Comparison Between Numerical and Experimental Static Performance and Sensitivity Study on a Tilting Pad Journal Bearing in a Load on Pad Configuration

Computational software based on Reynolds equation is widely used in the design of tilting pad journal bearings, its accuracy generally satisfactory for ordinary operating conditions. Computational models can be particularly useful to quickly quantify the sensitivity of certain bearing parameters to manufacturing tolerances, errors in assembly, and uncertain operating conditions. In this paper, predictions obtained from a thermoelastohydrodynamic model are compared against experimental data procured from static load tests. The flooded configuration test bearing is a four-pad, load-on-pad, with centered pivots and 0.3 pad preload. The shaft angular speed reached 12 krpm (surface speed 64 m/s) with a maximum unit load of 2.0 MPa. The supply oil flow rate varied from 50% to 150% of its nominal value. The model predictions with nominal input data fall short of reproducing many of the measured results. The largest differences amount to 8 ℃ in the loaded pad peak temperature, 23 μm in bearing eccentricity and 5.9 kW in power loss. A sensitivity study to changes in the pad geometry, clearances and fluid viscosity produced variations of up to 4 ℃ in pad peak temperature, 12 μm in bearing eccentricity, and 1 kW in drag power loss, hence reducing the gap between predictions and experimental results. The study reveals that, to improve the agreement between numerical predictions and experimental results, the model uncertainty should always be quantified.