Backlash fault detection in mechatronic system

In this paper, a fault detection and isolation model based method for backlash phenomenon is presented. The aim of this contribution is to be able to detect then distinguish the undesirable backlash from the useful one inside an electromechanical test bench. The dynamic model of the real system is derived, using the bond graph approach, motivated by the multi-energy domain of such mechatronic system. The innovation interest of the use of the bond graph tool, resides in the exploitation of one language representation for modelling and monitoring the system with presence of mechanical faults. Fault indicators are deduced from the analytical model and used to detect and isolate undesirable backlash fault, including the physical system. Simulation and experimental tests are done on electromechanical test bench which consists of a DC motor carrying a mechanical load, through a reducer part containing a backlash phenomenon. (C) 2007 Elsevier Ltd. All rights reserved.