Deformation prediction model based on an improved CNN + LSTM model for the first impoundment of super-high arch dams.

Autor: Yilun, Wei, Qingbin, Li, Yu, Hu, Yajun, Wang, Xuezhou, Zhu, Yaosheng, Tan, Chunfeng, Liu, Lei, Pei
Zdroj: Journal of Civil Structural Health Monitoring; Mar2023, Vol. 13 Issue 2/3, p431-442, 12p
Abstrakt: Herein, we propose a one-dimensional convolutional neural network (CNN) + long short-term memory (LSTM) model optimised by L1 regularisation and the dropout method to solve the problem of acquiring both computational speed and accuracy in a deformation prediction analysis model of a super-high arch dam's first impoundment. The calculation results of one class (OC) + LSTM, traditional LSTM, optimised LSTM, CNN + LSTM and multilayer perceptron are compared with the actual measurement results using deformation monitoring data from the first impoundment of a super-high arch dam in southwest China. The results show that the proposed OC-LSTM model can reduce the computational time without sacrificing computational accuracy, providing a new computational model for super-high arch dam deformation prediction during the first impoundment. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index
Externí odkaz: