Understanding and quantifying the influence of a damper parameters using FEM-based ANN and sensitivity analysis

The CAR1 friction-yielding damper, is the central subject of this investigation. These dampers are constructed from simple materials, yet they can provide predefined axial control forces and controlled seismic energy dissipation through an appropriate selection of design parameters consistent with the structural demands of buildings. The CAR1 damper includes 13 design variables, and optimal combinations of these variables can be identified using previous research. However, this methodology is constrained by the arbitrary pre-selection of certain parameters to derive the others. To address this limitation, this work proposes the integration of Artificial Neural Networks combined with a global sensitivity analysis based on the Fourier Amplitude Sensitivity Test (FAST). This approach aims to improve the robustness and reliability of the decision-making process for selecting the damper’s design parameters. The results of the sensitivity analysis indicate that the mechanical response of the system is strongly nonlinear.