Experimental vibration time-series data for rotor crack detection and condition monitoring

Abstract

Condition monitoring and fault diagnosis in rotating machinery require high-quality datasets that capture the progressive evolution of structural defects under controlled conditions. This paper presents a comprehensive dataset acquired using SpectraQuest MFS-Lite, designed to study the dynamic behavior of rotating shafts with transverse cracks. The dataset includes measurements from shafts with ten crack conditions focusing on both centrally and laterally located progressive shaft cracks: a healthy state and nine progressive crack levels. These fault levels correspond to relative crack depths ranging from 4.15% to 50.00% of the shaft diameter, which translates to a cross-sectional area reduction from 1.36% to 50.00%. Experiments were conducted under three different rotational speeds (20 Hz, 40 Hz, and 60 Hz) to capture the influence of operating conditions on the vibration response. Time-series data were recorded using a single uniaxial accelerometer at a sampling frequency of 6,000 Hz, with each individual signal containing 16,384 data points across a duration of 2.73 seconds. To guarantee statistical repeatability, a minimum of 1,000 independent repetitions were acquired for each crack level and speed configuration, and 2,000 repetitions for the healthy baseline, all archived in standard CSV format. For each crack level and rotational speed, vibration signals were recorded using a consistent experimental setup to ensure repeatability and comparability across measurements, comprising records from two distinct shafts to evaluate the physical influence of both middle and lateral crack positions. The progressive crack depths enable the analysis of fault severity and its effect on the dynamic response of the system, while the multiple rotational speeds allow the investigation of speed-dependent fault signatures. This controlled framework provides a valuable quantitative benchmark for developing, validating, and comparing signal processing, machine learning, and physics-based approaches for crack detection and severity estimation in rotating machinery.