I have created a model in Abaqus of a bistable unsymmetric lay-up [0/90]T composite laminate being forced to vibrate under harmonic base excitation. It is a square 150 x 150 mm laminate with [0/90]T lay-up and carbon fibre/epoxy material. The geometry is meshed with 60 x 60 S4 elements. Additionally, a point mass is positioned slightly inwards along the diagonal from each corner (four point masses in total) to aid snap-through. The laminate is clamped at it's centre (nodes contained within a square area with 5 mm edge length - similar to experimental set-up where the laminate is clamped to a stinger at its centre).
The model involves three steps: (1) general static step to simulate cool down from 135C to 22C, inducing curved deformation in the laminate and convergence to one of its two stable states; (2) general static step with gravity loading to account for static deflection due to point masses at each corner; (3) dynamic implicit analysis where a constant peak acceleration (8g) sinusoidal acceleration is applied in the z- translational DOF on the clamped nodes (z- DOF modified to be free in this step).
The sinusoidal excitation ramps from 0 to the peak amplitude and then remains at the constant peak amplitude sinusoidal excitation for the remainder of the step, as shown in the image below.
The simulation has no issues running; however, when analysing the results I have observed that the ETOTAL energy output decreases every time the laminate undergoes snap-through, as shown in the image below of the relative displacement of a point on the laminate and the ETOTAL.
I have run the model using the transient fidelity, moderate dissipation and specifying alpha=0 (no numerical damping), but the same thing happens in each model. I would really appreciate any help in identifying why the ETOTAL is decreasing and any means to prevent this from happening. Is it something to do with the fact that the laminate experiences negative stiffness during snap-through? Thanks.