A model-driven approach to automated diagnosis of industrial distributed I/O systems based on fieldbus technologies

Abstract

Fieldbus technologies have strongly contributed to the spreading of distributed control architectures in all industrial application domains as a viable and effective alternative to conventional centralized I/O solutions. In order to promote wide acceptance by application designers, the most vendors of these technologies propose powerful tools and run-time environments, adequately supporting both off-line network configuration and on-line real-time exchange of process data over the network. The focus of available facilities, however, is merely confined to the network side. The verification of wiring correctness and proper behavior of the plant sensorial/actuation equipment still relies on error-prone and time-consuming manual work. In order to overcome such limitations, a generally-applicable methodology is presented here, aiming to drastically reduce both operator direct intervention and time needed to ascertain that, not only network devices, but also plant devices fulfill the specified application-dependent requirements. Leaving intrinsic features of actuators and sensors out of consideration, typical plant devices are abstractly modeled in terms of the elementary services they may provide and the control/feedback signals they expose to request/notify execution of individual services. Such models give precise indications about devices that do not behave as expected, promptly reporting the cause(s) of each fault and suggesting possible remedies for fault recovery. The validity of the proposed methodology has been experimentally ascertained by carrying out real-life tests on a complex automated machine with a purposely developed diagnosis tool supporting multiple fieldbus technologies.