A Model of Magnetorheological Grease using Machine Learning Method

Abstract

Magnetorheological (MR) grease is a promising material to replace MR fluid because the advantage in term of stability and less possibility to leaking. To improve the material properties, an accurate model can be critical for reducing the time and cost of the development process. A model has been developed to predict MR fluid material properties by including the composition. However, the model may need adjustment and cannot predict other essential rheology parameters, such as viscosity, apparent viscosity, shear rate, and shear stress. Therefore, the technical novelty of this paper is to propose a model with composition as one of the inputs using extreme learning machine method. A scoring system is also introduced to quantify the significance of the composition effect toward the MR grease performance. Then, the model is simulated and compared with experimental data. The performance shows high accuracy estimation with normalized root mean square error about 1.25%.