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

Area of Science:

• Materials Science
• Mechanical Engineering
• Computational Science

Background:

• Optimizing composite material lay-up is crucial for automotive components.
• Balancing structural strength and failure safety in carbon fiber-reinforced plastic (CFRP) parts presents a multi-objective challenge.
• Traditional optimization methods can be computationally intensive.

Purpose of the Study:

• To propose and validate a novel multi-objective optimization framework for CFRP automotive components.
• To integrate deep neural networks (DNN) and gray relational analysis (GRA) for efficient lay-up configuration optimization.
• To simultaneously enhance structural strength and failure safety of CFRP components.

Main Methods:

• Development of a DNN surrogate model trained on finite element simulations of numerous lay-up sequences.
• Utilization of the DNN model to identify Pareto optimal solutions across all possible lay-up combinations.
• Application of GRA to select optimal configurations based on designer preferences.
• Experimental validation using a fabricated CFRP B-pillar under bending tests.

Main Results:

• The DNN-GRA method achieved high predictive accuracy for lay-up configurations.
• Optimized lay-up designs showed significant improvements in both structural strength and failure safety.
• Experimental validation confirmed the simulation results with less than 5% error.
• The approach demonstrated reduced computational effort compared to traditional methods.

Conclusions:

• The proposed DNN-GRA framework offers an efficient and flexible approach for multi-objective optimization of CFRP components.
• This method effectively balances competing design objectives like strength and safety.
• The validated framework provides reliable predictions for composite material design in automotive applications.