Composites FEA - Strain energy release rate (SERR)
A total of 4 compact tension composite specimens (with 2 different layups [45]6 and [+45/-45/+45]S) underwent cyclic loading in order to track their small crack growth. From crack nucleation all the way to losing visibility of the crack growth within a tomography window, the crack growth rate was tracked through 5802, 6500, 250 and 800 cycles. The amount of cycles depended on the layup (dependent in turn to the brittle/ductile nature of the layup).
Afterwards, a series of FEA simulations were run modelling one CT specimen at different stages of damage, in order to approximate the strain energy release rate (SERR) per number of cycles. The 3D modelled specimens were loaded at the maximum fatigue load, 432N, through couplings in the pin attachment of the top clevis grip.
From the results, the displacement at the top grip union was extracted from the master node for every simulation resulting in the approximation of the released energy by U = P v / 2
- The setup for the FEA simulations starts with the (A) detected damage profile from tomography, shown here at 1 cycle
- This damage profile is then processed and imported as a part into Abaqus as shown in (B), which is achieved by extracting the crack volumes slice by slice and projecting them onto a 2D sketch conformed by splines. Each sketch is then extruded and assembled to the previous slices.
- Given the high resolution of the tomography slices (images), staying true to their resolution resulted in a "part" that was impossible to mesh due to the scale and complexity of the tortuous 3D damage profile (crack) which meant reducing the true resolution of the crack in order to get a meshable part.
- In total, the version that was meshed successfully had a resolution of 18 pixels through the thickness of the specimen, with each ply spanning 3 voxels.
- While reducing the resolution led to a loss of detail, the crack length and growth remained consistent.
- After importing the damage profile, the CT specimen geometry is merged with the damage profile and is cut away at the notch, resulting in the assembly shown in (C), alongside the boundary conditions and loading used.
- The specimen was replicated with boundaries for each ply and the anisotropic material properties of a unidirectional 0 degree T650/5320 lamina were used, the material orientation was simply modified for each ply.
Using tomography as a guide, the total thickness of the specimen was set to 0.854mm.
The thickness for each ply approximated as follows:
Ply 1: 0.14904mm, Ply 2: 0.1311mm, Ply 3: 0.1449mm, Ply 4: 0.14766mm, Ply 5: 0.138mm, Ply 6: $0.14352mm
From manufacturer, the anisotropic material properties of a unidirectional 0 degree lamina were used:
$E_1 = 138,411~MPa$, $E_2 = 9,176~MPa$, $E_3 = E_2$, $\mu_{12} = 0.326 = \mu_{13} = \mu_{23}$, $G_{12} = 4,950~MPa = G_{13}$ and $G_{23} = 3,460~MPa$.
- A magnified view of the cracked specimen and ply layup is shown in (D).
- Then, for each tomography scan, a baseline geometry was created.
- For every ply that experienced crack growth, another model was created in which only the crack growth in that ply was modelled while the rest of the plies remained at the baseline length.
- This allowed for the per-ply strain energy release rate to be compared against the per-ply crack growth.
