Monitoring of Lubricating Conditions With a Neural Network Pattern Recognition Technique

This paper shows an application of neural networks for monitoring purposes regarding the lubrication within a squeeze film dampers for rotor support. A self-organizing map (SOM) is adopted for identifying the aspect of the pressure waves within the oil film. The information contained in the wave patterns can be put in relation with a number of factors which affect the damper operation, like rupture in the film, inertia effects, etc.. A prompt detection of possible changes in the characteristics of lubrication and damper operation turns out to be, in fact, an important feature of surveillance procedure, so as to allow, for instance, suitable choices in case theoretical models are adopted for reference within the same procedure. A benchmark program for testing the capabilities of the SOM has been carried out with recourse to a theoretical data set. A separate test with further adoption of experimental data has confirmed the potential of the present approach.