Self-Organising Features of Rotor-Dynamic Faults

Artificial Neural Networks offer an efficient platform for devising condition monitoring strategies in machinery and plants where the number of components and processes are too many and complex to be mathematically modeled appropriately. Self-Organizing Map (SOM) is an interesting artificial neural network algorithm which produces an ordered low dimensional representation of an input data space. The basic Self-Organizing Map (SOM) can be visualized as a sheet-like neural-network array, the cells (or nodes) of which become specifically tuned to various input signal patterns or classes of patterns in an orderly fashion. This work describes a study on application of such topographic mapping for rotordynamic faults. A variety of rotor faults like unbalance, bearing damage, cocked rotor etc., are simulated on an experimental rig in a controlled manner. Vibration Signals obtained from various sensors on the rig are processed to train a Self-Organising Neural Network. It is has been shown that input neurons from the same fault get mapped together in clusters in both two-dimensional and three-dimensional grid spaces and each individual fault occupies a distinct zone on the grid.