COMSOL Multiphysics^® 6.4 Release Highlights

For users of the Optimization Module, COMSOL Multiphysics^® version 6.4 introduces stop conditions for time-dependent optimization, a Parameter Optimization study step, and several new tutorial models. Read about these updates and more below.

Time-Dependent Optimization

Gradient-based optimization for time-dependent problems can now also be used to optimize the final time. This means it is possible to either minimize how long a process takes or let the solver choose the time that gives the best objective value. In the example below, the goal is to minimize the charging time for a lithium-ion battery. The model uses the Control Function feature, which now supports exporting the optimized result as either an analytic function or an interpolation function.

The Condition-based final time checkbox shown in the General Optimization study step settings. This example optimizes the charging profile for a lithium-ion battery with a constraint on the degradation. The model showcases how the Control Function feature now creates a variable for its average value.

Parameter Optimization

A Parameter Optimization study step designed specifically for gradient-free optimization has been introduced where the control scales are set based on the bounds, removing the need for manual definition. The feature also supports automatic creation of new parameter cases based on the optimized parameters.

Additionally, the user interface has been reorganized relative to the old Optimization study step, which has been renamed General Optimization. The General Optimization study step and the Parameter Optimization study step both support randomization of the initial values for bounded global controls so that different local minima can be identified.

This example demonstrates a ray optics analysis involving optimization of a lens system for different colors and field angles.

Gradient-Based Optimization Solvers

The SNOPT optimization solver has been discontinued; old models have been migrated to IPOPT, which is now the recommended way to achieve quadratic convergence (except for parameter estimation, where the Levenberg–Marquardt solver is expected to be superior). Additionally, choosing the MMA option in an optimization study step used to trigger the use of GCMMA (for the default solver configuration), but now both MMA and GCMMA are available at the study level.

The GCMMA and MMA optimization methods are now available for the optimization study steps, as demonstrated for this example model where shape optimization is used in the design of an RF filter. The model also demonstrates how to randomize the initial controls in order to identify different local optima.

Miscellaneous New Features and Improvements

• New P-norm and Standard Deviation features:
  
  
  • The P-norm feature can approximate the maximum value of a field in a way that is compatible with gradient-based optimization, as often used for stress constraints in structural problems.
  • The Standard Deviation feature can be used to homogenize a field over a selection.
• Far-field operators compatible with gradient-based optimization have been added to the Electromagnetic Waves interface.
• Optimization of eigenvalues now supports the use of nonanalytic operators in the objective expression so that it is possible to design for a target eigenfrequency considering only the real part.
• Support for extrusion constraints has been added to the Density Model feature.
• Performance improvements have been made for gradient-based optimization.

New and Updated Tutorial Models

COMSOL Multiphysics^® version 6.4 brings several new and updated tutorial models to the Optimization Module.

Time-Optimal Control for Heating of a Rod

The initial and optimized heating rates are plotted.

Application Library Title:
optimal_heating_control
Download from the Application Gallery

Designing a Metasurface Beam Deflector Using Shape Optimization

The pillars are colored based on how far they have been moved by the optimization.

Application Library Title:
metasurface_beam_deflector_optimization
Download from the Application Gallery

Minimizing the Charging Time of a Lithium-Ion Battery

The plot shows the charging rate and the electroplating current for the optimized charging profile.

Application Library Title:
lib_charge_rate_optimization
Download from the Application Gallery

Aperture Shape Optimization for Electroplating of a Printed Circuit Board

Initial and shape-optimized aperture and the corresponding deposition thickness on a printed circuit board.

Application Library Title:
pcb_aperture_optimization
Download from the Application Gallery

Wheel Rim — Stress Optimization with Fatigue Evaluation

The initial geometry is shown together with the optimized geometry. The optimization increases the fillet radius to decrease the stress, which also improves the fatigue properties. The fatigue usage factor is plotted in colors.

Application Library Title:
rim_fatigue_optimization
Download from the Application Gallery

Topology Optimization of a Metalens

The optimized design concentrates the wave energy in a point relatively close to the lens.

Application Library Title:
metalens_topology_optimization
Download from the Application Gallery

Minimizing the Drying Time of a Wood Particle

The moisture content is plotted for three fixed temperatures and one optimal controlled temperature.

Application Library Title:
superheated_steam_drying_optimization
Download from the Application Gallery