Study of the nonlinear behavior of prestressed concrete girders by a neural network

For the performance design of a bridge, improvement of the hysteresis model of prestressed concrete (PC) members is necessary in order to carry out checking and design in considering the superstructure nonlinearity. However, new functions and parameters must be investigated to appropriately express features of the PC superstructure hysteresis loop, for example 'the origin directionality by the prestress' and 'the asymmetry of the configuration of the strands', therefore, the modeling is very complicated. Then, it was shown that nonlinear hysteretic behavior of a PC member could be simply modeled using the ability of a neural network to approximate function in this study. The cyclic loading test using 1/8 test specimens of the superstructure of an actual bridge was conducted to acquire the data for learning of a neural network.