PSI - Issue 42

Pauline Herr et al. / Procedia Structural Integrity 42 (2022) 498–505 P. Herr et al. / Structural Integrity Procedia 00 (2019) 000–000 P. Herr et al. / Structural Integrity Procedia 00 (2019) 000–000

504

7

7

for the standard cohesive zone models implemented in Abaqus, which, all in all, makes our approach promising for structural simulations. a 0.030 005 for the standard cohesive zone odels i l e ted in baqus, which, all in all, makes our approach promising for structural simulations. a 0. 05 0.030 b

0.030 0.03

b

0.030

0.025

0.025 .0 0. 0.

0.025

000

0.005 Plastic Separation u pl [mm] 0.005 0.010 0.015 .020 l stic r ti pl [m ] 0. 5 0. 10 0.015 0.020 Plastic Separation u pl [m ] . 05 . 10 .015 . 20 Plastic Separation u pl [mm] . 0 .010 .01 .02 l stic r ti pl [m ] 0. 5 0. 0. 0. Plastic Separation u pl [m ] 0.0 0.010 0.015 0.020 0

0.000 0.005 0.010 S

Training Data AN r ining at FF FF r i in t AN r ining Data FF FF

0.000 .000 0.0 0.0

0. 00

0.000

0.0 0 0.005 0.010 0.015 0.020 0.025 0.030 Separation u [mm] 0. 0. 0.01 . 1 . 20 . .03 eparation u [mm] 0. 0 0. 5 .010 0 5 025 0. Sep a on u [mm] 0.0 0 0.005 0.010 0.015 0.020 0.025 0.030 Separation u [m ] 0. 0. 0.01 . 1 . 20 . 5 .03 Sep ration [ ] 0. 0 0. 5 .010 0 5 . S par t u [m ]

Fig. 3. (a) Plastic separation in dependency of separation: Training data and FFNN; (b) Comparison J -integral in DCB test and J c from the model. Fig. 3. (a) Plastic separation in dependency of separation: Training data and FFNN; (b) Comparison J -integral in DCB test and J c from the model.

P

θ

P

Fig. 4. Simulation of a DCB test with damage field. Fig. 4. Simulation of a DCB test with damage field.

4. Conclusion and outlook 4. Conclusion and outlook

In this work, we proposed an ML-based surrogate model for cohesive zone modeling, which is capable of consid ering the e ff ect of volume fraction of glass beads using results from a novel FFT-based homogenization schemes for cohesive zones. Within the context of this study, the following conclusions can be drawn: • The results from FFT-based homogenization show that with an increase in the volume fraction, the sti ff ness of the adhesive increases slightly, whereas the critical ERR decreases. This was qualitatively also observed similary in an experimental study, indicating the validity of the assumptions made concerning the RVE geometry and material model of the constituents. Nevertheless, there are still uncertainties regarding the assumptions of the FFT-based homogenization scheme for cohesive zones, as discussed in Bo¨deker et al. (2022) in more detail. • The proposed model is able to represent the TSL well for all volume fractions contained in the training data, and the computational times seem reasonable for an industrial virtual material development process. Nevertheless, volume fractions not contained in the training data cannot be covered, and the model is limited to mode I fracture at the current stage. In this work, we proposed an ML-based surrogate model for cohesive zone modeling, which is capable of consid ering the e ff ect of volume fraction of glass beads using results from a novel FFT-based homogenization schemes for cohesive zones. Within the context of this study, the following conclusions can be drawn: • The results from FFT-based homogenization show that with an increase in the volume fraction, the sti ff ness of the adhesive increases slightly, whereas the critical ERR decreases. This was qualitatively also observed similary in an experimental study, indicating the validity of the assumptions made concerning the RVE geometry and material model of the constituents. Nevertheless, there are still uncertainties regarding the assumptions of the FFT-based homogenization scheme for cohesive zones, as discussed in Bo¨deker et al. (2022) in more detail. • The proposed model is able to represent the TSL well for all volume fractions contained in the training data, and the computational times seem reasonable for an industrial virtual material development process. Nevertheless, volume fractions not contained in the training data cannot be covered, and the model is limited to mode I fracture at the current stage.

In order to clarify the unresolved points, we intend to undertake further research in the following areas: In order to clarify the unresolved points, e intend to undertake further research in the following areas:

• A basic, exemplary experimental validation is required to assess practical applicability and further perspectives. • A basic, exemplary experi ental ali ati is required to assess practical applicability and further perspectives. • The model could be extended to mixed-mode fracture and the e ff ect of di ff erent types of fillers and size dis tributions can be included as well. However, the size of the fillers is limited, s.t. both the adhesive layer and