A Hybrid Genetic Algorithm–Neural Network Framework for Optimizing Thermal and Mechanical Properties of 3D-Printed Components: An Empirical Study

Abstract

Additive manufacturing, particularly fused deposition modeling (FDM), enables the fabrication of complex polymer components; however, achieving simultaneous optimization of mechanical and thermal performance remains challenging due to the nonlinear and interdependent nature of process parameters. This study presents an empirical hybrid optimization framework integrating an Artificial Neural Network (ANN) with a Genetic Algorithm (GA) to optimize the mechanical and thermal properties of FDM-fabricated components. A quantitative experimental design was employed, generating an empirical dataset from 120 printed specimens produced across 40 experimental runs with systematic variation of layer thickness (0.12–0.28 mm), raster angle (0–90°), infill density (25–100%), print speed (35–75 mm/s), and extrusion temperature (195–235 °C). Mechanical testing results showed tensile strength values ranging from 43.7 ± 2.9 MPa to 52.8 ± 3.1 MPa across build orientations, flexural strength from 76.8 ± 4.7 MPa to 88.2 ± 4.5 MPa, and elastic modulus from 2.05 ± 0.07 GPa to 2.42 ± 0.08 GPa. Thermal measurements indicated thermal conductivity values between 0.26 ± 0.02 W/m·K and 0.31 ± 0.02 W/m·K, with warpage ranging from 0.42 ± 0.06 mm to 0.61 ± 0.07 mm. The ANN achieved high predictive accuracy, with R² values of 0.94 for mechanical properties and 0.92 for thermal properties. GA-based optimization identified parameter configurations that improved tensile strength by 8.4 MPa, flexural strength by 11.2 MPa, elastic modulus by 0.29 GPa, and thermal conductivity by 0.06 W/m·K, while reducing warpage by 0.17 mm and dimensional change by 0.28% relative to baseline conditions. Empirical validation confirmed close agreement between predicted and measured results, demonstrating the effectiveness of the hybrid GA–ANN framework for data-driven optimization of FDM process parameters.