PSI - Issue 54

V.M.G. Gomes et al. / Procedia Structural Integrity 54 (2024) 561–567 Author name / Structural Integrity Procedia 00 (2023) 000–000

566

6

In Fig. 7, the blue region represents the cloud of experimental points. Notice that the relationship between δ v and σ appears not to be linear, however, if a cycle-by-cycle analysis is made, one verifies the existence of hysteresis loops. These hysteresis loops have a higher probability of being due to the strain gauge’s hysteresis signal and not due to intrinsic structural behaviour. This hysteresis e ff ect tends to increase with the increase of loading frequency. Additionally, friction e ff ects do not a ff ect significantly the longitudinal stress measured. Comparing prediction curves with the first recorded point (red point), the numerical curve (green line) with a slope d σ / d δ v = 8.7MPa / mmcan be considered without a large error, 100 MPa, corresponding to 10 %.

M3 – Strain Gauge M31(E1L) b) Leaf Spring M3 – Strain Gauge M32(E2L) Fig. 7. Comparison of the numerical behaviour curve with the cloud of experimental data obtained after filtering, the first measurement point measured in the test and the UIC 517 standard curve.

4. Concluding Remarks

In this paper, a real-time monitoring campaign of the typical loadings and stresses generated in parabolic leaf springs of two-axle freight wagons was performed. Strain gauges and potentiometers were used to gather longitudinal strains and vertical displacements, respectively. The gathered data was used to compare and validate a full-scale numerical model which can be considered in future analysis. With respect to the stress analysis, initially, the data collected during the testing route was compared with the stress displacement curves suggested by the UIC and obtained by the numerical model. A slight deviation of the experimental results was observed both in relation to the UIC curve and the numerical curve, however, this deviation was not greater than 100 MPa, which is equivalent to less than 10 % for the tested load level and therefore the numerical curve σ − δ v is suitable for making future predictions, such as fatigue analysis under random loadings.

Acknowledgements

The authors thank to Research Projects:

• Doctoral Programme iRail - Innovation in Railway Systems and Technologies funding by the Portuguese Foundation for Science and Technology, IP (FCT) through the PhD grant (PD / BD / 143141 / 2019). • FERROVIA 4.0, with reference POCI-01-0247-FEDER-046111, co-financed by the European Regional Develop ment Fund (ERDF), through the Operational Programme for Competitiveness and Internationalization (COMPETE 2020), under the PORTUGAL 2020 Partnership Agreement;