Experimental study and gradient-based ensemble intelligent computing to investigate effect of ultrasound on rheological behavior of bio-based phase change materials

Abstract

The influence of ultrasound on the rheological behavior of new phase change material (PCM) is investigated. The PCMs are made from natural biological materials containing silicon carbide (SiC) nanoparticles or expanded graphite (EG) powders with weight fractions of 0, 0.05, 0.1, and 0.2 %. The rheological behavior of PCM composites is tested by a rheometer with shear rates of 10 up to 250 rpm. Surveys showed the smallest PCMs' viscosity at shear rates over 100 rpm is obtained for the highest sonication time. Results depict an ignorable change in the dynamic viscosity of PCMs with shear rate, and the shear thinning behavior reduces the PCMs' viscosity by raising the shear rate by 10–140 rpm. Results for EG/PCM and SiC/PCM composites showed non-Newtonian behavior at low shear rates. However, steadiness in viscosity at a shear rate higher than 140 rpm is observed for the Newtonian behavior of PCMs. The other outstanding consequence of this study is the development of novel robust machine learning, namely CatBoost, to simulate the rheological behavior of understudy nano-PCMs based on the weight fraction, sonication time, and shear rate. The multigene genetic programming (MGGP) as a comparative model is adopted to validate the primary model and provide a relationship for estimating the viscosity of each nano-PCMs. Simulation results revealed the superiority of the CatBoost (PCM-EG: R = 0.978 and RMSE = 1.739 mPa·s; PCM-SiC: R = 0.996 and RMSE = 0.374 mPa·s) to the MGGP (PCM-EG: R = 0.967 and RMSE = 2.103 mPa·s; PCM-SiC: R = 0.992 and RMSE = 2.368) in PCM composites.