Prediction of field dependent damping force as a function of magnetorheological damper design parameters using a machine learning method

This paper presents a machine learning approach to predict damping force as a function of its mechanical design parameters in a magnetorheological (MR) damper. The employed machine learning method is extreme learning machine. The studied MR damper is equipped by an MR valve with serpentine flux. The training data is firstly generated using FEMM (Finite Element Magnetic Method) software by varying several parameters. Then, the obtained magnetic flux density is translated into damping force by employing the steady state pressure drop equations. The results of the FEMM simulation and the calculation of the damping force are firstly evaluated to check the pattern. Then, the machine learning is applied. This performance design is built with extreme learning machine algorithms in Python. After simulation, hidden node number of 20 is selected because the simple neural network structure, high R-squared value, and low RMSE compared other hidden node numbers. In general, the R-squared value for hidden node number more than 10 is higher than 0.8 showing a good agreement between the reference data and the predicted values.