Neuro-intelligent performance prediction of fuzzy hybrid nanofluid heat transfer using Bayesian Regularization and Levenberg-Marquart

Fuzzy Set Theory (FST) offers a more practical framework for examining and simulating situations with ambiguous information because of the uncertainties present in real-life problems. The main goal of the present study is to create a model that uses triangular fuzzy numbers to analyze the impact of the fuzzy volume fraction of nanoparticles on the composite nanofluid flow. Fuzzy Set Theory is utilized to account for the ambiguity of the instances using a Williamson hybrid (Al 2 O 3 and Cu/EO) nanofluid. The findings reveal that adding nanoparticles of copper (Cu), and alumina (Al 2 O 3 ) significantly improves base fluid’s thermal conduction and stability. Employing the triangular fuzzy number (TFN), the percentage of nanomaterials affecting the thermophysical characteristics are considered as fuzzy parameters in order to compare the thermal profiles of composite nanofluids (Al 2 O 3 + Cu/EO) and nanofluids (Cu/EO or Al 2 O 3 /EO). According to the fuzzy logic study, hybrid nanofluids are more effective at transferring heat than nanofluids. In order to overcome the drawbacks of basic neural networks, including sluggish convergence, underfitting, and overfitting, a hybrid strategy that combines the Levenberg–Marquardt (LM) algorithm and Bayesian Regularization (BR) is used to analyze preliminary data and obtain key physical characteristics. An outstanding overlap across projected and targeted values demonstrates how this novel strategy greatly enhances the model’s extrapolation, improves training, and produces strong and trustworthy predictions even with little data. Neural networks and the fuzzy method together offer a foundation for better decision-making when it comes to nanofluid concentration and optimization.