Abstract: In order to accurately predict the deformation response of bridge structure, an EEMD-LSTM combined model was proposed, in which the iForest algorithm was used to denoise the original deformation data of the bridge, the Ensemble Empirical Mode Decomposition (EEMD) method was used to decompose the bridge deformation data, and the Long Short-Term Memory (LSTM) deep learning method was used to predict the obtained multi-scale deformation components. Taking Wuhan Zhuankou Yangtze River Bridge as the research object, RMSE, MAE, MAPE and R2 were selected as evaluation indexes to verify the proposed model. The research results show that compared with the single LSTM, SVM and Bayesian model, the EEMD-LSTM model has good robustness, applicability, and higher prediction accuracy.

Key words: bridge engineering; deep learning; iForest algorithm; EEMD-LSTM; deformation response

摘要： 为准确预测桥梁结构变形响应，提出了一种利用孤立森林(iForest)算法对桥梁原始变形数据进行降噪，集合经验模态分解法(EEMD)对桥梁变形数据进行分解，长短期记忆神经网络(LSTM)深度学习法对所得到的多尺度变形分量进行预测的EEMD-LSTM组合模型。以武汉沌口长江大桥作为研究对象，选取RMSE、MAE、MAPE和R2等参数作为评价指标，对该模型进行了验证。研究结果表明：与单一的LSTM、SVM和Bayesian模型相比，EEMD-LSTM模型有着良好的鲁棒性、适用性和更高的预测精度。

关键词: 桥梁工程；深度学习；孤立森林算法；EEMD-LSTM；变形响应