Optimization Module Updates

For users of the Optimization Module, COMSOL Multiphysics® version 5.6 brings a new optimization solver (IPOPT), the ability to save intermediate solutions for gradient-based optimization, and automatic objective function scaling. Read more about these and other optimization features below.

Design Parameter Optimization

When performing design parameter optimization with a derivative-free optimization solver, an objective table is produced by default. You can now right-click a row in such a table and choose Copy Selected Rows to New Parameter Cases. This simplifies the workflow when you want to perform further analysis on the optimized result.

The COMSOL Multiphysics version 5.6 UI with a bracket model in the Graphics window and the objective table shown with the Copy Selected Rows to New Parameter Cases option highlighted.
The multistudy_bracket_optimization model minimizes the mass of a bracket while constraining the eigenfrequency and the maximum stress. The optimized parameter values can be copied to a new parameter case by right-clicking the objective table.

Shape Optimization

There is a new Shape Optimization study step dedicated to gradient-based optimization. Additionally, there is a new gradient-based optimization solver, IPOPT, which uses an interior point algorithm. This solver is available for all optimization studies. The solver has properties similar to SNOPT, but IPOPT is an open source project under active development. The new study step and the IPOPT solver are both used in several tutorial models, including the Optimizing a Flywheel Profile model. In COMSOL Multiphysics® version 5.6, this model uses an axisymmetric component with a Solid Mechanics interface together with the built-in features for shape optimization.

The COMSOL Multiphysics version 5.6 UI with the Shape Optimization settings shown and the IPOPT method selected.
The IPOPT optimization solver is combined with the Free Shape Boundary feature to increase the moment of inertia for a flywheel without increasing the mass or the (approximate) maximum von Mises stress.

Models and applications that use IPOPT:

Topology Optimization

There is a new Topology Optimization study step dedicated to gradient-based optimization, with support for the new IPOPT optimization solver. Additionally, you have a new option to Keep solutions that allows you to save several intermediate optimization solutions. Choose between saving the Last N or Every Nth solution. The option is useful for finding mistakes in your model as well as for creating animations of the optimization progress.

Many optimization solvers work best if the objective function is well scaled, and the optimization study steps now feature an option called objective scaling that can be used to set a manual scale or a scale based on the initial solution. This simplifies the setup of many topology optimization problems. A new Topology Link material and model wizard tool automatically sets up the common case of volume-constrained compliance minimization for structural topology optimization.

The Topology Optimization study settings in COMSOL Multiphysics where the Keep solutions option is set to Every Nth and an optimized bracket is shown in the Graphics window.
The Keep solutions option can be set to save every Nth solution. If N = 1, all solutions will be saved, and then you can inspect the progress of the optimization in detail.

Models and applications that scale using automatic scaling of the objective function with the solution value:

New Tutorial Models

COMSOL Multiphysics® version 5.6 brings several new and updated tutorial models to the Optimization Module.

Optimizing a Flywheel Profile

A cylindrical blue disk with a third of it cut out to expose a red center
Shape optimization is performed on what is initially a cylindrical disk. The moment of inertia is maximized, so that the disk can be used as a flywheel.

Application Library Title:
flywheel_profile
Download from the Application Gallery

Optimization of a Photonic Crystal for Signal Filtering

Several pillars drawn as circles on a white background to represent a photonic crystal.
The position of the pillars is optimized to maximize the transmission at the desired frequency divided by the transmission at the undesired frequency.

Application Library Title:
photonic_crystal_filter_optimization
Download from the Application Gallery

Shape and Topology Optimization of Compliant Pliers

A model of optimized compliant pliers shown in the rainbow color table.
A compliant mechanism is optimized to work as a set of pliers and the resulting design is subjected to shape optimization.

Application Library Title:
compliant_pliers_mechanism
Download from the Application Gallery

Parameter Estimation for a Tensile Test

A 3D dogbone tensile stress specimen in rainbow color table and red arrows on either end.
The Young's modulus and Poisson ratio are estimated based on a tensile test. The test measures the tensile force and the radial displacement. The model is based on synthetic data generated in the model itself.

Application Library Title:
tensile_test
Download from the Application Gallery

Shape and Topology Optimization of an Extruded Beam

A gray and red extruded beam in 3D.
This model demonstrates 3D topology and shape optimization for an extruded geometry. The stiffness of a beam is maximized subject to a constraint on the mass, but the method can be used with any variant of the stationary and frequency domain solvers.

Application Library Title:
extruded_beam_optimization
Download from the Application Gallery

Optimization of a Truss Tower

Three truss tower models of various designs, going from more to fewer members.
The diameters of individual members in a truss structure are optimized such that the structure becomes stiffer without adding mass. The model demonstrates how the optimized design can be modified so that the tower can be built using just 3 types of members.

Application Library Title:
tower_optimization
Download from the Application Gallery