Investigating horn power and impact of sonication on TiO2@cotton composites with machine learning and computer vision

Abstract

An environmentally friendly sonication method is used to fabricate TiO₂@cotton composites. The process involves using the potential of high-frequency ultrasonic waves to effectively break down the agglomerates of nanoparticles, leading to improved interaction between fabric surface and nanoparticles. Hence, an effective and accurate prediction of sonication parameters (horn power, sonication time, sonication cycle) is of paramount importance for tailoring composite design, structure, and properties. This work is the first attempt to utilise machine learning models to determine particle size, nanoparticles loaded amount on the surface of a substrate, and nanoparticles dispersion index (measure of evenness). A neural network model is implemented to interpret a non-linear complex relationship between input conditions and output results. The prediction of particle size, loaded amount, and particles dispersion is verified by comparing predicted and experimental results, for proposed model’s effectiveness. The relationship between independent and dependent variables is reliably captured. It can be revealed from the results of the proposed models that nanoparticles size, loaded amount on the substrate and nanoparticles dispersion on cotton surface are significantly dependent on sonication attributes.