Hybrid approach and sensor fusion for reliable condition monitoring of a mechatronic apparatus

In this paper the results are presented of a methodology for identification of the best setting of a mechatronic device at its set-up and/or of fault occurrence in working conditions. The condition monitoring approach merges the information deriving from experimental data of sensors and from a simulation model. The methodology is used to identify the most suitable quantities to be measured, independently on location for sensors. The hybrid information suggests a synthetic set of features, able to identify specific statuses of the mechatronic system, different as for initial setting and/or for the occurrence of faults, to be processed by advanced algorithms like Artificial Neural Networks. The tests show that the methodology is able to realize a resolute and reliable condition monitoring of a mechatronic system in real scale, with reference to both the parameters of setting and to the occurrence of wear and lubrication problems in the kinematic linkage. Finally, the information content of data deriving from internal sensors to the controller is maximized, so reducing the need of external ones for reliable condition monitoring applications.