The Application of Fault Simulation to Machine Diagnostics and Prognostics

Abstract

The early development of machine diagnostics and condition monitoring was based on measurements from actual failures, but these cannot be predicted or arranged to occur when and where desired. In recent years it has become possible to make simulation models of a machine, such as a gearbox or engine, including the simulation of various faults of different types, severity, and location. There are a number of benefits from doing this, the first being to be able to produce sufficient representative signals to train automated fault recognition algorithms, such as artificial neural networks, as it is not economically viable to experience the number of actual failures required. Being able to produce signals from faults of different sizes and locations can be useful in the development of diagnostic and prognostic procedures — the latter, for example, by being able to develop appropriate trend parameters. Finally, the effects of faults in complex machines are often based on nonlinear interactions, which are difficult to foresee, and the simulation modelling of the whole machine can be very useful to obtain a physical understanding of these complex interactions. This paper illustrates these principles using examples of rolling element bearings, gears, geared systems (including bearings), and internal combustion engines, with favourable results in all aspects.