Reaching convergence In Collapse Analysis


Product(s):SACS
Version(s):ALL.
Area:Collapse / Collapse Advance

Theories supported by Collapse:

In difference about collapse standards which only support part of the basic formulation of Newton Raphson, Collapse advance implements two theories which are:

The Newton-Raphson Theory is used by default as a first resource to reach the convergency in this type of analysis, If this formulation is not sufficient to achieve convergence you can add the Arc-Leng solver once time that all possibilities have been exhausted using Newton-Raphson. You can Active the Arc-Leng solver in columns 44-45 of the card CLPOPT

What are the reasons why for Collapse solver stops the analysis?

Collapse solved could stop the analysis for many reasons but the most common and important are:

Recommendations for troubleshooting the convergency issues:

First of all, you need to resolve all solution problems related to elements, the common error is this

'*** WARNING: Solution for member ‘member name’ failed

When this error is presented a good way to resolve is by adding more segments at the elements, another way to resolve it is to consider the element as a secondary and specify it like Elastic using the card "GRPELA",  but if the element rooting this warning is not secondary and you need to know the plastic behavior, so, try to increase the number of subsegments until achieving this warning disappears using card "MEMSEG"

As a Second recommendation, Try to solve the non-convergence in the near vicinity of a limit load you can identify this issue by the following warning message:

*** WARNING: Load increment has failed to converge.

The following remedial actions can be considered:

To avoid this warning try reducing the user-defined load increment size by increasing the number of load steps on the LDSEQ input line for static analysis

Increase the maximum number of iterations, the convergence may be achieved by slightly increasing the maximum number of iterations in CLPOPT without significantly increasing the number of sub-incrementations.

use the card SUBINC to increase the number of the incrementation level in the load convergency

Using the card FRCTOL, the factor for default is 0.001, try with 0.01, 0.05, 0.1 until get the convergency

Finally, if after these recommendations the analysis stops in the middle of the process, and the analysis it's not able to reach finish, so, try fixing all elements with plasticity problem if the error solving any element is still presented, the typical warning is the next:

*** WARNING: Plasticity calculation error whilst solving for member ‘xxxx-xxxx’

The only way to overcome this warning is to override the Yield Stress is recommended using the command YSMGOV. it's probably that the element doesn't withstand the magnitude of force making it difficult to predict the behavior

Getting convergence using Newton Raphson:

Exceding the maximum sub incrementation level

It's normal for the analysis to be stopped when the sub-incrementation level has been exceeded, it's important to mention that the sub-incrementation level only works with the Newton Raphson theory and you can modify it by using SUBINC

Sub-incrementation may be used to improve the convergence by specifying ‘SI’ in columns 44-45 on CLPOPT
for the Newton-Raphson solver. The sub-incrementation solver automatically reduces load increment by
a factor of 2 until convergence is achieved.

Non-convergence caused by Elements

Sometimes we have a nonconvergency due to the large displacement of the imaginary nodes of the subsegmented elements, we can know when is happing this if you see the next legend in the Troubleshooting list.

Solution: When this warning is presented means that Sacs haven't achieved the convergency for the 8 Sub-segment elements set by default, to improve the convergency in the element you need to set a bigger value using the card MEMSEG or GRPSEG.,

see 4.2.4 Warning Messages in Members and 4.2.6 Warning Messages for Non-convergence in Collapse advance Manual

Non-convergence caused by the piles

When we have this legend during the process of solving of Collapse analysis this means that the piles start to fail and soon a pull out of the pile will present.

the previous warning will cause that the structure does not convert

Solution: When we have the two first errors in the Troubleshooting list, the analysis always will be stopped without means presenting the last two Warnings, as a solution, we will need to change the properties in piles to avoid plastification and with this avoid large deformations in the pile heads, if we have plastification in pile heads could be difficult to reach out at the convergency in our analysis

see 4.2.5 Warning Message in Piles in Collapse advance Manual

Getting convergence using Arc-Length:

once the convergency is not achieved using Newton Rapshon, SACS can use the Arc-Length method to get the convergency. There are two types of arc-length iteration methods available in Collapse Advanced. cylindrical (CYL) and spherical (SPH).

The arc-length solver may be used for post-buckling analysis by specifying the analysis option ‘AL’ in columns 42-43 in card CLPOPT.

The cylindrical method uses deflection increments to determine the arc-length parameter while the spherical arc-length method utilizes a factor (columns 31-36) to combine both force and deflection increment. The automatic arc-length parameter can be scaled by a factor in columns 24-29 in card ARCLEN (the default value is 1.0) – i.e. if the arc-length parameter is too large, a value smaller than 1 can be entered or vice versa.

The arc-length method also utilizes a sub-incrementation scheme if it is required. If this option is left blank (i.e. default), Collapse Advanced will attempt to solve using the following order:

For more information about the Arc-length Method check 6.6 Arc-length Method in Collapse Advance Manual.

In general for more information about mesages in Collapse Advance check 4.2 Collapse Advanced Messages in Collapse Advance Manual

Resolving the issue of local minimum:

Solution:

There are different ways to fix the local minimum issue. Among them, the most straightforward solution is :

First, increase the convergence using the FRCTOL line. Note that the default convergence tolerance for collapse advanced is 0.001 which is very small to ensure the most accurate solution.

Second, increase load increment step size. Although it seems unintuitive, increasing the load increment size increases the convergence chance by expanding the search area for the solver.
For some additional notes please see section 4 of Collapse Advanced user manual for troubleshooting and more specifically section 4.4.

Additional Note:

Collapse Advanced uses distributed plasticity, which uses the sub-segments and integration points within the sub-segments and cross-sections. The main advantage of distributed plasticity is that we don't limit failure to predetermined locations. Instead, we let the analysis and loading automatically determine the plastic region of a member.

Important:

If after following all these recommendations you aren't able to get the model up to the convergence, it's probably that the root of the issue will be related to this other issue

https://communities.bentley.com/products/offshore/w/wiki/68512/inconsistencies-using-members-with-the-same-joint-at-the-start-and-end-in-collapse-module

(members with the same joint at the start and end)