An application of extreme learning machine in a graphical user interface for magnetorheological fluid study

Abstract

Recently, magnetorheological (MR) fluid has been widely used in many fields because of its unique characteristics. It exhibits rapid, reversible and tunable transition from liquid to semi-solid state when external force such as magnetic field is applied. Numerous investigations have been carried out by researchers to understand the rheological properties of MR fluid. Various modeling techniques have been built to predict the fluid behavior for the development purposes. Those techniques can be parametric or non-parametric which includes rheological, polynomial and phenomenological models with their own application and limitation. In this stage, models are important to shorten the development time and saving the experimental cost. A Graphical User Interface (GUI) containing one of the models can be a beneficial tool to predict the fluid behavior easier during an experiment or investigation. However, a GUI needs a flexible modeling method that can be adjusted and updated easily. Therefore, machine learning as data to knowledge-model can be the best solution. In this paper, extreme learning machine (ELM) is proposed to train the datasets of MR fluid used for the GUI because of its advantage in terms of its faster training and better generalization. The best configuration of ELM is carefully investigated and applied in the GUI. The developed interface can act as prediction platform for the output of MR fluid based on the learning process through the given data.