Artificial neural network, Pareto optimization, and Taguchi analysis for the synthesis of single-walled carbon nanotubes

Abstract Floating-catalyst chemical vapour deposition (FC CVD) is a promising method for large-scale synthesis of single-walled carbon nanotube (SWCNT). The current work focuses on the comprehensive optimization studies with the objective of maximizing the yield and minimizing the diameter of SWCNTs considering the complex dependency and interaction effects between the process parameters. Data-analytics driven process parameter selection criterion was utilized for heat map generation of the eight process parameters. Orthogonal design of experiments encompassing Taguchi L-18 array and subsequent analysis of variance reveals strong interaction effects between the process parameters and demonstrates conflicting nature between the two objectives, yield and diameter of SWCNTs. Pareto optimization was used to capture the conflicting behaviour of the dual objectives in true sense. Artificial neural network (ANN) coupled with Taguchi analysis brings out the interaction effects of process parameters that could predict the output with more than 90% accuracy. The rigorous analysis concludes that the furnace temperature mostly affects the diameter, whereas, methane flow-rate determines the SWCNT yield. The optimum condition for higher SWCNT yield with minimum diameter was obtained at a lower methane and hydrogen flow while keeping furnace temperature and argon flow at the highest level and thiophene at the lowest level.