Prediction of prestress-force in prestressed concrete beams using natural frequency and machine learning

Abstract

Civil infrastructure plays an important role in our life and the development of our country. However, it is affected by many different factors during the operation and degraded after a long time of use. In order to ensure safe operation, the field of structural health monitoring (SHM) has been developed. In recent years, the explosion of artificial intelligence (AI) has contributed to all aspects of social life including the field of SHM. In the civil engineering field, prestressed concrete structures have been widely applied to many projects (i.e., bridges, tall buildings, ...). The common damages on prestressed concrete structures lead to the loss of prestress-force. This paper proposes a method to predict the prestress-force in prestressed concrete beams using natural frequency and machine learning. First, the natural frequencies of a prestressed concrete beam are achieved by a finite element model. The finite element model’s reliability is verified by the experimental results. Then, a machine learning model named polynomial regression algorithm is developed to predict the prestress-force in the beam using natural frequency. The proposed method is highly accurate in predicting the prestress-force for prestressed concrete beams. In addition, this study also examines and assesses the effect of combining natural frequencies of multi-mode to predict the prestress-force.