A supervised machine learning approach for accelerating the design of particulate composites: Application to thermal conductivity

A supervised machine learning (ML) based computational methodology for the design of particulate multifunctional composite materials with desired thermal conductivity (TC) is presented. The variables are physical descriptors material microstructure that directly link to material’s properties. sufficiently large and uniformly sampled database was generated on Sobol sequence. Microstructures were realized using an efficient dense packing algorithm, TCs obtained our previously developed Fast Fourier Transform (FFT) homogenization method. Our optimized ML method trained over establishes complex relationship between structure Finally, application model in inverse a new class materials, liquid metal (LM) elastomer, TC discussed. results show surrogate accurate predicting behavior respect high-fidelity FFT simulations, robust finding parameters according case studies.