PSI - Issue 58

J.R. Steengaard et al. / Procedia Structural Integrity 58 (2024) 61–67

65 5

J.R. Steengaard et al. / Structural Integrity Procedia 00 (2024) 000–000

Fig. 6. Strain values from the experiment, the original FE model, and the updated FE model. The curves show individual measurements. The lines are added for easier comparison.

3.2. FE model updating

To reduce the discrepancies between the FE model and the experimental data, parameter based FE model updating has been performed. Various parameters are updated, and the e ff ects of the updates are evaluated based on the MAE. The updated parameters and the resulting MAE can be seen in Tab. 1.

Table 1. Parameters used for FE model updating. Numeric parameters have been increased and decreased. Thus, lower and upper values for MAE are given in these cases.

Parameter

Description

MAE lower

MAEupper

Young’s modulus Spring sti ff ness

Decreased and increased

0.0747 0.0747 0.0578 0.0660 0.0740 0.0755 0.0744 0.0748 0.0651

0.0747 0.0747

Decreased and increased, all spring elements having the same sti ff ness

Spring face deformability

Spring attachment faces made deformable instead of rigid Attachment faces changed from inside to outside of beam

Spring faces

Connection, disc-beam Changed disc to beam connection face

Connection, feet-beam

Changed feet to beam connection type

Weld lengths

Partially welded instead of fully welded according to approximate measures

Thickness

Decreased and increased thickness of beam

0.0748

Strain gauge locations

Local strain gauges moved in FE model according to approximate measures

Normally, model updating is made as an optimization, where the parameters are optimized based on a given ob jective. In this case however, the parameters that significantly reduce the MAE are not numerical values, but instead binary options. Thus, the problem becomes combinatoric instead of continuous. For this reason, the FE model is up dated by implementing all updates that individually reduce the MAE significantly. Therefore, the implemented updates are: spring face deformability, spring faces, and strain gauge locations. Other parameters such as misalignment have not been investigated because no estimates of such are available to the authors. Also, the updated model is accurate after implementation of the mentioned updates. The attachment faces are changed from the inside to the outside of the cutterbar beam. This makes sense, as the bearings are clamped by tightening bolts that press on the outside faces of the cutterbar beam. Furthermore, changing the locations of the local strain gauges makes the FE model more accurate. The local strain gauges were specified to be placed relative to the weld toes, but the location of the weld toes varies from each produced unit. For this reason, the original locations of the local strain gauges in the FE model were not correct since the locations of the weld toes, and therefore the local strain gauges, were not known. The strain values from the updated model can be seen in Fig. 6 along with the experimental strains and the strain values from the original FE model. The MAE is reduced to 0.047 for the updated model, or 4.7 %. Based on the