Fig. 9. Calibration test. (a) Cylinder/cylinder. (b) Hertz contact pressures.
Table 1
Hertz contact theory cylinder/cylinder.
FH (N/mm) Force per unit length
lH (mm) Contact length lH =
4FHRs
1− 2
E
H (rad) Contact angle H ∈
−lH
Rs
, lH
Rs
H (MPa) Contact pressure H
rr
= 2FH
(lH)2
(lH)2 −
R2
s H
Table 2
Calibration coefficients.
Calibration coefficient Data acquisition
station (960 Hz)
Data acquisition
station (3378 Hz)
K45 4.7 1.44
K 1.4 0.96
Krr – –
the relative error. The computed calibration coefficients are listedin Table 2. It should also be mentioned that measured signals are notperfectly synchronized with computed strains at the inner radiusRm. There is a slight time difference between measured strains εm45and εm. This is likely due to cumulative effects of a slight mispo-sitioning of the circumferential optical fiber, that may have notbeen glued perfectly at the bottom of the hole, and a slight mis-alignment during the insertion of the plug inside the roll. Thus,FBGs (circumferential and at 45◦) are not perfectly aligned alongthe axial direction and do not enter the roll gap simultaneously asshown in Fig. 10. Calibration tests enable to quantify this misalign-ment and dmiss
1.74 mm where dmissdenotes the distance alongthe circumferential direction between FBGs measuring εm45and εm.Therefore measured signals can be shifted in order to correct thisFig. 10. Mispositioning of the circumferential optical fiber and plug misalignmentonce inserted into the work roll.
Table 3
Geometrical and material parameters.
Rs (mm) 195.5 Outer radius
Rm (mm) 193.325 Inner radius
d (mm) 2.175 Depth
E (MPa) 210,000 Young modulus
(–) 0.3 Poisson ratio
L (mm) 300 Roll width
l (mm) 100 Strip width
misalignment and obtain a good overlap between computationsand measurements.The comparison between direct computations and measure-ments for the test at 75 tons is presented in Fig. 11 for the dataacquisition station at 3378 Hz (very similar results are obtained for25 tons and 50 tons and for the other data acquisition station). It isclear that εm45and εmare consistent and well calibrated, the strainpeak is accurate and very good overlap between computations andmeasurements is obtained after shifting one signal in order to cor-rect the misalignment of the plug. However εmrrobviously presentsa lack of trueness (Joint Committee for Guides in Metrology, 2012)because the peak spreads on a much larger angular zone thanexpected. Probably, this FBG is not perfectly glued at its expectedposition in the hole, or residual air voids in the vicinity of the FBGcompromise measurements, but other explanations are still pos-sible. Most of the time, when inconsistent data are measured, theinverse interpretation needs a projection of the inconsistent inputson the set of consistent data. In the presented experimental tests,it has been considered simpler to remove the signal εmrrfrom theinputs because the simple adaptation of the inverse method pro-posed by Weisz-Patrault et al. (2011) presented in Section 2 wasavailable. The radial signal εmrris not used in the following.5. Pilot rolling testsSeveral tests have been performed at the ArcelorMittal researchcenter at Maizières-lès-Metz on a cold pilot rolling mill. For all tests,strips of low carbon steel grade (thickness = 2.8 mm) and strips ofaluminium killed grade (thickness = 0.75 mm) and hardened by aninitial cold rolling reduction have been used and lubrication wasmade by emulsion. During trials various rolling speeds, strip exittensions and thickness reduction ratios were tested. Comparisonsbetween measurement inverse calculation and LAM3 numericalsimulations (Hacquin, 1996) are proposed. The geometrical andmaterial parameters for all different tests are listed in Table 3. Itshould be noted that thermal conditions are very homogeneousand thermal compensation is negligible for these isothermal tests.However, thermal compensation will be necessary for different