FM-CGFCN

The experimental setup for collecting the Mechanical Fault Simulator (MFS) dataset is illustrated in Fig. 8. It primarily consists of a drive end, an acceleration sensor, an AC motor, and an signal controller. The sampling frequency is 15.6 kHz. This dataset records four bearing fault operation signals: ball fault signal, inner ring fault signal, outer ring fault signal, and combined inner and outer ring fault signal. Based on the location and size of the bearing damage, the dataset is categorized into four fault types, with the label correspondence provided in TABLE III. In this study, the vibration signal obtained under a rotation speed of 1130 r\/min is selected as the experimental data. To address limitations in sensor data acquisition, a sliding window of 1024 points is used to collect samples. We collect 800 samples from 4 fault types. Among them, 500 samples are used for model training, 150 samples are used for model testing, and 150 samples are used for model verification.