Vibration-based crack identification in rotor shafts using optimized experiments and physics-based simulation

This paper investigates crack detection in rotating machinery through integration of three complementary methods: Taguchi design of experiments (DOE) for parameter optimization, Fast Fourier Transform (FFT) spectral analysis for crack signature identification, and ANSYS finite element analysis (FEA) for experimental validation. Systematic experiments were performed on steel shafts with three diameters (20, 30, 40 mm) and crack depths (25%, 37.5%, 50% of diameter) across multiple crack configurations (1–3 coplanar cracks) at rotational speeds from 1,500 to 6,000 RPM. Key findings include: (1) shaft diameter dominates vibration response (38% effect magnitude via ANOVA); (2) counter-intuitively, multiple coplanar cracks reduce vibration amplitude through stiffness redistribution and critical speed shifting; (3) second harmonic (2X) amplitude scales quantitatively with crack severity, providing diagnostic metric; (4) FEA simulation matches experimental frequencies within 9% average error across 27 trials; (5) discrete Taguchi analysis masks resonance phenomena captured by continuous modal analysis, highlighting complementary nature of DOE and modal methods. The integrated approach reduced experimental trials by 67% (from 81 to 27 cases) while achieving 98.94% prediction accuracy via regression modeling. While rotational speed showed minimal main effect in Taguchi analysis (p = .854), detailed Campbell diagram analysis reveals critical speed ∼2,850 RPM with >10× amplitude magnification, emphasizing need for multi-method validation in complete dynamic characterization.