A data-driven physics-informed neural network for predicting the viscosity of nanofluids

Nanofluids have been applied in various fields, such as solar collectors, petroleum engineering, and chemical due to their superior properties compared traditional fluids. Among the thermophysical of nanofluids, viscosity plays a critical role thermal applications involving heat transfer fluid flow. While several conventional machine learning (ML) techniques proposed predict viscosity, these models require many experimental measurements be optimized make accurate predictions. This study reports novel ML method using multi-fidelity neural network (MFNN) accurately nanofluids by incorporating physical laws into model. The MFNN correlates low-fidelity dataset derived from prediction theoretical model with high-fidelity dataset, which consists measurements. It is shown that can recover rheology outperforms artificial underlying physics