Researchers from Pusan National University Develop a Rapid Yet Accurate Approach for Composite Material Homogenization
BY Composights
Published: 30 Apr 2025
South Korea-based Pusan National University has
developed a novel approach to address the challenges of predicting composite
material performance, which traditionally relies on time-consuming experimental
testing or computationally expensive numerical analysis.
Predicting the performance of a composite material in
real-world conditions can be challenging. Engineers usually rely on
experimental testing or numerical analysis to predict homogenized properties
like thermal conductivity and elasticity, but these approaches can be
time-consuming or computationally expensive.
Composite materials, engineered by combining distinct
components like fibers, and matrices, require accurate prediction of
homogenized properties such as thermal conductivity and elasticity.
Conventional numerical homogenization approximates these macroscopic properties
by solving microscale partial differential equations (PDEs), but this process
often demands significant computational resources.
The numerical homogenization process approximates composite material properties at a macroscopic scale by solving partial differential equations (PDEs) at a microscopic scale. The team, led by Professor Kyunghoon Lee from Pusan National University, proposed a Reduced Basis Homogenization Method (RBHM) to speed up numerical homogenization. RBHM decreases the computational cost by performing numerical homogenization on a reduced basis space and allows one to easily change fiber and matrix materials to obtain desired composite material properties. Their work was published online in the International Journal of Mechanical Sciences on November 15, 2024.
.jpg)
“Through the use of the RBHM, we can rapidly yet accurately generate the solutions of microscale PDEs. We then use these solutions to quickly evaluate the homogenized properties of a periodic composite material as we try various combinations of fiber and matrix materials,” explains Professor Lee. RBHM significantly expedites numerical homogenization, particularly for parametric analyses requiring multiple simulations, while providing prompt evaluations of macroscopic properties with minimal error.
RBHM achieved computational speeds up to 1,030 times faster than the FEM for evaluating thermal properties and 670 times faster for elastic properties, while maintaining accuracy comparable to that of the FEM. RBHM predictions matched not only the FEM predictions but also the experimental data, providing confidence in the method’s reliability. For instance, RBHM produced homogenized thermal conductivity and Young’s modulus with errors of less than 5% and less than 3%, respectively, compared with experimental results.
“Our method also allows for easy adjustments to fiber and matrix properties, enabling engineers to swiftly explore and test new fiber and matrix combinations,” adds Professor Lee. This feature is crucial for industries where the virtual testing and design of a composite material is needed.
RBHM can reduce overall computation time by up to 70%, which not only improves efficiency but also ensures scalability for large-scale industrial applications. Looking to the future, the researchers aim to expand RBHM to handle even more complex materials and use cases, including nonlinear elastic behavior and thermoelastic coupling, further broadening its applications in cutting-edge material science.
Source - Pusan National University
