Debugging Issues

[blaiman]

We have been working on developing a new element which incorporates warping as a 7th DOF and now in the process of verifying the results with some "toy" problems to see if the outputs match up with analytical solutions. For the most part, running a simple cantilever case with a unit load applied at the tip with one DOF at a time shows very accurate results. However, when it came to running a case where there was axial load which was 67% of the buckling limit along with a transverse load, convergence issues started to arise. The problem seemed to be "handled" when using 100 elements with unit length each element and using 1000 steps through the loadcontrol integrator, but the output obtained through OpenSees is very off from the second-order solution. We have also made corresponding geometrical transformation files which get used by this new element. Does anyone have any suggestions on how to approach or try to find where the problem is with our new formulations? Any input would be greatly appreciated. Thanks!

[kostast88]

In a brief experience I had with modifying the source code, I had similar problems. My cantilever test was a success, everything else was close but not close enough to the analytical solution, and convergence was an issue. The problem pointed out newton raphson failure due to loss of equilibrium,so my guess would be, check if model is in equilibrium at every step.