Czesław Żółkiewski

2025-01

Vibration fatigue assessment for engine mount brackets — computed stress spectra from NASTRAN SOL 111 frequency response and converted to equivalent fatigue cycles using Dirlik rainflow counting approximation. Assessed against fatigue allowables.

Challenge: Dirlik approximation for PSD-based fatigue is only valid for narrowband or relatively broadband random loading — the engine mount excitation spectrum had spectral peaks (tonal components) superimposed on broadband. The hybrid tonal+broadband spectrum required separate peak and broadband fatigue damage computation with appropriate summation.

Learned: Vibration fatigue for engine-induced excitation with both tonal and broadband components requires separate damage computations for each component type. The Dirlik method is for broadband random loading only — applying it to a mixed tonal+broadband spectrum underestimates damage from the tonal peaks.

2024-05

Engine-induced vibration assessment for pylon and engine mount — computed forced harmonic response using NASTRAN SOL 111 with unbalance excitation at 1P, 2P, and fan blade passing frequencies. Identified resonance of pylon second bending mode at 1.8P.

Challenge: Pylon second bending resonance at 1.8P fell within the engine operating range at approach power. Tuned the pylon mass distribution (added a tuned mass damper at the pylon tip) to shift the resonance frequency above 2.0P and outside the operating range.

Learned: Engine-induced pylon vibration assessments must cover the full engine operating speed range, not just the design point. Resonances discovered only at approach power (a condition sometimes excluded from early analysis) can drive expensive late-stage structural modifications.

2023-10

GVT correlation — compared NASTRAN SOL 103 modal predictions against LMS Test.Lab GVT measurements. Performed MAC (Modal Assurance Criterion) analysis for 18 correlated modes. Updated FEM to improve correlation for 4 modes with MAC < 0.9.

Challenge: Wing first bending mode frequency from GVT was 7.3% lower than NASTRAN prediction — outside the 5% tolerance for flutter analysis. Root cause: fuel in the wing tank was not modeled in the NASTRAN model as a structural mass, only as gravity load. Adding fuel mass as non-structural mass corrected the frequency to within 1.8%.

Learned: GVT FEM correlation must match the exact mass distribution of the test configuration. Fuel mass, payload, and any instrumentation masses must be included as non-structural mass elements — modeling them only as applied forces is insufficient for frequency and mode shape correlation.

2023-03

Ground vibration test (GVT) planning for a new aircraft variant — defined accelerometer and force transducer placement for 280-channel measurement campaign. Optimized sensor placement using effective independence (EI) method to maximize mode shape observability.

Challenge: Effective independence sensor placement optimization required running 280-choose-N subset evaluations — computationally prohibitive for large N. Implemented a greedy forward selection algorithm that added sensors iteratively to maximize the Fisher information matrix determinant, achieving near-optimal placement in 15 minutes vs. 6 hours for exhaustive search.

Learned: Optimal sensor placement for GVT using exhaustive subset selection is computationally intractable for large sensor networks. Greedy EI forward selection achieves 92-97% of the exhaustive solution quality in a fraction of the time — the standard approach for large-scale GVT sensor placement.