Abstract
Several applications and improvements of swarm intelligence based metaheuristics are presented. Τhe visited algorithms are Particle Swarm Optimization (PSO), Chaotic Accelerated Particle Swarm Optimization (CAPSO) and Brain Storm Optimization (BSO). These algorithms or their novel variants are utilized for the solution of realistic engineering optimization problems in the fields of Optics and Electromagnetics. A literature-based analysis of some fundamental terms and subjects is also provided. First, an electromagnetic cloaking problem is examined. The scattering cross section which describes a layered spherical medium is tackled by PSO and CAPSO. The design variables are the radii, permeabilities and permittivities describing the shells of metamaterials which are surrounding either a PEC or dielectric spherical core. PSO provided promising results. For CAPSO, the results showcase a variety of feasible structure designs with perfect or almost perfect cloaking behaviour for several set- ...
Several applications and improvements of swarm intelligence based metaheuristics are presented. Τhe visited algorithms are Particle Swarm Optimization (PSO), Chaotic Accelerated Particle Swarm Optimization (CAPSO) and Brain Storm Optimization (BSO). These algorithms or their novel variants are utilized for the solution of realistic engineering optimization problems in the fields of Optics and Electromagnetics. A literature-based analysis of some fundamental terms and subjects is also provided. First, an electromagnetic cloaking problem is examined. The scattering cross section which describes a layered spherical medium is tackled by PSO and CAPSO. The design variables are the radii, permeabilities and permittivities describing the shells of metamaterials which are surrounding either a PEC or dielectric spherical core. PSO provided promising results. For CAPSO, the results showcase a variety of feasible structure designs with perfect or almost perfect cloaking behaviour for several set-ups. Both algorithms are described in detail, while the results are organized and visualized thoroughly. A nanorod configuration is then examined regarding its design as a polarization switch. The structure is described geometrically by five layers of two alternating materials constituting the nanorod, which is excited by a line source. The design variables are the layers’ thicknesses. The optimization goals concern the power absorption regarding the transverse electric/magnetic fields. The desired effect is polarization switching for which a suitable metric was devised. For each alternative design, a plasmonic and a dielectric material were chosen. The line source’s wavelengths were chosen in the visible spectrum. A novel CAPSO variant which employs Opposition-Based Learning (OBL) was developed. OB-CAPSO utilizes its populations’ opposites for solution improvement. It does so via opposition-based initialization and opposition-based generation jumping. Numerous feasible structure designs were obtained, and the most effective pairs of materials were highlighted. The new algorithm is thoroughly presented, accompanied by numerical results and visualizations. Source placement analysis and robustness tests are examined. The obtained nanorod designs can be used as components in polarization-controlled photonic integrated systems for various applications. Finally, a novel hybrid algorithm is presented. The BSO and CAPSO hybrid, BSO-CAPSO, is devised to exploit both the algorithms’ strengths. According to literature, BSO has shown greater global exploration capabilities if randomly initialized, compared to PSO which seems to locally exploit more accurately when initialized with a predefined seed. For BSO-CAPSO, the algorithm initially runs as BSO with random initialization, and then continues as CAPSO which is familial to PSO. Thus, BSO’s population is utilized as CAPSO’s initial population. CAPSO is computationally light and designed with quick convergence in mind. This is considered beneficial since the hybrid algorithm runs as CAPSO for most of its iterations. BSO-CAPSO was tested via benchmarking functions against its parent algorithms. These functions vary in complexity, while they were also examined for different numbers of dimensions. The hybrid showcases improved behaviours. Its most important characteristic is the ability to discover a promising, high quality region in the search space early in the optimization process, making it a promising tool for engineering optimization or further hybridization. Parameter selection guidelines are also provided.
show more
