17. Experimental Variability on Modal Characteristics of an In-Situ Pump

Considering that experimental data, usually considered as the reference in test-analysis correlation and model updating purpose, must also be concerned with uncertainty quantification, the experimental variability on modal characteristics of two nominally-identical pumps in industrial environment is observed, via a benchmark between 5 independent teams. Free choice is let to determine the number and the type of sensors, the excitation means, the number of excitation points and the modal identification method. Variabilities relative to the operator and pump specimen are addressed; moreover specific variability relative to the modal identification phase can be quantified using the same experimental test runs.

Due to differences on experimental meshes, from 27 to 138 measurement points, a SEREP expansion on a finite element mesh is first performed, allowing the comparison of experimental modal bases. Global variability observed, based on half paired among identified eigenmodes (within 10 % frequency gap and 50 % MAC criteria), shows a 2 % mean frequency gap and 32 % mean damping ratio gap. Unit-to-unit variability, essentially attributed to different actual bolted torque levels at component connections and pipe-to-pump connections, and realization of the frame fixation on concrete, shows 10 % mean frequency gap and 28 % mean damping gap. Specific variability due to modal identification operator, as a component of the global variability, is still significant with a number of identified modes that varies by a factor of 2; the so paired modes show 1 % mean frequency gap and 24 % mean damping ratio gap.