Experimental determination of both steady-state and dynamic performance of a journal bearing requires the use of a high precision spindle with a vanishingly small range of run-out. This was achieved by first eliminating the mechanical run-out of the spindle by grinding the journal specimen while rotating in place. Once the mechanical run-out was removed, the electrical run-out sensed by the displacement proximity-probe-transducers was also removed. Using this procedure the mechanical and electrical run-outs of a research spindle were reduced to less than 0.2 micron (10 μin.), which is better than the resolution of the data acquisition system, 1 micron (50 μin.).

This paper presents laboratory experiments that confirm and further explore earlier computer simulations related to oil-whip and rub-impact. Both phenomena were explored using Spectrum Analysis, and Chaos tracking techniques (e.g. Poincare’ maps). The results show that oil-whip induces quasi-periodic rotor vibrations, while rub-impact generates complex dynamical behavior, such as chaotic vibrational motion. The nonlinear oil-whip hysteresis loop results presented confirm phenomena uncovered in previous computer simulations. A Sommerfeld number-consistent “instability threshold load” was experimentally observed, along with its corresponding hysteresis loop similar to the classic instability speed hysteresis loop, but probably the most important experimental discovery was a second Hopf bifurcation-Saddle node instability hysteresis loop (i.e., at higher speed than the classical hysteresis loop), separated by a region of stable-in-the-small operation. Another result of especially high practical importance, was the discovery that spectrum analysis of speed-up, or coast-down rotor-vibration data can be effectively used to locate the saddle node speed.

This work formulates the problem of incipient fault detection and diagnostics for rotating machinery in a statistical model-based framework. This includes problem description, modeling of rotating machinery and fault mechanisms, formulation of the detection and diagnostics problem, and an implementation of the proposed scheme in a simulation environment to test the feasibility of this approach. More specifically, a multiple model nonlinear filtering algorithm is proposed for fault detection and diagnostics in a statistical framework. A simulation study, which includes normal and different fault modes, illustrates the potential of the proposed approach, especially in the presence of measurement noise and process uncertainty.