Experimental Study on Smart Sloped Rolling-type Bearings Integrated with Early Prediction of Peak Velocity

Abstract

The use of sloped rolling-type bearings (SRBs) to protect critical equipment from seismic damage has been widely adopted. However, maintaining SRB displacement responses within safe limits often requires a large damping force, which results in unacceptable accelerations. This highlights the trade-off in seismic isolation: reducing displacement typically comes at the cost of increased acceleration, and vice versa. Although previous numerical studies have shown that adjusting the SRB damping force based on peak velocity (PV) predictions from an earthquake early warning (EEW) system enables better control over acceleration and displacement responses, experimental validation is necessary. This study employs magnetorheological (MR) dampers to apply the damping force determined by the predicted PV. Performance tests are conducted to characterize the physical properties of the MR dampers and develop their mathematical representation. Numerical simulations incorporating the MR damper and SRB models under seismic conditions are used to establish a control law that correlates PV with the voltage required by the MR dampers. To predict PV, the first few seconds following primary seismic wave (P-wave) detection undergo a multi-scale Short-Term Fourier Transform (STFT) to generate a spectrogram. After feature enhancement, it is fed into a Convolutional Neural Network (CNN)-based model, which predicts the impending earthquake's PV. This prediction, alongside the established control law, enables rapid adjustment of the SRB damping force shortly after P-wave detection. The effectiveness of the proposed approach is evaluated through shake table testing. Results indicate that the approach significantly mitigates isolation displacement responses under major earthquakes, demonstrating its potential for improving seismic isolation performance.