Design optimization of electric-fish-inspired power sources using statistical and machine learning approaches

Hydrogel-based power sources inspired by electric fish offer a biocompatible solution for flexible electronics, but suffer from low energy density, restricted power output, and poor cycling stability. To overcome these challenges, we integrate a validated physics-based model with data-driven tools to map and optimize performance across a broad design space. A parametric sweep of experimentally tunable features was followed by principal component analysis (PCA) to identify dominant trends, and Gaussian process regression (GPR) to accurately predict performance metrics. The GPR model captured nonlinear relationships and enabled inverse design, identifying input configurations that meet user-defined performance targets without extensive trial-and-error experimentation. A Mann–Whitney U[jls-end-space/]-test revealed statistically significant thresholds across key input–output combinations, defining practical design ranges to improve energy density, power output, and cycling stability. Experimental validation confirmed model predictions, supporting the identified design thresholds. Our integrated framework unites simulation, statistical approaches, machine learning, and optimization to extract actionable design rules and deepen mechanistic understanding. As we demonstrate the value of multivariate analysis for understanding tradeoffs and tailoring device behavior, our findings reveal that a small subset of parameters primarily control overall performance. This work demonstrates the power of physics-informed, data-driven tools to accelerate the development of soft-material power sources.