SKF Heterogeneous Test-rig Bearing Load Sensing Dataset
收藏NIAID Data Ecosystem2026-05-02 收录
下载链接:
https://zenodo.org/record/14959000
下载链接
链接失效反馈官方服务:
资源简介:
SKF Heterogeneous Test-rig Bearing Load Sensing Dataset
This dataset is developed to estimate bearing loads under various operating conditions (rotational speed, axial and radial loads) using data from temperature and vibration sensors. These sensor modalities provide complementary information: vibration signals indicate the magnitude of the load, while temperature measurements reveal the spatial distribution of the load within the bearing. The dataset emulates a real-world deployment scenario of a virtual sensor, mirroring scenarios where a physical sensor's operational life is limited, such as when a sensor roller collecting field data experiences premature battery depletion. It contains 55 unique operating conditions, defined by axial load (Fx), radial load (Fy), and rotational speed.
Experimental Setup and Data Collection
Data for this study were collected at the SKF Sven Wingquist Test Centre (SWTC) using a face-to-face test rig featuring two identical single-row tapered roller bearings (TRBs). The key characteristics of the TRBs are as follows:
Dimensions:
Outer Diameter: 2,000 mm
Inner Diameter: 1,500 mm
Width: 220 mm
Design:
Each TRB bearing incorporates a rotating inner ring and is equipped with 50 rollers.
Sensor Configuration:
Sensor position:
TB1 and TB2 prefixes indicate the sensor's location on TRB bearings 1 or 2, respectively.
Angular sensor positions are defined relative to the bearing's top position (referred to as 0 degrees).
Temperature Sensors: Ten sensors per bearing. The naming convention for temperature sensors is as follows: TB[1/2]_T_OR/IR_[degrees]. For example, TB1_T_OR_022.5 indicates a temperature sensor on TRB1, located on the outer ring (OR) at 22.5 degrees.
Eight sensors are uniformly distributed on the outer ring (OR).
Two sensors are located on the inner ring (IR).
Temperature is recorded at a sampling rate of 1 Hz with a precision of 0.05°C.
Vibration Sensors: Three sensors per bearing are mounted on the outer ring, capturing both axial (AX) and radial (RA) vibrations. The naming convention for vibration sensors is as follows: TB[1/2]_VB_[degrees]_[AX/RA]. For example, TB1_VB_006_AX represents vibration in the axial direction, measured by a sensor installed 6 degrees.
For the radial direction, sensors are placed at the top and bottom of the bearing housing.
For the axial direction, sensors are placed on the side face of the bearing
Vibration data is aggregated to 1 Hz using Root Mean Square (RMS) calculations.
Load Measurements:
Multiple load cells measure and control Axial (AX) and radial (RA) loads.
Aggregated load values in both directions serve as ground truth for this study.
Data Preprocessing
Data Filtering:The raw sensor data was filtered to retain only continuous, stationary operating periods lasting between 20 minutes and 1 hour, resulting in a final dataset comprising 164,948 samples.
Temperature Gradient:To isolate immediate load and speed changes from cumulative thermal effects and ambient temperature influences, we calculated temperature gradients. These gradients, preprocessed before data filtering, provide an instantaneous response to operational variations. A 1-minute moving average filter was applied to reduce noise, and the rate of temperature change was calculated over 5-minute periods to capture the immediate impact of load variations. Temperature gradient data is identified by the DIFF_ prefix, e.g., DIFF_TB1_T_OR_022.5 for the sensor TB1_T_OR_022.5.
Temporal Train-Test Split:The dataset was split by time while maintaining representation across operating conditions. Within each operating condition:
The initial 50% of sequential data was allocated to the training sets.
The remaining 50% formed the test set, representing the period during which the virtual sensor would be actively predicting bearing loads.
创建时间:
2025-03-10



