A model-based technique to identify lubrication condition of hydrodynamic bearings using the rotor vibrational response

The paper clearly formulates the four-feature identification (fb1, fb2, φ1, φ2), uses a 3×3 multi-start grid, bounds Q in [0.25 QT, 2 QT], builds a central-difference Jacobian, and solves a generalized Newton search, reporting errors ≤ 3% at SNR ≈ 12 dB with both in-bearing and out-of-bearing measurements . It also documents why the fb ratio is preferred (monotone) and why φ is complementary but can have multiple roots and even create a spurious φ-only intersection that must be rejected by the total error criterion . However, the paper does not provide a rigorous identifiability proof, nor does it specify precisely how the overdetermined Newton step [J] s = −S is solved (least squares, pseudoinverse, etc.) despite using four errors (n = 4) with two unknowns . The model’s solution supplies a clean mathematical route (Gale–Nikaidō via P-matrices, Gauss–Newton/LM, a posteriori certificate using σ_min(J) and residual), but its injectivity claim for the φ-submap is inconsistent with the paper’s observed closed φ-level sets and multiple φ-root intersections; those contradict global P-matrix conditions that the model assumes without verification. The certificate also relies on unverified Lipschitz/small-noise constants. Thus, the paper is methodologically sound but theoretically incomplete; the model adds valuable theory but over-asserts injectivity and leaves key assumptions unvalidated.