Field applicability of a machine learning–based tensile force estimation for pre-stressed concrete bridges using an embedded elasto-magnetic sensor

Junkyeong Kim,Seunghee Park

DOI: https://doi.org/10.1177/1475921719842340

2019-04-17

Structural Health Monitoring

Abstract:It has been proposed that pre-stressed concrete bridges improve load performance by inducing axial pre-stress using pre-stress tendons. However, the tensile force of the pre-stress tendons could not be managed after construction, although it directly supports the load of the structure. Thus, the tensile force of the pre-stress tendon should be checked for structural health monitoring of pre-stressed concrete bridges. In this study, a machine learning–based tensile force estimation method for a pre-stressed concrete girder is proposed using an embedded elasto-magnetic sensor and machine learning method. The feedforward neural network and radial basis function network were applied to estimate the tensile force of the pre-stress tendon using the area ratio of the magnetic hysteresis curve measured by the embedded elasto-magnetic sensor. The feedforward neural network and radial basis function network were trained using 213 datasets obtained in laboratory experiments, and trained feedforward neural network and radial basis function network were applied to a 50-m real-scale pre-stressed concrete girder test for estimating tensile force. Nine embedded elasto-magnetic sensors were installed on the sheath, and the magnetic hysteresis curves of the pre-stress tendons were measured during tensioning. The area ratio was extracted and inputted to the trained feedforward neural network and radial basis function network to estimate the tensile force. The estimated tensile force was compared with the reference tensile force measured by the load cell. According to the result, the estimated tensile force can represent the actual tensile force of the pre-stress tendon without calibrating tensile force estimation algorithms at the site. In addition, it can measure the actual friction loss by estimating the tensile force at the maximum eccentric part. Based on the results, the proposed method might be a solution for the structural health monitoring of pre-stressed concrete bridges with field applicability.