Introducing an integrated self-organizing map and radial basis function network for accurate prediction of water- Fe3O4 nanofluid viscosity versus solid volume fraction and temperature

Abstract

Nanofluids, colloidal suspensions of nanoparticles in base fluids, promise improved thermal conductivity and energy efficiency. Predicting nanofluid viscosity, influenced by factors like temperature and nanoparticle concentration, is crucial for optimizing performance. This research introduces an innovative approach combining Self-Organizing Map (SOM) and Radial Basis Function (RBF) networks to accurately predict viscosity. The SOM guides RBF center placement for enhanced accuracy, validated with water-Fe₃O₄ nanofluid data. Temperature significantly affects viscosity, with adjustments from 10 to 50 °C showing viscosity decreases from around 5 to lower than 2 mPa.s. The integrated SOM-RBF model achieves high precision (max absolute error 0.1078 mPa.s, 4 % relative error), suggesting potential for further enhancement. This novel neural network combination enhances efficiency in nanofluid applications, advancing predictive modeling in nanofluid engineering for broader industrial and scientific innovation.