【摘要】 针对服役梁桥易受材料和环境等因素影响出现易损性问题，以一座三跨连续梁桥为研究对象，提出一种基于灰关联广义比例柔度曲率差(GMGPFCD-A)构建的动力指纹和人工大猩猩群体优化算法(GTO)-一维卷积神经网络(1D CNN)-双向长短期记忆网络(BiLSTM)预测模型的梁桥分级损伤识别方法；该方法以低阶模态参数构建的广义柔度矩阵和比例柔度矩阵为基础，结合灰色关联分析(GRA),构建动力指纹识别结构的损伤位置，并将该指标输入到1D CNN-BiLSTM损伤预测模型中进行量化分析，引入GTO优化预测模型的超参数以提高其对结构损伤程度的预测性能。研究结果表明：该模型不仅能在无需测得外部环境激励的情况下准确识别结构的损伤位置，并且在噪声水平10%以内具有一定的抗噪性；经GTO优化后的预测模型对识别出的损伤部位的损伤程度准确率达93.548%;提出模型收敛速度更快、更稳定，且具有较高预测准确率和较强鲁棒性。更多还原

【Abstract】 Aiming at the vulnerability of serving beam bridges to the influence of factors such as materials and environment, a three-span continuous beam bridge was taken as the research object, and a beam bridge graded damage identification method based on dynamic fingerprint constructed by grey relational generalized proportional flexibility curvature difference(GMGPFCD-A) and artificial gorilla group optimization algorithm(GTO)-one-dimensional convolutional neural network(1D CNN)-bidirectional long short-term memory network(BiLSTM) prediction model was proposed. Firstly, the generalized flexibility matrix and proportional flexibility matrix constructed by low-order modal parameters were combined with grey relational analysis(GRA) to construct dynamic fingerprint to identify the damage position of the structure, and the index was input into the 1D CNN-BiLSTM damage prediction model for quantitative analysis. Finally, the hyperparameters of the GTO optimization prediction model were introduced to improve its prediction performance of the structural damage degree. The research results show that the model can not only accurately identify the damage location of the structure without measuring the external environmental excitation, but also has a certain noise resistance within 10% of the noise level; the prediction model after GTO optimization has an accuracy rate of 93.548% for the damage degree of the identified damaged parts; the proposed model converges faster, is more stable, and has higher prediction accuracy and stronger robustness. 4 tabs, 10 figs, 28 refs.更多还原