Experimental Study on the Flow Characteristics of Oil-Based Carbon Nanotubes Nanofluids and Establishment of Viscosity Prediction Model Based on KAN.

Nanofluids, recognized as innovative heat transfer media, exhibit high thermal conductivity, which enhances heat transfer efficiency and reduces energy dissipation. Investigating the flow characteristics of the nanofluids is essential for optimizing system design, minimizing energy consumption, improving stability and expanding their applicability across various industrial sectors. However, research on the flow characteristics of nanofluids remains limited. In this paper, carbon nanotubes were dispersed in engine oil to formulate nanofluids, and their flow characteristics were experimentally examined. A key innovation is the development of a viscosity prediction model using Kolmogorov–Arnold Networks (KANs), which complements traditional prediction models. The experimental results revealed that the viscosity of the nanofluids decreases with increasing shear rate and temperature, while stability significantly declines with higher volume fractions. The KAN model, trained and tested with experimental data, demonstrated high accuracy in predicting viscosity. This research offers valuable insights into the flow behavior of nanofluids, paving the way for their effective implementation in advanced thermal management applications. In this paper, oil-based carbon nanotubes nanofluids were prepared and their flow characteristics were experimentally examined. A viscosity prediction model was developed using Kolmogorov-Arnold Networks (KAN). A portion of the experimental data was used to train the KAN model, with the remainder reserved for testing, demonstrating that the KAN-based viscosity prediction model achieves high accuracy. [ABSTRACT FROM AUTHOR]