Prediction of long-term strain in concrete structure using convolutional neural networks, air temperature and time stamp of measurements

A data prediction method for long-term strain measurements from concrete structures based on the strong correlation between air temperature and structural response is proposed. A convolutional neural network (CNN) is employed to capture and define the relationship between the structural response and air temperature. The CNN is trained using measurements of air temperature and strain collected before the data interruption. To reflect the time-dependent long-term behavior of a concrete structure, the air temperature and corresponding time information are simultaneously utilized in the input layer of the proposed CNN. The trained CNN is then used to estimate the strain in the structure using only the air temperature data from the weather station in the event of a data loss from the structure's sensors. The presented method is validated using long-term data from fiber optic sensors embedded in a concrete footbridge at Princeton University and air temperature data from a nearby weather station.

Automated summary

The proposed method utilizes a convolutional neural network to establish the relationship between air temperature and structural response in concrete structures, predicting strain using only temperature data in case of sensor data loss. Validation was done using long-term data from fiber optic sensors in a concrete footbridge at Princeton University and temperature data from a nearby weather station.