聚类分析
计算机科学
水准点(测量)
人工智能
人工神经网络
替代模型
机器学习
集合(抽象数据类型)
数据集
数据挖掘
大地测量学
程序设计语言
地理
作者
Yuchen Liao,Ruiyang Zhang,Gang Wu,Hao Sun
出处
期刊:Journal of Engineering Mechanics-asce
[American Society of Civil Engineers]
日期:2023-09-01
卷期号:149 (9)
被引量:3
标识
DOI:10.1061/jenmdt.emeng-6812
摘要
Machine learning–based methods, especially deep learning methods, have achieved great success in seismic response modeling due to their exceptional performance in capturing nonlinear features. However, imbalanced features of a limited training data set can significantly decrease the prediction accuracy of machine learning models. Therefore, this study proposes a novel frequency-based clustering approach for ground motion selection to generate a balanced training data set to improve the data-driven surrogate modeling of bridges. The hierarchical clustering method was developed to suppress the redundant information on the basis of a wavelet analysis of ground motion records. The proposed method was validated by a benchmark finite-element model of a girder bridge, in which long short-term memory (LSTM) neural network was used to predict the seismic responses given ground motion excitations. Specifically, the prediction performances of LSTM surrogate models trained using different data sets have been compared, while the influence of time-frequency characteristics of ground motions has been discussed in detail. The results indicated that the proposed method can provide a balanced training data set with a uniform distribution of time-frequency characteristics and effectively improve the prediction accuracy of deep learning–based surrogate models.
科研通智能强力驱动
Strongly Powered by AbleSci AI