Experimental Study and Neural Network Predictions of Early-Age Behavior of Microexpansion Concrete in Large-Diameter Steel Tube Columns

Abstract

The improved mechanical performance of concrete-filled steel tube (CFST) components has led to their widespread application in megastructures. Therein, CFST hybrid arch bridge has become an optimal selection of large span bridges. Nonetheless, for massive concrete poured in CFST members with large diameter, the thermal cracking and sustained rising temperature caused by early hydration heat of core concrete are urgently required to be studied. The two large-diameter CFST columns in this work, each having a diameter of 2.1 m, had their temperature and strain fields recorded in-situ. An existing CFST arch bridge served as the model for the two CFST columns’ design. Additionally, early-age characteristics of several scaled CFST columns with varying diameters were documented. A multi-field finite element (FE) model that combines linked chemical (hydration), heat, and mechanical fields is suggested in order to properly characterize the evolutions of temperature and strain fields. The model is validated by comparing the in-situ measurements with the numerical results. Finally, to investigate the affect factors on the hydration temperature of core concrete in CFST columns, early-age hydration behaviors of CFST columns was simulated using the validated FE models input parameters as water to cement ratio (w/c), cement dosage, heat release of cement and diameter of CFST columns. Based on the numerical results under the input parameters mentioned, the LSTM neural network was constructed, and the hydration temperature variances computed by FE models were selected as the input dataset. Afterwards, the temperature variance of core concrete of CFST columns was predicted using the established LSTM network. It is discovered that the LSTM neural network that was previously constructed was able to predict the peak temperature of CFST columns as well as the hydration temperature of CFST specimens with respect to time.