The invention discloses a fault diagnosis method based on displacement vibration of a
machine tool feed shaft, and belongs to the technical field of
machine tool fault diagnosis. According to the method, acceleration sensors are installed on key parts of a
machine tool, sampling frequency is set to collect data, after mean filtering preprocessing, displacement is calculated through integration, data are segmented with the displacement as a horizontal coordinate according to key nodes of a
machining technology or a movement track,
time domain,
frequency domain and time-
frequency domain characteristic parameters are extracted, and differences are compared and analyzed. Meanwhile, when a
machine tool normally operates and a fault is simulated, displacement is used as an index to construct a sample
library to
label a fault type and a displacement interval, a
convolutional neural network is adopted to input a characteristic parameter training model containing displacement information, and a fault characteristic mode is learned to realize accurate diagnosis of the feed shaft fault. According to the method, faults such as bearing abrasion, lead screw looseness and poor guide rail
lubrication can be accurately recognized, the accuracy and real-time performance of
machine tool fault diagnosis are effectively improved, the operation reliability and stability of a
machine tool are enhanced, and the production loss is reduced.