Figure. Difference between solid model and configured model optimization.
In ProTOp any free region of a part can be configured in order to enforce certain geometrical features. For this purpose, three configurator types are available as follows:
Each configurator type enforces certain design restrictions on the optimization domain. In the mathematical sense this actually means that the design space becomes more limited. As a consequence, the optimization process will not be able to deliver as good designs as if configuration would not be enforced.
It should be noted however, that in engineering sense an optimal design is also expected to have certain properties that can not be easily incorporated into the mathematical formulation of the optimization problem. Design configuration offers here a very good opportunity to get more robust and more reliable optimal designs as they would be obtained with an ordinary solid model optimization.
The figure below illustrates the difference between optimization of a solid model and a configured model. Conventional solid model optimization yields the best possible result because the corresponding design space is not restricted in any way. On the other hand, configured model optimization yields the best possible design only within the limited design space. In the mathematical sense the former result is always better than the latter.
Figure. Difference between solid model and configured model optimization.
It might be worth noting that by removing the design limitations and running further optimization on the optimized configured design, the result should ultimately be the optimized solid design. Obviously, the optimized configured design is less optimal than the optimized solid design. However, under some circumstances, the former one might be expected to be more robust and more reliable.
Configurators are often used to generate relatively slender structures: thin shells and lattices with thin bars. At this point one has to remember that the underlying solid model is meshed by solid finite elements. Since configurators operate on the numerical model, this means that the finite element mesh must be fine enough to accommodate at least a few finite elements within the slender regions of the structure. If this condition is not met, the configured design will not be usable.
Figure. Enforcing configurators on a coarse FE mesh will yield unusable results.
Figure. The FE mesh is fine enough only after at least a few elements fit within the most slender regions.
According to this, care should be taken to engage any configurators only after the FE mesh is fine enough. This can be achieved by engaging the mesh refinement tool until the average FE edge lengths within the configured region is about 3~4 times smaller than the minimal thickness of a shell or lattice bar.
IMPORTANT. Before using configurators, refine the FE mesh until the FE edge lengths are about 3~4 times smaller than the minimal thickness of a shell or lattice bar.
In order to engage the configurators properly, it is important to understand how they operate. First of all, it is important to note that if no configurators are defined, ProTOp will automatically engage a default solid configurator.
The default solid configurator removes any restrictions in the design space. Any point of the free region can become either material or void. The initial value of the material function is set to a small positive value so that the whole free domain is actually close to switching to void. This ensures a relatively fast progress of the optimization process.
Once the user defines at least one configurator, the default configurator is dropped and only the defined configurators are engaged. This process starts by closing down completely the whole design space, which means that within the free region no point can become a material point. After that individual configurators are applied sequentially, where each configurator is only allowed to open up the design space, while closing down is prohibited.
To illustrate this process, let us consider an example structure - a washing machine drum holder. Let the CAD model of the holder be a solid model, which, after import into ProTOp, looks as follows. Note that immediately after import, there were no custom configurators defined. So, the default solid configurator was engaged.
Figure. Example model: washing machine drum holder - a solid model imported into ProTOp.
Now let suppose, we want to optimize this structure, but not as a full solid model. Instead, we want this structure to be of a shell/lattice type. We might start by creating a shell by adding one shell configurator. In this way the holder can be quickly reconfigured into the form shown in the figure below. Note that this step closed down the majority of the design space; only those points belonging to the shell (within maximal thickness) can become material points. All other points can only be void.
Figure. A solid/shell model after engaging an adequate shell configurator.
After the shell is created, we may proceed by filling the interior with lattices. In our case, we will add four lattice configurators, each of them with corresponding data. After adding the first lattice configurator, the structure looks as follows.
Figure. A solid/shell/lattice model after using the first lattice configurator.
The addition of the second and third lattice configurators results in
Figure. A solid/shell/lattice model after using the second lattice configurator.
and
Figure. A solid/shell/lattice model after using the third lattice configurator (ring).
Finally, after adding the fourth lattice configurator, we get the final configured design.
Figure. A solid/shell/lattice model after using the fourth lattice configurator (pattern).
At this point it is important to note that the configured model is immediately ready for FEA and optimization without any additional work.
Figure. The configured model is immediately ready for topology optimization
The main purpose of the configuration procedure is to enforce certain design limits in the design space. Apart from this, however, there is another reason to engage design configuration that might be quite useful in certain situations. This alternative purpose is the preparation of the initial design.
To illustrate this let us consider the example part shown in the figure below. The interior of the oil pan has to be optimized so that the lowest eigenfrequency will be maximal.
Figure. Oil pan example part with the free region configured by the default solid configurator (full material).
Conventionally this optimization task is started from full material design as configured by the default solid configurator. This means that the optimization process will start from full material design and the RAM and CPU consumption for the FEA will be at maximum for this example.
At this point it is worth noting that the part is well supported and numerically valid even if the whole free region becomes void. This means that one can simply add a custom solid configurator and define the free region as initially void, as shown in the figure below. Note that although the free region is initially void, it can become material later in the optimization process.
Figure. Oil pan example part with the free region configured by a custom solid configurator (initially void).
Starting the optimization from void means that the RAM and CPU consumption for FEA will be substantially smaller than this would be the case if started from full material design. This is because ProTOp engages special semi-active finite element procedures to benefit from void regions. Consequently, a much faster optimization process can be expected.
Figure. Oil pan example part after a few optimization cycles when started from void free region.
The figure above illustrates the current design after a few optimization cycles when the process started from full void design. For the oil pan example shown on the figures the initial FEA of the void design required only 22% of RAM and 16% of CPU time compared to the FEA of the full material design. Thus, efficiency increased by a factor of 5.
NOTE. Proper initial design configuring can make the optimization process much faster. To define an adequate initial design any combination of custom configurators can be used.