Methodology

What this research is trying to solve.

CPW-fed antenna design normally depends on slow simulation sweeps. This project builds a reproducible shortcut: train models from CST exports, use them for fast design exploration, and verify inverse suggestions with a forward model.

01

CST exports

Raw curves from 12 CPW antenna templates.

02

Normalization

S11, gain, efficiency, and sweep metadata become one master table.

03

Forward models

Geometry predicts scalar metrics and full S11 signatures.

04

Inverse tandem

Target behavior proposes antenna ID, Lp, and Wp.

05

Verification

Forward model checks the synthesized design against the requested target.

ProblemExploring every antenna geometry in CST is expensive and hard to repeat.The sweep has thousands of runs across Lp and Wp, with full S11 curves and scalar metrics for each design.
ApproachNormalize CST output into a master dataset and train forward and inverse models.Forward models predict resonance, bandwidth, gain, efficiency, S11 minimum, and S11 signatures from geometry.
ResultA user can synthesize a candidate antenna from RF targets and inspect verification errors.The inverse output is not treated as truth; it is checked by the trained forward stack and shown with nearest examples.
Research contract
DatasetFinal CST-derived master dataset with 6,072 rows.
Geometry inputFamily, antenna ID, Lp, and Wp.
Forward outputScalar RF metrics plus a 151-point S11 response.
Inverse outputCandidate antenna ID, Lp, Wp, and forward verification.
UI roleMake the research traceable for a thesis reviewer and interactive for live demos.
Acceptance checklist
Final dataset loads in trained mode Forward lab renders 2D and 3D geometry before prediction Inverse lab reports requested vs verified errors Evaluation page exposes model and dataset evidence Light and dark themes remain readable