Review of Available Options and Possibilities

General Information for Sizing Optimization

• Since the design variables for sizing optimization are the elemental thicknesses, the elemental stiffness changes during the optimization process. This can lead to larger deformations and strains compared to the initial simulation. Therefore, the material laws need to be defined for all material states, especially plasticity laws. Continuous material laws such as Johnson-Cook are preferred for nonlinear optimizations since they are defined for all strain values.
• As the structure softens during the optimization, the recommendation is to adapt applied loads in such a manner that they are reasonable for the final, softer structure with less material, for example 20% of the original material. Otherwise, too much plastic deformation occurs.
• If possible, convert loading to prescribed displacements to improve the robustness of the optimization.
• For shell elements, it is possible that elements form new contacts over the course of the sizing optimization when they are initially not in contact and vice versa, because their shell thickness changes during the process.

Tosca Structure Options

In Tosca Structure, all the following described options are set in the OPT_PARAM block of the *.par file.

Option:	Description:
THICKNESS_MOVE	This option sets the maximum possible move limit on design variables per optimization iteration. Smaller values of this parameter help to overcome optimization convergence issues, but tend to slow down the convergence rate. This is a relative value of the difference between the minimum and maximum thickness. The default value is THICKNESS_MOVE=0.25, meaning that the maximum change in thickness per iteration is 25%. In cases where it is observed that either the objective function or some of the constraint functions are oscillating during the optimization, setting THICKNESS_MOVE=0.1 is the recommendation.
THICKNESS_UPDATE	This option affects the thickness update behavior of CSA (Algorithm based on Convex Separable Approximations). There are three settings available: CONSERVATIVE, NORMAL, and AGGRESSIVE. The default setting is THICKNESS_UPDATE=NORMAL, whereas CONSERVATIVE and AGGRESSIVE are used for complex and unstable or simple and stable problems, respectively. In cases where it is observed that either the objective function or some of the constraint functions are oscillating during the optimization, the recommendation is to set THICKNESS_UPDATE=CONSERVATIVE.
AUTO_FROZEN	This option keeps elements with loaded nodes and boundary conditions from being optimized, respectively. There are four settings available: LOAD, SPC, BOTH, and OFF. LOAD and SPC constrain elements with loaded nodes or boundary conditions, respectively, from being optimized, while BOTH simultaneously freezes all elements with either load and boundary condition. The default setting is AUTO_FROZEN=LOAD.

Solver Settings: Abaqus

The Abaqus solver gives the possibility to define settings to improve the convergence for nonlinear models. See Guide for Nonlinear Models.