PU Dataset Overview

Dataset Introduction

The PU Dataset, provided by Paderborn University, contains bearing fault signal data of various types, such as inner‑ring, outer‑ring, and rolling‑element faults. Its distinguishing feature is the inclusion of extensive motor‑drive‑system fault data, making it a comprehensive experimental platform for bearing fault diagnosis research.

Dataset Structure

The PU Dataset is divided into multiple subsets, each containing different fault types and experimental conditions. Key subset descriptions:

• Faults originate from manual causes and accelerated life testing.
• The modular test rig consists of five main components: torque shaft, bearing test module, flywheel, motor, and load motor.

Experimental Conditions

Experimental conditions such as N15_M01_F10 denote a bearing speed of 1500 rpm, radial load of 1000 N, and system load torque of 0.1 Nm. Each condition is repeated 20 times, sampled at 64 kHz, with each acquisition lasting 4 seconds.

Data Collection Method

Vibration signals are collected using high‑precision accelerometers and recorded via a data acquisition system. Various operating conditions (speed, load, lubrication, temperature) are set to ensure data diversity and representativeness. All data are pre‑processed to remove noise and interference.

Data Format

Data are typically stored in MAT files, each containing one or more channels of vibration signal data. Each channel presents a time series of acceleration amplitudes. Files also include metadata such as fault type, speed, and load.

Application Scenarios

The PU Dataset is widely used for rotating‑machinery fault diagnosis, condition monitoring, and predictive maintenance. Researchers develop efficient fault‑diagnosis algorithms by analyzing vibrational features, enhancing equipment reliability and safety. Recently, the dataset has also been employed to train and test machine‑learning and deep‑learning models, contributing to the advancement of intelligent fault‑diagnosis technologies.