PSI - Issue 57

Ewelina Czerlunczakiewicz et al. / Procedia Structural Integrity 57 (2024) 743–753 / Structural Integrity Procedia 00 (2019) 000 – 000

752 10

After conducting an extensive exploration of the parameter space, several matching parameters were identified that represented similar damage values and failure types. Figure 10 shows the population of verified parameters during the optimization process. Nearly 5,000 combinations were calculated, from which the most promising values were selected to accurately replicate nonlinearities using linear spring-dampers. One of the optimization objectives was to achieve stress levels at the examined points similar to tho se from the nonlinear analysis, as they directly influence product fatigue and estimated product life. It is important to note that these results refer to a specific case. However, building upon the success of this particular case, a response surface was generated. The underlying idea is to develop a surrogate model that can rapidly and accurately identify new stiffness and damping parameters for connectors in future projects. The chosen method for this purpose was Kriging, which is capable of estimating errors and uncertainties. This is particularly crucial when seeking optimal solutions within tight time constraints, especially in the early stages of a project. By employing Kriging, the surrogate model can efficiently provide valuable insights while considering the potential for errors and uncertainties. The final results obtained from the parametric optimization strongly indicate that finding a linear approximation for a non-linear contact case is a challenging task. The primary difficulty lies in accurately representing the stress spectrum while preserving favorable overall behavior. The linearization of non-linear cases is a significant area of study that will surely be further developed in future research attempts. Conclusions & Further Works The paper has demonstrated that using a non-linear approach in random load dynamic finite element simulations yields different results compared to the linear approach. The comparison revealed that tested linear approaches, whether directly integrated or mode-based, consistently provided comparable results. The non-linear approach allows for a more accurate representation of the behavior of the system, for instance when dealing with significant non-linear effects like contact interactions. It provides a more realistic prediction of the response and can capture phenomena that cannot be adequately captured by linear analysis. The analysis of the most damaged elements from different components revealed varying trends. In some cases, the non-linear model predicted significantly higher damage, particularly in areas with clipping connections. However, in other cases, the non -linear model predicted lower damage compared to the linear model, especially in regions where stress hot spots were observed away from clipping connections. However, it is important to note that the non-linear approach introduces increased computational complexity and requires more resources for analysis. Additionally, defining non-linear connections can be complex and requires expertise, as the results are sensitive to the parameter values used in the connection definition. Therefore, careful consideration of computational resources and expert knowledge is essential when employing the non-linear approach. Further investigations will be still in the framework of catching design issues related to vibration fatigue in the most reasonable way, i.e. by finding a trade off between calculation time and accuracy, in line with automotive constraints and time to market for our products. One main investigation involves the search for representative boundary conditions to orient towards simple linear or need for non linear methods. A correlation task is on-going to verify the validity of the model, especially in the case 4.