DOI RECORD
Nature-inspired antenna design using hybrid ANN and Bayesian optimization techniques for biomedical applications
Abstract
Abstract This paper is presenting a systematic approach for designing of antenna resonating at 3.18 GHz with an impedance bandwidth of 3.10–3.285 GHz which lies at the lower edge of the UWB spectrum used in IEEE 802.15.6-based Wireless Body Area Network (WBAN) research. The proposed compact antenna becomes more suitable for WBAN-assisted healthcare monitoring and microwave biomedical sensing applications. The antenna designing is covered in two sections. First one deals with the non-linear regression analysis with correlation analysis to achieve final design for patch. The surrogate modelling for reflection coefficient parameter for complete 50 samples of the patch design is presented and compared with other techniques. The non-linear regression based surrogate modelling along with correlation analysis by Pearson and Spearman techniques helps in making prediction of the patch design structure by selecting most featured and significant design parameters. The second section caters the process of ‘Artificial Neural Network (ANN)’ modelling with optimization techniques to achieve optimized design of ‘Defected Ground Structure (DGS)’ structure. The optimization is performed by ‘Non-dominated Sorting Genetic Algorithm II (NSGA-II)’, Multi-Objective Particle Swarm Optimization (MOPSO)’, ‘Grey Wolf Optimization (GWO)’, and ‘Bayesian Optimization using Gaussian Process Regression (BO + GPR)’ techniques. The optimized antenna having size of $$0.2968 {\lambda}_{0} \text{m}\text{m} \times 0.318 {\lambda}_{0} \text{m}\text{m}$$ at resonant frequency of 3.18 GHz, is fabricated on FR-4 substrate. The excellent agreement is observed between simulated results and practical results which confirms the robustness of the proposed systematic approach for the antenna designing and making it suitable for biomedical application.