Modeling of a magneto-rheological damper using modified FNN without force sensor input

This paper proposes a model for magneto-rheological (MR) dampers using feed-forward neural network (FNN) that does not require force information. Existing methods that utilize FNN to predict the damping force generated by an MR damper use previous time-histories of displacement, voltage and force as inputs. The resulting predicted force is accurate but requires that real-time force measurements be available through the installation of force sensors to the system. Unlike the previous FNN model, the proposed model does not require these force measurements, therefore leading to a simpler and more economical implementation of the MR damper model. The system is designed using time-histories of displacement, current and velocity to predict damping force. This paper shows results of proposed model compare to previous FNN model using training and validation data sets generated by experimental MR damper system.