【摘要】 为准确识别桥梁结构的损伤程度，制作了桥梁的关键构件——预应力钢筋混凝土梁，进行三点弯曲加载试验．收集了损伤破坏全过程的声发射（AE）信号，通过AE信号参数分析，将梁的损伤破坏过程划分为4个典型阶段．构建了长短时记忆神经网络（LSTM）模型，根据经验设置LSTM模型的超参数容易导致网络陷入局部最优而影响了分类结果，提出采用Sine混沌映射和自适应权重来改进鲸鱼优化算法（WOA），对LSTM进行超参数寻优．设计了IWOA-LSTM算法模型，训练识别试验梁各损伤阶段的AE信号特征参数．定型网络结构，并识别同种工况下其他梁的AE信号．结果表明：IWOA-LSTM算法模型识别准确率均超过或接近92%，相较于普通LSTM模型，IWOA-LSTM模型识别准确率提高了约7%.更多还原

【Abstract】 To accurately identify the damage degree of bridge structure, the key bridge component of prestressed reinforced concrete beams was fabricated, and the three-point bending loading experiment was carried out. The acoustic emission(AE)signals were collected during the whole damage failure process, and the damage and failure process of the beam was divided into four typical stages through AE parameter analysis. The long short-time memory(LSTM)neural network was constructed. To solve the problem of falling into local optimum by setting hyper parameters of LSTM model according to the experience, the improved whale optimization algorithm(WOA) based on sine chaotic map and adaptive weight was proposed to optimize hyper parameters of LSTM. The IWOA-LSTM algorithm model was designed to train and identify AE signals characteristic parameter data of experimental beam at various damage stages. The network structure was finalized, and the AE signals of other beam under the same working conditions were identified. The results show that the recognition accuracies all exceed or approach 92%. Compared with the common LSTM model, the recognition accuracy of IWOA-LSTM model is improved by about 7%.更多还原