SCA bearing dataset

The dataset contains vibration measurements from a pulp mill collected between 2019 and the end of 2022 that describe different health states of bearings. It comprises 11 cases each, except case 11, describing a bearing fault before the bearing was changed. Case 11 contains a fault that is not related to the bearing which can be used to test the tendency of methods to generate false alarms. Each case includes two files: one with normal data called “train” and one with only a fault or both a fault and normal data depending on fault development time named “test”. The files contain approximately four months of data with one or more measurements per day. The files are in .mat format and include the vibration data, sampling rate, bearing type, the fault frequency of each bearing, the time for each measurement, the rotating speed of the shaft and the labels. The possible values of the label are -1, which means that the machine is off or that the speed data of the shaft is missing, 0 for normal conditions, 1 for inner ring fault, 2 for ball fault and 3 for outer ring fault. For more information, we refer to the published article. When referring to this dataset, please consider citing both the article and the dataset.

Note regarding version 2:
We discovered that the non-bearing related fault scenario (case 11) might not be related to any fault in the equipment. Because of this, we decided to change it to a confirmed fault scenario diagnosed as shaft misalignment. Since this change is not related to the bearing fault scenarios and the labelling is set to 0 for both versions 1 and 2 it should have no impact on research that has used the dataset for bearing fault diagnosis.