Applications of Neural Networks in Fault Detection of Rotating Machinery.

摘要

The purpose of this research was to design a neural network based fault diagnosis system that is capable of detecting and classifying incipient faults in rotating machinery. A neural network is essentially a pattern recognition system which produces a mapping from a set of input data to a set of output data. Neural networks are unique in that this mapping is created autonomously, based on a learning algorithm that the user specifies. In this research, the mapping is from a set of measured parameters (e.g., vibration spectrum) of the rotating machinery to a classification of the system's condition (e.g., worn bearings). Limited success has been achieved in this area over the past decade. Research to date indicates that neural networks have the ability to recognize faults in machinery. This research focused on the following objectives: (1) creating a system that can recognize basic fault conditions based on the fault's vibration signature; (2) improving the ability of the neural network to recognize transient conditions as normal rather than classify them as faults; (3) examining the effect that disturbances have on the neural network's output; and (4) developing a system that will enable detection of faults which are not included in the training set. Neural networks, Fault detection, Transients, Backpropagation.