Neural Network-Based Surrogate Modeling for Buckling Performance Optimization of Lightweight-Composite Collapsible...

Area of Science:

• Aerospace Engineering
• Structural Mechanics
• Computational Science

Background:

• Collapsible tubular masts (CTMs) are crucial structural elements in space applications.
• Ultra-thin CTMs are susceptible to localized buckling under axial and bending loads, challenging traditional design.
• Predicting buckling behavior is complex due to the intricate relationship between cross-section geometry and failure modes.

Purpose of the Study:

• To develop accurate and computationally efficient surrogate models for predicting the nonlinear buckling behavior of CTMs.
• To enable multi-objective optimization (MOO) of CTM cross-sections for improved stability and load-bearing capacity.
• To accelerate the design cycle for ultra-thin CTMs by reducing analysis time.

Main Methods:

• Utilized finite element analysis (FEA) to generate buckling load data for CTMs.
• Developed neural network (NN)-based surrogate models trained on FEA data.
• Integrated NN surrogate models with the non-dominated sorting genetic algorithm II (NSGA-II) for MOO.

Main Results:

• NN surrogate models achieved high prediction accuracy for buckling loads (R2 values up to 0.9987).
• Prediction time was reduced from minutes (FEA) to fractions of a second, drastically improving computational efficiency.
• Identified 1000 non-dominated CTM cross-sectional configurations optimizing buckling performance through MOO.

Conclusions:

• NN-based surrogate models offer a highly accurate and efficient method for predicting CTM buckling behavior.
• The proposed approach facilitates the design of structurally optimized and stable ultra-thin CTMs.
• This methodology enables informed design decisions for advanced aerospace structures.