To determine the degree of fire damage in reinforced concrete (RC) T-beams, a damage identification method based on an improved support vector machine and firefly algorithm (FA-SVM) is proposed herein. First, based on the fire test of 10 simply supported beams, a refined finite element model of simply supported T-beams was established. A modal analysis of the simply supported beams was performed to obtain a sample library of the input and output parameters for the FA-SVM identification network. Subsequently, the trained FA-SVM identification network was used to predict the fire exposure duration of the samples. The sectional temperature during the fire exposure duration could be calculated by using the finite element model, and the bending stiffness and bending capacity of the simply supported beams after exposure to the fire were calculated. The test sample results were similar to the experimental results of the simply supported beams exposed to fire for 60, 90, and 120 min, which demonstrated the feasibility and effectiveness of the proposed method. Finally, a three-step method for fire damage identification suitable for RC continuous beams was developed based on the FA-SVM identification network. An example calculation analysis of three-span continuous beams was performed, and the results demonstrated accuracy of the identification results. The identification sample magnitude was significantly reduced using this method, which can be conveniently used in practical engineering applications.

中文翻译：

---

基于支持向量机和萤火虫算法的钢筋混凝土T梁火灾损伤模态识别方法

为了确定钢筋混凝土（RC）T型梁的火灾损伤程度，本文提出了一种基于改进支持向量机和萤火虫算法（FA-SVM）的损伤识别方法。首先，基于10根简支梁的耐火试验，建立了简支T梁的精细有限元模型。对简支梁进行模态分析，以获得 FA-SVM 识别网络的输入和输出参数的样本库。随后，经过训练的 FA-SVM 识别网络用于预测样本的火灾暴露持续时间。利用有限元模型可以计算出火灾暴露期间的截面温度，并计算出简支梁暴露于火灾后的抗弯刚度和抗弯承载力。试件结果与简支梁暴露于火中60、90和120分钟的实验结果相似，证明了该方法的可行性和有效性。最后，基于FA-SVM识别网络，开发了适用于RC连续梁的火灾损伤识别三步法。对三跨连续梁进行了实例计算分析，结果证明了识别结果的准确性。使用该方法显着降低了识别样本量，可以方便地应用于实际工程应用中。证明了所提出方法的可行性和有效性。最后，基于FA-SVM识别网络，开发了适用于RC连续梁的火灾损伤识别三步法。对三跨连续梁进行了实例计算分析，结果证明了识别结果的准确性。使用该方法显着降低了识别样本量，可以方便地应用于实际工程应用中。证明了所提出方法的可行性和有效性。最后，基于FA-SVM识别网络，开发了适用于RC连续梁的火灾损伤识别三步法。对三跨连续梁进行了实例计算分析，结果证明了识别结果的准确性。使用该方法显着降低了识别样本量，可以方便地应用于实际工程应用中。