Table 5 Antenna III: modeling results and benchmarking.

From: Two-stage variable-fidelity modeling of antennas with domain confinement

Modeling method

Number of training samplese

50

100

200

400

800

Kriging

Modeling error

61.4%

50.7%

39.8%

32.8%

31.8%

Model setup costb

50

100

200

400

800

RBF

Modeling error

65.3%

51.8%

43.2%

37.1%

33.6%

Model setup costb

50

100

200

400

800

Nested kriging68

Modeling error

17.9%

13.3%

7.5%

5.4%

4.5%

Model setup costa

1949

1999

2099

2299

2699

No-reference-design modeling73

Modeling error

10.8%

8.4%

7.1%

5.9%

5.0%

Model setup costc

242

292

392

592

992

Two-stage variable-resolution surrogate (this work)

NBf = 50

Modeling error

8.7%

7.2%

6.7%

6.5%

6.0%

Model setup costd

118.5

131.9

158.5

211.9

318.5

NBf = 100

Modeling error

6.8%

6.1%

5.4%

5.3%

5.2%

Model setup costd

168.5

181.9

208.5

261.9

368.5

  1. aThe cost includes acquisition of the reference designs, which is 1899 EM high-fidelity simulations of the antenna.
  2. bThe cost refers to high-fidelity simulations.
  3. cThe cost includes generation of random trial points, here, 192 high-fidelity simulations in total to yield Nr = 50 accepted samples.
  4. dThe cost includes generation of random trial points at low-fidelity level, here, 207 simulations in total to yield Nr = 50 accepted samples.
  5. eIn the case of the proposed model, the numbers 50 through 800 refer to low-fidelity EM simulations.