Semi-active dampers are widely used in various fields due to their greater flexibility compared to passive and active dampers. Among these, Magnetorheological (MR) dampers, which contain MR fluid, can adjust damping characteristics in real time by changing the magnetic field through the supplied current. Accurate modeling of MR dampers is essential for control systems, as it enables precise prediction of damping force based on input conditions, allowing controllers to make informed decisions and dynamically adjust current to achieve desired performance. MR damper modeling can be approached using either parametric or non-parametric methods. However, modeling is challenging due to dynamic hysteresis effects, variations in magnetic fields, and nonlinear behavior. Parametric approaches often involve complex mathematical models with numerous parameters, making them difficult to implement and tune. To address these challenges, this research proposes an integrated differential evolution with one-dimensional Long Short-Term Memory (DE-1DLSTM) model to capture MR damper behavior. The performance of the DE-1DLSTM model was evaluated using datasets from the Network for Earthquake Engineering Simulation (NEES) and experimental data. The proposed model achieved accurate predictions for both datasets, with the predicted force-displacement and force-velocity graphs closely matching actual MR damper behavior. Thus, implementing the DE-1DLSTM model effectively reduces the complexity of MR damper modeling, enhances prediction accuracy, and supports the development of intelligent control systems.
SAUFI et al. (Mon,) studied this question.