ABSTRACT

Mathematical modelling and computer simulation (MMCS) (Bartelmus 1994, 1996, 1998b, 2000b, 2001a) has proved to be very important tool for supporting diagnostic inference (SDI) (Bartelmus and Zimroz 1998a, 2000a, 2001b, 2002, 2003a, 2003b, 2003d, 2003e, Zimroz 2002). We may also say that the aim of MMCS is to detect faults in a system. In the paper the term fault is used rather than failure to denote a malfunction rather than a catastrophe. The term failure suggests complete brake down of a system component of function, while the term fault may be used to indicate that malfunction may be tolerated at its present stage. A fault must be diagnosed as early as possible even it is tolerable at its early stage, to prevent any serious consequences. Development of SDI based on MMCS has been going though many sages. For SDI of gearboxes condition, models have been developed starting with creation a system with one stage-gearbox incorporated into the system consisting of an electric engine, flexible coupling and driven machine (Bartelmus 1998a, 2000a, 2001a). For modelling such system and using for SDI many factors has to be taken into consideration. The factors can be divided into four groups namely: design factors (DF), technology factors (TF), operation factors (OF), condition change factors (CCF); (collectively) DPTOCCF. All the mentioned factors have influence to dynamic behaviour of the system. Investigating influence of

DPTOCCF to vibration generated by the system with a gearbox it is possible to infer relation between gearbox condition and symptoms given by vibration signal. Big progress in SDI is given by model development for system with a two-stage gearbox (Bartelmus 2000b, 2001a, 2003a, 2003d, 2003e). All the publications are presenting DPTOCCF based way for gearbox diagnostic inference. Current challenge in diagnostic method developments using mathematical modelling and computer simulation (MMCS) is to give the background for inferring process automation using neural networks (Bartelmus et al. 2003c). The alternative way is to use MMCS for model-based fault detection, for which is a need to create what is called a robust model. The robust model is used in the process of diagnostic automation as analytical redundancy (Patton & Frank 1999; Patton et al. 2000).