Machine learning based models to investigate the thermoelectric performance of carbon nanotube-polyaniline nanocomposites

Thermoelectric materials have been widely recognized as a simple approach for harnessing green energy by converting thermal gradients into electrical energy. However, the intricate and conflicting interplay between electrical conductivity (σ), Seebeck coefficient (S), and thermal conductivity (k) in known materials present a challenge for improving their thermoelectric conversion efficiency. To overcome this challenge, various data-driven machine-learning techniques such as correlation matrix (CM), multiple linear regression (MLR), polynomial regression (PR), principal component analysis (PCA), and artificial neural network (ANN) have been utilized to identify the impact of different structural and compositional factors on the thermoelectric performance in carbon nanotube (CNT)-polyaniline (PANI) based nanocomposites. Our findings suggest that the thermoelectric figure of merit, (ZT=S2σκT) of these nanocomposites can be positively influenced by proper choice of doping agent of PANI and by using SWCNT, since these two parameters influence the thermoelectric outputs in the desired manner. The other input variables have conflicting impacts on the thermoelectric performance, although optimal solutions for maximized performance can be extracted. The investigation provides valuable insights about designing a PANI-CNT nanocomposite system with superior thermoelectric performance.