Optimization Module Updates

For users of the Optimization Module, COMSOL Multiphysics® version 5.5 includes a more simplified setup of shape optimization with new built-in features, a filter dataset for smoothing topology optimization results, and strict enforcement of design constraints for derivative-free optimization. Read more about these and other optimization features below.

Shape Optimization

Setting up shape optimization problems is now easier. As of version 5.5, you can select parts of a geometry for shape optimization using the new Free Shape Domain feature. Boundaries to be optimized are specified using Polynomial Boundary or Free Shape Boundary features, which have built-in support for regularization. You can now also perform shape optimization on shells using the new Free Shape Shell feature. The Symmetry/Roller feature specifies points or edges that can slide on a flat boundary. All of the features interact with each other to ensure continuity, and the default plot makes it easy to visualize the result.

Models that use this feature:

A wrench is modeled in COMSOL Multiphysics and the Free Shape Domain settings are shown.
The new Free Shape Domain shape optimization feature Settings window available under component definitions.

Topology Optimization

The default plot for 3D topology optimization now uses the new Filter dataset, from which you can directly create a mesh part. That way, the optimized geometry can be imported into a new component retaining all selections associated with the topology optimization geometry. This reduces the work needed to set up a verification simulation of a topology optimization result. A new Prescribed Density feature interacts with the Density Model feature, improving the regularization filter behavior near boundaries. This can improve the robustness when creating a mesh part based on the filtered field.

Models that use these features:

The geometry of a topology optimized hook design imported from a mesh part pointing to a Filter dataset in COMSOL Multiphysics.
The selections from the old component can be recycled when the geometry is imported from a mesh part pointing to a Filter dataset.

Parameter Estimation

When solving a parameter estimation study with the Levenberg–Marquardt method, you can compute the confidence intervals for the parameters based on a user-defined confidence level. The computation of the intervals assume that the measurements are independent and that the error is normally distributed. You could then combine the confidence intervals with synthetic data generation to estimate the required sample size to achieve a certain accuracy for the parameters.

The COMSOL Multiphysics UI showing confidence intervals for a model and the Optimization study step settings.
The computation of confidence intervals in parameter estimation can be enabled in the Optimization study step.

Parameter Optimization

The derivative-free optimization solvers now have an option for strict enforcement of design constraints, improving robustness and reducing computational time. The model will not be evaluated for control parameter sets that violate constraints depending only on the control parameters. The option is enabled by default for COBYLA and Nelder–Mead. You can see this new feature demonstrated in the multistudy_bracket_optimization model.

A bracket model in COMSOL Multiphysics with the Optimization settings showing design constraints for preventing topological changes.
Parameter optimization can often be made more robust by defining design constraints that prevent topological changes. Strict enforcement of design constraints can improve the behavior of such models.

New Tutorial Models and Applications

Version 5.5 brings several new and updated tutorial models and applications.

Shape Optimization of a Shell

Visualization of von Mises stress in the optimized and initial designs of a shell.
The von Mises stress is plotted for the optimized (left) and initial (right) shell design.

Application Library Title:
shell_shape_optimization
Download from the Application Gallery

Shape Optimization of a Wrench

A wrench model showing an optimized design in red and the original geometry in gray.
The volume of the optimized wrench is shown in red together with the original geometry in gray.

Application Library Title:
wrench_shape_optimization
Download from the Application Gallery

Shape Optimization of a Tesla Microvalve

A model of a shape optimized Tesla microvalve, half in color, half as a gray mesh.
The mesh generated in the deformed configuration (green lines) is plotted together with the backwards flow velocity in a Tesla microvalve.

Application Library Title:
tesla_microvalve_shape_optimization

Download from the Application Gallery

Shape Optimization of an MBB Beam

The shape of an MBB beam where the optimized shape is red and black and the initial design is shown in gray.
The optimized shape is plotted in black/red, while the initial design (from topology optimization) is shown in gray.

Application Library Title:
mbb_beam_shape_optimization

Download from the Application Gallery

Optimization of an Extruded MBB Beam

A shape-optimized MBB beam design shown in red with the topology optimization shown in gray.
The result of the topology optimization is shown as a gray volume together with the final shape-optimized design in red.

Application Library Title:
mbb_beam_extruded_optimization

Download from the Application Gallery

Optimization Tutorials

An optimization model using the quadratic function in COMSOL Multiphysics.
It is possible to perform discrete optimization by converting a continuous control to a discrete one using the round function. However, this gives rise to a function that is locally flat, which makes the problem ill suited to gradient-based optimizers.

Application Library Title:
optimization_tutorials

Download from the Application Gallery

Maximizing the Buckling Load of a Diagonal Brace

Two versions of a diagonal brace design, where the optimized design maximizes the buckling load.
The shell displacement is plotted for the critical load factor with the lowest absolute value for the initial (bottom) and optimized (top) design.

Application Library Title:
diagonal_brace_buckling_optimization

Download from the Application Gallery

Parameter, Shape, and Topology Optimization of a Beam

A beam design is optimized three ways for minimized weight.
The weight of the beam is minimized using parameter (top), shape (middle), and topology (bottom) optimization. The beam is subject to a distributed load and a displacement constraint.

Application Library Title:
beam_optimization

Download from the Application Gallery

Impedance Tube Parameter Estimation with Data Generation

A 1D plot showing pressure as a function of frequency for two sensors in an impedance tube.
The measured and fitted real pressure is plotted as a function of the frequency for two sensors.

Application Library Title:
impedance_tube_parameter_estimation_data

Download from the Application Gallery

Shape Optimization of an Acoustic Demultiplexer

An acoustic demultiplexer model visualizing the sound pressure level distribution and intensity.
Sound pressure level distribution and intensity streamlines plotted at the center of both frequency bands.

Application Library Title:
demultiplexer_shape_optimization

Download from the Application Gallery

Bracket — Topology Optimization

One gray bracket design and one topology optimized bracket design that uses much less material.
Original design space and optimized structure with deformation from one of eight load cases.

Application Library Title:
bracket_topology_optimization_stl
Download from the Application Gallery