基于动力测试的钢筋混凝土梁火灾损伤识别方法

为获得混凝土梁的受火损伤程度,提出了基于小波神经网络技术以等效爆火时间为指标的损伤识别新方法。首先建立了简支梁的火灾损伤识别方法,并用数值模拟对其进行了验证;然后建立了的适用于混凝土连续梁火灾损伤识别的三步定位新方法,以三跨连续梁为例对其应用进行了详细说明,数值模拟结果表明该方法准确度较高;最后设计4根足尺寸钢筋混凝土简支梁L1 -L4,分别对LI -L4进行60 min、90 min、120 min、150 min的火灾试验及灾后承载力试验,实测了火灾前、后及过程中的结构模态信息及灾后荷载-位移曲线,基于修正后的精细化模型,利用前2阶不完备模态信息构造小波神经网络输入参数,等效爆火时间作为输出参数进行损伤识别,实测值与识别预测值吻合较好,验证该方法的可靠性。

Identification method for fire damage of RC beams based on dynamic tests

Here, in order to acquire fire damage degree of RC beams, a new damage identification method based on wavelet neural network ( WNN) was proposed taking the equivalent fire time as the damage index. Firsdy, a fire damage identification method for a simply supported beam was established and verified with numerical simulation. Then, a new 3-step positioning method was established for fire damage of RC continuous beams, and it was described in detail taking a 3-span continuous beam as an example. Finally, 4 full scale RC beams (LI L4) were designed, fire tests were conducted for LI L4 within 60 min, 90 min, 120 min and 150 min, respectively, and after firing their load-bearing capacities were tested. These beams' structural modal features before and after firing as well as their load-displacement curves after firing were measured. Based on LI L4 modified refinement models, input parameters of WNN were constructed based on LI L4 first 2 orders incomplete modal information, respectively and their corresponding equivalent fire times were taken as output parameters of WNN to do LI L4 damage recognitions. The results showed that the actual measured values agree better with the predicted ones using WNN, so the reliability of the proposed method is verified.