In this work, design and validation techniques of two soft sensors for the estimation of the motorcycle vertical dynamic have been proposed. The aim of this work is to develop soft sensors able to predict the rear and front stroke of a motorcycle suspension. This kind of information are typically used in the control loop of semi‐active or active suspension systems. Replacing the hard sensor with a soft sensor, enable to reduce cost and improve reliability of the system. An analysis of the motorcycle physical model has been carried out to analyze the correlation existing among motorcycle vertical dynamic quantities in order to determine which of them are necessary for the development of a suspension stroke soft sensor. More in details, a first soft sensor for the rear stroke has been developed using a Nonlinear Auto‐Regressive with eXogenous inputs (NARX) neural network. A second soft sensor for the front suspension stroke velocity has been designed using two different techniques based respectively on Digital filtering and NARX neural network. As an example of application, an Instrument Fault Detection (IFD) scheme, based on the rear stroke soft sensor, has been shown. Experimental results have demonstrated the good reliability and promptness of the scheme in detecting different typologies of faults as losing calibration faults, hold‐faults, and open/short circuit faults thanks to the soft sensor developed. Finally, the scheme has been successfully implemented and tested on an ARM microcontroller, to confirm the feasibility of a real‐time implementation on actual processing units used in such context. [edited by Author]

2017 - 2018

XXX ciclo