I’ve been running an implicit dynamic analysis on a single element with a uniaxial sinusoidal load in the x-direction. The material model is elastic (density = 7860; E = 197E9; nu = 0.3). The boundary conditions are x-displacements = 0 on the back side of the element (opposite the load) and one corner is pinned. The sinusoidal load has a magnitude of -2.6E8. The amplitude has a circular frequency of 125.66 rad/s (= 20 Hz), a starting amplitude of 0, and a starting time of 0. The element is a fully-integrated hex (C3D8).

When I run the analysis, the S11 stress looks just fine. It is the S22 and S33 stresses which are large in the beginning, about 700,000 depending on the time step you use (I got down to 0.00025 time-step), but gradually go to zero during the analysis. Obviously, those stresses should be zero from the beginning and remain so throughout.

I have tried full integration, reduced integration, implicit, explicit, really small time intervals to capture the nature of the loading, etc. I tried breaking up the cube into multiple elements, but the result was even worse with stresses as high as 11.6E6 at one point. I even tried the quadratic brick (C3D20). All had large S22 and S33 stresses at the beginning.

I did a dynamic simulation where the element was loaded monotonically using a ramped amplitude. With that simulation, the S22 and S33 stresses *did* start and remain zero throughout.

Can anyone tell me what I am doing wrong with the sinusoidal load case?

Note: I am a student and a fairly new user of FEA software.