A Virtual ANN-Based Sensor for IFD in Two-Wheeled Vehicle

In the context of automotive and two-wheeled vehicles, the comfort and safety of drivers and passengers is even more entrusted to electronic systems which are closed-loop systems generally implementing suitable control strategies on the basis of measurements provided by a set of sensors. Therefore, the development of proper instrument fault detection schemes able to identify faults occurring on the sensors involved in the closed-loop are crucial for warranting the effectiveness and the reliability of such strategies. In this framework, the paper describes a virtual sensor based on a Nonlinear Auto-Regressive with eXogenous inputs (NARX) artificial neural network for instrument fault diagnosis of the linear potentiometer sensor employed in motorcycle semi-active suspension systems. The use of such a model has been suggested by the particular ability of NARX in effectively take into account for the system nonlinearities. The proposed soft sensor has been designed, trained and tuned on the basis of real samples acquired on the field in different operating conditions of a real motorcycle. The achieved results, show that the proposed diagnostic scheme is characterized by very interesting features in terms of promptness and sensitivity in detecting also “small faults”.