Modeling Particle Sensing and Sorting Using Bio-Inspired Cilia in Low Reynolds Number Flows

Biological cilia, found in many living organisms, perform diverse functions, ranging from sensory perception to actuation. In this study, a two-dimensional numerical model is proposed to mimic both functionalities of biological cilia in a flow system. Particles transported in a channel are detected by cilia placed along the walls, which sense their characteristics through deformation during transit. Based on this detection, particles are sorted into either half of the channel by motile cilia actuated at the channel ends. The adapted framework uses Kirchoff rod theory to represent cilia, while particles are modeled as neo-Hookean massless solids and solved using the Finite Element Method. Coupling between the cilia, particles, and the surrounding fluid is achieved through the Immersed Boundary Method. It is demonstrated that, at low Reynolds numbers, a minimal number of cilia, characterized by their sperm number, can be strategically arranged to disrupt the motion of elliptical particles. The particles, initially tending to align their semi-major axes with the mean flow, are continuously perturbed by the cilia, causing deformation and encoding size and shape information into the cilia. A particle sorting mechanism is also implemented, where motile cilia at the channel ends are actuated to rigid or deformable states based on predefined threshold criteria. This mechanism sorts particles into the top or bottom half of the channel based on their detected characteristics. Multiple simulations were conducted using particles that varied in both size and aspect ratio. For each particle type, additional simulations were run with different initial orientations and initial positions along the vertical to generate comprehensive training data. This data was then utilized to train a Recurrent Neural Network model. The trained model demonstrated reliable performance in predicting particle size and aspect ratio. These predictions, combined with user-defined threshold values, were employed to actuate cilia gates, enabling the sorting of particles as desired.