An Experimental Setup to Detect the Crack Fault of Asymmetric Rotors Based on a Deep Learning Method

Crack is a common fault of rotor systems. The research on crack detection methods mainly divided into numerical and experimental studies. In research, the current algorithms based deep learning are mostly applied to bearings gearboxes, there few studies diagnosis. rotors used in an experiment single-span rotors. However, complex structures such as multi-span systems actual industrial field. Thus, that have been successfully not verified To obtain signal close asymmetric shaft system validate method, platform designed built independently. We measure vibration signals three channels under five working conditions establish intelligent method for location residual network. factors influence performance analyzed, laws discussed. Results show accuracy anti-noise proposed higher than those commonly machine learning. average 100% when SNR (signal-to-noise ratio) greater or equal −2 dB, 98.2% −4 dB.