Why an Unsymmetric Laminate Warps: the B Matrix
Two cross-ply laminates, the same tension. One stretches. The other bends.
Problem
Take a two-ply [0/90] laminate of T300/5208 carbon/epoxy and a four-ply [0/90]s laminate of the same material. Load each with Nx = 100 N/mm of tension and nothing else: no moment, no shear. What are the midplane strains and curvatures, and what shape does a 200 mm square plate take?
Ply 1 on top, 0.125 mm each. In [0/90] the 0° ply sits above the midplane and the 90° ply below it.
Inputs
| Layups | [0/90] and [0/90]s, ply 1 on top |
| Ply thickness | 0.125 mm |
| Nx | 100 N/mm |
| E1 | 181000 MPa |
| E2 | 10300 MPa |
| G12 | 7170 MPa |
| ν12 | 0.28 |
See the plate bend
Out-of-plane deflection from the curvatures, measured from the plate center:
Move Nx and the [0/90] edge lifts in proportion. The [0/90]s line stays on zero at every load.
At 100 N/mm the linear answer puts the plate edge 555 mm out of plane, more than the plate is wide. That is far outside the small-deflection range CLT assumes, so read the shape and its sign there, not the size. Slide down to a few N/mm for deflections a real plate would show.
Nx = 100 N/mm
Plan view of the [0/90] plate, darker where |w| is larger. The [0/90]s plate stays flat.
| [0/90] | [0/90]s | |
|---|---|---|
| εx0 (µε) | 10360 | 2084 |
| κx (1/mm) | −0.1109 | 0 |
| w at x = ±100 mm | 554.6 mm | 0 mm |
At Nx = 100 N/mm the [0/90] plate bends into a cylinder about the y axis: its edges at x = ±100 mm sit 554.6 mm off the center. The [0/90]s plate only stretches, by 2084 µε, and stays flat.
Results
The [0/90] laminate has a non-zero B matrix, so pulling on it in x also curves it in x. The [0/90]s laminate has twice the plies, the same ply angles, and B = 0. It only stretches.
Classical Lamination Theory ties the running loads and moments to the midplane strains and curvatures through the ABD matrix:
A in N/mm, B in N and D in N·mm. The B block (shaded) couples stretching and bending. Each row of the table is one equation.
[0/90]
| εx0 | εy0 | γxy0 | κx | κy | κxy | |
|---|---|---|---|---|---|---|
| Nx | 24019.7 | 724.231 | 0 | 1339.57 | 0 | 0 |
| Ny | 724.231 | 24019.7 | 0 | 0 | −1339.57 | 0 |
| Nxy | 0 | 0 | 1792.5 | 0 | 0 | 0 |
| Mx | 1339.57 | 0 | 0 | 125.102 | 3.77204 | 0 |
| My | 0 | −1339.57 | 0 | 3.77204 | 125.102 | 0 |
| Mxy | 0 | 0 | 0 | 0 | 0 | 9.33594 |
[0/90]s
| εx0 | εy0 | γxy0 | κx | κy | κxy | |
|---|---|---|---|---|---|---|
| Nx | 48039.3 | 1448.46 | 0 | 0 | 0 | 0 |
| Ny | 1448.46 | 48039.3 | 0 | 0 | 0 | 0 |
| Nxy | 0 | 0 | 3585 | 0 | 0 | 0 |
| Mx | 0 | 0 | 0 | 1670.6 | 30.1763 | 0 |
| My | 0 | 0 | 0 | 30.1763 | 331.034 | 0 |
| Mxy | 0 | 0 | 0 | 0 | 0 | 74.6875 |
The A matrix of [0/90]s is exactly twice that of [0/90], because it has twice as many of the same plies. B is where they part. In [0/90] the stiff 0° ply sits on one side of the midplane and the soft 90° ply on the other, so B11 = 1339.57 N and B22 = −1339.57 N. Mirror the stack and every ply above the midplane has a twin below it at the same angle, so each term of B cancels.
Invert the 6 x 6 matrix and multiply by the load vector (Nx, 0, 0, 0, 0, 0). Strains in µε, curvatures in 1/mm:
| Laminate | εx0 | εy0 | γxy0 | κx | κy | κxy |
|---|---|---|---|---|---|---|
| [0/90] | 10360 | −312.3 | 0 | −0.1109 | 0 | 0 |
| [0/90]s | 2084 | −62.82 | 0 | 0 | 0 | 0 |
The [0/90] laminate stretches 4.97 times as much as [0/90]s. Half the thickness accounts for a factor of 2. The rest comes from B: a laminate that is free to bend also gives more in its own plane. And it bends, κx = −0.1109 1/mm, with no moment applied. κy and κxy are zero to rounding, so the plate bends about the y axis only, into a cylinder.
What it means
Along x the 0° ply is about 18 times stiffer than the 90° ply, so the stiffness of [0/90] is centerd in the top ply, above the midplane. Nx acts at the midplane, below that center. The pull is off-center, the soft bottom ply stretches more than the top, and the plate curves with the 90° ply on the convex side. That is B at work: in-plane load in, curvature out. In a symmetric stack the stiffness is centerd on the midplane, so the same pull only stretches it.
This is the linear CLT answer to a mechanical load. A real thin [0/90] plate also curls on its own when it cools from the cure temperature, before any load goes on, because the two plies shrink by different amounts across the fibers. The hygrothermal docs cover that thermal and moisture response.
So if a part has to stay flat under in-plane load, make the layup symmetric. If it cannot be symmetric, check B before you trust a membrane-only answer: the strains and the ply stresses both change once the laminate bends.
Related docs:
References
- Jones, R.M., Mechanics of Composite Materials, 2nd Ed., Taylor & Francis, 1999.
- Barbero, E.J., Introduction to Composite Materials Design, 3rd Ed., CRC Press, 2017.