Optimizing solid microneedle design: a comprehensive ML-augmented DOE approach

Abstract

Microneedles (MNs), that is, a matrix of micrometer-scale needles, have diverse applications in drug delivery, skincare therapy, and health monitoring. MNs offer a minimally invasive alternative to hypodermic needles, characterized by rapid and painless procedures, cost-effective fabrication methods, and reduced tissue damage. This study explores four MN designs, cone-shaped, tapered cone-shaped, pyramidal with a square base, and pyramidal with a triangular-shaped base, and their optimization based on predefined criteria. The workflow encompasses three loading conditions: compressive load during insertion, critical buckling load, and bending loading resulting from incorrect insertion. Geometric parameters such as base radius/width, tip radius/width, height, and tapered angle tip influence the output criteria, namely, total deformation, critical buckling loads, factor of safety (FOS), and bending stress. The comprehensive framework employing a design of experiment approach within the ANSYS workbench toolbox establishes a mathematical model and a response surface fitting model. The resulting regression model, sensitivity chart, and response curve are used to create a multiobjective optimization problem that helps achieve an optimized MN geometrical design across the introduced four shapes, integrating machine learning (ML) techniques. This study contributes valuable insights into a potential ML-augmented optimization framework for MNs via needle designs to stay durable for various physiologically relevant conditions.