Applying Self-Recursive Neural Network Prediction to Compensate for the Delay of Real-Time Substructure Experiment

Self-recursive neural network is used to predict structural dynamic responses and compensate for the delay of the hydraulic servo actuator which is the major problem of real-time substructure experiment and cause a direct influence on the stability and veracity. In this paper, the experimental setup is established consisting of D-space real-time simulator, hydraulic actuator, measuring system, data collecting system and measure the value of the delayed time of actuator. On the basis of that, the self-recursive neural network is trained and used to compensate for the delay, so that the numerical model and the experimental substructure can be coordinated and transfigured. Finally, a real-time substructure experiment is performed on a three-storied structure under seismic excitation, which proves the validity of this method.