PSI - Issue 46

J. Bialowas et al. / Procedia Structural Integrity 46 (2023) 49–55 J. Bialowas et al. / Structural Integrity Procedia 00 (2021) 000–000

53

5

distorted by the angle θ , as indicated in Fig. 2 (b). This distortion axially offsets the exit and entry point by exactly that portion of the feed that results from deep rolling of the remaining axle. This measure allows a continuous change of the feed, which can be selected independently of the mesh and component size. A Python script controls the creation of the entire model with all geometry and process parameters. An implemented search algorithm allows the coupling of individual nodes, which change their position depending on the selected geometry and feed. The python script couples the process parameter feed f to the geometry parameter distortion angle of the cylinder sector θ , resulting in a different geometry with an altered feed. This setup allows creating models with different parameter combinations and performing parameter studies for the deep rolling of cylindrical components, not only wheelset axles. 3. Results and Discussion Fig. 3 (a) shows a stress contour plot of a first attempt to model the deep rolling process with the possibility to roll over the edges of a cylinder sector with a total aperture angle of 45°. This model is built up from two parts, a plastic region of interest in the center surrounded by an elastic part. The roller touches down on the elastic part and rolls over the entire plastic region before lifting off again on the elastic part on the other side. This model is considered as a reference for the updated model described in this article. Fig. 3 (b) shows the stress distribution of the model described in chapter 2 with an aperture angle of 6°. The distribution of the residual stresses in tangential direction is more homogeneous compared to the reference model with a 45° aperture angle in Fig. 3 (a). Furthermore, the number of elements is significantly smaller (200,000 for the reference model and 80,000 for the updated model), although the mesh size in tangential and radial direction could be decreased by approx. 50% per side. Thus, the mesh of the updated model is finer and can better reproduce the high stress and strain gradients. The biggest difference between the two models is the computation time. It is reduced from about 24 hours per rollover for the 45° aperture angle reference model to less than one hour for the updated model with 6° aperture angle using an Intel Xeon CPU with 64 GB memory for both computations. To obtain a steady state after deep rolling, it is necessary to reach a certain number of rollovers in the simulation. The evaluation path in Fig. 3 is located at half the length of the cylinder sector, which is also half of the rolled-over length in the axial direction (i.e. the number of rollovers times the feed rate). The stress state in this path is already influenced before the work roller comes into direct contact with this path and continues to change after the work roller rolls over the zone behind the path. Considering the material model and possible values of the process parameters for

(a)

(b)

Fig. 3 Contour plots of the residual σ zz stresses, the position of the evaluation path and the modeled movement of the work roller: (a) Reference model with 45° aperture angle of the cylinder sector and (b) the corresponding updated model with 6° aperture angle with an additional detail view.