A Method of Bearing Fault Feature Pattern Recognition Based on Improved ACGAN

Abstract

As a key component of rotating machinery, bearing plays an irreplaceable role in the operation of rotating machinery. The ability to identify bearing faults effectively and timely can ensure the safe operation of the equipment. In this paper, a logic diagnosis method framework of bearing fault feature pattern recognition was proposed by using ACGAN model structure. With the same excellent learning efficiency, the multi-layer convolution layer structure was used to ensure the learning ability of the network. Finally, a series of experiments were conducted. Experiments’ results indicated that compared with a single CNN, the improved ACGAN network architecture had better learning ability and fault state recognition rate.