ABSTRACT
Most dangerous highway bridges are caused by cracks, and crack width is the key object of crack monitoring. To effectively prevent the occurrence of false alarms in the current monitoring system, it is often within the range of the early warning threshold. Based on the long short-term memory neural network, this paper establishes a bridge crack monitoring data model and analyzes the prediction results. Through the comparison of the measured values and the predicted values, mean square error (MSE), and fitting disturbance (R-square), it is found that the amount of data with an error of less than 20% between the calculated data from the proposed model and the measured value is approximately 97.85%, which proves the feasibility of the model for bridge crack width prediction. It also shows that the LSTM model is suitable for bridge crack prediction and can effectively prevent false alarms in early warning.
