Study of rough set fault diagnosis system based on wavelet packet analysis

When to construct the intelligent diagnosis system of rotary machine, vibration signals are always flooded by other running spare parts' vibration signals and massive random noises. The wavelet packet analysis was adopted at first in acquisition of fault diagnosis system feature parameters. Each bandwidth of wavelet packet was identical after disassembled. In order to reduce the structure complexity, the application of rough sets in parameter attribute optimization was explored. In order to acquire some knowledge about the reduct set, a hybridization strategy based on genetic algorithm was studied. After reducing the train sample set with rough set method, the unnecessary condition attributes were eliminated, and the inner redundancy of fault diagnosis system condition attributes is revealed. The result of attribute reduction is given finally.