Prediction method of structural health monitoring data based on CNN-LSTM

Abstract

The health monitoring of architecture structure is critical for the construction safety and stable operation of architecture structure during operation. In order to fully understand the structural safety of architecture structure, this paper analyzes the actual bridge monitoring data, and proposes a neural network prediction method based on the combination of a convolutional neural network and long-term and short-term memory network, that is, a hybrid model (CNN-LSTM). Thereby, this paper discusses the application characteristics, analysis flow, and modeling process of the CNN-LSTM neural network prediction model, analyzes the actual bridge monitoring data by using this method, and explores the change rule of training model sample size and model prediction accuracy. Firstly, CNN is used to extract the local spatial feature information of time series data, which effectively solves the problem of gradient explosion of the LSTM network when the time series data is too long. At the same time, the experiment on the monitoring data of a bridge in China shows that compared with the traditional method using CNN or LSTM, this method improves all kinds of evaluation indexes and has a better prediction effect, which can provide technical support for bridge structure safety evaluation and damage identification.