Dessi, D., D Orazio, D.

CORRELATION OF MODEL-SCALE AND FULL-SCALE DATA: SENSOR VALIDATION AND ELASTIC SCALING EVALUATION Dessi, D., D Orazio, D. INSEAN-CNR Rome - Italy 1

Project structure hydroelastic side This work was funded by the Italian Navy within the cooperative research project MOU-6dof-RANS, whose aim was the development of codes for the prediction of the ship behavior in wavy seas Full-scale trials Scaled model tests Main investigation areas 3D-FE Analysis Equivalent beam simulations

Project structure hydroelastic side Full-scale trials 3D-FE analysis response prediction & comparison Scaled model tests Equivalent beam simulations

Motivation of this work Full-scale measurements will be more available in the future as long as monitoring systems are installed on-board. Is it possible to validate the codes, or at least the scaled physical modes, with the full-scale measurements? In this presentation, with reference to global loads (VBM) we will focus on: full-scale data validation and/or error recovering analysis of intrinsic differences between the ship and the scaled model

Full-scale measurement Motion reference unit for 6dof motions GPS for trajectory meas. WAVEX radar for wave field recongnition 3 sg x 5 sections for bending moment distribution Italian Navy patrol vessel Lpp = 80 m Wave finder for relative wave meas. Bow panel strain gages for slamming impulse loads

Reduction to 1D ship beam Real ship Real Ship 3D FE model (PATRAN) 1D Equivalent Structure Reference Ship Data mass distribution bending stiffness distribution shear area distribution2 Ship data for validation Segmented model modo.jpg natural frequencies, mode shapes (NASTRAN SOL 103)

Scaled model-tests

Comparison between full-scale and scaled meas. Under which assumptions, scaled tests and full-scale trials can be compared? Compared quantities has to be homogeneous Strains can not be the same, better use of VBM in the same locations Ship has to be in the same loading conditions Unexpected heel angle of about 2 degs could not be reproduced with the experimental set-up. Speed and encountered sea are the same The wave-maker reproduced the same sea-states in the towing tank. However the linear towing-tank basin imposed to ensure head and following wave conditions in the ship trials by cruising at given angles with the prevalent sea direction. Use of RAO allows to minimize 2 uncertainties in the test input conditions (RAO = ) M h B f s L pp

Practical determination of bending moments However, unlike rigid-body motion, their determination is not straightforward at full-scale and may be affected by several error sources. Elastic model strain-gage calibration Theoretical shear force and bending mom. M f Calibration factor k =M f /V Given force input V Meas. output voltage

Practical determination of bending moments Ship strain-gage calibration We have to rely on the virtual calibration of the FE model k =M f /V

Practical determination of bending moments Thermal strain Local modes Global modes other than vert. bending ones Strain-gage Calibration rod sg Strain gage V k 1 Device block sg Structural element loc glob Physical block vbend vbend k 2 M f FEM model y measured vbend k =M f /V = k 1 k 2 M E I / I yy z P F d FP z yy sg calculated / P Strain gage position

Previous results Preliminary results were not so satisfactorily [HYEL09], even if: VBM is dependent on the longitudinal ship loading that, in the same sea state, should be close between model and full scale. Let us observe: M E I / z y vbend yy sg M F d if P sg y FP I z F d / yy P FP P RAO (Response Amplitude Operators) were computed using preliminary wave spectrum data in low sea states ensuring linear response regime However, in practice: Bending moment is dependent on sensor calibration and expected deformations Computation of RAO is sensitive to inappropriate wave spectrum model

Comparison between full-scale and scaled meas. Determination of k 1 (sensor calibration) does not present any particular difficulty Assessment of the disturbances and the strain - VBM relationship (k 2 ) must be investigated Thermal stresses due to reasonable thermal loading induce negligible structural deformations T=15-30 Effects of global unsymmetric loading with respect to the longitudinal vertical plane are partially filtered out taking the average between the port and the starboard strains

In the past, we have used several techniques to identify ship elastic modes based on output-only approach (no-input measurement, ambient excitation) that: Frequency domain decomposition [JSR 2008] Proper Orthogonal Decomposition [JFS 2012, ] In this case, a less straightforward technique in the time domain known as Stochastic Subspace Identification is used because of noise on the data (it requires a certain expertise on modal analysis and a feeling on expected results) Identification technique

Strain-modes analysis Modes obtained with SSI in terms of strains: points in the top side refer to upper deck measurements points in the bottom side refer to the keel measurements Modes are poorly described

Comparison between full-scale and scaled meas. Can we trust on the data? Is there any chance for an internal data validation? The idea is to find a tool for independent validation / recovering of the data based on data internal coherence Criteria for internal coherence of strain-gage / VBM results: if strain measurement are correct, strain modes should appear an acceptable VBM RAO variation along the ship sections has to be preserved

Sensor validation / Data recovering Timoshenko beam parameters from 3D-FE model Experimental strain mode identification with SSI Strain vibration mode extraction from 1D Timoshenko beam Mode shape evaluation on strain-gage points Same eigenvectors normalization Comparison between Num. & Exp. 1-st bending mode Correction of strain-gage calibration constants (only on faulty sensors) Correlation between full-scale and model scale experiments is improved Check box Re-extract experimental modes Orthogonality among exp. modes has improved? Comparison between Num. & Exp. Is enhanced?

Strain-modes analysis via SSI Next, it is assumed that the local strain effect is proportional to the structural loading, i.e., to the global response represented by the VBM strain locmodes Thus, since,, using the previous relationship yields: sg vbend sg locmodes vbend ( 1 ) vbend vbend It is reasonable to assume that other effects can be included in the calibration coefficients ( / Volt) of the strain gages. vbend The correction factor for the calibration coefficients is defined as vbend s (1 ) sg sg k First mode ( num) sg (exp) sg ( x) ( x) ˆ (exp) sg Overall strain gage time-history ( x i, y i, z i, t) k (exp) sg ( x i, y i, z i, t)

Strain-modes analysis via SSI Modes shapes Modal assurance criterion Validation Check

Comparison between full-scale and scaled meas. The comparison in terms of the response amplitude operator of the vertical bending moment is more reliable than the direct comparison of the outputs M h B f s L 2 pp f * e f ( L / g) e pp 1/ 2

Comparison between full-scale and scaled meas. The comparison in terms of the response amplitude operator of the vertical bending moment is more reliable than the direct comparison of the outputs M h B f s L 2 pp

Comparison between full-scale and scaled meas. The comparison in terms of the response amplitude operator of the vertical bending moment is more reliable than the direct comparison of the outputs M h B f s L 2 pp

From scaled to full-scale tests Also error sources inherent to the intrinsic differences in the physical models has to be considered using a chain of numerical models representing just one difference at time from the segmented elastic model to the real ship Analysis of the effects due to the load segmentation Analysis of the effects due to the difference between the model backbone and the ship beam Analysis of the effects due to the differences between the ship-beam and the 3D-FE model

Num. vs. exp. (true ship) Lack of correspondence may depend on several reasons: The ship is a quite complex and huge structure that may suffer of: design modifications difficult in modelling some ships details Fot the wet real ship, estimation of damping is a critical point Ship on-board measurement 3D-FE simulation Non-dimensional VBM 0 0 0 5 10 15 20 25 30 35 Non-dimensional time In this comparisons, quite the same load has been applied.

Concluding remarks Use of 3D finite element modeling: Extraction of Timoshenko beam parameters Strain-gage calibration Vibration modes Time-domain simulations Sensor validation / Data recovery by: comparison with numerical model on 1 st mode shape checking over the other identified modes (shape and orthognality)

Concluding remarks Thank you for your attention References: 1. Coppotelli, G., Dessi, D., Mariani, R. and Rimondi, M. (2008). "Output-only analysis for modal parameters estimation of an elastically scaled ship" in Journal of Ship Research. 2. Dessi, D. and D'Orazio, D. (2010). "Analysis of the vessel global loads via ship trial investigations and segmented-hull tests in 28th Symposium on Naval Hydrodynamics, Pasadena (CA). 3. Mariani, R. and Dessi, D. (2012). "Analysis of the global bending modes of a floating structure using the proper orthogonal decomposition" in Journal of Fluids and Structures.