Convergence problems can arise in a large deflection analysis for many reasons.
When this occurs, you will see the following messages in the large deflection analysis job progress table:
* CONVERGENCE FAILURE *
or
* DIVERGENCE OCCURRED *
When you see these warnings, look at the left column of the table, which shows the different strategies (KSTRA) used during the analysis. The strategies are described in the section Solution Methods.
If the last value of KSTRA in the table is in the range 0-4, then the solution will be satisfactory. If it is 5, it means the structural or warpage analysis programs have given up equilibrium iterations, and the results will probably be unreliable.
You can assess the accuracy of the solution by plotting a load-deflection graph (request a load deflection history when reading in results). After a few steps in strategy 5, the graph will usually become erratic, indicating that the solution is not accurate. Strategy 5 is used because, for some non-linear problems, the program may recover after a few steps at strategy 5 and return to lower strategies.
When using the "load control" load incrementation method, the analysis cannot traverse a limit point. As the point is approached, the analysis takes smaller and smaller steps and will attempt higher non-linear strategies. After reaching the limit point, the analysis will probably remain in its highest strategy (KSTRA=5).
Although in many cases you can increase the "factor controlling maximum step size" to speed up the analysis, where there is a true limit point in the load path it is wise to limit the steps to about 5%, otherwise the analysis may overshoot the limit load. It is easy to spot a limit point by tracing the history of a relevant node. The slope of the load-deflection graph will approach zero at a limit point.
If such a situation (prolonged increments in strategy 5) occurs, it doesn't always indicate a limit point. It could indicate a problem with the analysis itself. To find out, follow this procedure:
1. Run a small deflection analysis (if you haven't already). Check that the response is reasonable. A modeling error (or unreasonable shrinkages in the case of a warpage analysis) could cause failure of the non-linear solution.
2. Check your constraints. In a warpage analysis, if you have over-constrained the model so that the shrinkage strains are in conflict with constraints, this will often cause the large deflection analysis solution to fail, even if the linear result is reasonable.
3. Look at the load-deflection graph of some relevant nodes. If there is a true limit point or simply a very non-linear region, then the slope of the graph should gradually decrease until it is nearly horizontal. Alternatively, look at the deflected shapes of several steps just before convergence trouble occurs (this will require you to read several results). A highly non-linear region in the load path (that is buckling of the plastic part) usually shows up as a significant change in shape over a small change in load.
4. Sometimes, if the load steps are too large, the gradual decrease in load-deflection slope is not clear (especially if a limit point was overshot). To look in more detail, re-run the analysis as follows:
View the analysis log and examine the Job Progress table. Note the load level (RFAC) at the step before the strategy (KSTRA) becomes 5, call this value RFAC*.
Now re-run a structural or warpage analysis, but this time force the analysis to take smaller steps in the region of RFAC*. For example, if trouble occurs at RFAC* = 0.55, then type in a series of steps like the following as load factor increments:
0.1 , 0.1 , 0.1 , 0.1 , 0.1 .
Now set the "Maximum load factor increase per step" to about 0.005. When the analysis takes over the load stepping (after RFAC = 0.5), all steps will be limited to a maximum of 0.005. Then repeat step 3 above, which should show the response in detail near RFAC*.
If this investigation shows that the solution failed because of a highly non-linear region in the load path, then this is strong evidence of a buckling problem with your plastic part. In fact, this is stronger evidence of a problem than any result obtained from the buckling analysis. Usually, however, you would use the buckling analysis for design purposes because it is considerably faster.