Universal Machinery Fault Diagnosis Dataset (UMFDD) A Unit Normalized Multi Signal Repository for Machinery Fault Diagnosis

The Universal Machinery Fault Diagnosis Dataset (UMFDD) is a multi signal repository created for machinery fault diagnosis. This dataset integrates multiple publicly available and experimentally collected datasets into standardized units allowing for cross domain learning and evaluation. All signals are converted into a physically consistent representation through unit normalization. Vibration signals originally measured in acceleration m/s^2 are transformed into velocity m/s while acoustic signals are converted from voltage to sound pressure and subsequently to particle velocity. This allows for comparability across different datasets. The dataset is organized into four primary fault categories 0_healthy 1_bearing_faults 2_gearbox_faults 3_induction_motor_faults Each data file is stored in CSV format with a consistent structure. Column 1 represents Source 1 and Column 2 represents Source 2. Metadata includes dataset origin, sampling frequency, and operating conditions. Signals are structured to support consistent multi source analysis across different machinery types. The dataset allows for the evaluation of machine learning models under varying operating conditions, sensor configurations, and fault types within a normalized dataset. The datasets used to create this dataset include Case Western Reserve University (CWRU) [1], University of Ottawa Electric Motor Dataset Vibration and Acoustic Faults under Constant and Variable Speed Conditions (UOEMD) [2], University of Ottawa Rolling Element Bearing Dataset (UORED) [3], Huazhong University of Science and Technology Bearing Dataset (HUST) [4], Induction Motor Vibration and Acoustic Cellular Device Dataset (IM-VACD) [5], Multi mode fault diagnosis datasets of gearbox under variable working conditions [6]. References [1] Case Western Reserve University Bearing Data Center. Bearing Data Center Dataset. Case Western Reserve University. Link: https://engineering.case.edu/bearingdatacenter/download-data-file [2] University of Ottawa Electric Motor Dataset Vibration and Acoustic Faults under Constant and Variable Speed Conditions. Mendeley Data. Link: https://data.mendeley.com/datasets/msxs4vj48g/2 [3] University of Ottawa Rolling Element Bearing Dataset Vibration and Acoustic Fault Classification. Mendeley Data. Link: https://data.mendeley.com/datasets/y2px5tg92h/5 [4] Huazhong University of Science and Technology Bearing Dataset. Mendeley Data. Link: https://data.mendeley.com/datasets/cbv7jyx4p9/3 [5] Induction Motor Vibration and Acoustic Cellular Device Dataset. Mendeley Data. Link: https://data.mendeley.com/datasets/yc8yhg5xjd/1 [6] Multi mode fault diagnosis datasets of gearbox under variable working conditions. Mendeley Data. Link: https://data.mendeley.com/datasets/p92gj2732w/2

搜集汇总

背景与挑战

背景概述

Universal Machinery Fault Diagnosis Dataset (UMFDD) 是一个多信号机械故障诊断数据集，通过单位归一化将来自多个公开数据集的振动和声学信号转换为物理一致表示，包含健康、轴承故障、齿轮箱故障和感应电机故障四类，支持跨域学习与机器学习模型评估。