Online Cyber-Physical Neural Network Model for Real-Time Hybrid Simulation

Real-time hybrid simulation (RTHS) is an experimental testing methodology that divides a structural system into an analytical and an experimental substructure. The analytical substructure is modeled numerically, and the experimental substructure is modeled physically in the laboratory. The two substructures are kinematically linked together at their interface degrees of freedom, and the coupled equations of motion are solved in real-time to obtain the response of the complete system. A key challenge in applying RTHS to large or complex structures is the limited availability of physical devices, which makes it difficult to represent all required experimental components simultaneously. The present study addresses this challenge by introducing Online Cyber-Physical Neural Network (OCP-NN) models–neural network-based models of physical devices that are integrated in real-time with the experimental substructure during an RTHS. The OCP-NN framework leverages real-time data from a single physical device (i.e., the experimental substructure) to replicate its behavior at other locations in the system, thereby significantly reducing the need for multiple physical devices. The proposed method is demonstrated through RTHS of a two-story reinforced concrete frame subjected to seismic excitation and equipped with Banded Rotary Friction Dampers (BRFDs) in each story. BRFDs are challenging to model numerically due to their complex behavior which includes backlash, stick-slip phenomena, and inherent device dynamics. Consequently, BRFDs were selected to demonstrate the proposed framework. In the RTHS, one BRFD is modeled physically by the experimental substructure, while the other is represented by the OCP-NN model. The results indicate that the OCP-NN model can accurately capture the behavior of the device in real-time. This approach offers a practical solution for improving RTHS of complex structural systems with limited experimental resources.