量化不同失效模式钢筋混凝土柱损伤状态的方法。通过引入贝叶斯神经网络模型，建立了适用于不同失效模式钢筋混凝土柱的位移角预测公式。研究结果表明：能量等效原则可将钢筋混凝土柱的骨架曲线等效为一个理想弹塑性模型，不仅便于定义屈服点、峰值点和极限点，而且便于在同一框架下对比不同失效模式钢筋混凝土柱的抗震力学性能差异。根据屈服点、峰值点和极限点的位移角，钢筋混凝土柱的损伤状态可以划分为：“基本完好”、“轻度破环”、“中等破坏”和“严重破坏”四个等级。贝叶斯神经网络模型可以准确预测不同失效模式钢筋混凝土柱屈服点、峰值点和极限点的位移角。传统的峰值承载力经验折减系数法在预测屈服点的位移角时偏于保守，在预测极限点的位移角时偏于不安全。

关键词:钢筋混凝土柱;失效模式;损伤状态;能量等效;贝叶斯神经网络

Abstract:

Reinforced concrete （RC） columns have different failure modes under earthquake excitations. The damage state quantification methods of RC columns under different failure modes have problems including inconsistent quantification procedure， and poor prediction accuracy. Therefore， an energy equivalence principle was proposed in this study to quantify the damage states of RC columns under different failure modes using quasi-static experimental data of 246 RC columns with rectangular cross-sections. The Bayesian neural network was introduced to develop the drift angle prediction models for RC columns independent of failure patterns. The research results indicated that the energy equivalence principle can transfer the backbone curves of RC columns into an idealized elastic-plastic model. This not only facilitates the definition of yield point， peak point and ultimate point， but also benefits to compare the seismic performance of RC columns under different failure modes in a unified framework. The damage states of RC columns under different failure modes can be categorized into four groups， namely basic intact damage stage， slight damage state， moderate damage state and sever damage state using the drift angle at yield point， peak point and ultimate point. The Bayesian neural network model can predict well the drift angles of RC columns independent of failure modes at yield point， peak point and ultimate point. The traditional empirical factor method by reducing the peak strength tends to underestimate the drift angle at yield point， and overestimate the drift angle at ultimate point.

Key words:reinforced concrete column;failure mode;damage state;energy equivalence;Bayesian neural network