A neuro-intelligent heuristic approach for performance prediction of triangular fuzzy flow system

Abstract

Due to uncertainties found in real-world problems, fuzzy set theory (FST) suggests a more realistic framework for analyzing and modeling scenarios with imprecise information. This study’s primary focus is to develop a model for analyzing the significance of the fuzzy volume fraction of nanomaterials on the hybrid nanofluid flow using triangular fuzzy numbers. The two nanomaterials Al₂O₃ and Cu are entrenched in engine oil (EO) host fluid. To account for the vagueness involved in the phenomenon, fuzzy set theory (FST) is employed. To build a comparison of thermal profiles for hybrid nanofluids (Al₂O₃ + Cu/EO) and nanofluids (Cu/EO), (Al₂O₃/EO), the concentration of nanoparticles influencing the thermophysical properties will be taken as an uncertain parameter using triangular fuzzy number (TFN) [0, 0.1, 0.4]. The fuzzy logic analysis points out that hybrid nanofluids Al₂O₃ are more operational than nanofluids (Cu/EO) (Al₂O₃/EO) for heat transfer. To address the limitations of simple neural networks, such as overfitting, underfitting, and slow convergence, a hybrid approach combining Bayesian Regularization (BR) and the Levenberg–Marquardt (LM) algorithm implemented to examine initial data and extract main physical properties. This innovative approach has significantly improved the model generalization, accelerated training, and provided robust and reliable predictions, even with limited data as confirmed by an exceptional match between projected and targeted values. A comparison analysis indicates that the magnetic parameter effect lessens the Cu/EO nanofluid’s velocity more actively than hybrid nanofluid Al₂O₃+Cu/EO and raises the temperature of hybrid nanofluid more readily then nanofluid.