PSI - Issue 57

Sven Maier et al. / Procedia Structural Integrity 57 (2024) 731–742

741

11

S. Maier et al. / Structural Integrity Procedia 00 (2023) 000±000

For the sake of clarity, values below the failure criterion are not highlighted in red in this graph. Furthermore, the experimental results are also visualised within the plots to show the enhanced accuracy of the improved simulation run. It can be clearly seen that the scatter band of the distributions for the oscillation and fatigue prediction could sig nificantly be reduced due to lower standard deviations of the parameters obtained by measurements. Attribute scatter of the investigated parts is low and uncertainty originating from manufacturing is low. Simulation fidelity is signifi cantly improved using experimentally determined parameters. Simulation without measured uncertainty considerably overestimates durability for most simulations while simulation with consideration of measured uncertainty shows bet ter agreement with experiments. This improvement can mainly be explained by the measured damping characteristic, as the damping was assumed to be significantly higher (3% instead of 0.23-0.6%) but has a strong influence on the fatigue of the critical areas. Moreover, a correlation between the 1st eigenfrequency f 1 of the first mode and the dam age is observable, especially in Figure 9. Both simulation runs with and without measured data show for a decreasing eigenfrequency lower fatigue lifetimes. This is because of the used PSD load which contains a higher energy content with decreasing frequency. This correlation can also be seen in the di ff erences between the variants. Variant 2 and 3 have lower eigenfrequencies and also lower fatigue lifetimes than the nominal variant.

Fig. 9. Comparison between MCS (1200 sample) results with assumed and measured scatter bounds for input parameter. The 1st eigenfrequency f 1 and fatigue lifetime for each simulation run for all three variants (V1, V2, V3) is displayed. Variants have same color type. Simulation results with assumed and measured scatter are illustrated by dots and triangles respectively. Besides the experimental results are visualised by crosses.

5. Conclusion

In this study a probabilistic approach for the vibration fatigue analysis of a mass force excited component in fluenced by uncertain model parameter was successfully implemented. It was shown that a linear modelling and frequency based fatigue analysis can be applied for metallic structures with joints and create a decent accuracy for the vibration and fatigue behaviour. Good agreements between simulation and experimental model were found for the eigenfrequencies, MAC-value and frequency response functions. The scatter of the sensitive uncertain parameters mass, modal damping, thickness of mounting and plate were determined by means of measurements and an exper imental modal analysis on a pronounced sample base. It was shown that the scatter of these uncertain parameters due to deviations of manufacturing are considerably lower than expected from literature. For the investigated system a damping lower than 0.6% was observed. A comparison of Monte-Carlo-Simulations with measured and assumed scatter of uncertain input parameters shows a significantly improvement in the fidelity of the computed vibration and fatigue behaviour for the simulation run with measured data. This demonstrates an excellent correspondence with the experimental results. Furthermore, a correlation between the frequencies of the modes and fatigue lifetime could