PSI - Issue 80

Yichen Zhang et al. / Procedia Structural Integrity 80 (2026) 289–298

296

8

Author name / Structural Integrity Procedia 00 (2019) 000 – 000

Five elements in thickness

(e) (f) Fig. 6. Damage modes at final failure for the 3D and 2D models with 90° fibre direction: (a) 3D model with single-layer element, (b) 2D model with single-layer element, (c) 3D model with two-layer element, (d) 2D model with two-layer element, (e) 3D model with five-layer element, (f) 2D model with five-layer element. 4. Influence of through thickness number of elements on ultimate torque While the 1-layer 3D model yields a significantly lower ultimate torque, the 2-layer and 5-layer models demonstrate identical ultimate torque values. This performance difference can be attributed to local buckling when only one integration point exists in the thickness direction. For all fibre orientations (0°, 45°, 90°), we can observe that if only one element layer is used through the thickness, the 3D model produces an incorrect ultimate torque compared to its 2D counterpart. Despite both models having a single element layer, the 2D model enforces three integration points through the element thickness, while the 3D model has, as a result of the reduced integration in 3 directions, only one. For that reason, the 3D model with single reduced element in the thickness direction is unable to effectively withstand torsion due to local buckling. This leads to an incorrect reduction in its ultimate torque, and an apparent local buckling. On the other hand, when the number of layers of 3D models is increased to two and then to five, the apparent local buckling is mitigated, and the 3D model’s ultima te torque is equal to that of the 2D model. As a result, the 3D model, when used with at least 2 through thickness elements, correctly predicts the ultimate torque, while providing far more reasonable crack patterns, compared to the built-in 2D model.