A new constitutive model of a magneto-rheological fluid actuator using an extreme learning machine method

Abstract

In this work, a new constitutive model of a magneto-rheological fluid (MRF) actuator is proposed using an extreme learning machine (ELM) technique to enhance the prediction accuracy of the field-dependent actuating force. After briefly reviewing existing rheological constitutive models of MRF actuator, ELM algorithm is formulated using a single-hidden layer feed-forward neural network. In this formulation, both the magnetic field and measured shear rates are used as inputs variables, while the shear stress predicted from the ELM training is used as an output variable. Subsequently, in order to validate the effectiveness of the proposed model, the target defined as the error between the prediction and measured data is set. Then, the fitness of the training and prediction performances is evaluated using a normalized root mean square error (NRMSE) method. It is shown that the shear stress estimation based on the ELM model using sinusoidal activation function is more accurate than conventional rheological constitutive models such as Herschel-Bulkley model. It is also demonstrated that the proposed model is capable of predicting the field-dependent yield stress which is defined as an actuating force of the MRF actuator without causing significant errors.