Vibration time-frequency images of wind turbine planetary gearboxes

Citation Author(s):
The School of Mechanical Engineering, Shenyang University of Technology
The School of Mechanical Engineering, Shenyang University of Technology
Submitted by:
pinyang zhang
Last updated:
Mon, 05/09/2022 - 05:22
0 ratings - Please login to submit your rating.


The vibration signals used in this dataset were collected from 1.5 MW wind turbines in June 2020. The wind farm locates in eastern China. The planetary gearbox contains two-stage planetary and one-stage parallel transmission mechanisms. Vibration signals were acquired using a piezoelectric accelerometer (VIB 6.140) mounted on the ring of the second-stage planetary mechanism of the gearbox by a DAQ system with a sampling frequency of 16384 Hz. There are four health conditions in the dataset: normal, ring gear tooth pitting, planetary gear meshing misalignment, and pitting-misalignment concurrence. All considered fault conditions occurred on the second-stage planetary mechanism. Each health condition corresponds to four gearboxes. We used the generalized S-transform algorithm to generate vibration time-frequency maps and labeled each one of them.


Original vibration signals last for 10 seconds. We randomly intercepted ten segments from each vibration signal. The generalized S-transform (generalized factor = 2) was used to generate 512×512 size time-frequency spectrums from signal segments. The characteristic patches or bands were highlighted with a gray level of 255. The gray level of other parts was set to 0. All images in 'time_frequency_image' folder correspond to those in 'label' folder one by one.