Figure. Unstable process due to eigenmode switching: design and eigenmode in two subsequent optimization cycles differ substantially.
In ProTOp the topology optimization process is considered to be a continuous process which can be either in a
The optimization process can be controlled by a set of basic and a set of advanced parameters. The advanced parameters can be accessed via the Optimization parameters dialog by selecting the Parameters ... command from the Advanced controls drop-down menu. In a typical optimization task, these parameters do not require any adjustments.
The set of basic parameters is accessible directly from the optimization dialog. The following parameters are available:
Maximization objective
Range: [0, 100]
This parameter sets the weight factors for structural stiffness and structural lowest eigenfrequency in the definition of the objective function. By setting this parameter to 0, the structural stiffness is maximized. By setting this parameter to 100, the structural lowest eigenfrequency is maximized. Intermediate values optimize the design with respect to both objectives, by considering the corresponding weight factors.
Target
Range: [Volume part | Ref boundary stress | Max displacement]
This parameter sets the desired optimization target as follows:
Material redistribution rate
Range: [Low, High]
This parameter sets the rate by which material is ether removed or restored. Note that low rates tend to produce more fragmented optimal designs, while high rates tend to produce less fragmented optimal designs.
Details filtering intensity
Range: [Low, High]
This parameter sets the intensity of filtering out the details of the FEA fields involved in the optimization process. Note that low intensities tend to produce more fragmented optimal designs, while high intensities tend to produce less fragmented optimal designs.
Optimization cycles: Minimal requested
Range: [0, ~]
This parameter sets the minimum requested number of optimization cycles to run, irrespective of convergence indicators.
Optimization cycles: Maximal requested
Range: [0, ~]
This parameter sets the maximum allowed number of optimization cycles to run, irrespective of convergence indicators.
Several run FEA and pause commands are available from the Run cycles drop-down menu:
Run FEA only
This runs the FEA of unsuppressed static (stress) and modal (eigenfrequency) load cases and pauses without doing any optimization step. This might be useful for checking the structural response without doing any optimization.
Run ALL stress FEA cases
This runs the FEA of all (suppressed and unsuppressed) static (stress) load cases and pauses without doing any optimization step. This might be useful for checking the structural response of all stress load cases, including the disabled ones.
Run ALL eigenfrequency FEA cases
This runs the FEA of all (suppressed and unsuppressed) modal (eigenfrequency) load cases and pauses without doing any optimization step. This might be useful for checking the structural response of all eigenfrequency load cases, including the disabled ones.
Several other commands and options are available from the Additional controls drop-down menu:
Parameters ...
This command opens the Optimization parameters dialog, which provides access to advanced optimization parameters.
Load cases ...
This command opens the Optimization load cases dialog, which enables monitoring and management of currently active load cases.
Record cycles designs
If checked, design data and picture frames are recorded to enable later design recovery and generation of animations. Note. Picture frames are recorded only if the graphical window is visible (not minimized).
Topology optimization problems may quickly become numerically unstable, which may result in numerous vain optimization cycles. This situation is more often observed in eigenfrequency maximization and less often in stiffness maximization.
The figure below illustrates a typical unstable situation in an eigenfrequency maximization process: the instability is caused by switching between various eigenmodes from one cycle to another.
Figure. Unstable process due to eigenmode switching: design and eigenmode in two subsequent optimization cycles differ substantially.
To resolve such a situation the following steps are recommended.
1. Reduce material redistribution rate
2. Increase FEA response dump factor that corresponds to the engaged FEA type: either stress or eigenfrequency
