On the Applicability of Several Models of Cylindrical Roller Bearings, an Experimental Study

Purpose: Equivalent modeling and nonlinear dynamics of cylindrical roller bearings under variable stiffness conditions are investigated, with a focus on evaluating the applicability of classical and alternative stiffness models. Methods: Experimental measurements of stiffness for the NU1006 cylindrical roller bearing are conducted. Multiple models (Hertz model with 10/9 exponent, exponential model, polynomial model) are established and compared against experimental data. Nonlinear factors including bearing clearance and variable stiffness are incorporated. System vibration characteristics are analyzed through stiffness curve fitting and rotor system resonance frequency prediction. Results: The accuracy of the Hertz model within a certain load range was validated. When the external load does not exceed 40% of the rated load, the Hertz model, exponential model, and polynomial model can accurately describe the system’s vibration characteristics. However, when the external load exceeds 40% of the rated load, the stiffness curves of both the Hertz model and exponential model show significant distortion, whereas the polynomial model can still accurately fit the experimental data and precisely capture the primary resonance frequency of the rotor system during varying compliance (VC) resonance. Conclusion: Linearized models inadequately represent cylindrical roller bearing stiffness characteristics. For external loads ≤40% of rated load, the Hertz, exponential, and polynomial models all effectively describe vibration behavior. Above 40% loading, the polynomial model proves essential for accurate stiffness characterization and VC resonance prediction. These findings establish load-dependent guidelines for bearing model selection in dynamic analyses.