NEURAL APPROXIMATORS FOR FAULT DETECTION OF ACTUATORS IN THE PRESENCE OF FRICTION: THE CASE OF THE DAMADICS BENCHMARK PROBLEM

摘要

The problem of actuator fault detection (FD) for mechanical systems with friction phenomena is addressed. A novel methodology based on an on-line neural approximation scheme is applied to the DAMADICS benchmark problem. The FD algorithm is based on the well known dynamic LuGrc model characterizing mechanical friction effects. This friction model is suitable for use in the simulation model of the DAMADICS benchmark which is developed in order to approximate the industrial process in a sugar factory located in Lublin (Poland). The approximation scheme makes it possible to evaluate on line suitable thresholds for the detection of incipient or abrupt faults regarding the friction and the spring models of the considered actuator.