openEMS Parametric Design & Optimization Framework

Who is this for?

For engineers, researchers and students

This resource is intended for RF and microwave engineers, antenna and electromagnetic designers, radar and sensing engineers, academic researchers, engineering students, and developers interested in computational electromagnetics and automated design optimization.

01 · Foundations

Why parametric electromagnetic design?

Electromagnetic structures are often designed by changing dimensions, materials, feed locations or other physical parameters and repeatedly checking the resulting response. When this process is performed manually, each experiment can become a separate simulation setup, making comparisons harder to reproduce and optimization difficult to automate.

A parametric approach describes the structure through variables and relationships. The same simulation workflow can then generate many candidate geometries while changing only the parameters under investigation.

This framework applies that idea to full-wave 3D electromagnetic simulation. It combines YAML-based design definitions with Python, openEMS and CSXCAD, then adds objective functions, optimization engines and post-processing.

03 · Design Definition

Fixed geometry versus parametric geometry

The difference between a fixed design and an optimization-ready design is easiest to understand by looking at the complete YAML structure. In a fixed design, the geometry and port coordinates are explicit numbers. In a parametric design, the same physical description can be expressed through variables and mathematical relationships, allowing the framework to generate many candidate geometries automatically.

primitives:
  - type: "polygon"
    material: "PEC_Patch"
    points:
      x: [0.001, 0.161, 0.161, 0.001]
      y: [0.0005, 0.083, -0.083, -0.0005]
    elevation: 0
    length: 0.0005
    norm_dir: "z"

  - type: "polygon"
    material: "PEC_Patch"
    points:
      x: [-0.001, -0.161, -0.161, -0.001]
      y: [0.0005, 0.083, -0.083, -0.0005]
    elevation: 0
    length: 0.0005
    norm_dir: "z"

ports:
  - id: 1
    type: "lumped"
    resistance: 50.0
    direction: "x"
    excite: 1
    start: [-0.001, -0.0005, 0]
    stop:  [0.001, 0.0005, 0.0005]

simulation:
  unit: 1.0
  fstart: 0.2
  fstop: 1.2
  n_pts: 1001

variables:
  arm_length:
    type: "float"
    start: 0.15
    stop: 0.30
    step: 0.01
    default: 0.20

  root_width:
    type: "float"
    start: 0.004
    stop: 0.020
    step: 0.002
    default: 0.010

  tip_width:
    type: "float"
    start: 0.010
    stop: 0.40
    step: 0.005
    default: 0.30

  feed_gap:
    type: "float"
    start: 0.002
    stop: 0.002
    default: 0.002
    optimize: false

points:
  x:
    - "feed_gap / 2.0"
    - "arm_length"
    - "arm_length"
    - "feed_gap / 2.0"
  y:
    - "root_width / 2.0"
    - "tip_width / 2.0"
    - "-tip_width / 2.0"
    - "-root_width / 2.0"

ports:
  - id: 1
    type: "lumped"
    resistance: 50.0
    direction: "x"
    excite: 1
    start:
      - "-feed_gap / 2.0"
      - "-root_width / 2.0"
      - 0
    stop:
      - "feed_gap / 2.0"
      - "root_width / 2.0"
      - 0.001

simulation:
  unit: 1.0
  fstart: 0.3
  fstop: 1.0
  n_pts: 301
  end_criteria: 1e-2
  max_timesteps: 50000
  target_s11_db: -15.0

What can become variable?

Any design quantity exposed through the configuration can potentially become a design variable: dimensions, gaps, widths, positions and other geometry parameters. Variables can be bounded and quantized with a step size, while parameters such as a feed gap can remain fixed when they should not participate in the search.

05 · Objectives

Turn electromagnetic performance into an objective

Optimization requires a numerical measure of how good a candidate design is. The framework therefore keeps objective functions separate from geometry and solver code.

Keeping objectives modular means that a design study can be changed from matching optimization to another engineering criterion without rebuilding the geometry or solver workflow.

07 · Verification & Analysis

Do not stop at the optimizer

Optimization is a search process, not a substitute for engineering verification. The selected design should be inspected and then simulated with higher-accuracy settings before its results are used for further engineering decisions.

The post-processing workflow extracts S11, input impedance and VSWR, and can calculate far-field patterns, directivity, realized gain and efficiency. The resulting plots provide a way to inspect the electromagnetic behavior rather than relying only on a single objective value.

Recommended learning sequence

1. Inspect the default parametric geometry.
2. Change one design variable and observe the generated geometry.
3. Run a single simulation and inspect its electromagnetic response.
4. Study the objective function and understand what it rewards.
5. Run a small optimization experiment.
6. Inspect the selected parameters and repeat the simulation at higher accuracy.
7. Compare the optimized design against the original design.

09 · Questions & Feedback

Questions or feedback?

If you have a question, encounter an issue, or have a suggestion regarding the ETROYL framework, feel free to contact ETROYL. For questions about the openEMS solver itself, please use the official openEMS project resources.

Contact ETROYL →