Bandpass Filter Design Using Deep Neural Network and Differential Evolution Algorithm
ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING, cilt.47, sa.11, ss.14343-14354, 2022 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 47 Sayı: 11
- Basım Tarihi: 2022
- Doi Numarası: 10.1007/s13369-022-06769-7
- Dergi Adı: ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Aerospace Database, Communication Abstracts, Metadex, Pollution Abstracts, zbMATH, Civil Engineering Abstracts
- Sayfa Sayıları: ss.14343-14354
- Anahtar Kelimeler: Microstrip band pass filter, Deep learning Surrogate modeling, Differential evolution, Structural parameter optimization, MICROWAVE, OPTIMIZATION
- Süleyman Demirel Üniversitesi Adresli: Evet
Özet
In this study, we have used a hybrid approach to design parallel-coupled microstrip bandpass filters. It can take a long time to design a parallel-coupled microstrip bandpass filter within the desired constraints. We developed a two-phase approach to achieve an efficient design process. We chose 3 GHz as the center frequency of the designed filter. The 3 GHz center frequency is a standard frequency used in radar, maritime, and radio navigation applications. To optimize the structural parameters, we first created the surrogate model of the filter with a deep neural network. For this, we created our dataset with the EM simulator using five different structural parameters. Our dataset consists of the simulation output of S-11 and S-21 values in the specified frequency range between 2.5 GHz and 3.5 GHz. After creating the surrogate model, we optimized the structural parameters using the differential evolution algorithm. We tested our method by designing filters with different structure parameters in the optimization phase. We optimized the structural parameters for different bandwidths. The simulation results show that our method is accurate and reliable.