A prediction method for compressive strength of biomimetic lattice structures based on stacked ensemble learning
Abstract
With the advancement of metal additive manufacturing technology, bionic lattice structures with high specific strength and excellent energy absorption characteristics have been realized, which have broad application prospects in aerospace and biomedical fields. However, the complex nonlinear mapping relationship between geometric parameters and mechanical properties poses significant challenges to traditional empirical formulas and high-fidelity finite element analysis. To address the challenge of predicting the compressive strength of bionic structures under small sample conditions, this study proposes a stacked ensemble learning framework. Based on 81 sets of full-factor simulation data, this framework integrates ridge regression, random forest, and XGBoost as base learners and achieves the optimal fusion of prediction results through a meta learner. Further, the SHAP analysis quantitatively decouples the independent influence mechanisms of each design variable. The framework proposed in this paper provides a high precision and interpretable surrogate model for the mechanical property prediction of bionic lattice structures, which can effectively support subsequent structural optimization design.