Effect of rod electrode shape on the performance of a plate and rod electrohydrodynamic pump – An optimization study using ANN

This paper presents a numerical investigation of a mini-channel electrohydrodynamic pump, utilizing an elliptical rod electrode positioned between two parallel plate electrodes. The study aims to enhance pumping performance through geometric optimization while maintaining a constant electrode surface area. A numerical model is developed to simulate electrohydrodynamic conduction pumping by solving the linearized Onsager model using COMSOL Multiphysics® software. In the first phase, baseline optimization for the channel height-to-rod size ratio is conducted for an equiaxial elliptical rod using response surface methodology. The optimal configuration predicted an average velocity of 16.246 mm/s, closely matching the simulation results with a deviation of <0.17%. In the second phase, the rod electrode's aspect ratio, eccentricities, and orientation angle are varied while maintaining a constant electrode surface area across all configurations. An artificial neural network trained on simulation data is coupled with the Bonobo metaheuristic algorithm to determine optimal design parameters. The global optimum is achieved at an aspect ratio of 1.13 and at zero-degree orientation, resulting in a maximum velocity of 16.621 mm/s, a 2.48% improvement over the baseline optimum electrode geometry. The results indicate that geometric refinements can lead to improved electrohydrodynamic pumping. The combined use of response surface methodology and artificial neural network-based metaheuristic optimization provides a systematic and effective framework for designing efficient electrohydrodynamic systems.