Differences between shape algorithms

The user may choose between two shape optimization algorithms in Tosca Structure.shape. The algorithms have different ways to find the solution and their differences will be discussed in this chapter.

The algorithm used for the controller based shape optimization is based on an optimality criteria. Usually an optimality criteria is based on a set of rules that guide to a more optimal solution if certain prerequisites are fulfilled. This leads to fast specialized algorithms typically usable for one class of problems with the drawback that they are not applicable if the prerequisites are not fulfilled. For more information about the optimality criteria used with the controller-based shape optimization, see Controller based shape optimization.

SHAPE_CONTROLLER

The controller based shape optimization approach is very robust and reliable in removing stress peaks that lie in the design area. On the other hand this includes that stress peaks outside the design area are in general not attackable. One drawback is that by default only stress based design responses can be used. These can be used as objectives only and not even as constraints.

  • for standard problems (stress minimization and volume constraint)
  • just volume constraint feasible
  • only one constraint for each task
  • manufacturing restrictions are possible
  • just about 15 iterations necessary
  • homogenization of stress distribution by adding material at points of high stress and removing material at points of low stress

SHAPE_SENS

The sensitivity approach is much more flexible in terms of what should be optimized i.e. what design responses are usable in the objective function and in constraints. Typically this higher flexibility is bought with longer run times. Tosca specific disadvantages of the sensitivity based approach compared to the controller based approach cover the amount of usable element types and the number of supported manufacturing constraints.

  • extended problems (combination of different design response)
  • up to 45 iterations, sometimes more
  • uses sensitivities (derivatives) of design responses
  • more types of design responses feasible (displacements, volume, reaction forces etc.)
  • possible to define several constraints for one task
  • possible to solve problems with long range effects
  • manufacturing restrictions not implemented correctly yet