Whats NEW IN TOSCA 2021

After few blogs on Abaqus 2021 enhancements we now move over to TOSCA. Before getting into each feature update, lets recapitulate at the history of enhancements of TOCSA since 2017. The image below is self explanatory. One noticeable improvement is handling of large groups in design responses about which we will discuss more.

  • Easier interpretation of results:

We know that in case of topology optimization, sensitivity based in particular, output has elements of variable densities ranging from fully solid to fully void. That means on a normalized scale, density varies between 0 and 1. Tosca uses interpolation schemes to compute some intermediate densities. A new unified material interpolation scheme has been introduced for a clearer solid-void boundary.

Interpolation schemes have potential to create interwoven solid and void elements. This is called checkerboarding. Filtering is required to minimize this effect. Filtering enforces minimum length scale and prevents checkerboarding. Please note that filtering is different from minimum member size constraint as it is applied to whole model. It can be adjusted in optimization set up. A unified filtering method has been introduced along with interpolation to further enhance clear output.

  • New algorithm for large group operation:

Existing algorithm limits the number of nodes allowed due to amount of adjoint loads generated. New algorithm allows usage of large group nodes in design responses thereby reducing total number of internal constraints that reduces solver time. Supports MAX, MIN, AVE, SUM operators from 2021xFP2042. This enhancement is particularly useful when displacement responses are defined on large number of nodes. An example would be mass minimization under displacement constraints.

Consider a bead optimization case with displacement constraints on oil pan. The location of max displacement is not known before hand as it is a function of bead pattern. However, MINMAX displacement constraint is required. Total number of nodes in model is more than 10K. So ideally displacement constraint should be applied on all nodes.

Earlier only way to solve such a problem was through maximize stiffness approach that would eventually minimize displacement. But a hard constraint on displacement could not be applied as location is unknown. This problem can be accurately solved now because of efficient handling of very large groups. The difference between max stiffness vs. MINMAX displacement approach is shown below.

  • Manufacturing constraints enhancements:

There are two straightforward enhancements. First, milling constraint has been added in topology optimization. This has been achieved by combining multiple casting constraints internally. Second, rotational symmetry constraint in sensitivity based shape optimization. This works for regular as well as unstructured meshes.

Leave a Reply