Multi-Objective Optimization Using Deep Neural Network and Grey Relational Analysis for Optimal Lay-Up of CFRP Structure

Abstract

This paper proposes a multi-objective optimization method that integrates deep neural networks (DNN) with gray relational analysis (GRA) to optimize lay-up configurations for carbon fiber-reinforced plastic (CFRP) automotive components. Specifically, a lab-scale CFRP B-pillar structure was investigated to simultaneously maximize structural strength and failure safety. A DNN surrogate model was trained using finite element simulations of 2000 random stacking sequences to achieve high predictive accuracy. The trained model was then used to evaluate all possible lay-up combinations to derive Pareto optimal solutions. Gray relational analysis was subsequently employed to select the final optimal configurations based on designer preferences. The selected lay-up designs demonstrated improvements in both strength and failure safety. To validate the proposed framework, laboratory-scale CFRP B-pillar was fabricated using a prepreg compression molding process and subjected to bending tests. The experimental results confirmed an error below 5% and failure trends consistent with the simulation results, thereby validating the reliability of the proposed method. The proposed DNN-GRA approach enables efficient multi-objective optimization with reduced computational effort and flexibility in reflecting different engineering priorities.