Hello guys, can you help me in this please
I had a convergence test failure after applying the gravity load. I have two questions; why we need to apply the gravity load in steps? And why there is a convergence issue here, convergence about what? its a gravity load
some parts of the code;
geomTransf PDelta 1;
element dispBeamColumn 1 1 2 10 100 1;
set Tol 1.0e-8; # convergence tolerance for test
constraints Plain; # how it handles boundary conditions
numberer Plain; # renumber dof's to minimize band-width (optimization)
system BandGeneral; # how to store and solve the system of equations
test NormDispIncr $Tol 6; # determine if convergence has been achieved
algorithm Newton; # use Newton's solution algorithm
integrator LoadControl 0.1; # determine the next time step for an analysis
analysis Static; # define type of analysis static or transient
analyze 10; # apply gravity in 10 steps (0.1*10= 1 load factor)
After checking. it was because I used Concrete02 Material -- Linear Tension Softening, when I changed it to Concrete01 Material -- Zero Tensile Strength, it worked!
But still, anyone knows why there was a convergence issue there?
When you initiate the analysis, in the very first step, all the loads are calculated and felt. In the other words, the constraints sense the model. How I always imagine it is that the model is falling from the sky and suddenly it hits the boundary conditions. This sudden event or impact applies unwanted acceleration and velocity to the model. Because of these acceleration and velocity the model bounces. To stop bouncing and reaching an stable balanced condition, we go through a few steps of gravity with usually very high numerical damping. Through these steps, the model bounces and damps the acceleration and velocities to reach balanced conditions. Basically, there is no new load application, it is only damping out the movements.
After this stage, your model is built. What can help you more is to look at acceleration and displacement recordings in your gravity analyses.
There might be a few reasons behind convergence failure at gravity step. It might be because your model is not built correctly and it bounces out of balance (e.g. to very large deformations). It also might be because of improper setting of analysis option. If your model does not work even in very small time steps, the latter reason probably is the cause. But if you can run the analysis for a few steps then it goes off, the former reason probably is the cause (e.g. for example improper definition of constraints).
Finally, I noticed that you are defining the integrator as LoadControl 0.1. If you have a load embedded in a time-series, solver marches with 0.1 time steps and calculates the load at that time step and applies it to your system. This extensively helps your convergence by reducing impact bouncing problems. For example, say you want to apply 100 kN on a node in gravity step. if you define the load as a constant load, it will suddenly appear to the system and impact the structure, possibly giving lots of acceleration to it. However, if you put it in a time series of Load = 10*time and run the model for 100 steps, at each step the model adds (10*0.1=1 kN) to your model to the end of 100 steps which results 100 kN on the system. The overall, load on the node is the same, but because you have applied it in smaller magnitudes (here 1 kN by 1 kN), it impacts the system considerably less. This can help your convergence substantially.
Unfortunately, I haven't worked with concrete materials. However, the type of problem you are talking about, seems to me as a problem in defining the material rather than analysis settings. Perhaps, Jeena can help you better. If I saw her in the department, I'll ask her about it. Meanwhile, why don't you send a new post about concrete02 material and mention how you are defining it.
