Data and code for "Numerical Investigation of Bio-inspired Branching Flow Channels in AEM Electrolyzers Using Topology-Aware Graph Neural Networks and Multi-objective Optimization"

This repository contains the datasets and code supporting the manuscript "Numerical Investigation of Bio-inspired Branching Flow Channels in AEM Electrolyzers Using Topology-Aware Graph Neural Networks and Multi-objective Optimization" (Chen and Chen, National Chin-Yi University of Technology, Taiwan; funded by NSTC 114-2637-E-167-015). It includes the 3D COMSOL design-of-experiments dataset (74 cases, 51 converged), the TA-PIGNN and Optuna-tuned baseline training and evaluation scripts, SHAP feature-importance and GAT attention outputs, the NSGA-II Pareto front, and the COMSOL model-builder script for reproducibility.