Weifeng Shang
Weifeng Shang has a total of 2 co-authors. Here are the top co-authors by number of shared publications:
- Dongliang Hu
Hohai University
2 shared publications Huatao Chen
1 shared publication
Latest Publications
- Open Access
- 22 July 2026
- Published by Elsevier in SSRN Electronic Journal
Non-parametric Dynamic Modeling of Magnetorheological Dampers Based on Fourier Neural Operators
Magnetorheological (MR) dampers have emerged as prominent semi-active actuators in the field of structural vibration control, owing to their millisecond-level responsiveness and continuously adjustable damping characteristics. However, their inherent strong nonlinearity, dynamic hysteretic behavior, and path dependency pose significant challenges for precise dynamic modeling. To address these issues, this paper introduces a non-parametric modeling approach based on the Fourier Neural Operator (FNO). By leveraging operator learning, the proposed method establishes an infinite-dimensional function-space mapping from excitation inputs to damping force outputs. Unlike traditional neural network models, FNO parameterizes the integral kernel in the frequency domain. This enables the model to capture the underlying continuous dynamics of the physical system while maintaining discretization invariance and high computational efficiency. Validated on a comprehensive public experimental dataset, the proposed FNO model achieves a Root Mean Square Error (RMSE) of 10.39 N and a coefficient of determination of 0.9998 on the test set, significantly outperforming baseline models such as LSTM and CNN-LSTM. Furthermore, the study reveals that the proposed model exhibits distinct advantages in characterizing the highly nonlinear roll-off region of force-velocity curves, handling complex multi-frequency excitations, and demonstrating superior extrapolation capabilities across unseen operating conditions.
- 22 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessNon-parametric Dynamic Modeling of Magnetorheological Dampers Based on Fourier Neural Operators
Magnetorheological (MR) dampers have emerged as prominent semi-active actuators in the field of structural vibration control, owing to their millisecond-level responsiveness and continuously adjustable damping characteristics. However, their inherent strong nonlinearity, dynamic hysteretic behavior, and path dependency pose significant challenges for precise dynamic modeling. To address these issues, this paper introduces a non-parametric modeling approach based on the Fourier Neural Operator (FNO). By leveraging operator learning, the proposed method establishes an infinite-dimensional function-space mapping from excitation inputs to damping force outputs. Unlike traditional neural network models, FNO parameterizes the integral kernel in the frequency domain. This enables the model to capture the underlying continuous dynamics of the physical system while maintaining discretization invariance and high computational efficiency. Validated on a comprehensive public experimental dataset, the proposed FNO model achieves a Root Mean Square Error (RMSE) of 10.39 N and a coefficient of determination of 0.9998 on the test set, significantly outperforming baseline models such as LSTM and CNN-LSTM. Furthermore, the study reveals that the proposed model exhibits distinct advantages in characterizing the highly nonlinear roll-off region of force-velocity curves, handling complex multi-frequency excitations, and demonstrating superior extrapolation capabilities across unseen operating conditions.
- 18 February 2026
- Published by Elsevier in Mechanical Systems and Signal Processing
- Vol. 248
Stochastic stability of a two-degree-of-freedom vibro-impact system driven by non-Gaussian colored noise
- 18 February 2026
- Published by Elsevier in Mechanical Systems and Signal Processing
- Vol. 248
Stochastic stability of a two-degree-of-freedom vibro-impact system driven by non-Gaussian colored noise