Intelligent monitoring and prognosis of degradation of rotating machinery

The most economic way to do maintenance of rotating machinery is to justify the maintenance actions based on measurements i.e. maintenance actions are carried out when there is a need that has been indicated by these measurements. The reason for this is that the need for maintenance so clearly depends on how the machinery in question has been used, has there been overloading, poor lubrication etc. that has initiated a wear process that will stop the machine in the end. The paper describes a principal wear function that mimic the wear of rotating machinery which in practise takes place very progressively. The use of higher order regression function with limited number of terms is suggested for statistical processing of condition monitoring parameters. This approach enables the saving of the history of measured parameters and also the prognosis of the further development of these measured values and consequently the expected lifetime of the component in question. In order to beable to do the prognosis automatically the use of simplified fuzzy logic is suggested together with hierarchically constructed sub-models which can also include neural networks for handling various process and loading conditions. The successful results of bearing tests are described as an example.