Dynamic rheological behavior and ANN model with Bayesian optimization for elastosil-based magnetorheological elastomers

Abstract

Magnetorheological elastomers (MREs), which can change their rheological properties under magnetic stimuli, have seen a surge of interest for their utility in sophisticated vibration isolation systems. This study investigates the impact of carbonyl iron particle (CIP) concentration and magnetic field strength during the curing process on the rheological properties of MREs. An Artificial Neural Network (ANN) model using Bayesian optimization was also developed to predict properties and its accuracy was confirmed by comparing it with the experimental results. MRE samples with varying volumes of CIPs (5, 10, 15, and 20%) were prepared using a matrix of room temperature vulcanized silicon elastomer, specifically Elastosil 4511. During the curing process, MRE samples were subjected to different magnetic field strengths (0, 0.15, 0.3, 0.5 T). The rheological behavior was analyzed using a dynamic shear rheometer (DSR), conducting amplitude, frequency, and magnetic sweep experiments. The study reveals that higher CIP concentrations initially increase MRE stiffness, but a pronounced Payne effect emerges at higher strains. Furthermore, elevated curing magnetic fields and magnetic sweeps significantly enhance MRE stiffness and response. The optimized ANN model, developed through the Bayesian method, demonstrated a marked potential in accurately predicting the behavior of MREs under various conditions.

Graphical abstract